Seemings and Epistemic Justification: How Appearances Justify Beliefs (SpringerBriefs in Philosophy) 3030433919, 9783030433918

Citation preview

SPRINGER BRIEFS IN PHILOSOPHY

Luca Moretti

Seemings and Epistemic Justification How Appearances Justify Beliefs 123

SpringerBriefs in Philosophy

SpringerBriefs present concise summaries of cutting-edge research and practical applications across a wide spectrum of fields. Featuring compact volumes of 50 to 125 pages, the series covers a range of content from professional to academic. Typical topics might include: • A timely report of state-of-the art analytical techniques • A bridge between new research results, as published in journal articles, and a contextual literature review • A snapshot of a hot or emerging topic • An in-depth case study or clinical example • A presentation of core concepts that students must understand in order to make independent contributions SpringerBriefs in Philosophy cover a broad range of philosophical fields including: Philosophy of Science, Logic, Non-Western Thinking and Western Philosophy. We also consider biographies, full or partial, of key thinkers and pioneers. SpringerBriefs are characterized by fast, global electronic dissemination, standard publishing contracts, standardized manuscript preparation and formatting guidelines, and expedited production schedules. Both solicited and unsolicited manuscripts are considered for publication in the SpringerBriefs in Philosophy series. Potential authors are warmly invited to complete and submit the Briefs Author Proposal form. All projects will be submitted to editorial review by external advisors. SpringerBriefs are characterized by expedited production schedules with the aim for publication 8 to 12 weeks after acceptance and fast, global electronic dissemination through our online platform SpringerLink. The standard concise author contracts guarantee that • an individual ISBN is assigned to each manuscript • each manuscript is copyrighted in the name of the author • the author retains the right to post the pre-publication version on his/her website or that of his/her institution.

More information about this series at http://www.springer.com/series/10082

Luca Moretti

Seemings and Epistemic Justification How Appearances Justify Beliefs

123

Luca Moretti Department of Philosophy University of Aberdeen Aberdeen, UK

ISSN 2211-4548 ISSN 2211-4556 (electronic) SpringerBriefs in Philosophy ISBN 978-3-030-43391-8 ISBN 978-3-030-43392-5 (eBook) https://doi.org/10.1007/978-3-030-43392-5 © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2020 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Acknowledgements

Portions of this book draw on previous published work of mine. Some parts of Chap. 2 are based on ‘Phenomenal conservatism’, Analysis (2015) 75, 296–309. Chapter 3 uses material of ‘Cognitive penetrability of perception and epistemic justification’, Internet Encyclopedia of Philosophy (2019), co-authored by Christos Georgakakis. Chapter 4 includes a revised version of ‘In defence of dogmatism’, Philosophical Studies (2015) 172, 261–282. Some parts of Chap. 5 draw from ‘Phenomenal conservatism and the problem of reflective awareness’, American Philosophical Quarterly (2018) 55, 265–280 and ‘Inferential seemings and the problem of reflective awareness’, Canadian Journal of philosophy (2019) 49, 253–271. Various colleagues have commented on drafts or presentations of parts of this manuscript at various stages of their development. I would like to thank particularly: Michael Bergmann, Francesco Berto, Berit Brogaard, Peter Brössel, Daniel Burnston, Catrin Campbell-Moore, Adam Carter, Lorenzo Casini, Michel Croce, Richard Dawid, Dylan Dodd, Anna-Maria Eder, Filippo Ferrari, Christos Georgakakis, Stephan Hartmann, Christopher Kelp, Federico Luzzi, Alessia Marabini, Anna Mathani, Kevin McCain, Matthew McGrath, Samir Okasha, Orestis Palermos, Jim Pryor, Thomas Raleigh, Martin Smith, Karim Thebault, Chris Tucker, Filippo Vindrola, Crispin Wright, Elia Zardini and anonymous reviewers of American Philosophical Quarterly, Canadian Journal of Philosophy, Philosophical Studies, Internet Encyclopedia of Philosophy and SpringerBrief. A special thanks to Tommaso Piazza for his invaluable feedback over the years about many of the theses defended in this work. Finally, I’m in debt to Christos Georgakakis and Alessia Marabini for proofreading the manuscript. I’m also grateful to the (no longer operative) Northern Institute of Philosophy of the University of Aberdeen, the Tilburg Center for Logic and Philosophy of Science, the Munich Center for Mathematical Philosophy, the Eidyn Research Centre of the University of Edinburgh, and the Emmy Noether Research Group ‘From Perception to Belief and Back to Perception’ of the Ruhr University of Bochum for hosting me and providing a stimulating atmosphere to conduct various stages of this investigation.

v

vi

Acknowledgements

The research leading to this book was supported by a Visiting Fellowship from the Tilburg Centre for Logic and Philosophy of Science, two Visiting Fellowships from the Munich Center for Mathematical Philosophy, a Carnegie Grant from the Carnegie Trust for the Universities of Scotland, a Supporting Grant from the University of Aberdeen, and a Visiting Fellowship from the Emmy Noether Research Group ‘From Perception to Belief and Back to Perception’ of the Ruhr University of Bochum.

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 What I do in this Book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 1 3

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

5 5 8 12 16 18 18

3 Cognitive Penetrability . . . . . . . . . . . . . . . . . . . . . . . 3.1 Characterizing Cognitive Penetrability . . . . . . . . . 3.2 The Epistemic Problem of Cognitive Penetrability 3.3 The Reliabilist Account . . . . . . . . . . . . . . . . . . . 3.4 The Inferentialist Account . . . . . . . . . . . . . . . . . . 3.5 Taming Cognitive Penetrability . . . . . . . . . . . . . . 3.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

23 23 27 29 31 46 54 54

4 The Bayesian Objection . . . . . . . . . . . . . . . . . . . . . . . 4.1 White’s Objection . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Responding to White’s Objection . . . . . . . . . . . . . 4.3 Perceptual Appearances and Reflective Justification 4.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

57 57 60 65 68 69

2 Phenomenal Conservatism . . . . . . . . . . . . . . 2.1 The Basics of Phenomenal Conservatism . 2.2 The Nature of Seemings . . . . . . . . . . . . . 2.3 Supporting Phenomenal Conservatism . . . 2.4 Criticism of Phenomenal Conservatism . . 2.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

vii

viii

Contents

. . . .

71 71 77 80

....... ....... .......

85 87 87

6 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

89 91

5 Antiscepticism and Easy Justification . . . . . . . . . . . . . . . . . 5.1 Antiscepticism and Reflective Awareness . . . . . . . . . . . . 5.2 Easy Justification Objections . . . . . . . . . . . . . . . . . . . . . 5.3 Answering the Easy Justification from Closure Objection 5.4 Answering the Easy Justification from Bootstrapping Objection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

Chapter 1

Introduction

Abstract In this introduction I present the topic of the investigation carried out in this book and the central theses defended in it. I also clarify some assumptions of my research, specify the intended audience of this book and summarize its structure. Keywords Phenomenal conservatism · Phenomenal dogmatism · Seemings · Appearances · Epistemic justification · Epistemic internalism and externalism

1.1 What I do in this Book This work aims to introduce, dissect and evaluate the important but controversial epistemological position called phenomenal conservatism. Variants or special versions of this position have been lingering in the views of various philosophers for the last forty years, until Michael Huemer (2001) introduced phenomenal conservatism officially into epistemology, using this name.1 Ever since, the popularity of this position has constantly increased. Phenomenal conservatism maintains that our appearances or seemings—the ways things appear to us to be—have the inherent power to justify our beliefs. According to the phenomenal conservative, for example, if it appears to you, say, that it is raining outside, that 15 – 7 = 8, or that Hitler was a wicked man, you thereby have a good reason to believe these things. This reason or justification is nevertheless defeasible—it can be destroyed by further evidence indicating that the appearance is unreliable or the belief false. Appearances are conceived of by the phenomenal conservative as experiences—so, not as beliefs or other doxastic states—provided with propositional content. Phenomenal conservatism looks very natural and comes with a number of apparent epistemological benefits. For instance, it supplies a clear account of where our justification—perhaps all our epistemic justification—basically comes from: it originates from our seemings. It also illuminates the rationality of ordinary epistemic practices in which we take ourselves or others to have reasons to entertain beliefs 1 James Pryor (2000, 2004) has simultaneously introduced a very similar view, though less general,

called dogmatism. © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2020 L. Moretti, Seemings and Epistemic Justification, SpringerBriefs in Philosophy, https://doi.org/10.1007/978-3-030-43392-5_1

1

2

1 Introduction

just because of how things appear to be. Phenomenal conservatism also shields us from sceptical threats, for we don’t need antecedent guarantee that our appearances are reliable to get justification from them. As other interesting philosophical views, phenomenal conservatism has been praised for its merits but also targeted with various objections. My research aims to show that phenomenal conservatives can dismiss some of the most worrisome challenges raised against their view. In particular, I will argue that if seemings were penetrated (i.e. partly caused) by other cognitive states of the subject, they would not lose their inherent justifying power. So, against the claims of certain epistemologists, the possibility of cognitively penetrated appearances is not a threat to phenomenal conservatism. Furthermore, I will show that it hasn’t actually been proven that phenomenal conservatism clashes with Bayesian methodology. Hence, in spite of what various philosophers think, phenomenal conservatism isn’t objectionable in this sense. I will also show that phenomenal conservatism doesn’t sanction suspicious procedures that appear to produce justification in an excessively easy way. Thus, in contrast to an apparently forceful and recurring criticism, phenomenal conservatism isn’t problematic in this sense either. I will nevertheless contend that phenomenal conservatism has an important limit: seeming-based justification is elusive: it fades away when the subject becomes reflectively aware of the relevant seeming. I will describe some ways in which this fact limits the actual explanatory role of phenomenal conservatism and its antisceptical bite. Phenomenal conservatism could virtually be connected to indefinitely many issues and debates in philosophy. In this short monograph I have selected only some of the issues discussed in current literature, and I have introduced some novel questions. In the final part of the book, I suggest further areas of investigation that scholars interested in phenomenal conservatism might want to explore. As other philosophical investigations, my work rests on some assumptions. One of the most crucial is this2 : I assume that justification is an internalist notion. In other words, I assume that when a subject S has justification for entertaining some propositional attitude, what produces this justification is a factor reflectively accessible to S, or a mental state of S.3 There may be various reasons to endorse internalism about justification. My view is that the new evil demon argument (Lehrer and Cohen 1983; Cohen 1984) gives it a strong support.4 Although I will generally introduce and clarify the philosophical notions that I use in my analyses and arguments, the discussion carried out in the next chapters will typically be rather “technical” and so quite advanced. Accordingly, this book is 2 These

are other assumptions of my research: I work with an invariantist, non-relativist and non-pragmatically encroached notion of epistemic justification (although these assumptions might ultimately not be necessary to the soundness of my arguments). 3 Although I’m very sympathetic to accessibilism, I prefer to leave open the possibility that internalism could find its best characterization when interpreted as mentalism. (For a characterization of accessibilism and mentalism see Chap. 2.). 4 Littlejohn (forthcoming) and Williamson (forthcoming) have challenged this argument. See however Madison (2017)’s rejoinder.

1.1 What I do in this Book

3

suitable for an audience of postgraduate students and scholars of analytic philosophy who have already a background in epistemology and want to be introduced to phenomenal conservatism and/or intend to go deeper into some of its more or less problematic features and implications. This is the structure of the book. In Chap. 2 I present phenomenal conservatism and the notion of seeming or appearance. I also review asserted epistemic merits of phenomenal conservatism and some preliminary difficulties of it. In Chap. 3 I discuss the conjecture that appearances are cognitively penetrable and evaluate and reject a number of objections to phenomenal conservatism hinging on this conjecture. In Chap. 4 I criticize and reject a celebrated argument to the effect that phenomenal conservatism is incompatible with Bayesianism. I also contend that seeming-based justification is elusive in the way described before. In Chap. 5 I argue that since seeming-based justification is elusive, the antisceptical bite of phenomenal conservatism is limited but phenomenal conservatism isn’t actually affected by easy justification problems. In Chap. 6, I draw the conclusions of my work.

References Cohen S (1984) Justification and truth. Philos Stud 46:279–295 Huemer M (2001) Skepticism and the veil of perception. Rowman and Littlefield, Lanham, MD Lehrer K, Cohen S (1983) Justification, truth, and coherence. Synthese 55:191–207 Littlejohn C (forthcoming) A plea for epistemic excuses. In: Dutant J, Dorsch F (eds) The new evil demon. Oxford University Press, Oxford Madison B (2017) On justifications and excuses. Synthese 195:4551–4562 Pryor J (2000) The skeptic and the dogmatist. Nous 34:517–549 Pryor J (2004) What’s wrong with Moore’s argument? Philos Issue, Epistemology 14:349–378 Williamson T (forthcoming) Justifications, excuses, and sceptical scenarios. In: Dutant J, Dorsch F (eds) The new evil demon. Oxford University Press, Oxford

Chapter 2

Phenomenal Conservatism

Abstract In this chapter I introduce and analyse the tenets of phenomenal conservatism, and discuss the problem of the nature of appearances. After that, I review the asserted epistemic merits of phenomenal conservatism and the principal arguments adduced in support of it. Finally, I survey objections to phenomenal conservatism and responses by its advocates. Some of these objections will be scrutinized and appraised in the next chapters. Keywords Phenomenal conservatism · Epistemic justification · Epistemic internalism · Seemings · Appearances · Phenomenal force · Rational commitment · Huemer’s self-defeat argument · Meta-justification

2.1 The Basics of Phenomenal Conservatism Phenomenal conservatism (also called dogmatism)1 holds that appearances or seemings—namely, ways things appear to be—are sources of defeasible justification.2 According to phenomenal conservatism, for instance, you can have defeasible justification for believing that the sun is rising simply because it seems visually so to you, or you can have defeasible justification for believing that 5 + 7 = 12 because this appears to be a priori true to you.3 A key intuition of the phenomenal conservative is that one should grant that things are the way they appear to be unless one has reason to doubt it.

1 See

for instance Tucker (2013). introductions to phenomenal conservatism are Tucker (2013) and Huemer (2019). 3 Phenomenal conservatism must not be confused with doxastic or epistemic conservatism, which says that one’s mere believing P gives one some defeasible justification for continuing doing so. 2 Good

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2020 L. Moretti, Seemings and Epistemic Justification, SpringerBriefs in Philosophy, https://doi.org/10.1007/978-3-030-43392-5_2

5

6

2 Phenomenal conservatism

Phenomenal conservatism is chiefly associated with Michael Huemer’s work. Other authors have nevertheless proposed similar views, though less articulated or narrower in scope. (For instance, Chisholm 1989; Audi 1993; Tolhurst 1998; Swinburne 1998; Pollock and Cruz 1999; Bealer 2000; Pryor 2000, 2004; Chudnoff 2011). Huemer (2007) states the tenet of phenomenal conservatism as follows: (pc) If it seems to [a subject] S that P, then, in the absence of defeaters S thereby has some degree of justification for believing that P. (30)

Huemer (2001) defends a stronger schema: (PC) If it seems to S that P, in the absence of defeaters S thereby has justification sufficient to believe P. (Cf. 99)

Huemer drops (PC) because he finds it too general: it is implausible that a hazy, wavering or fleeting seeming that P could give S justification sufficient to believe P (cf. 2007: 30n1). Huemer seems to concede, however, that if S’s seeming that P is clear and firm, in the absence of defeaters, S has justification sufficient to believe P. Throughout this book I concentrate on (PC), rather than (pc), on the assumption that S’s seeming that P is clear and firm. (PC) so interpreted expresses the view actually at stake in most discussions on phenomenal conservatism. Hereafter, whenever I speak of a seeming or appearance, I always mean clear and firm seeming. Let’s dissect (PC). ‘It seems’ in (PC) must not be read in a way that expresses, for instance, a doubt (‘it seems that P but I’m not so sure’), deception (‘it seems that P but I know it isn’t actually the case’), or a comparison (‘when your dog barks it seems to me he’s a wolf’ meaning that the dog resembles a wolf).4 The expression ‘it seems’ in (PC) is just meant to pick up a mental state of a general type, which we are supposed to be intimately familiar with, that comes about whenever things appear to us to be in a definite way. A mental state of this type is what phenomenal conservatives call seeming or appearance. ‘Justification’ in (PC) is intended as ultima facie (or all things considered) justification. (PC) can easily be re-phrased in terms of prima facie justification as follows: If it seems to S that P, S thereby has prima facie justification sufficient to believe P.

Although (PC) is about propositional justification, it can be re-formulated as follows to concern with doxastic justification: (PCD ) In the absence of defeaters, S’s belief that P is justified if it is based on S’s seeming that P. (Cf. Huemer 2019)5

4 In

the literature on appearances the semantic ambiguity of ‘appears’ was probably first noticed in Chisholm (1957). Huemer (2013) denies that ‘appears’ is semantically ambiguous. 5 S has propositional justification for P just in case S has a justifier of P whether or not S actually believes P on the basis of the justifier or at all. S has doxastic justification for P just in case S has propositional justification for P and bases her belief that P on the justifier of P (cf. Kvanvig 2003). On these standard characterizations, (PC) entails (PCD ).

2.1 The Basics of Phenomenal Conservatism

7

The ‘thereby’ in (PC) signals that S’s justification for P rests solely on S’s seeming that P. So, S’s justification for P isn’t based (even in part) on S’s justification for entertaining the seeming that P6 or any belief of S—for instance, the belief that S’s cognitive faculties are working properly, or that no sceptical conjecture is true. Since S’s justification for P rests solely on S’s seeming that P, it is both immediate (cf. Pryor 2005) and non-inferential (cf. Huemer 2019) justification. Various sorts of things can count as defeaters of S’s seeming-based justification for P. This justification can be destroyed or weakened by, for instance, S’s evidence in favour of ¬P,7 evidence that P’s truth is not ascertainable by S’s relevant faculty, evidence that this faculty is malfunctioning, and so on, where ‘evidence’ refers to another justified belief or a seeming of S (cf. Huemer 2006). In the next chapters, I will argue that S’s entertaining a reflective belief that she has a seeming that P has the effect to destroy S’s seeming-based justification for believing P. Internalism about epistemic justification is the view that justification is entirely determined by factors “internal” to the subject S. More precisely, accessibilism holds that the only factors relevant to justification are those that S can find to be present (or absent) by mere reflection. Whereas mentalism holds that justification supervenes on (non-factive) mental states of S. Externalism about epistemic justification, on the other hand, just denies both accessibilism and mentalism (cf. Pappas 2014). For example, according to an influential form of externalism called process reliabilism, what makes a belief of S justified is its being produced by a statistically reliable process (where the reliable process need not be reflectively accessible or mental). Phenomenal conservatism—at least the version defended by Huemer—is an internalist (both accessibilist and mentalist) view of justification.8 In fact (PC) fits with, though doesn’t entail, the assumption that S’s justification depends only on mental factors reflectively accessible to S—namely, S’s seemings and the absence of defeating evidence. Huemer (2006) defends internalism and argues that it is the position that better than any other accounts for the central internalist intuitions.

6 Supposing

there could be such a justification. Indeed, the claim that a seeming conceived as an experience (see below) can be epistemically justified appears incongruous (cf. Pryor 2005). 7 Where ‘¬’ is logical negation. 8 For externalists variants of phenomenal conservatism see Goldman (2008) and Bergmann (2013a).

8

2 Phenomenal conservatism

2.2 The Nature of Seemings Seemings are a genus that includes many species. (PC) or close variants have been invoked to account for the justification of beliefs of different types—for instance, perceptual (Pryor 2000; Huemer 2001; Tucker 2010), introspective (Huemer 2007; Audi 2013), mnemonic (Pollock and Cruz 1999; Audi 2013), a priori (Bealer 2000; Chudnoff 2011, 2013) and moral (Huemer 2005). These accounts assume that there exist perceptual, introspective, mnemonic, intellectual (or rational), and moral seemings.9 I gave examples of perceptual and intellectual seemings at the beginning of Sect. 2.1. These are examples of the remaining types: it seems to me that I’m cheerful (introspective); it seems to me to recall that Mary went to China (mnemonic); it seems to me that torturing animals for fun is abominable (moral). Philosophers have taken at least four different views about the general nature of seemings. According to the belief view—endorsed for instance by Lycan (1998: 165– 66), Hanna (2011) and somewhat Swinburne (2001: 141–142)—a seeming that P is nothing but a spontaneous, non-inferential belief that P. This view is incompatible with (PC) because if seemings were beliefs, they couldn’t supply the subject with non-inferential or immediate justification. The basic problem of this view is that S’s having a seeming that P doesn’t require S’s believing P (cf. Huemer 2007, 2019; Cullison 2010). Suppose for example S is aware of having an optical illusion that P. S will have an appearance that P but won’t believe P. (See Lyons 2009: 71–72 and Hanna 2011 for responses, and McCain 2012; McAllister 2017 for counterarguments.) This difficulty doesn’t afflict the disposition view, which identifies a seeming that P with a conscious disposition or inclination to believe P. This view is endorsed by, for example, Swinburne (2001: 141–142), Sosa (2007: 258–259), Rogers and Matheson (2011) and possibly Taylor (2015). The disposition view doesn’t support (PC) because S’s mere conscious inclination to believe P cannot give S justification for believing P. The disposition view appears implausible for independent reasons (cf. Tolhurst 1998; Huemer 2007, 2019; Cullison 2010). To begin with, it seems possible that S could be aware of having a misleading appearance thatP and be so accustomed to it that she may even lack the inclination to believe P. (Imagine S is a psychologist that routinely tests patients with the same optical illusions.) Moreover, S might be consciously disposed to believe a proposition that doesn’t actually seem true to her—e.g. out of her wishful thinking. Finally, S’s seeming that P could explain non-trivially S’s conscious inclination to believe P. If S’s seeming that P were the same as S’s conscious inclination to believe P, this would be impossible. (See Earlenbaugh and Molyneux 2009; Taylor 2015 for responses, and McAllister 2017 for counterarguments.) Another conception of the nature of seemings, which Tucker (2013) calls evidence-taking view, holds that S has a seeming that P just in case S believes or is consciously inclined to believe that she has a mental state M that counts in favour of P’s truth (see Tooley 2013; Conee 201310 ). On this view, for instance, it could seem 9 Tucker (2011) also posits religious appearances and Brogaard (2018) auditory-semantic seemings. 10 Conee

only discusses but doesn’t endorse this view.

2.2 The Nature of Seemings

9

to S that there is a tiger before her in the sense that S could believe or be consciously inclined to believe that her sensory image of a yellow–brown coat striped with black before her is evidence that there is a tiger. The evidence-taking view doesn’t sustain (PC). For S’s mere believing or being consciously inclined to believe that she has a mental state that supports P couldn’t give S justification for believing P (cf. Conee 2013). The evidence-taking view isn’t beset with the difficulties discussed before. Nevertheless, it has problems on its own. One is that it takes S’s having a secondorder mental state or conscious disposition to be a necessary condition for S’s having an appearance, but it is intuitively plausible that S could have an appearance without having any second-order mental state or conscious disposition—and even without being able to have them.11 The evidence-taking view looks thus over-intellectualized and farfetched. Another problem is that S could come to believe or consciously tend to believe that her mental state M supports P even if P doesn’t actually seem to be true to S. This might happen when S wishfully believes that P on the basis of M. It also appears possible that S could have a seeming that P without believing or consciously tending to believe that any mental state of her counts in favour of P. Think of a committed sceptic who admits that it seems to her that there are external things, though she neither believes nor is consciously inclined to believe that her mental states count in favour of the existence of these things (cf. Huemer 2013; see McAllister 2017 for further criticism). The most popular conception of the nature of seemings is the experience view. Its advocates include Bealer (2000), Pryor (2000, 2004), Huemer (2001, 2007, 2013), Cullison (2010, 2013), Chudnoff (2011), Brogaard (2013),12 Lycan (2013), Markie (2013), McGrath (2013), Skene (2013), Tucker (2010, 2013) and McAllister (2017). On this view, seemings are experiences with propositional content (some prefer to say sui generis propositional attitudes), a mind-to-world direction of fit and a distinctive phenomenology. Perceptual, introspective, mnemonic, a priori, and moral appearances have all these features. A first motive for supposing that seemings have propositional content is that we refer to them by using that-clauses (e.g. we say ‘it seems to me that…’). It would be odd if expressions such as ‘S desires that’ and ‘S believes that’ expressed a relation between a subject and a proposition but the expression ‘it seems to S that’ did not. A more decisive reason to think that seemings have propositional content is that they are the sorts of things that can be accurate. Suppose it seems to you that the cat is on the mat, but the cat is a hologram. Your appearance is inaccurate. Your seeming couldn’t be inaccurate if it didn’t embed something capable of being true or false. Since propositions are the primary bearers of truth and falsehood, seemings probably have propositional content (cf. Cullison 2013).13 11 Suppose

S is a small child. be accurate Brogaard contends that certain seemings, which she calls epistemic, are beliefs. 13 Crane (2009) argues that the fact that perceptual experiences have accuracy conditions is insufficient to show that they have propositional contents. A parallel case could probably be made for seemings. See Cullison (2013) for a response. Importantly, Siegel (2013) emphasizes that even if the contents of perceptual appearances and the contents of beliefs differed in structure, they could often be close enough to assure that the first provides immediate justification for the latter. For discussion see Siegel and Silins (2015). 12 To

10

2 Phenomenal conservatism

The content of a mental state—including the one of seemings—may fit or fail to fit how things are. If a seeming’s content fails to fit how things are (so it is false or inaccurate), the seeming is considered defective. The same happens with beliefs. That’s why seemings and beliefs have a mind-to-world direction of fit. Intentions and desires, on the other hand, have a world-to-mind direction of fit, for they are taken to be defective if the world fails to fit their contents (cf. Tolhurst 1998). Part of what makes seemings different from and irreducible to any other representational state—like hoping, believing or imagining—is their distinctive phenomenology. Tolhurst characterizes it as ‘felt veridicality’ or ‘the feel of truth’. According to Tolhurst, Seemings have the feel of truth, the feel of a state whose content reveals how things really are. (1998: 298–299)

Pryor describes this phenomenology as ‘the feeling of seeming to ascertain that a given proposition is true’ (2004: 357). He stresses that. Our [appearances] represent propositions in such a way that it ‘feels as if’ we could tell that those propositions are true—and that we’re perceiving them to be true—just by virtue of having them so represented. (2000: 547)

In sum, in virtue of their particular phenomenology, seemings represent their content as actualized or verified. This special feature is called phenomenal force by Pryor (2000), forcefulness by Huemer (2001), and assertiveness by Tucker (2010). The forcefulness (or phenomenal force or assertiveness) is present, for example, when it seems to you that 1 + 1 = 2,14 or when it seems to you that the cat is on the mat. But it is absent when, for example, you conjecture that 854 – 77 = 777, or when you imagine that your phone is ringing. The last two mental states—as the first two—do have a propositional content, but they don’t give you the feeling of ascertaining that this content is true. Importantly, belief as such also lacks the forcefulness of seemings. For you can believe something—perhaps based on calculation or testimony—that doesn’t feel true or even feels false to you (cf. McAllister 2017). The experience view of appearances is immune to the problems described before that afflict rival views. For instance, if seemings possess forcefulness, this explains— non-trivially—why they normally incline us to believe their content (cf. McAllister 2017). The experience view supports (PC) because it isn’t implausible that experiences with propositional content are capable of justifying beliefs (cf. Tucker 2013). Some advocates of phenomenal conservatism (or cousin views) maintain that the source of the justifying power of appearances is their phenomenal force. The intuition is that if you have the distinct feeling of ascertaining that P, you do have a prima facie epistemic reason to believe P (cf. Pryor 2000, 2004; Berghofer 2018).

14 Note

that forcefulness doesn’t seem to require any sensible representation. In other words, it is characterized by cognitive qualia but not necessarily non-cognitive qualia. The latter involve colours, sounds, flavours, and so on. The former are qualia present in certain types of mental processes, such as thinking and understanding.

2.2 The Nature of Seemings

11

A complaint sometimes raised against the experience view says that the notion of forcefulness (or phenomenal force or assertiveness) can be illuminated only partly through metaphors and examples, but it is actually unanalysable. Hence, this notion is obscure and suspect (cf. Conee 2013; Tooley 2013). I don’t see this as a problem. The forcefulness of appearances has been acknowledged as quite a familiar feature of these mental states by very many epistemologists and philosophers of mind (cf. McAllister 2017). Also, it is worth stressing that we can talk intelligently about things without having a precise or complete analysis of them. (A prominent philosophical example is the notion of knowledge.) We shouldn’t forget that not any notion is suitable to philosophical analysis—some notions are just fundamental. They can be illuminated only ostensively by examples and by disclosing relations they entertain with other notions. Forcefulness might be one of these basic notions. In the remaining of this book I presuppose the experience view of appearances. Before concluding this section, let’s dwell on perceptual appearances. Although seemings of this kind are those frequently mentioned in examples and arguments, there are divergences about their composition even among advocates of the experience view. Tolhurst (1998), Huemer (2001, 58–79) and, more recently, Chudnoff and DiDomenico (2015)—among others—insist that perceptual seemings are just sensations or perceptions.15 This is the unified view of perceptual appearances. According to these authors, for instance, when it visually seems to you that (R) the phone directory is on the desk, your seeming that R is your perception that R. On the other hand, Tucker (2010, 2011), Brogaard (2013) and Cullison (2013)—among others—contend that seemings are distinct from sensations or perceptions, and that what we call ‘perceptual seemings’ are in fact constituted by perceptions and seemings.16 This is the disunified view of perceptual appearances. For instance, these authors would claim that when it visually seems to you that R, you actually entertain a perception—e.g. a visual image of two rectangularly shaped objects with different sizes, colours and locations17 —together with a seeming—i.e. a propositional attitude with phenomenal force having R as its content.18 Brogaard (2013) suggests that perceptual seemings 15 Throughout

this work ‘perception’ and ‘to perceive’ refer to a non-factive state or act. (2010) thinks that perceptual seemings and sensations might come apart. For example, in (scientifically tested) cases of blind sight, the subject would have seemings of objects before her accompanied by no visual sensation. The reverse would also happen: subjects affected by certain pathologies that make them incapable of recognizing very ordinary things could have standard visual sensations but lack the usually correlated seemings. 17 How perceptions should be conceived of is also controversial within this view of perceptual seemings. Cullison (2013) thinks of perceptions as raw sensations without propositional content. Whereas Tucker (2010) and Brogaard (2013) think of them as states capable of having contents on their own, which don’t necessarily match the contents of the correlated seemings. For example, Brogaard writes: ‘If I am giving a talk to fifty-four people, my perceptual experience… represents fifty-four people in the room but it doesn’t… seem to me that there are fifty-four people in the room. At best, it seems to me that there are many people’ (2013: 276, my emphasis). 18 I have used an example of visual seemings but this view also applies to other types of perceptual seemings. Visual seemings involve—according to these authors—a seeming and a visual image. Auditory seemings would involve a seeming and a mental sound, and so on for the other senses. A view of this type could be applied in other domains. For instance, one might argue that recalling a directly experienced event involves a seeming accompanied by a mental image of that event, or 16 Tucker

12

2 Phenomenal conservatism

are kind of interpretations of correlated sensations. Brogaard (2013), Cullison (2013) and Tucker (2013) also think that perceptual seemings can be caused by appropriate sensations. These philosophers argue that telling apart perceptual seemings and sensations helps us explain some psychological phenomena, such as the one of the enhanced perceptual skills of experts, and solve epistemological puzzles such as the speckled hen one (see mainly Tucker 2010). For counterarguments see Chudnoff and DiDomenico (2015). The theses and arguments that I put forward in the next sections and chapters of this work are neutral with respect to this debate on perceptual seemings.

2.3 Supporting Phenomenal Conservatism Phenomenal conservatism is philosophically alluring for many reasons. To start with, (PC) provides a rationale for widespread epistemic practices. Our reasons for holding many ordinary beliefs don’t seem to include thoughts about, say, the trustworthiness of our experiences or the reliability of our faculties (cf. Tucker 2013). Rather, we seem to take ourselves to have justification for entertaining many ordinary beliefs just because of how things appear to us to be. Furthermore, we often attribute reasons for entertaining beliefs to small children, who couldn’t have the complex thought that their faculties are reliable or that scepticism is false (cf. Pryor 2000). Phenomenal conservatism is perceived to be attractive also because it offers a unified account of the non-inferential justification of beliefs of very different types (e.g. perceptive, introspective, a priori, mnemonic and moral). Indeed, (PC) has been asserted to constitute the general basis of fallible foundationalism—a view that many epistemologists find natural and plausible in itself (cf. Huemer 2001: 102). The claim is that (PC) explains how beliefs of very different types can be basically—i.e. noninferentially—justified. It is our seemings, according to this view, that put an end to the regress of justified beliefs when we search for a basis for the justification of all our beliefs.19 Another celebrated virtue of phenomenal conservatism is that it affords us the means of a thoroughgoing response to the sceptic—specifically, the sceptic who assumes that a subject S must have independent justification for ruling out any relevant sceptical alternative in order to possess justification for believing ordinary things (cf. Huemer 2001, 2019). Epistemologists agree that this assumption fosters virulent forms of scepticism (cf. Pryor 2000; Wright 2004; Schiffer 2004). The phenomenal conservative can adduce (PC) and argue that S doesn’t need the independent justification the sceptic assumes S must have. For instance, suppose S entertains a visual that moral judgments are typically based on seemings accompanied by emotions such as revulsion or admiration (cf. Tucker 2013). 19 The acquaintance theory of justification (see for instance Fumerton 1995) provides an alternative explanation in some cases. See Huemer (2006, 2007, 2011)’s arguments against the acquaintance theory. Smithies (2019) argues that (PC) cannot actually unify all types non-inferential justification. I return on this issue in Chap. 6.

2.3 Supporting Phenomenal Conservatism

13

experience as if (P) there is a hand. The sceptic might argue that S’s experience as if P gives S justification for believing P only if S has independent justification for ruling out that (SH) S is a disembodied soul in an immaterial world with the hallucination of a hand caused by a demon. The sceptic will insist that since S cannot possess this independent justification, S doesn’t have justification for believing P. Suppose, however, that S’s experience as if P is a seeming that P. If (PC) is true, S has prima facie justification for believing P even if S has no independent reason to rule out SH (cf. Huemer 2001; Pryor 2000, 2004). Mnemonic scepticism can be addressed in similar way. Whenever S seems to remember that P, if (PC) is true, S possesses prima facie justification for believing P even if S has no independent justification for ruling out, say, the Russellian conjecture that the universe popped into existence an instant ago. Other forms of scepticism can be addressed along these lines.20 Quite independently of these asserted virtues, phenomenal conservatism appears plausible once one adopts a broadly internalist perspective about epistemic justification. Internalists tend to see a very tight connection between the attitudes that are epistemically justified for a subject and those that are epistemically rational from her standpoint (cf. Huemer 2001: 22). When justification and rationality are tightly linked in this way, (PC) looks true. McGrath has explained why: Suppose it seems to you that P and you have no defeaters (i.e. no good evidence for ¬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 ¬P? Withhold judgment on P? It does seem to you that P, and you lack evidence for ¬P and for the unreliability of the seeming with respect to P. The only reasonable attitude to take is belief. (2013: 226, edited)

Provided that epistemic rationality (or reasonability) is or entails epistemic justification, McGrath’s conclusion strongly support (PC). I find this argument quite persuasive. Huemer (2001: 103–104) has a similar argument that relies, more specifically, on Foley (1993)’s instrumentalist conception of epistemic rationality. According to Foley, it is epistemically rational for S to do X, if doing X would appear, from S’s standpoint, to be an effective way of satisfying the central epistemic goal of believing the true and not believing the false. Suppose it seems to S that P and S has no reason to doubt P. From S’s standpoint, believing P would appear to be an effective means of pursuing the central epistemic goal. Hence, S’s believing P would be epistemically rational, and so epistemically justified. A drawback of this argument—which doesn’t afflict McGrath’s—is that it cannot substantiate important applications of (PC). As said, a reason of (PC)’s appeal is that it vindicates attributions of epistemic justification to small children who couldn’t conceive of sceptical alternatives. A problem of Huemer’s argument is that small children would be incapable of grasping the complex notion of an epistemic goal.21

20 See 21 A

Tucker (2010) for further asserted merits of phenomenal conservatism. similar concern is raised in Hasan (2013: 133–134).

14

2 Phenomenal conservatism

Tucker (2013) doesn’t find McGrath’s case convincing.22 He suggests that since ‘reasonable attitude’ in McGrath’s conclusion can be interpreted as ‘non-incoherent attitude’—rather than ‘justified attitude’—the argument doesn’t support (PC). Suppose for instance you believe Q (whether or not you have justification for Q). If you also disbelieved or withheld belief about the disjunction Q v R, these attitudes would be incoherent with your attitude towards Q. So they would be—in this sense— irrational. The only non-incoherent and—in this sense—rational attitude for you to have towards Q v R is believing it. This type or rationality that depends on coherence is called rational commitment. Since you believe Q, you are rationally committed to believing Q v R. Rational commitment to believing a proposition doesn’t entail epistemic justification for doing it. For example, if you believe Q without justification, you are still rationally committed to believing Q v R, but this doesn’t give you justification for believing Q v R. Tucker thinks that the conclusion of McGrath’s argument doesn’t rule out this possibility: your having a seeming that P and no defeater only rationally commits you to believing P without providing epistemic justification for it. Therefore—Tucker contends—McGrath’s argument doesn’t actually vindicate (PC).23 This ingenious objection is flawed. The problem is that there are defeaters that, if undefeated, necessarily harm the type of rationality described in McGrath’s case, but they may not harm rational commitment. So it is implausible that the type of rationality described in McGrath’s case is identical to rational commitment. The defeaters I have in mind are evidence indicating that the basis of your justification may be inappropriate. Suppose you initially believe Q with justification. This gives you justification for believing Q v R and commits you to believing Q v R. Imagine that you then acquire evidence D that Q is false with probability ½. If undefeated, D will defeat your justification for believing Q, and thus your justification for believing Q v R. However, D won’t necessarily defeat your rational commitment. If after acquiring D, you obstinately keep believing Q (perhaps out of wishful thinking), your justification for Q v R is destroyed, but your rational commitment to believing Q v R stands undefeated. This shows that D, if undefeated, necessarily harms your justification for Q v R but not necessarily your rational commitment to believing Q v R. Now take again McGrath’s argument. If you have a seeming that P and no defeater, it is rational for you to believe P. Imagine that you then acquire evidence D* that your seeming is inaccurate with probability ½. You still have that seeming, but at this point it is for you no longer rational—in McGrath’s sense—to believe P. Note that the defeated rational support must be epistemic justification. For if it were rational commitment, it would stand undefeated.24 This shows that seemings do supply epistemic justification.

22 See

also Ghijsen (2016). (2013) gives this criticism a response, which Tucker (2013: 12n27) finds dubious. 24 If seemings also produce rational commitment, there must be a sense in which it is for you still rational to believe P. 23 McGrath

2.3 Supporting Phenomenal Conservatism

15

To conclude the section, let me dwell on Huemer’s infamous self-defeat argument (see mainly Huemer 2007), which aims to show that the belief that (PC) is false cannot be doxastically justified. This is a reconstruction of it: (1) All our beliefs (with a few irrelevant exceptions)25 are based on our seemings. (2) If a belief that P is based on something that doesn’t constitute a source of propositional justification for P, then the belief is doxastically unjustified. (3) Therefore, if no seeming confers justification on the proposition that constitutes its content, then no belief is justified. (4) But if (PC) is false, then no seeming confers justification on the proposition that constitutes its content. Therefore, (C) If (PC) is false, then no belief is doxastically justified, including any belief that (PC) is false.26

While Huemer (2019) acknowledges that this argument is controversial, other supporters of phenomenal conservatism have explicitly distanced themselves from it (e.g. Tucker 2013: 9 n22; Lycan 2013). The self-defeat argument appears to be valid. However, some of its premises are contentious. For instance, Huemer claims that (1) is empirically true but DePaul (2009) and Markie (2013) complain that he doesn’t give enough evidence for believing so. Note that if a belief that (PC) is false could be based on something other than appearances, against (C), the belief might prove doxastically justified. Some have attacked (1) contending that our doxastically justified beliefs aren’t or might not be based on seemings. For instance, DePoe (2011), Hasan (2013) and Tooley (2013) insist that the facts with which the subject S is directly acquainted with, rather than S’s seemings, are the bases of the S’s beliefs in general or in some cases.27 Conee (2013) suggests that the doxastic justification of all our beliefs may depend on an evidentialist condition that assigns no basing role to appearances. Another problem of (1) is that it may be questioned that inferential beliefs and testimonial beliefs are based on seemings (cf. Conee 2013).

25 The exceptions concern cases of self-deception and leaps of faith—these are irrelevant because opponents of (PC) would not base their belief that (PC) is false on such sources. (cf. Huemer 2007: 39n14). 26 As Huemer (2007) and Markie (2013) note, from (C) we can get to the conclusion that (PC) is true by adding the premise that our beliefs aren’t generally unjustified. 27 The acquaintance theory of justification holds, in its most elementary version, that S’s belief that P is doxastically justified if S is acquainted with the fact that P. Acquaintance is a kind of direct awareness of facts. Acquaintance is factive. So S’s being acquainted with the fact that P—unlike S’s having a seeming that P—entails that P is true. Possible objects of acquaintance include facts about one’s own mental states and abstract entities.

16

2 Phenomenal conservatism

A different group of objections targets (4). DePaul (2009), Tooley (2013) and Markie (2013) contend that (4) is false because if (PC) were false but a restricted version of (PC) were true, seemings could still confer defeasible justification on their contents. For instance, Markie (2013) rejects (PC) and proposes a restrained variant according to which S’s seeming that P gives S prima facie justification for believing P only if it results from a special type of knowledge-how that S has. (I return to this in Sect. 3.4.) In accordance with this view, most of our seemings still confer defeasible justification on their contents, and the belief that (PC) is false can be doxastically justified without incoherence. For further objections see Mizrahi (2014) and Beillard (2016). Although Huemer (2007, 2009, 2011, 2013, 2014, 2019) and Skene (2013) have endeavoured to addressed many of these challenges, my impression is that the self-defeat argument remains quite shaky.

2.4 Criticism of Phenomenal Conservatism Phenomenal conservatism has been struck by diverse objections. Philosophers—e.g. Siegel (2017)—have argued that seemings don’t actually have intrinsic justifying power because they might be cognitively penetrable. Formal epistemologists—e.g. White (2006)—have contended that phenomenal conservatism is flawed because it is incompatible with Bayesianism. Other epistemologists—e.g. Cohen (2002)—have claimed that phenomenal conservatism is problematic because it yields versions of the problem of easy justification (or easy knowledge). I myself have argued that that the antisceptical bite of phenomenal conservatism is weaker than expected, for phenomenal conservatism is affected by what I’ve called the problem of reflective awareness (cf. Moretti 2018). I’ll return to these four objections in the next chapters. Let me now touch upon a number of somehow minor criticisms of phenomenal conservatism. Byerly (2012) contends that phenomenal conservatism doesn’t provide an adequate explanation of the source of non-inferential or immediate justification. For there is no good reason to rule out that what terms like ‘seeming’ and ‘appearance’ refer to is, not one kind of sui generis propositional attitude, but a range of heterogeneous mental states. So there is no good reason to exclude that ‘seeming’ and ‘appearance’ denote a disjunctive kind.28 As McAllister (2017) has noted, however, the virtues of ontological simplicity and unification do supply this reason: ceteris paribus, we should rationally prefer a simple and unified conception of seemings to a less simple and non-unified conception of them.

28 Note

that this criticism doesn’t say that (as suggested in Sect. 2.1) ‘appearance’ and ‘seeming’ can be used in a variety of different senses—for instance, dubitative, comparative or epistemic. The existence of these alternative usages has no bearing on the possibility that there is one type of mental state that is always present when something seems to be the case in one particular sense.

2.4 Criticism of Phenomenal Conservatism

17

DePaul (2009), Markie (2005, 2013) and Tooley (2013) maintain that phenomenal conservatism is unacceptable because it enables any sort of odd or crazy appearances to justify their own contents. Markie (2005) imagines the following possibility: as S sees a walnut tree, it just seems to S that (P) the walnut was planted on April 24, 1914. This is not so because S reads a date-of-planting sign or recalls the date. S just happens to have this seeming. If we accept (PC), we must conclude that S has prima facie justification for P. But this conclusion—Markie claims—looks counterintuitive. Littlejohn (2011) and Tooley (2013) suggest that phenomenal conservatism enables seemings to justify dangerous beliefs. Imagine it seems to S that (Q) God wants her to kill all non-believers, or that (R) cannibalism is morally permissible. If we endorse (PC), we must conclude that S has prima facie justification for Q and R. But this looks absurd. A way to play down the unpalatable implications of these thought experiments is emphasizing that S’s ordinary background knowledge would probably defeat S’s seeming-based justification for propositions like P, Q and R. S would know, for instance, that people can’t normally tell when a tree was planted by just looking at it, that mystic experiences of the type described are symptomatic of mental disorder, and that cannibalism is unanimously thought of as incompatible with civilization (cf. Huemer 2019). One could retort that it is intuitive that in the envisaged scenarios S shouldn’t possess even prima facie justification for believing P, Q and R. Phenomenal conservatives may respond that since scenarios like these are highly unusual and thus hard to imagine vividly, people’s intuitions about them are untrustworthy (cf. Huemer 2019). Phenomenal conservatives could insist that when it comes to strange scenarios like these, our intuitions can be taken to indicate, at very best, that S lacks some epistemic status but not necessarily justification. S might for instance lack knowledge or warrant.29 We tend to believe that Q and R are false. Since knowledge is factive, it is no surprise that we think that S cannot know Q or R. Furthermore, it would be unnatural to suppose that in scenarios like Markie’s there is reliable link between S’s belief that P and P’s truth. Hence, it is quite natural that we think that S doesn’t know P. In conclusion, it is dubious that these thoughts experiments show that phenomenal conservatism is flawed or seriously problematic. Let’s turn to a last criticism. Some contend that phenomenal conservatism is false because if S’s belief that P were based on S’s seeming that P, this wouldn’t suffice to give S prima facie justification for believing P. S should also possess metajustification—namely, justification for taking her seeming to be a reliable indicator of P’s truth. This is a reason often adduced: if S’s seeming-based belief that P proved true but S hadn’t metajustification, P’s truth could at best count as an accident from S’s viewpoint, which is incompatible with P’s being justified for S (cf. Bonjour 2004; Bergmann 2013b).30

29 This

is additional property that a true belief needs to have to become knowledge.

30 Bergmann (2013a, b) formulates a dilemma for the phenomenal conservative (which is special ver-

sion of Bergmann 2006’s dilemma for the internalist). According to it, the phenomenal conservative either sticks to the false belief that the justifying power of seemings needs no metajustification, or she

18

2 Phenomenal conservatism

It is far from clear, however, that if P appeared true to S and this caused a belief that P in S’s mind, P’s truth could at best count as an accident from S’s perspective if S hadn’t metajustification.31 Many epistemologists—including myself—don’t have this intuition; they think that in this case P’s truth would be non-accidental from S’s perspective even if S hadn’t metajustification (cf. Rogers and Matheson 2011; Moretti and Piazza 2016; Gage 2016). See Huemer (2013, 2019) for an alternative response.

2.5 Conclusions In this chapter I have introduced and analysed phenomenal conservatism, the internalist view to the effect that appearances, conceived of as sui generis propositional attitudes, supply defeasible justification for our beliefs. I have shown that phenomenal conservatism is intuitive, prima facie plausible and promissory of significant philosophical achievements, such as affording a unified theory of non-inferential justification and defeating scepticism. I have suggested that phenomenal conservatism survives a number of initial objections. To carry out a thorough appraisal of phenomenal conservatism, in the next chapters I will inspect further and more complex objections.

References Audi R (1993) The structure of justification. Cambridge University Press, Cambridge Audi R (2013) Doxastic innocence: phenomenal conservatism and grounds of justification. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 181–201 Bealer G (2000) A theory of the a priori. Pac Philos Q 81:1–30 Beillard J (2016) Phenomenal conservatism, reflection and self-defeat. Logos Epistem 7:187–199 Berghofer P (2018) Towards a phenomenological conception of experiential justification. Synthese OnlineFirst. https://doi.org/10.1007/s11229-018-1744-5 Bergmann M (2006) Justification without awareness. Oxford University Press, New York Bergmann M (2013a) Externalist justification and the role of seemings. Philos Stud 166:163–184

admits that the justifying power of seemings requires metajustification. The second option commits the phenomenal conservative to a potential infinite regress leading to scepticism. 31 To support this thesis, Bergmann (2013a, b: 171–172) adduces a thought experiment in which (i) S has a seeming that P but (ii) due to a cognitive impairment, S is incapable of thinking of her seeming as a reliable indicator of P’s truth and thus S cannot have metajustification, finally (iii) S forms the true belief that P independently of her seeming and without any actual reason for doing it. Bergmann claims that it is intuitive that the truth of S’s belief that P would be accidental from S’s perspective in this case. I agree. Note that this result cuts no ice because, due to (iii), S’s belief that P is not caused by S’s appearance that P (cf. Moretti and Piazza 2016).

References

19

Bergmann M (2013b) Phenomenal conservatism and the dilemma for internalism. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 154–178 BonJour L (2004) In search of direct realism. Philos Phenomenol Res 69:349–367 Brogaard B (2013) Phenomenal seemings and sensible dogmatism. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 270–289 Brogaard B (2018) defense of hearing meanings. Synthese 195:2967–2983 Byerly RT (2012) It seems like there aren’t any seemings. Ratio 40:771–782 Chisholm R (1957) Perceiving: a philosophical study, vol 9. Cornell University Press, Ithaca Chisholm R (1989) Theory of knowledge. Prentice-Hall, Englewood Cliffs, NJ Chudnoff E (2011) The nature of intuitive justification. Philos Stud 153:313–333 Chudnoff E (2013) Intuition. Oxford University Press, Oxford Chudnoff E, DiDomenico D (2015) The epistemic unity of perception. Pac Philos Q 94:535–549 Cohen S (2002) Basic knowledge and the problem of easy knowledge. Philos Phenomenol Res 65:309–329 Conee E (2013) Seeming evidence. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 52–68 Crane T (2009) Is perception a propositional attitude? Philos Q 59:452–469 Cullison A (2010) What are seemings? Ratio 23:260–274 Cullison A (2013) Seeming and semantics. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 33–51 DePaul M (2009) Phenomenal conservatism and self-defeat. Philos Phenomenol Res 78:205–212 DePoe J (2011) Defeating the self-defeat argument for phenomenal conservatism. Philos Stud 152:347–359 Earlenbaugh J, Molyneux B (2009) Intuitions are inclinations to believe. Philos Stud 145:89–109 Foley R (1993) Working without a net. Oxford University Press, New York Fumerton R (1995) Metaepistemology and skepticism. Rowman & Littlefield, Lanham, MD Gage LP (2016) Phenomenal conservatism and the subject’s perspective objection. Acta Anal 31:43– 58 Ghijsen H (2016) The real epistemic problem of cognitive penetration. Philos Stud 173:1457–1475 Goldman A (2008) Immediate justification and process reliabilism. In: Smith Q (ed) Epistemology: new essays. Oxford University Press, Oxford, pp 63–82 Hanna N (2011) Against phenomenal conservatism. Acta Anal 26:213–221 Hasan A (2013) Phenomenal conservatism, classical foundationalism, and internalist justification. Philos Stud 162:119–141 Huemer M (2001) Skepticism and the veil of perception. Rowman and Littlefield, Lanham, MD Huemer M (2005) Ethical intuitionism. Palgrave Macmillan, New York Huemer M (2006) Phenomenal conservatism and the internalist intuition. Am Philos Q 43:147–158 Huemer M (2007) Compassionate phenomenal conservatism. Philos Phenomenol Res 74:30–55 Huemer M (2009) Apology of a modest intuitionist. Philos Phenomenol Res 78:222–236 Huemer M (2011) Phenomenal conservatism and self-defeat: a reply to DePoe. Philos Stud 156:1–13 Huemer M (2013) Phenomenal conservatism Über Alles. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 328–350 Huemer M (2014) Alternative self-defeat arguments: a reply to Mizrahi. Logos Epistem 5:223–229 Huemer M (2019) Phenomenal conservatism. Internet encyclopedia of philosophy. https://www. iep.utm.edu/phen-con/. Accessed 31 May 2019 Kvanvig J (2003) Propositionalism and the perspectival character of justification. Am Philos Q 40:3–18 Littlejohn C (2011) Defeating phenomenal conservatism. Anal Philos 52:35–48 Lycan WG (1988) Judgment and justification. Cambridge University Press, Cambridge

20

2 Phenomenal conservatism

Lycan WG (2013) Phenomenal conservatism and the principle of credulity. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 293–306 Lyons J (2009) Perception and Basic Beliefs. Oxford University Press, New York Markie P (2005) The mystery of direct perceptual justification. Philos Stud 126:347–373 Markie P (2013) Searching for true dogmatism. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 248–268 McAllister B (2017) Seemings as sui generis. Synthese 195:3079–3096 McCain K (2012) Against Hanna on phenomenal conservatism. Acta Analytica 27:45–54 McGrath M (2013) Phenomenal conservatism and cognitive penetration: the ‘Bad Basis’ counterexamples. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 225–247 Mizrahi M (2014) Phenomenal conservatism, justification and self-defeat. Logos Epistem 5:103– 110 Moretti L (2018) Phenomenal conservatism and the problem of reflective awareness. Am Philos Q 55:267–280 Moretti L, Piazza T (2016) Phenomenal conservatism and Bergmann’s dilemma. Erkenntnis 80:2071–2090 Pappas G (2014) Internalist vs. externalist conceptions of epistemic justification. In: Zalta EN (ed) Stanford encyclopedia of philosophy. https://plato.stanford.edu/archives/fall2017/entries/justepintext/ Pollock JL, Cruz J (1999) Theories of knowledge. Rowman and Littlefield, Lanham, MD Pryor J (2000) The skeptic and the dogmatist. Nous 34:517–549 Pryor J (2004) What’s wrong with Moore’s argument? Philos Issue, 14, Epistemology:349–378 Pryor J (2005) Is there immediate justification? In: Steup M, Sosa E (eds) Contemporary debates in epistemology. Blackwell, Oxford, pp 181–202 Rogers J, Matheson J (2011) Bergmann’s Dilemma: exit strategies for internalists. Philos Stud 152:55–80 Schiffer S (2004) Skepticism and the vagaries of justified belief. Philos Stud 119:161–184 Siegel S (2013) The epistemic impact of the etiology on experience. Philos Stud 162:697–722 Siegel S (2017) The rationality of experience. Oxford University Press, Oxford Siegel S, Silins N (2015) The epistemology of perception. In: Matthen M (ed) The Oxford handbook of philosophy of perception. Oxford University Press, Oxford Skene M (2013) Seemings and the possibility of epistemic justification. Philos Stud 163:539–559 Smithies D (2019) On the global ambitions of phenomenal conservatism. Anal Philos 60:206–244 Sosa E (2007) A virtue epistemology, vol 1. Oxford University Press, Oxford Swinburne R (1998) Providence and the problem of evil. Oxford University Press, Oxford Swinburne R (2001) Epistemic justification. Oxford University Press, Oxford Taylor SA (2015) What seemings seem to be. Episteme 12:363–384 Tolhurst W (1998) Seemings. Am Philos Q 35:293–302 Tooley M (2013) Michael Huemer and the principle of phenomenal conservatism. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 306–327 Tucker C (2010) Why open-minded people should endorse dogmatism. Philos Perspect 24:529–545 Tucker C (2011) Phenomenal conservatism and evidentialism in religious epistemology. In: Clark KJ, Van Arragon RJ (eds) Evidence and religious belief, Ch 4. Oxford University Press, Oxford

References

21

Tucker C (2013) Seemings and justification: an introduction. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 1–29 White R (2006) Problems for dogmatism. Philos Stud 131:525–557 Wright C (2004) Warrant for nothing (and foundations for free)? Aristot Soc. Sup. 78:167–212

Chapter 3

Cognitive Penetrability

Abstract In this chapter I introduce the thesis that perceptual appearances are cognitively penetrable and analyse cases made against phenomenal conservatism hinging on this thesis. In particular, I focus on objections coming from the externalist reliabilist camp and the internalist inferentialist camp. I conclude that cognitive penetrability doesn’t yield substantive difficulties for phenomenal conservatism. Keywords Phenomenal conservatism · Perceptual seemings · Cognitive penetrability · Perceptual justification · Reliabilism · Inferentialism · Proper function

3.1 Characterizing Cognitive Penetrability Preformationism was the biological theory that organisms are fully formed in miniature within the ovum or the spermatozoon. Some microscopists who endorsed preformationism claimed that they actually saw embryos in the sperm cells (cf. Siegel 2012). This historical datum has more than one possible explanation. Perhaps these biologists didn’t actually have visual experiences of embryos; they mistakenly described the things that they saw as embryos, or their visual experiences as experiences of embryos. Another possibility is, however, that they had—literally— visual experiences of embryos. If the content of one’s experience can be cognitively penetrated by one’s beliefs, it isn’t implausible that biologists who believed in preformationism actually had visual experiences of embryos in sperm cells. The thesis of cognitive penetrability of perception holds—roughly—that the content of one’s perceptual experiences or appearances can partly be determined directly by one’s prior or concurrent cognitive states—for example, beliefs, conjectures, expectations, suspicions, desires, hopes, moods, emotions, traits, decisions, accepted values, pursued goals, and attitudes or abilities acquired through experience or training. It isn’t easy to clarify what ‘directly’ means in this characterization (cf. Tucker forthcoming). A shared intuition is, however, that not just any kind of influence on one’s perceptual appearances by one’s cognitive states would qualify as cognitive penetration. A cognitive state of mine might influence my perceptual appearance simply because it causes a change in the spatial location or orientation of my sense © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2020 L. Moretti, Seemings and Epistemic Justification, SpringerBriefs in Philosophy, https://doi.org/10.1007/978-3-030-43392-5_3

23

24

3 Cognitive Penetrability

organs, so that I receive new perceptual stimuli. Imagine I intend to watch a painting. Accordingly, I turn my head towards the painting. My perceptual experience changes from representing, say, a white wall to representing the painting. This kind of influence doesn’t count as direct. Intuitively, in this case, there is an indirect or external link in the causal chain from my intention to my experiential state. So there is no cognitive penetration (cf. Stokes 2012). People have tried to supply more precise characterizations of cognitive penetrability. (See for instance Pylyshyn 1999; Siegel 2012; Macpherson 2012; Stokes 2012; Wu 2013; Raftopoulos 2019). This is a basic characterization that focuses on visual appearances. Take subjects S 1 and S 2 , where S 1 and S 2 can be the same subject in different times or different counterfactual circumstances. (CP) Visual appearances are cognitively penetrable just in case the following scenario is nomologically possible: When S 1 ’s and S 2 ’s organs of sight function equally well and are oriented towards the same distal stimulus, under the same external conditions, S 1 and S 2 have visual appearances with different contents because of S 1 ’s and S 2 ’s concurrent or antecedent cognitive states. (Cf. Macpherson 2012; Siegel 2012)

Suppose the distal stimulus is an apple. In the scenario envisaged in (CP), S 1 and S 2 have organs of sight that work equally well. Since both subjects watch in the direction of the apple, they have the same sensory inputs. Furthermore, S 1 and S 2 are in the same external conditions. This entails, for instance, that the quantity and quality of light available to S 1 and S 2 is the same. Finally, S 1 and S 2 have visual appearances with different contents partly caused by their different cognitive states. If all these conditions are satisfied, visual appearances are cognitively penetrable. (CP) comes together with a characterization of cognitively penetrated perceptual appearance. Suppose that in the above scenario, S 1 has a cognitive state that S 2 lacks. For instance, only S 1 but not S 2 has the expectation that the apple is green. If S 1 ’s visual seeming is different from S 2 ’s—e.g. only S 1 ’s experience represents the apple as green—because of S 1 ’s expectation, this state penetrates (the content of) S 1 ’s appearance. To familiarize with the notions introduced, let’s consider two imaginary cases of cognitive penetration often discussed in the literature. Angry Jack Jill believes for no good reason that Jack is angry. This belief (or the arising expectation) affects the way Jack looks to her, producing in her a visual seeming that Jack is angry. Jill thus believes, even more firmly than before, that Jack is angry. (Cf. Siegel 2012) Prospectors Gus and Virgil are gold prospectors. Gus is an expert. He has learned to do so through training and experience. He began with a list of identification criteria and consciously applied them. He then reached the point where he could “see” that a nugget is gold. Virgil is a novice. He has a general sense of what gold looks like, but he isn’t very good at its visual identification. While searching for gold, they find a pebble. Gus has the visual appearance that the pebble is gold as a result of his expertise. So he believes that the little stone is gold. Virgil also has a visual appearance that the pebble is gold, but only as a result of his desire of being rich. Like Gus, Virgil comes to believe that the pebble is gold. (Cf. Markie 2013)

3.1 Characterizing Cognitive Penetrability

25

Consider again (CP). In Angry Jack the distal stimulus is Jack’s face, S 1 corresponds to Jill, and S 2 corresponds to Jill in a counterfactual situation in which she watches Jack’s face in the same external conditions but she doesn’t have the initial belief that (P) Jack is angry.1 Jill’s belief that P is involved in causing Jill’s appearance that P. In the counterfactual situation, however, Jill doesn’t have the belief that P and, consequently, she doesn’t have the appearance that P. If all this is nomologically possible, in accordance with (CP), experience is cognitively penetrable. This happens because Jill’s seeming that P can be penetrated by her belief that P. Take now Prospectors. In this case the distal stimulus is the pebble. Let’s first identify S 1 with Virgil, and S 2 with Virgil in a counterfactual situation in which he inspects the pebble under the same external conditions but he doesn’t crave for becoming rich. Virgil’s desire to become rich is involved in causing his seeming that (Q) the pebble is gold. In the counterfactual situation, however, Virgil has no desire to be rich2 and, accordingly, he doesn’t experience that Q. If all this is nomologically possible, in accordance with (CP), perceptual appearances are cognitively penetrable. This happens because Virgil’s desire to become rich can penetrate his experience that Q. Let’s now identify S 1 with Gus, and S 2 with Gus in a counterfactual situation in which he still inspects the pebble under the same external conditions but he has no perceptual expertise. Gus’ perceptual expertise is involved in causing his seeming that Q. In the counterfactual situation, however, Gus has no perceptual expertise and thus no appearance that Q. If all this is nomologically possible, appearances are cognitively penetrable, for Gus’ perceptual expertise can penetrate his appearance that Q. As we have seen in Sect. 2.2, many philosophers think that any perceptual appearance is a disunified compound of two elements: a sensory state and a seeming (where the seeming is a sort of interpretation of the sensory state). If this conception is correct, there are two prima facie possible ways in which a subject S might come to have a perceptual seeming that P cognitively penetrated by a state M of S. M might directly cause or change a sensory state of S, and this might in turn produce an appropriate seeming that P in S. (For instance, under the influence of her belief that Jack is angry, Jill could have a sensory impression of two slanted eyes and a frowning mouth while watching Jack’s face. This sensory state could then cause Jill’s seeming that Jack is angry.) Alternatively, M might directly cause the seeming that P in S without affecting S’s sensory states. (For instance, Jill might have the visual sensation of a non-angry face, but her belief that Jack is angry could directly produce her seeming that Jack is so.) In this case, there might be a mismatch detectable by S between her sensory states and S’s interpretation of them (cf. Huemer 2013b). If the unified conception of perceptual seemings proved correct (see above Sect. 2.2), the contents of perceptual seemings might include representations of low-level properties of things (such as shapes and colours) and representations of high-level properties of things (such as being a dog or being a computer). In this 1 Or 2 Or

in which she has a belief that P that has no causal bearing on her seemings. he has a desire to be rich that has no causal bearing on his seemings.

26

3 Cognitive Penetrability

case too there are two prima facie possible ways in which S might come to have a perceptual seeming that P cognitively penetrated by a mental state M. M might directly produce a representation of low-level properties in S that in turn elicits a representation of high-level properties, or M might directly cause a representation of high-level properties in S. I have characterized cognitive penetrability in terms of (CP), but the adequacy of this characterization can be questioned. Philosophers could contend that (CP) doesn’t discriminate between the genuine cognitive penetrability of one’s perception and the ability of one’s covert attention to select some features of the distal stimulus as opposed to some others. S’s attention to something X (in the external environment) is overt if it is implemented by S’s bodily behaviour—e.g. by S’s pointing her eyes in the direction of X. S’s attention to X is covert if its implementation only involves a selection process internal to S that makes no bodily difference (cf. Mole 2015). Suppose that S 1 ’s and S 2 ’s organs of sight function equally well and are oriented towards the same distal stimulus X. Imagine that under the same external conditions, S 1 and S 2 have visual seemings with different contents because of S 1 ’s and S 2 ’s previous cognitive states make them covertly attend to different features of X. If X is an apple, S 1 might focus on its shape, and S 2 on its colour. On (CP), we should conclude that perceptual appearances are cognitively penetrable. Yet many would deny this. Many would insist that a mental state can produce cognitive penetration only if it affects S’s perceptual processing itself. If a state of S causes a change in the content of S’s appearance only because it determines the focus of S’s covert attention before she has the experience, there seem to be at best an indirect link from S’s cognitive state to S’s appearance. This doesn’t look like cognitive penetration (cf. Siegel 2012, 2013a; Silins 2016; Tucker forthcoming).3 To settle this difficulty, we might supplant (CP) with a principle of this sort: (CP*) Visual appearances are cognitively penetrable just in case the following scenario is nomologically possible: When S 1 ’s and S 2 ’s organs of sight function equally well and are oriented towards the same distal stimulus X, and S 1 and S 2 covertly attend to the same features of X, under the same external conditions, S 1 and S 2 have visual appearances with different contents because of S 1 ’s and S 2 ’s concurrent or antecedent cognitive states. (Cf. Siegel 2012; Silins 2016)

A problematic aspect of the notion of covert attention presupposed in (CP*) and in our discussion so far is that it assumes that this type of attention is pre-experiential. Covert attention is thought of as a process external to the relevant experience, for the allocation of attention to a spatial location is taken to occur before the experience occurs. Although such a conception seems to be taken for granted by some authors—e.g. Pylyshyn (1999), Siegel (2012, 2013a, 2017) and Firestone and Scholl (2016)—other scholars explicitly reject it—e.g. Mole (2015) and Wu (2017). The latter describe findings of psychology and cognitive sciences that strongly suggest 3 The phenomenon of selection by covert attention (as opposed to cognitive penetration) might

have by itself problematic implications for phenomenal conservatism, which I won’t investigate in this work. On this issue see Tucker (forthcoming).

3.1 Characterizing Cognitive Penetrability

27

that the selection processes of covert attention are constitutive elements of the relevant appearances. If this is correct, the selective ability of covert attention may count as a type of cognitive penetrability, and (CP*) should be dropped as inadequate.4 Because of these complications, in the remainder of this chapter I will simply assume that some principle in the neighbourhood of (CP) and (CP*) does provide an appropriate characterization of cognitive penetrability. If a more specific characterization is adopted, this won’t affect the adequacy of my analyses and arguments in the next sections.

3.2 The Epistemic Problem of Cognitive Penetrability Although the cognitive penetrability of perception remains to date a controversial empirical hypothesis, it has often been adduced to challenge the views that hold that perceptual experiences have inherent justifying power. Phenomenal conservatism has been the target par excellence (see for instance Lyons 2011; Siegel 2012; Brogaard 2013; Markie 2013; McGrath 2013a; Teng 2016). Those who adduce examples like Prospectors and Angry Jack to challenge phenomenal conservatism argue that it is intuitive that in these situations Jill and Virgil cannot have justification for holding their perceptual beliefs. So Virgil’s seeming that the pebble is gold would be unable to give him justification for believing that the pebble is gold, and Jill’s seeming that Jack is angry would be unable to give her justification for believing that Jack is angry.5 As we will see, there are various ways to unpack these intuitions. The epistemologists who challenge phenomenal conservatism by adducing examples of these types insist that the perceptual seemings of the subjects in these situations cannot justify their contents whether or not the subjects are or can be aware that the seemings are cognitively penetrated. So these examples are not thought of as cases of defeat. The contention is that the subjects in question cannot acquire even prima facie justification for their beliefs from the cognitively penetrated seemings (cf. Lyons 2011; Markie 2013; McGrath 2013a; Siegel 2013b). This thesis is incompatible with (PC). As we have seen in Sect. 2.1, (PC) entails that if it seems to S that P, S thereby has prima facie justification for believing P.6 4 For opinionated overviews of this multifaceted debate see for instance Gatzia and Brogaard (2017)

and Raftopoulos (2019). a variant of Angry Jack, Jill starts with a justified belief that (P) Jack is angry, which causes in Jill a seeming that P. Siegel (2012) adduces this variant to contend that phenomenal conservatism is false or implausible. The argument is that Jill’s justification coming from her appearance that P would need to boost Jill’s original justification for believing P, which is intuitively impossible in this case. So Jill cannot have justification from her appearance. The central problem of this argument is that it presupposes that a justification based on an appearance can boost a justification already available for a belief even if the subject ignores whether the appearance is independent of the belief. (Jill, for instance, ignores this.) But it is dubious or unclear that the phenomenal conservative is committed to this principle (cf. Tucker 2013; Huemer 2013a). 6 Although the current dispute mainly focuses on visual appearances, cognitive penetrability might affect other types of perceptual appearance. It is also in principle conceivable that phenomena analogous to cognitive penetrability could affect, for example, mnemonic, rational or moral seemings. 5 In

28

3 Cognitive Penetrability

Not just all cases of cognitive penetration (real or supposed) are deemed to be cases of bad cognitive penetration—i.e. situations in which the subject has no prima facie justification for the relevant belief. Some think that there are cases of good cognitive penetration—i.e. situations in which the subject does have prima facie justification (or more prima facie justification) for a perceptual belief just because she has cognitively penetrated seemings. Perceptual learning is often deemed to produce good cognitive penetration. Perceptual learning is a process based on training and experience that ends up causing changes in the subject’s perceptual-discriminating abilities. An example could concern the cognitively penetrated visual appearances of a trained radiologists who inspects X-rays to detect a bone fracture (cf. Siegel 2012). Another example could be Gus’ cognitively penetrated seeming that the pebble is gold in Prospectors. Some epistemologists suggest that there may be cases of good cognitive penetration independent of perceptual learning. Lyons (2011) imagines a scenario in which a subject—let’s call him John—gets his snake-detection skills sharpened up because his perceptual appearances are penetrated by his fear that there are snakes around. Suppose for example that, due to his fear, all snake-shapes stand out in John’s visual experience, so that he can see the reptiles despite their camouflages.7 The authors who appeal to cognitive penetrability to challenge phenomenal conservatism primarily adduce alleged counterexamples based on cases of bad cognitive penetration. Some also insist that phenomenal conservatism is problematic because it cannot spell out the divergent epistemic consequences of bad and good cognitive penetration (cf. Lyons 2011). The objections to phenomenal conservatism that adduce cognitive penetrability come from both the externalist and the internalist camp. That internalists level objections of this type is remarkable. There is in fact a striking similarity between a perceptual appearance that S would have if S were in a sceptical scenario (such as the Cartesian demon scenario or the Matrix scenario), and an appearance of S cognitively penetrated. Both these appearances would have anomalous aetiologies. In the first case, the aetiology would be anomalous because the distal cause of S’s appearance would be unnatural (e.g. if S’s visual experience of a cat were caused by a demon, the distal cause of S’s appearance would be the demon rather than a cat). In the second case, the aetiology would be anomalous because a mental state of S would interfere with S’s normal causal chains that produce perceptual appearances (e.g. in Angry Jack, Jill’s belief that Jack is angry interferes with Jill’s normal visual processes). Internalists generally agree that if S were in a sceptical scenario, the anomalous aetiologies of S’s seemings would not affect the justifying power of these appearances. A frequently adduced reason is that the aetiologies wouldn’t be reflectively accessible to S. One might thus expect that internalists also agree that if S’s appearances were cognitively penetrated, their anomalous aetiologies wouldn’t affect their justifying power. For these aetiologies (at least segments of them) wouldn’t be reflectively accessible to S. However, certain self-avowed internalists—called by

7 Note

that this might be argued to be a mere attentional effect unrelated to cognitive penetration (cf. Siegel 2017: 123–125).

3.2 The Epistemic Problem of Cognitive Penetrability

29

Lyons (2016) inferentialists—insist that if S’s seemings were badly cognitively penetrated, the flawed aetiologies of these appearances would undermine their justifying power. Since these internalists conceive of the aetiologies of perceptual appearances as sequences of mental states, they are committed to embracing a mentalist variant of internalism (cf. Lyons 2016). In the next two sections, I criticize externalist and internalist cases against phenomenal conservatism that appeal to cognitive penetrability. I will dedicate much more attention and space to internalist arguments because—as clarified in Chap. 1— in this work I presuppose that internalism is correct.8 Before starting off, let me call attention to a crucial premise of this debate. As said, the cognitive penetrability of perceptual appearances is a controversial empirical hypothesis. Even so, phenomenal conservatives couldn’t hope to dismiss objections that appeal to cognitive penetrability by simply emphasizing that it is an unattested phenomenon. The problem is that phenomenal conservatives conceive of (PC) as an a priori true principle (cf. Huemer 2007). Hence, even if our actual hardwiring ruled out cognitive penetrability, the mere conceptual possibility of a rational subject whose seemings lack justifying power because they are cognitively penetrated would be incompatible with (PC) (cf. Markie 2013; Georgakakis and Moretti 2019; Tucker forthcoming).

3.3 The Reliabilist Account Externalists typically explain the asserted intuition that cognitive penetration alters the justifying power of appearances by invoking reliabilism. They contend that the malformed aetiologies of cognitively penetrated appearances make the belief production processes based on those appearances unreliable. Since reliabilism in essence equates the justification of a belief with the reliability of the process that has produced it, the conclusion is that beliefs based on cognitively penetrated appearances are unjustified. More accurately, externalist reliabilists contend that bad cognitive penetration affects reliability—and thus justification—negatively, by destroying or reducing it, whereas good cognitive penetration affects reliability—and thus justification—positively, by generating or enhancing it (see mainly Lyons 2011, 2016; Ghijsen 2016).9

8 In

epistemology, cognitive penetrability has also been approached from perspectives that don’t fit well the dichotomy internalist/externalism. (I include Brogaard 2013 in this group. Brogaard describes herself as an epistemic internalist but the view she defends—sensible dogmatism—has a marked reliabilist component.) For lack of space I cannot analyse these interesting accounts here. For an overview see Georgakakis and Moretti (2019). 9 There are two basic forms of reliabilism about justification: process reliabilism (outlined in Sect. 2.1) and indicator reliabilism. The latter states that S’s belief that P is prima facie justified just in case S bases it on a mental state M that reliably indicates P; where M reliably indicates P just in case in most of the closest possible worlds in which a subject has M, P is true. I focus on process reliabilism but my claims can be re-phrased to apply to indicator reliabilism.

30

3 Cognitive Penetrability

For example, reliabilists would maintain that the reason why it is counterintuitive that, in Angry Jack, Jill’s perceptual belief is prima facie justified is the following: the belief production process deployed by Jill is unreliable because it encompasses the initial ungrounded belief of Jill that Jack is angry. Reliabilists would also maintain that the reason why it is intuitive that in Prospectors Gus’ perceptual belief is prima facie justified while Virgil’s identical belief is not is the following: the belief production process used by Gus is reliable, while the one used by Virgil is unreliable. Gus’ belief production process in fact exploits detection skills that Gus is endowed with, whereas Virgil’s belief production process is ultimately based on a mere desire of Virgil. Cases of good cognitive penetration independent of perceptual learning— such the snake-detection one—could be handled likewise. Reliabilists could claim that in the situation imagined by Lyons, John’s perceptual beliefs that there are snakes in the trail are prima facie justified because John’s fear has sharpened up his snake-detection skills, and so enhanced the reliability of the processes that yield the correlated beliefs. The reliabilist contends that phenomenal conservatism cannot explain the intuition that cognitive penetration affects negatively or positively the ability of appearances to supply prima facie justification. For these effects of cognitive penetration essentially rest—according to the reliabilist—on reflectively inaccessible features of seeming production processes (i.e. their reliability), whereas (PC) makes prima facie justification depend on reflectively accessible factors only (see mainly Lyons 2011). If this is true, the adequacy of phenomenal conservatism is in peril. But phenomenal conservatives have good counterarguments at hand. Firstly, they can put reliabilists under pressure by directly attacking the adequacy of the reliabilist conception of epistemic justification, which is notoriously afflicted by longstanding problems such as the new evil demon and the generality one (see Goldman and Beddor 2016 for useful discussion). More interestingly for this chapter’s theme, phenomenal conservatives can adduce thought experiments to show that our epistemic intuitions about good and bad cognitive penetration don’t get really explained or clarified by considerations about the reliability of one’s belief production processes. Take again Virgil from Prospectors but consider this variant of it. Virgil is unable to recognize the distinctive features of gold. However, it is true that whenever he entertains an appearance that a pebble is gold cognitively penetrated by his desire to become rich, very often the pebble is gold. Suppose that a benevolent demon incessantly intervenes to make this possible (cf. Fumerton 2006: 80). In this new scenario, Virgil’s perceptual belief production process is reliable when Virgil’s belief is based on an appearance that an object is gold cognitively penetrated by his desire to be rich. Suppose Virgil has one of these appearances and bases a belief that the object is gold on it. Many would say that Virgil’s belief still looks epistemically defective despite the belief production process used by Virgil is reliable. This suggests that the reliabilist account of our intuitions about the epistemic consequences of bad cognitive penetration misses the target. One can criticize the reliabilist account of our intuitions about the different epistemic consequences of good and bad cognitive penetration by exploiting Markie

3.3 The Reliabilist Account

31

(2013)’s variant of Prospectors.10 Suppose again Gus is an expert prospector and Virgil a novice. Unbeknownst to them, however, they have always been living in the Matrix. Imagine both Gus and Virgil have a perceptual seeming that a given nugget is gold. Gus’ appearance is caused by the external stimuli due to the Matrix11 and partly by his expertise and background knowledge. This looks like a case of good cognitive penetration. Virgil’s seeming is caused by the same external stimuli from the Matrix and partly by his desire to become rich. This looks like a case of bad cognitive penetration. Imagine that both Gus and Virgil come to believe that the nugget is gold on the grounds of their seemings. The epistemic standing of Virgil’s belief is intuitively worse than the epistemic standing of Gus’ belief. But—note— the perceptual belief production processes used by both subjects are now completely unreliable. So reliability cannot be adduced to explain this difference. This example suggests that the reliabilist account of our intuitions about the divergent epistemic consequences of bad and good cognitive penetration is flawed. In conclusion, the reliabilist explanation of our epistemic intuitions about cognitive penetrability—at least in its basic form—appears flawed. Some reliabilists have put forward sophisticated variants of their basic conception of epistemic justification in which the link between one’s justification and the reliability of one’s mental processes is more indirect and complex (see for instance Goldman 1979; Comesaña 2002, 2010; Bergmann 2006). Reliabilists might turn to these views to try to illuminate the epistemic intuitions about cognitive penetrability elicited by thought experiments like those above. A concern is, however, that these more sophisticated conceptions of justification would produce more complicated and less immediate and straightforward explanations of those intuitions, which ultimately wouldn’t be preferable to internalist explanations. For more specific objections see Tucker (2014a).12

3.4 The Inferentialist Account Let’s turn to the family of internalist mentalist views that Lyons (2016) calls inferentialism—namely, Siegel (2012, 2013b, 2017)’s process inferentialism, and McGrath (2013a, b)’s and Markie (2013)’s versions of evidence inferentialism. Inferentialism aims to account for our epistemic intuitions about cognitive penetrability by appealing to the rationality of the subject’s cognitive processes rather than their reliability. Its pivotal assumption is that when S has a cognitively penetrated appearance, the

10 See

also McGrath (2013b). stimuli are identical to those that his brain would receive in the real world. 12 See also Tucker (2014b), Vahid (2014) and Siegel (2017: Chap. 6). 11 These

32

3 Cognitive Penetrability

appearance always or typically results from S’s doing.13 A consequence is that the epistemic standing of this appearance and its justifying power can be appraised by assessing whether S has produced it in a rational way. This can be done in roughly the same way in which we appraise the epistemic standing and the justifying power of a belief of S by assessing how S has inferred it from other beliefs (cf. Lyons 2016). Thus, whether a cognitively penetrated appearance can justify believing its content hinges on whether the appearance’s aetiology is epistemically rational. The inferentialist opposes phenomenal conservatism by insisting that since (PC) doesn’t make the justifying power of appearances depend on their aetiologies, phenomenal conservatism cannot explain the epistemic bearing of cognitive penetrability. Inferentialism isn’t vulnerable to the objections I have raised against the reliabilist account of cognitive penetrability. For the rationality of the aetiologies of perceptual appearances (supposing they can be rational) is logically independent of their reliability. For instance, the inferentialist could claim that Virgil’s belief that his pebble is gold based on a badly cognitively penetrated appearance would be unjustified because its aetiology is irrational even if the appearance were yielded by an actually reliable process thanks to a benevolent demon’s intervention. Process Inferentialism Siegel offers the most developed inferentialist account to date. She accepts a very liberal notion of inference, according to which inferring is just a distinctive type of responding to an informational state that produces a conclusion epistemically dependent on that state. The informational state may be unconscious and can consist of a belief (or a similar doxastic state), an appearance, a fear, a desire, or a combination of these elements. The conclusion can be a doxastic state or an appearance. The informational state and the response may occur simultaneously. Finally, the process of inferring doesn’t have to feel like anything, it need not leave any mark in consciousness (cf. Siegel 2017: Chaps. 2–8). Siegel doesn’t analyse any further this general characterization. Yet she provides examples to distinguish this type of transition from those that fail to respond informational states (e.g. bypassing evidence), those that respond to non-informational states (e.g. associating thoughts), and those that respond non-inferentially to informational states (e.g. directing one’s attention) (cf. 2017: Chap. 5).14 Let’s examine how an inference thus conceived and performed by a subject S would bear on the epistemic standing of both its conclusion and S herself. Siegel dubs epistemic charge the property of a mental state of S in virtue of which the state and S are epistemically appraisable (cf. 2017: Chaps. 2–3). Epistemic charge is a property more general than epistemic justification. In the case of belief, justification and epistemic charge go hand-in-hand; no belief has one of these unless it has both. But mental states that cannot be justified—like appearances—can also have epistemic 13 Perhaps Markie (see below) would claim that this assumption is true only when cognitive penetration is good, and that bad cognitive penetration typically doesn’t result from S’s doing. 14 Siegel doesn’t consider Helmholtz-style “unconscious inferences” to be proper inferences, for these transitions are typically conceived of as mere causal processes that don’t redound on the subject’s rational standing (cf. Siegel 2019).

3.4 The Inferentialist Account

33

charge. Like justification, epistemic charge can be transmitted from a mental state to another. Depending on the sign of the charge, a mental state is enabled to justify other mental states. If a seeming that P of S has a positive charge, S and the seeming enjoy a positive epistemic status, so the seeming can prima facie justify S’s believing P. Conversely, if S’s seeming that P has a negative charge, S’s and the seeming’s epistemic statuses are downgraded, so the seeming cannot even prima facie justify S’s believing P. The epistemic charge of a mental state can be modulated by the rationality of the mental process that has produced the state. In particular, the epistemic charge of an appearance can be modulated by the rationality of the inference drawn by S that yields it.15 A good inference concluding with a seeming bestows on it a positive charge. A poor inference concluding with a seeming bestows on it a negative charge (cf. 2017: Chaps. 2, 4 and 6–7).16 For Siegel, good cognitive penetration results from a good inference by S, and bad cognitive penetration results from a bad inference by S. When S’s experiences are badly cognitively penetrated, S’s overall outlook on the world sustains itself circularly; in the sense that it creates illusory appearances that represent the world in the way the outlook characterizes it. This is why Siegel says that badly cognitively penetrated appearances are hijacked by the subject’s outlook (cf. 2017: Chap. 1).17 Siegel (2017: Chaps. 6 and 8) offers some examples of poor inferences ending with appearances that would instantiate epistemic flaws such as circularity, inherited inappropriateness or jumping to conclusions. For ease of illustration, let’s dwell on an example of the last type. Suppose you infer a mental state B from a mental state A, but that the content of A doesn’t support the content of B. This poor inference can downgrade the epistemic status of B and disable its justifying power. All this is uncontroversial if A and B are beliefs. Siegel contends that jumping to conclusions can also be instantiated by an inference from a perceptual appearance to another perceptual appearance. An example would be a variant of Angry Jack in which Jill entertains an appearance B that Jack is angry by jumping to conclusions from her appearance A that Jack has a neutral (non-angry) face, due to Jill’s unfounded fear that Jack is angry. Siegel concedes that since A and B would occur in Jill’s mind simultaneously, what Jill’s fear brings about is not a temporal transition from A to B but a dependence relation of B on A (such that, presumably, B exists in virtue of A).18 Thanks to this dependence relation, B would qualify as a response by Jill to the informational state A. So Jill can be described as inferring B from A. Since the content of A doesn’t actually support the content of B, this is a bad inference that downgrades B’s epistemic status and makes B unable to justify Jill’s belief that Jack is angry (cf. 2017: 117–119). 15 Inferences

modulate but don’t generate the epistemic charge of appearances. Siegel takes the epistemic charge of appearances to stem from what I have called, in Sect. 2.2, their phenomenal force (cf. 2017: 43–47). 16 For Siegel there are various ways in which experiences can be subject to rationality—being the conclusion of an inference is one of them (cf. 2017: 21). 17 Siegel includes among the hijacked appearances also those non-cognitively penetrated that arise from S’s biased selection processes. 18 From now on, I will use ‘transition’ to refer to atemporal dependence relations of this type too.

34

3 Cognitive Penetrability

In relation to this example, Siegel makes two additional claims. She suggests that a subject (e.g. Jill in another variant of Angry Jack) could have a perceptual appearance as a result of jumping to conclusions from an unconscious perception (cf. 2017: 123, see also 101–103 and 114). Siegel also contends that fears and desires could legitimately work as inputs to inferences (like beliefs and perceptual appearances), rather than counting as “disturbing factors” (cf. 2017: 123 and Chap. 8). Although original and interesting, Siegel’s account of cognitive penetrability is troubled by serious difficulties. One turns on Siegel’s claim that to infer a state B form a state A, S doesn’t need to take A to support B—where S takes A to support B just in case S appreciates (in some non-factive sense of this term) that the content of A is evidence for the content of B (cf. 2017: 95). Siegel’s claim is worrisome because many if not most philosophers have long accepted the taking condition: (TC) S infers B from A only if S comes to entertain B because of her taking A to support B.

See for instance Locke (1689/2008: book IV, Ch. 17.2), Frege (1979), Peirce (1905), Russell (1920), Stroud (1979), Broome (2014), and Boghossian (2019). There are various reasons why (TC) looks plausible. To begin with, it helps us produce a straightforward explanation of why inferring is more than associating thoughts: when we merely associate two thoughts, we don’t take one to support the other. We take a thought to support another when we infer the latter from the former. Siegel has an alternative explanation. She proposes that when we associate two thoughts, we just respond to concepts involved in the first thought. In other words, the second thought is triggered only by concepts, and not by predicativelystructured (or truth-evaluable) segments of the first thought. (Imagine for example that while observing at dusk that the sky is growing dark, I recall that I need to buy lightbulbs. For Siegel, this would be a mere association of thoughts elicited by, say, my linking the concept of darkness with the concept of light.) On the other hand, when we genuinely infer a thought from another—according to Siegel—we respond to predicatively-structured segments of the first thought (cf. 2017: 87). I don’t think Siegel’s proposal clarifies the difference between associating and inferring. Imagine for example that my thought that (P) Mickey Mouse has a dog draws my attention to the fact he has no cat, and so it triggers my thought that (Q) Mickey Mouse has no cat. Here it is my thought that P as a whole that triggers my thought that Q. So I’m responding to the predicatively-structured information that P by entertaining Q. But this is not an inference. (TC) also supplies a clear answer to the question why inferring is, intuitively, a form of epistemic agency. If (TC) is satisfied, it is the agent who carries out the inference once she appreciates that the premises support the conclusion. This appeal to

3.4 The Inferentialist Account

35

the agent also helps us illuminate why inferences are more or less rational. Inferences are so because the agent’s doing can be more or less epistemically appropriate.19 One could try to reject the intuition that inferring is a form of epistemic agency or the intuition that the rationality of inference arises from the agent’s doing. Yet Siegel grounds her explanation of cognitive penetrability in both intuitions. Since Siegel rejects (TC), it is unclear how her explanation could account for these intuitions. Appealing to (TC) also offers a neat explanation of what distinguishes an arational mental transition from a poor inference: while an arational transition doesn’t satisfy (TC), a poor inference does. For example, in a poor inference, the subject may carry out the transition in virtue of her mistakenly appreciating that the premises support the conclusion.20 One might try to deny that there is a substantive difference between arational transitions and poor inferences. But Siegel couldn’t. For she explicates bad cognitive penetration in terms of poor inferences, which she tells apart from arational transitions. Since she rejects (TC), it is unclear why, in her account, arational mental transitions are substantially different from poor inferences.21 Some philosophers think that (TC) is implausible because it involves overintellectualisation. These philosophers observe that small children and perhaps higher animals can draw inference though they lack conceptual resources to grasp notions— like that of reason, rational support, premise and conclusion—which the subject must possess, according to these authors, to comply with (TC) (cf. McHugh and Way 2016). Another criticism states that (TC) cannot be satisfied because attempting to do so would elicit a vicious infinite regress (cf. Boghossian 2014 and McHugh and Way 2016). My view is that these objections miss the target, for they interpret the claim that S should appreciate that the content of A is evidence for the content of B as stating that, substantially, S should rationally believe or accept by reflection that the first content supports the second. But there is no need for such a demanding reading. These difficulties fade away when this claim is interpreted as the more modest proposition that it should seem to S that the content of B is likely or plausible in light of the truth of A.22 Another possible objection to (TC) comes from the observation that inferences we routinely draw don’t seem to satisfy it. As Boghossian (2019) suggests, however, phenomenology can be deceptive. For instance, non-occurrent beliefs are an important 19 Interestingly,

Siegel (2019) countenances that a subject S might somehow fulfil (CT) at subpersonal level. However, she acknowledges that this not the standard view, for the fulfilment of (CT), in this case, could no longer be adduced to explain the rationality of inference as resulting from S’s doing. 20 As Georgakakis has noted, a subject S may alternatively happen to correctly appreciate that P supports Q, while mistakenly coming to believe an unsupported proposition R instead of Q. I think this would count as an incorrect inference only if S’s appreciating that P supports Q were an essential part of the deviated causal chain that produces her belief that R. So, it would still be true that S believes R because of her appreciating thatP supports Q. 21 See Boghossian (2019) for further reasons in support of (CT). 22 McHugh and Way (2016: 319–320) insist that this won’t solve any problem. But their arguments are just sketched. My impression is that the view presented in Huemer (2016) isn’t affected by the difficulties they allude to.

36

3 Cognitive Penetrability

part of our mental and rational life, though they come without any phenomenology. We become aware of these beliefs only if we are stimulated with questions or external circumstances. I think that the apparent unfulfilment of (TC) can in most cases be explained along similar lines. When we genuinely draw an inference, irrespective of the phenomenology, if asked why we have come to believe the conclusion, we would normally realize that we have done so because we have taken the premises to support it.23 Siegel attempts to reject (TC) by adducing examples of genuine inferences that would not satisfy this constraint. Let’s consider one of them. (Siegel’s cases aren’t essentially dissimilar, so by criticizing one, I’ll criticize all of them.) Kindness The person ahead of you in line at the Post Office is finding out from the clerk about the costs of sending a package. Their exchange of information is interspersed with comments about recent changes in the postal service and the most popular stamps. As you listen you are struck with the thought that the clerk is kind. You could not identify what it is about the clerk that leads you to this thought. Nor could you identify any generalizations that link these cues to kindness. Though you don’t know it, you are responding to a combination of what she says to the customer, her forthright and friendly manner, her facial expressions, her tone of voice, and the way she handles the packages. (2017: 95)

In Kindness you are meant to reach a conclusion B—your belief that the clerk is kind—from a specific set A of cues that you possess. Siegel insists that this is a genuine inference (cf. 2017: 96). She stresses that in the imagined circumstances, nevertheless, you cannot identify the A-cues you are responding to. So you don’t believe that they indicate kindness. Nor do you have a seeming or an intuition that these A-cues indicate kindness. Finally, you have no disposition to judge that these A-cues are cues for kindness. You might be disposed to say ‘people who act like that are kind’, but you cannot understand what ‘like that’ picks out. Siegel concludes that in Kindness, you infer that the clerk is kind but you don’t fulfil (TC). For ‘there is no X such that [you] know [you are] responding to X in forming [your] belief’ (2017: 97).24 I think Kindness offers no evidence for believing that inferring doesn’t require satisfying (TC). The basic problem is that the transition described in Kindness doesn’t look like a process that most of us would name ‘inference’. If one doesn’t already presuppose a very liberal conception of inference that rejects (TC)—like Siegel’s— one won’t consider this transition to be an inference. Many would rather claim that Kindness depicts a typical case of non-inferential transition: one from a seeming to a belief. Indeed, it is natural to think that in Kindness, you would come to believe that the clerk is kind because it would seem to you that the clerk is so.25 (You could have such a seeming even if you were unable to individuate specific features indicating 23 I wrote ‘normally’ because some psychological findings suggest that in certain cases, when people think they are describing the bases of their beliefs and decisions, they are instead merely theorizing or confabulating (see for instance Johansson et al. 2006). 24 Siegel (2019) returns to Kindness but her conclusion remains unchanged. 25 Cf. Boghossian (2019).

3.4 The Inferentialist Account

37

that the clerk is kind.) To forestall criticism of this type, Siegel (2019) stipulates that, in Kindness, the clerk’s kindness isn’t represented in your experience. So ‘you are struck with the thought that the clerk is kind’ though it doesn’t seem to you that the clerk is kind. I wonder, however, if a case of this sort is psychologically possible and, if it is, whether the belief that the clerk is kind shouldn’t be taken to be a pathological and so irrational state. Siegel doesn’t accept (TC) presumably because she sees that its implementation would disqualify cognitively penetrated perceptual appearances from counting as conclusions of inferences. Consider her account of Jill’s fear-penetrated appearance that (Q) Jack is angry. Jill is said to infer it from her appearance that (P) Jack has a neutral face. But it is implausible that Jill could have the appearance that Q because she appreciates that P is evidence for something. It is strongly intuitive, instead, that in the scenario described by Siegel, the psychological processes yielding Jill’s appearance that Q are secluded from whatever Jill might concurrently appreciate about the evidential force of P. As Lyons (2016) observes, the processes that produce perceptual experiences are fast, automatic, determined by domain specific principles proprietary to the perceptual system in question, relatively immune to cognitive penetration from occurrent beliefs and goals and the like, largely unconscious in [their] inner workings, and performed by systems that are relatively independent and isolable from other cognitive mechanisms. (16)

So whether Jill appreciates that P supports Q or another proposition, or she has no attitude about these issues, these facts would have no impact on the processes in charge of the production of Jill’s perceptual appearance that Q. These considerations expose another problematic feature of Siegel’s account. Siegel insists that cognitively penetrated appearances result from “inferences” that the person carries out ‘covertly, silently, and unreflectively’ (2017: 17). Yet Lyons’ description strongly suggests that these transitions, if existing at all, must take place at subpersonal level. The subject as such doesn’t perform them—rather, it is some perceptual system of the subject that makes these transitions for her.26 Given this, it is hard to understand how the subject could modulate the justifying power of her seemings by drawing relevant “inferences”, along the lines outlined in Siegel (2017). It is plausible that the subject cannot do it. Because of this problem, Siegel’s account of the epistemic consequences of cognitive penetration appears flawed. Let me make a final objection to Siegel’s account. Siegel claims that whether or not a cognitively penetrated seeming that P of S prima facie justifies S’s believing P depends on whether the seeming has positive or negative epistemic charge. Saying that a seeming of S has positive or negative epistemic charge is substantially the same as saying—for Siegel—that S holds the seeming with a high or low degree of rationality, as the conclusion of a good or bad mental transition (cf. 2017: Chap. 2). Siegel is thus committed to maintaining that cognitively penetrated appearances can be held by S (more or less) rationally. Since the type of rationality Siegel speaks about is epistemic, and epistemic rationality is responsive to defeating evidence, Siegel is 26 I

follow Lyons (2016) and Tucker (forthcoming) in interpreting ‘personal’ as ‘attributable to the subject’ and ‘subpersonal’ as, roughly, ‘attributable only to parts of the whole subject’.

38

3 Cognitive Penetrability

also committed to the claim that any cognitively penetrated appearance that P held by S rationally is responsive to defeating evidence. This means that S must in principle be able to revise her appearance in light of any evidence E that discredited P’s accuracy or the reliability of the process that has produced the appearance. (Similarly, if a belief that P is held byS rationally, S should in principle be capable of revising it in light of any defeating evidence.) However, it is very dubious that S could revise— i.e. change or stop having—an appearance that P in light of E.27 As observed in Sect. 2.2, our seemings typically don’t change even when we become aware that they are deceptive or unreliable (cf. Brogaard 2013), and it would be only capricious to insist that cognitively penetrated seemings behave differently.28 Since S could not revise any cognitively penetrated appearance in light of defeating evidence, S could not hold any such appearance rationally—S could entertain cognitively penetrated appearances only arationally.29 This flies in the face of Siegel’s contention that cognitively penetrated appearances can be held rationally. Siegel (2017: 34–35) is aware that her account lays itself open to criticism of this type and attempts a response. She reminds us that delusional beliefs—such as the ones in Capgras syndrome or in schizophrenia—cannot be revised by the subject in light of defeating evidence. For these pathologies prevent the subject from doing it. Siegel stresses that these delusional beliefs nonetheless appear to be paradigms of irrationality—namely, poor rationality—rather than arationality. She infers from this that epistemic rationality doesn’t require revisability. I’m not convinced by this response. The intuition that S’s holding a belief with some degree of rationally requires S’s capability of revising it in the presence of defeaters is very widely shared in epistemology. If a plausible interpretation of Siegel’s apparent counterexample compatible with this entrenched intuition were available, we should embrace it. Let me offer such an interpretation. Siegel would seem to confuse the notion of rationality as epistemic rationality—the one at stake here—with the notion of rationality as coherence between propositional attitudes (as characterized above in Sect. 2.3). In accordance with the second notion, a subject S can be said to be irrational for having two incoherent propositional attitudes simultaneously—e.g. a belief that P and an appearance that ¬P. S’s being irrational in this sense doesn’t seem to require S’s being able to revise her incoherent attitudes. I suggest that people affected by unrevisable delusional beliefs are irrational just because they have incoherent propositional attitudes (e.g. a belief that P and evidence of some type that ¬P). But they are not irrational in the sense of holding their delusional beliefs with a low degree of epistemic rationality. When we focus on 27 S

could certainly look away or cover her eyes. But the resulting perceptual change wouldn’t intuitively count as a rational revision (cf. Siegel 2017: 34). 28 One might suspect that if they are revisable, they aren’t proper seemings but doxastic states. 29 Siegel insists that although S couldn’t make herself stop having a given perceptual appearance in light of defeating evidence, S can normally “revise” the appearance in the sense of ceasing to rely on it in reasoning and action. Siegel also suggests that through perceptual learning, S might arrive at controlling the conditions that tend to give rise to a given type of appearance (cf. 2017: 35–37). All this looks plausible. However, it doesn’t seem to me that these forms “revisability” and control are good diagnostics of the fact that the relevant appearances are held by S rationally.

3.4 The Inferentialist Account

39

epistemic rationality, we should probably conclude that these subjects entertain their delusional beliefs arationally, since they cannot revise them in light of recalcitrant evidence. Evidence Inferentialism McGrath’s and Markie’s versions of evidence inferentialism explain the epistemic downgrade of badly cognitively penetrated appearances specifically in terms of the basing relation. These authors think that an appearance can be based on another appearance or experience in the same way as a belief can be based on another belief. The downgrade would happen when the content of the experience on which the appearance is based isn’t adequate evidence for the content of the appearance. McGrath (2013a, b) suggests that there are two types of perceptual seemings: receptive, which represent only low-level properties, and non-receptive, which represent higher-level properties (cf. 2013b). McGrath submits that we produce nonreceptive seemings from receptive seemings through a sort of inferential process. Receptive seemings are the inputs and non-receptive seemings are the outputs of quasi-inferences. According to McGrath, (QI) 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.30 (2013a: 237)

A transition counts as an inference by the person—for McGrath—only if (i) its input and output states are mental states of the person (rather than a sub-personal system) and (ii) there is an explanation ‘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’ (2013a: 238).31 (ii) very strongly suggests that McGrath accepts that inferences should comply with (TC). McGrath thinks that the tenet of phenomenal conservatism (PC) is true of all receptive seemings—hence, any receptive seeming that P of S provides S with prima facie justification for believing P.32 Yet he submits that (PC) is false when applied to non-receptive seemings. The reason being that while receptive seemings are given to us, non-receptive seemings are produced by us through quasi-inferences. Because of this, McGrath contends—following Siegel—that non-receptive seemings can give us justification for believing their contents only if the relevant quasi-inferences that we perform are good. Good and bad quasi-inferences can be characterized by a comparison with good and bad inferences. McGrath’s grounding intuition is that a good quasi-inference from a seeming that P to a seeming that Q transmits justifying power from the first to the second seeming in the same way as a good inference from a belief that P to a belief that Q transmits justifying power from the first to the second belief. Justifying power 30 For

McGrath quasi-inferences also include mere dependence relations between seemings. suggests such an explanation could appeal, for instance, to the person’s grasp (good or faulty) of the support that one proposition gives to the other, the person’s background information, or the cognitive states that make the person jump to conclusions. 32 I’m assuming that S’s seeming that P is clear and firm. 31 McGrath

40

3 Cognitive Penetrability

transmits—whether we are speaking of seemings or beliefs—just in case the first state has justifying power and P sufficiently supports Q. McGrath suggests that all poor quasi-inferences are cases of jumping to conclusions caused by some interfering factor. They are cases in which S quasi-infers a seeming that Q from a seeming that P, where P doesn’t support Q, because S is misguided by other mental states—e.g. fears, desires, expectations, etc. (cf. 2013b).33 McGrath explains our intuitions about the epistemic bearing of bad cognitive penetration by appealing to quasi-inference: the epistemic downgrade of a cognitively penetrated seeming happens when the quasi-inference that has produced it is poor. Take Angry Jack for example. McGrath suggests the following rendition: Jill has a receptive seeming that (P) Jack’s face displays certain neutral features. Nevertheless, under the influence of her unjustified belief that Jack is angry, Jill quasi-infers from the first appearance a non-receptive seeming that (Q) Jack is angry. This is a bad quasi-inference because P isn’t evidence that Q. So Jill’s seeming that Q cannot justify Jill’s belief that Q. McGrath’s framework could be used to explain why good and bad cognitive penetration seem to have divergent epistemic consequences. Take again Prospectors. Gus is the expert and Virgil the novice. McGrath could argue that Gus’ belief that (Q) this pebble is gold is prima facie justified by Gus’ non-receptive seeming that Q because the latter seeming results from a good quasi inference. Specifically, Gus would entertain a receptive seeming that (P) this pebble has features Fs. From this, Gus would quasiinfer the non-receptive seeming that Q. Since P in conjunction with Gus’ background information supports Q, this is a good quasi-inference. What said doesn’t apply to Virgil, who doesn’t possess Gus’ background information. Virgil’s quasi-inference from his receptive seeming that P to his non-receptive seeming that Q would thus be a bad one. For P alone isn’t evidence that Q. This would explain why Virgil’s belief that Q isn’t justified by his seeming that Q. McGrath’s evidence inferentialism can be seen as a special version of Siegel’s process inferentialism, for it concentrates on one type of “inferences” (the appearanceto-appearance ones) and one way in which these transitions can fail (jumping to conclusions). Given its limited ambitions, McGrath’s account isn’t exposed to all the problems that afflict or may afflict Siegel’s. For instance, McGrath doesn’t claim that quasi-inferences are genuine inferences. Nor does he endorse controversial theses such as that there are unconscious experiences or that fears and desires are proper inputs to inference. McGrath’s account is nevertheless problematic in various respects. For example, note that quasi-inferences involve two states (i.e. two seemings) both with propositional content. Cognitive penetration might nevertheless depend on a transition from a state without propositional content to a seeming. Suppose for instance that, in Angry Jack, Jill has no receptive seeming about Jack’s face. She has some raw 33 McGrath (2013a) says that these are cases of free enrichment—i.e. cases in which a non-receptive

seeming is freely enriched due to an interfering state.

3.4 The Inferentialist Account

41

visual impressions lacking representational content which, together with her unjustified belief that (Q) Jack is angry, brings about her seeming that Q (cf. Lyons 2016). McGrath’s inferentialism couldn’t explain cases of this type. Another limitation of it hinges on the possibility that cognitive penetration could directly affect receptive seemings (or raw sensations). Suppose that, because of her unjustified belief that Q, Jill has a receptive seeming that (R) Jack’s face has anger features. If this is what actually happens in Angry Jack, McGrath should conclude that Jill is justified in believing that Q on the basis of her non-receptive seeming that Q. For this nonreceptive is supported by the receptive seeming that R (cf. Lyons 2016). Some would find this counterintuitive. McGrath’s evidence inferentialism is afflicted by more serious difficulties. McGrath is convinced—correctly, in my view—that if a person S performs an inference, S must treat the content of the input state as supporting the content of the output state. This presumably means that S must produce the output state because she appreciates that its content is supported by that of the input state. So McGrath endorses (TC) as a necessary condition for inference. Given this, the application range of McGrath’s framework cannot but be drastically limited. For only a very few seeming-to-seeming transitions, if any, qualify as quasi-inferences on (QI). Let me explain why. Consider that if S has been subjected to perceptual learning, the fact that S has a seeming that Q on the basis of a seeming that P might partly depend on her possessing a background belief that (RI) P’s truth is a reliable indicator of Q’s truth. Imagine for instance that P and Q respectively describe low-level and higher-level properties of birds, and that S has acquired the background belief (RI) in a birdwatching class before doing any fieldwork. Suppose that this background belief together with some fieldwork has eventually endowed S with a disposition to have a seeming that Q upon having a seeming that P. A seeming that Q acquired by S thanks to this disposition would be cognitively penetrated by S’s background belief (RI). Take now a transition from a seeming that P to a seeming that Q of this type in S’s mind. One could insist that if the seeming that P and the seeming that Q were replaced respectively by a belief that P and a belief that Q, the resulting belief-to-belief transition in S’s mind would satisfy (TC). For S would believe Q, via this transition, in virtue of her believing (RI), and so in virtue of takingP to support Q. If this conclusion is correct,34 McGrath’s framework might account for cases of cognitive penetration of this type. However, note that background beliefs of the sort just described are seldom possessed even by persons who have undergone various types of perceptual learning. We very rarely have beliefs about how low-level properties and higher level properties 34 This

might perhaps be questioned. One might rejoin that in these circumstances, though S would actually take P to support Q, it is unclear that S would come to believe Q in virtue of her taking P to support Q. For it is still unclear that in these circumstancesS would come to believe Q through her agency rather than a mere subpersonal mechanism of association partly shaped by (RI).

42

3 Cognitive Penetrability

represented in our experiences are linked with one another, although a large amount of information of this sort is implicitly and unconsciously stored in our subpersonal perceptual systems (cf. Lyons 2016). In all cases of cognitive penetration in which the relevant seeming-to-seeming transition doesn’t depend on some background belief of the type considered, if the seemings were replaced with corresponding beliefs, the resulting belief-to-belief transition would not satisfy (TC). Consider in fact that— as emphasized before—any transition from a seeming that M to a seeming that N in S’s mind would take place independently of any concurrent attitude of S about the relation of support between M and N. Thus, if the seemings were replaced with beliefs, the resulting belief-to-belief transition would also take place independently of any concurrent attitude of S about the relation of support between M and N. Hence, (TC) wouldn’t be satisfied. These seeming-to-seeming transitions wouldn’t qualify as quasi-inferences on (QI). McGrath’s framework appears unable to account for all cases of cognitive penetration of this type, which are the largest majority. Here is a last, probably lethal difficulty. McGrath is persuaded that since (QI) models quasi-inferences on proper inferences (between beliefs), and the latter are transitions by the person, quasi-inferences are also transitions by the person (cf. 2013a: 238). McGrath is indeed committed to the last claim because his explanation of the justifying power of non-receptive seemings rests—as we have seen—on the assumption that these seemings are produced by us. But this appears false. As Lyons (2016) has suggested, it is plausible that all transitions leading to the production of perceptual seemings happen at subpersonal level. Consider for example Gus in Prospectors. Despite Gus possesses accurate background information about how gold looks like and is able to detect gold successfully, it would be very odd to say that it is Gus—rather than his visual cognitive system—that works out an appearance that this nugget is gold from his appearance that this nugget has such and such features. (Although it would be fully appropriate to say that it is Gus who works out a belief that this nugget is gold from his belief that this nugget has such and such features.) If Lyons is right—and I think he is—quasi-inferences like those hypothesized by McGrath, which are attributable to the subject’s doing, simply don’t exist.35 Let’s turn to Markie’s version of evidence inferentialism. Markie (2013) presupposes a disunified view of perceptual appearance according to which perceptual appearance is a compound of sensations and seemings. To develop his account of cognitive penetrability, Markie mainly analyses Prospector. The impression that some have is that Gus’ seeming that (Q) this nugget is gold prima facie justifies Gus’ belief that Q, whereas Virgil’s (assumed) identical seeming that Q doesn’t even prima facie justify Virgil’s belief that Q. Markie thinks that the best way to explain this impression is the following: Gus does know how to visually identify gold, and his seeming that P is just an instance of this knowledge-how. That’s why Gus’ seeming does justify the corresponding belief. Contrarily, Virgil doesn’t know how to visually identify gold. So Virgil’s seeming that Q cannot be an instance 35 For

further criticism see Lyons (2016), Huemer (2013a) and Siegel (2013c).

3.4 The Inferentialist Account

43

of any such knowledge-how. That’s why Virgil’s seeming doesn’t even prima facie justify the corresponding belief. This diagnosis leads Markie to reject (PC) and endorse a variant according to which: (PC*) If S has an appearance that P that constitutes an instance of S’s knowledge-how to perceptually ascertain that P, S has thereby prima facie justification for believing P. (Cf. 2013: 250 and 262)

Markie calls epistemically appropriate the perceptual appearances that satisfy the left-hand side of (PC*) and so have justifying power. He elucidates the notion of knowledge-how to perceptually ascertain that P36 as follows: (KH) S knows how to perceptually ascertain that P if S has a disposition to entertain an appearance that P in response to S’s attending to certain features Fs of her experience, and S has this disposition in virtue of her having justified background information that these Fs indicate that P. (Cf. 2013: 263–264)

A few remarks are in order. According to Markie, S’s knowledge-how to perceptually ascertain that P need not come with S’s reliable practice. (In the Matrix, Gus would still know how to ascertain that a pebble is gold, though he would fail to ascertain it reliably.) Furthermore, S’s background information that Fs indicate that P need not consist of S’s conscious beliefs. S might just have evidence (e.g. experiences, justified beliefs) that provides S with prima facie justification for believing that if Fs are instantiated, then P. Finally, saying that S has the mentioned disposition in virtue of having this background information is not saying that this information causes S to have the disposition. It is only saying that the information helps to determine the character of the disposition and to sustain it. Suppose now S has a disposition to entertain an appearance that P in response to her attending to features Fs of her experience, and she has this disposition in virtue of having justified background information that these Fs indicate that P. Thanks to (KH), any so-generated appearance that P would count as an instance of S’s knowledge-how to perceptually ascertain that P. Furthermore, thanks to (PC*), any such appearance would be epistemically appropriate, to the effect that it would give S prima facie justification for believing P. We have seen how this framework applies to Prospectors. Markie suggests that it can be used to illuminate why Jill’s seeming that (P) Jack’s is angry, in Angry Jack, cannot justify Jill’s belief that P. In Markie’s interpretation, when Jill examines Jack’s face, she attends to certain features Rs. In the envisaged scenario, Jill has no background information that these Rs indicate that the subject—Jack, in the case in point—is angry. So Jill has no background information that these Rs indicate that P. However, Jill’s irrational belief that P causes her having a seeming that P. This seeming isn’t epistemically appropriate because it is not an instance of S’s knowledge-how to perceptually ascertain that P. Thus, this seeming cannot justify S’s belief that P. 36 Markie

(2013: 264) in fact focuses on the notion of knowledge-how to perceptually identifying something as being Q. The property I’m considering is an innocuous generalization of Markie’s.

44

3 Cognitive Penetrability

We saw that a possible drawback of McGrath’s evidential inferentialism is that it wouldn’t explain cognitive penetration depending on a transition from a state without propositional content to a seeming. Markie’s version of evidential inferentialism isn’t subject to this problem, for in Markie’s account, S’s seemings are thought to be based on experiences of S that don’t need to have propositional content. Yet Markie’s account is afflicted by another difficulty that troubles McGrath’s evidential inferentialism too: cognitive penetration could directly affect the features Fs of experience that S attends to, rather than the correlated seeming. If this happened, S’s seeming would count as epistemically appropriate, though it might appear intuitive that it isn’t so. (Suppose Jill attends to anger features, directly caused by her irrational belief.)37 Markie’s account isn’t hostage to problems related to (TC), as it doesn’t appeal to inferential transitions between beliefs. A serious difficulty of Markie’s account is, however, that what Markie (2013) considers to be knowledge-how doesn’t actually appear to be knowledge-how. Specifically, (KH) takes S’s having a disposition to entertain an appearance that P in given circumstances to be a condition sufficient for S’s having knowledge-how to perceptually ascertain that P. This principle is quite dubious. The problem is that knowledge-how is customarily conceived of as a type of goal-directed ability possessed by an agent (see for instance Carr 1981; Markie 2006; Cath forthcoming), but S’s disposition is not a goal-directed ability possessed by S as an agent. S’s disposition appears to be a mere “mechanistic” or “automatic” function of S’s perceptual cognitive system. Because of this problem, Markie’s account of the epistemic effects of cognitive penetration proves implausible. Let me expand on this. Whenever S knows how to do something G (e.g. riding a bike, counting, walking, etc.), S is capable of adopting a behaviour directed to achieve the goal G. Accordingly, S must have a correct understanding of G and be able to follow the norms that allow her to pursue it. The most basic of these norms have presumably this form: to attain G, in circumstances C, do D. For instance, if S knows how to ride a bicycle, S must be able to behave in a way directed to the goal of riding a bicycle. S must thus correctly understand this goal and be able to conform to the correlated norms. One of these norms—let’s call it N—might state this: in order to ride a bicycle, if you lose your balance while cycling, regain your balance by leaning in the opposite direction. Note that S’s following N doesn’t require S to make an explicit reasoning that appeals to N to decide what to do (cf. Markie 2006). The only thing required of S in order to follow N is presumably this: whenS finds herself off balance while cycling, as long as S intends to ride the bicycle, S should somehow see or feel 38 that she needs to lean in the opposite direction to regain balance, and she should act accordingly. (This seems to apply, mutatis mutandis, to any norm of knowledge-how in general.)

37 Markie 38 Cath

way.

(2013: 266) bites the bullet and insists that this upshot is perfectly acceptable. (2012) for instance suggests that S should have a seeming that she ought to act in a certain

3.4 The Inferentialist Account

45

When S exercises her knowing how to do something G, if S were asked why she has behaved in a given way, in some cases S could correctly explain, upon reflection, why she has behaved that way by giving an approximate description of the norms she has followed. For instance, if asked why she has leant left while riding her bike, S might say that she has done so because she needed to regain balance (cf. Markie 2006). In other cases, S’s vocabulary may not be rich enough to describe the norms S has actually followed. Imagine the question asked to S concerns her knowing how to whistle a melody.39 Furthermore, S might lack reflexive or linguistic abilities necessary to supply explanations altogether. Imagine S is a small child who knows how to clap.40 Suppose now that S has an appearance that P as a result of her exercising a disposition of the type described in (HK). Markie would maintain that this appearance is an instance of S’s knowing how to perceptually ascertain that P. This seems false. It is very plausible that if S were asked why she has brought about the appearance that P upon attending to features Fs of her experience, S could offer no description of the norms she has followed. This wouldn’t depend on S’s inadequate vocabulary or lack of introspection. The reason would simply be that the seeming that P would pop up in S’s mind mechanically, without S following any norm to generate it (not even in the undemanding sense described before). The production of this seeming would not be accompanied in S’s mind by the characteristic phenomenology of goal-directed acts. All this strongly suggests that the production of this appearance would not involve S as an agent. Since S’s seeming that P would not result from an agential ability of S, it would be inappropriate to take it to be an instance knowledge-how of S (cf. Lyons 2016).41 In conclusion, the central problem of Markie’s inferentialism is the same as the central problem of McGrath’s and Siegel’s versions of it. These three authors are convinced that whether or not a cognitively penetrated appearance has justifying power essentially depends on whether and how the subject, conceived of as an agent or person, has brought about the appearance. This thesis is quite implausible, for appearances in general are not brought about by persons or agents. Markie, McGrath and Siegel have supplied no good reason to believe the contrary. Since this crucial presupposition of the internalist accounts of the epistemic import of cognitive penetration defended by inferentialists is implausible, internalists had better look for an alternative account.

39 Also, the norms reported by S might prove uninformative, for they might just refer to inexpressible

feelings. Suppose the question asked to S concerns her knowing how to ascertain that a number of objects on the table is higher than 10 without counting them. 40 Even if S didn’t lack these abilities, she might make mistakes of various type in reporting her internal states. 41 For further criticism see Ghijsen (2016) and Lyons (2016).

46

3 Cognitive Penetrability

3.5 Taming Cognitive Penetrability Consider any perceptual appearance that P of S cognitively penetrated such that S couldn’t detect it by simply inspecting the appearance. (I expand on this notion below.) Phenomenal conservatives can contend that any such perceptual appearance retains the power to prima facie justify S’s believing P. This contention should go hand in hand with explaining away the intuitions and impressions incompatible with it that many epistemologists claim to have.42 Phenomenal conservatives have all the trumps to pursue both tasks successfully. Let’s start with the first. As we have seen, a perceptual seeming that P of S might happen to be cognitively penetrated by a state of S because this state might directly causes in S a high-level representation that P. In these situations, there might be a mismatch detectable by S between this high-level representation (or proper seeming) and other lower-level representations (or bare sensations), component parts of the same perceptual seeming. Sometimes S might be actually aware of the mismatch. (Suppose S is a microscopist who endorses preformationist and observes a sperm cell. S might actually realize that although she doesn’t perceive certain parts of the cell as components of an embryo, she sees their sum as an embryo. Another case might concern a variant of Angry Jack. Jill might actually be aware that although she cannot spot any element expressing anger in Jack’s face, she sees the face as angry.) In other cases it might happen that S isn’t aware of any incoherence within her cognitively penetrated seeming, though it would be easy for her to spot a discrepancy if she only attended to some of its features. In both these general cases, S’s perceptual appearances would be cognitively penetrated in a way that S could detect an effect of it by simply inspecting the appearances. Let’s say that in these cases S’s perceptual seemings would be detectably cognitively penetrated. In other cases of cognitive penetration, nevertheless, it might be impossible for S to find any incoherence in her perceptual seemings. These cases are those in which cognitive penetration directly affects S’s low-level perceptual contents (or bare sensations), ending up affecting S’s perceptual seemings as wholes. In these cases S’s perceptual seemings would be cognitively penetrated in such a way that S couldn’t detect any effect of it by simply inspecting the perceptual appearances. Let’s say that in these cases S’s perceptual seemings would be undetectably cognitively penetrated. Detectable cognitive penetration affects negatively perceptual justification but it doesn’t seem to raise particular difficulties to phenomenal conservatism. The phenomenal conservative can acknowledge that if S’s perceptual seeming that P were detectably cognitively penetrated, S’s belief that P wouldn’t be perceptually justified. Huemer (2013a, and more explicitly 2013b: 345) suggests that, in this case, S’s justification for P based on S’s high-level content that P would be defeated by some concurrent lower-level content of S’s experience that doesn’t support P. I suggest an alternative diagnosis: S’s incoherent perceptual seeming that P wouldn’t provide S with even prima facie justification for believing P. In fact recall that on (PC)—as 42 It

is important to stress that certain epistemologists—e.g. Lycan (2013) and Huemer (2013a)— don’t share these intuitions.

3.5 Taming Cognitive Penetrability

47

characterized in Sect. 2.1—only a clear and firm seeming that P of S could give S prima facie justification for believing P. Let’s turn to undetectable cognitive penetration. Hereafter, whenever I use the expressions ‘cognitively penetrated seeming(s)’ or ‘cognitively penetrated appearance(s)’ I always refer to undetectably cognitively penetrated seeming(s) or appearance(s). The best argument I can think of to conclude that any cognitively penetrated seeming supplies the subject with prima facie justification for believing its content draws from McGrath’s case for (PC) detailed in Sect. 2.3. It appears very plausible that if S had a cognitively penetrated seeming that P and no defeater, the only epistemically rational doxastic attitude from S’s own standpoint that S could have towards P is believing P. If we rely on the internalist intuition that S’s believing P is epistemically rational only if it is epistemically justified, we must conclude that in this case S has justification for believing P. Precisely, we must conclude that if S has a cognitively penetrated appearance that P and no defeating evidence, S has both prima facie and ultima facie justification for believing P. Thus,S has prima facie justification for believing P (cf. Huemer 2013a, b). Here is a concrete application of this reasoning, described in Huemer (2013b). Suppose that a fearful subject S suspects that (G) there is a gun in the refrigerator. In fact the fridge contains only a banana. However, when S goes to check, S’s fear causes her to see the banana as a gun. Imagine that S has a cognitively penetrated seeming that G and no reason to suspect that the seeming is inaccurate or unreliable. It is clear that S would be justified in believing G. As Huemer emphasizes: This does not strike me as a difficult or borderline case; it strikes me as a perfectly clear case of epistemic justification. If things look that way to you, and you have no reason to doubt your eyes, then you would be crazy not to think there is a gun in the refrigerator. (2013b: 743)

To lend further support to the thesis that our cognitively penetrated appearances give us prima facie justification for believing their contents, Huemer (2013b) adduces another thought experiment. The experiment aims to exemplify the following argument: since a subject S wouldn’t be able to find any epistemically significant difference between a cognitively penetrated and a non-cognitively penetrated appearance if S inspected both of them, the appearances of these two types cannot differ in terms of their intrinsic justifying power. Hence, both of them must provide S with prima facie justification for believing their contents. Let G be ‘there is a gun in the fridge’ and let E be ‘there is a carton of eggs in the fridge’. In this new scenario, S has a perceptual appearance that G and E. Suppose that the part of S’s appearance that represents G is cognitively penetrated by S’s fear of guns (the gun is actually a banana), whereas the part that represents E isn’t cognitively penetrated. Imagine that after scrutinizing her appearance and considering all information available to her, S is unable to find any reason to distrust any part of the appearance.43 As a result, S comes to believe E. In this case, S’s believingE appears to be rational—or at least prima facie so. However, suppose that S responds to her perceptual appearance and inability to find reasons to distrust it 43 Suppose

S doesn’t suspect that her fear could penetrate her experience.

48

3 Cognitive Penetrability

by refusing to believe G—i.e. by disbelieving G or suspending judgment about it. This attitude of S towards G is absurd. For since ‘S would have no rational way of explaining why S believed E while refusing to [believe] G, then S [is] irrational to believe E while refusing to [believe] G’ (746). Clearly, in these circumstances, S’s believing G would be rational—at least prima facie—just as S’s believing E is. This involves that S’s believing G would be prima facie justified, and it would be so because of S’s having the cognitively penetrated appearance (or appearance-part) that G. While I find this thought experiment straightforward, Siegel (2013c) finds it is questionable. She thinks that Huemer, in his reasoning, makes use of a principle like this: (R1) A doxastic attitude of S is rational only if S has a rational way to explain why she adopts it. (Cf. 753)

Siegel claims that (R1) is quite dubious. To clarify why, she appeals to an example originally due to McGrath (2013b). Imagine that S has an occurrent apparent memory that (P) her own child plays the piano better than any other child. Suppose that S also has an unretrieved memory that (Q) another child plays the piano better than her own child. If S refused to believe P in these circumstances, S’s attitude would be— according to Siegel—rational. For S’s memory that Q is—intuitively—a defeater of S’s support for P based on S’s apparent memory that P. Yet since S’s memory thatQ is unretrieved, if S refused to believe P, S would have no rational way to explain her attitude. This clashes with (R1).44 An austere internalist accessibilist might respond that given that S’s memory that Q is actually unretrieved, it doesn’t intuitively count as a defeater. But this response would certainly be controversial. What makes Siegel’s objection flawed is that (R1) is too general a principle, which Huemer is not committed to. Huemer’s thought experiment can legitimately be interpreted as relying on, not (R1), but this more specific principle: (R2) A doxastic attitude A of S is prima facie rational in virtue of an occurrent experience E of S only if S could rationally explain why she adopts A by appealing to E.

(R2) looks straightforward. In the young pianist example, if S refused to believe P, this attitude would not be prima facie rational in virtue of S’s occurrent memory that P.45 This is attested by (R2): in the envisaged situation, S couldn’t rationally explain why she would refuse to believe P by appealing to her apparent memory that P. On the other hand, if S believed P, this attitude would intuitively be prima facie rational in virtue of S’s apparent memory that P. Note that in this case, in harmony with (R2), S could rationally explain why she would do so by adducing her apparent memory that P. Importantly, once (R1) is replaced with (R2), Huemer’s thought experiment looks impeccable. In the situation imagined by Huemer, if S refused to believe G, this 44 Siegel (2013c: 753) offers an additional example. My reply to the young pianist example can easily be adapted to respond to it. 45 I assume that this apparent memory is an experience.

3.5 Taming Cognitive Penetrability

49

attitude would not be prima facie rational in virtue of the part G of her appearance. This is certified by (R2): S couldn’t rationally explain why she would refuse to believe G by appealing to the part G of her appearance. Contrarily, if S believed G, this attitude of S would intuitively be prima facie rational in virtue of the part G of S’s appearance. In this case, S could rationally explain why she would do so by appealing to the part G of her appearance, in accordance with (R2). To bring to completion the task of defending (PC) from cognitive penetration objections, the phenomenal conservative can concede that there is something wrong with beliefs based on appearances that are badly cognitively penetrated. Beliefs like these look in some sense epistemically defective. The phenomenal conservative should insist, however, that it is quite another issue as to whether these defects concern prima facie justification. More explicitly, the phenomenal conservative should contend that the intuitions that seem to suggest that beliefs of this type lack prima facie justification are misconstrued, for these beliefs actually lack a different epistemic property that can be confused with prima facie justification. Phenomenal conservatives might be allured by the view that in every case of bad cognitive penetration, S’s seeming that P doesn’t supply S with all things considered justification for believing P because some mental state of S is a defeater of S’s prima facie justification. In accordance with this view, the cases of bad cognitive penetration are situations in which S has some belief, memory, intuition or a similar state that counts for S as a reason to take her appearance to be untrustworthy. For example, in Prospectors, Virgil would probably be aware that he lacks experience and training. This would give him a reason to suspect that his perceptual appearance that the nugget is gold is untrustworthy (cf. Tucker 2010: 539). The fact that Virgil lacks ultima facie justification would explain why his belief that the pebble is gold looks epistemically defective, and why it appears to be epistemically worse than Gus’ identical belief, which is all things considered justified (cf. Georgakakis and Moretti 2019). Unfortunately, this account cannot generalize to all cases of bad cognitive penetration because there is actually no guarantee that S would always have a defeater of her seeming-based justification in these circumstances. This is evident as we consider cases different from Prospectors. Take Angry Jack. It is possible and not implausible that in a scenario of this type, Jill could be unaware that her appearance might be caused by some of her prior beliefs. It appears false, therefore, that Jill would inevitably have a reason to take her appearance to be untrustworthy in these circumstances.46 Markie indicates a deeper problem of this account: it seems to locate 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. [The impression] is that he should not have his seeming experience in the first place’ (Markie 2013: 258). Tucker (2011) and Huemer (2013b) make a different proposal to the effect that whenever a seeming that P of S is badly cognitively penetrated, S’s belief that P 46 See

Siegel (2012) for further discussion.

50

3 Cognitive Penetrability

based on it is prima facie justified but not warranted. Since warrant is the property that must be added to true belief to become knowledge, according to this proposal, whenever S’s seeming that P is badly cognitively penetrated, S’s belief that P based on it cannot result in S’s knowledge that P. A warranted belief is typically thought of as one that isn’t just accidentally true, for it is properly connected with the external world by a reliable or truth-tracking belief production mechanism (cf. Plantinga 1993a, b). It is intuitively plausible that neither Jill’s perceptual belief in Angry Jack nor Virgil’s perceptual belief in Prospectors is caused by a reliable belief-forming process. On the other hand, Gus’ perceptual belief sustained by his expertise seems to be produced by a reliable belief-forming process. Markie (2013) has worked out a variant of Prospector that casts doubts on this explanation. Suppose both Virgil and Gus are in the Matrix and are fed with sensory inputs identical to those that they would have in the real world. Gus’ seeming that his pebble is gold is still partly caused by Gus’ skills and background information, so it still appears to be affected by good cognitive penetration. On the other hand, Virgil’s identical seeming is still partly caused by Virgil’s desire to become rich, so it still appears to be afflicted by bad cognitive penetration. Hence, Virgil’s seemingbased belief still looks epistemically worse than Gus’ identical belief, despite both Virgil and Gus lack perceptual knowledge and warrant.47 For both Virgil’s and Gus’ beliefs are disconnected from the world, due to the sceptical scenario. This suggests that what explains the epistemic inadequacy of a belief based on a badly cognitively penetrated seeming isn’t the fact that the subject lacks warrant for it. Pressed by Markie’s objection, Tucker (2010, 2013) has offered a new explanation. In Tucker’s new account, when a seeming that P of S is badly cognitively penetrated, S possesses prima facie justification for believing P. However, S is epistemically blameworthy for having produced her seeming; thus S would also be epistemically blameworthy for believing P on the basis of it.48 This proposal assimilates the situations in which S has a badly cognitively penetrated seeming to situations in which S fabricates her own evidence. The following is a case of evidence fabrication imagined by Huemer (2013a: 344–345): suppose S has a brain-manipulation device invented by a scientist. Imagine that for some reason S intends to make herself believe that (Q) there is a cat before her. So S pushes a button on the device that causes S to have a hallucination that Q while simultaneously erasing S’s memory of her having used the brain-manipulation device. As a result, S believes Q.49 In this example S appears to be epistemically blameworthy for producing her evidence that Q. S is also epistemically blameworthy for having a belief that Q, since she believes Q on the basis of her fabricated evidence. Still, S’s evidence provides S with prima facie (and all things considered) justification for believing Q. For, from S’s viewpoint at the time she has 47 Markie

only mentions Virgil’s and Gus’ lack of knowledge, but he would agree that they both also lack warrant. 48 Tucker specifically focuses on wishfully produced appearances (like Virgil’s appearance in Prospectors). I follow McGrath (2013b) in extending this proposal to cover virtually all cases of bad cognitive penetration. 49 See Tucker (2010) and McGrath (2013b) for alternative examples.

3.5 Taming Cognitive Penetrability

51

the hallucination, there is nothing about the hallucination that distinguishes it from a perfectly ordinary experience (cf. Tucker 2010, 2013; Huemer 2013a). Tucker thinks that something very similar holds true in the cases of bad cognitive penetration: S is epistemically blameworthy for producing a cognitively penetrated appearance that P, and S would also be epistemically blameworthy for believing P if she did so on the basis of her appearance. Nevertheless, S’s belief would be prima facie justified by S’s appearance. All this would explain—according to Tucker—why Virgil’s prima facie justified belief that P is intuitively epistemically worse than Gus’ prima facie justified identical belief, even when both subjects are in the Matrix. Virgil ‘is blameworthy for his false belief and his inappropriately-caused seeming and [Gus] isn’t’ (Tucker 2010: 541). Markie (2013) complains that Tucker’s new proposal leaves it unspecified the nature of the negative upshot for which S would be epistemically blameworthy when entertaining a badly cognitively penetrated appearance. In particular, Tucker doesn’t clarify the sense in which a badly cognitively penetrated appearance would be inappropriately produced by S. I do have an independent concern: it is normally assumed that S can be blameworthy for doing something X only if X is a voluntary act ofS or, at least, S has some control on X. Yet since the processes that make S’s seemings badly cognitively penetrated (supposing cognitive penetration exists) take place at subpersonal level, it is implausible that S could control them. S may be able to revise a belief that she has formed on a seeming cognitively penetrated by her desires or expectations, if S discovered that the belief has these features. Yet it is dubious that S could prevent previous or concurrent desires or expectations that she might have from penetrating her seemings.50 As we have already seen, it is equally dubious that S could revise her cognitively penetrated appearances. In conclusion, the claim that S would be blameworthy for entertaining a badly cognitively penetrated seeming appears implausible. This casts doubts on the acceptability of Tucker’s proposal. Let me outline an alternative account that aims to improve on Tucker (2011) and Huemer (2013b)’s proposal. Consider a belief that P of S based on a seeming that P of S that looks badly cognitively penetrated. Tucker and Huemer maintain that we have the impression that this belief is epistemically defective because we take it to be unwarranted. I suggest that we have this impression because—more precisely—we judge that the belief isn’t produced by a properly functioning cognitive faculty of S. This is why we find the belief unwarranted. If this is correct, Markie’s objection can be answered. For in the Matrix variant of Prospectors, only Virgil’s belief is epistemically defective in this specific sense, not Gus’. 50 If perceptual learning actually produces good cognitive penetration, it might be correct to say that S can control factors that produce good cognitive penetration by controlling these learning process. Perhaps S’s appearances might also be subject to bad cognitive penetration depending on (bad) perceptual learning that S is responsible for. It might thus be correct to maintain that, in these cases, S is epistemically blameworthy for having the relevant appearances. However, note that these cases would be very different from the typical examples of bad cognitive penetration—such as Angry Jack and Virgil’s case in Prospectors—discussed in the literature.

52

3 Cognitive Penetrability

To appreciate how this account works, let me expand on the notion of proper function. Consider an example from Plantinga (1993a: 195–198 and 205–207). Imagine S has a brain lesion that elicits various cognitive processes generating random beliefs. One of these processes—call it PR—produces only the true belief that (Q) One has a brain lesion Suppose that PR is very reliable because it would always result in one’s having a true belief that Q if PR took place in one’s brain. Despite PR’s high reliability, it is intuitive that S’s belief that Q cannot result in S’s knowing that Q. Plantinga suggests that we have this intuition because we realize that S’s belief that Q isn’t formed through a properly functioning cognitive faculty. Plantinga concludes from this example that any reliabilist account of warrant must include a condition that states that a belief of a subject S can be warranted only if it is formed through cognitive faculties of S that function properly. Note that this is just a necessary condition. The proper functioning of S’s cognitive faculties doesn’t suffice for warrant: it is possible that S’s faculties be functioning properly but S’s beliefs still lack warrant. This happens when S’s faculties and the environment in which S is are not properly attuned (cf. Plantinga 1993b: 7). Imagine for example that S is in a sceptical scenario. S’s cognitive faculties could in this case function properly, though S’s perceptual beliefs would remain unwarranted. The notion of proper function presupposes the one of a design plan, which specifies the way in which a thing is supposed to function in various circumstances to achieve certain goals. In the case in point, a design plans specifies the ways in which S’s cognitive faculties are supposed to function in different circumstances to produce true beliefs. The existence of a design plan doesn’t necessarily require the existence of a conscious designer, such as God (cf. Plantinga 1993b: 21). Millikan (1984), for instance, gives a fully naturalistic-evolutionary account of this notion. Plantinga (1993b: 237) ultimately endorses a theistic explanation. Suppose now that S has a cognitively penetrated perceptual seeming that P and believes P on that seeming. I suggest that our intuitive judgments about the epistemic standing of S’s belief that P result from a process of this type: we (implicitly) attribute a design plan to the cognitive faculties of S involved in producing the belief that P. Then, we check whether the seeming that P has been formed through the proper functioning of those faculties. If the answer is positive, we conclude that the cognitive penetration of the seeming is good, and that S’s belief that P is not epistemically defective.51 This is what presumably happens when we judge that Gus’ belief in Prospector isn’t epistemically defective. If the answer is instead negative, we conclude that the cognitive penetration of S’s seeming that P is bad, and S’s belief that P is epistemically defective as originated from faculties that aren’t properly functioning. This is what presumably happens when we judge that Virgil’s belief in Prospector and Jill’s belief in Angry Jack are epistemically defective.

51 As

far as the bearing of cognitive penetration is concerned.

3.5 Taming Cognitive Penetrability

53

This account enables us to respond to Markie’s criticism. When we scrutinize the Matrix variant of Prospectors described before, we acknowledge that both Virgil’s and Gus’ perceptual beliefs are unwarranted because they are secluded from the external world. Nevertheless, since we attribute to Virgil and Gus the same design plan that we would attribute to them if they were in the ordinary world, we do perceive Virgil’s belief to be epistemically worse than Gus’. For we reckon that the belief isn’t formed through properly functioning cognitive faculties, whereas Gus’ belief is. Markie (2013) thinks that an account of this type misses the target because: We can modify the [Matrix variant of Prospectors] 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... [But even so, Virgil] 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. (259–260)

I don’t find these sketchy considerations convincing. Let’s try to produce a viable variant of Prospectors similar to the one that Markie has in mind. In this variant, Gus is an expert gold prospector, whereas Virgil barely knows the most basic features of gold. While Gus lives in the ordinary world, Virgil inhabits another world W. That world is almost identical to ours, with one difference: inW a strong desire that a pebble is gold is very likely to turn the pebble into gold. (Perhaps there are wandering demons that satisfy that desire, using their powers.) Suppose Virgil’s cognitive faculties have been designed (by the nature or God) in a way to enable him to acquire many true beliefs. Given the mentioned peculiarity of W, Virgil’s faculties are such that his strong desire that a pebble is gold can easily penetrate his perception of the pebble, to elicit in Virgil a (probably true) belief that the rock is gold. One day, Virgil and Gus are abducted and put in the Matrix by the crazy scientist. The scientist erases all traces of the abduction from their memories, so that Virgil and Gus keep thinking they are in their original worlds. In the virtual reality of the Matrix, Virgil and Gus come to inspect the same pebble and have an identical cognitively penetrated seeming that (P) the pebble is gold. Virgil’ seeming is penetrated by his desire, while Gus’ is penetrated by his background information and skills. Thereby, they both believe P. In this thought experiment, the design plan of Virgil requires his desire that P to penetrate his appearance that P. Virgil’s belief that P is thus produced by faculties that are functioning properly. Markie insists that, even in this case, Virgil’s belief that P is intuitively epistemically worse than Gus’ belief that P. I candidly confess that I lack this intuition, and I trust that many readers may lack it as well. In the envisaged scenario, Virgil and Gus acquire their respective perceptual beliefs that P while they are secluded from the external world. So both their beliefs are unwarranted and cannot constitute knowledge. Furthermore, both beliefs are produced by faculties that function properly, and they are both prima facie justified by identical perceptual appearances. I see no clear sense in which one of these beliefs could be epistemically worse than the other.

54

3 Cognitive Penetrability

3.6 Conclusions I have scrutinized arguments against phenomenal conservatism, levelled by epistemic externalists and internalists, which appeal to the controversial thesis that the contents of perceptual appearances can be penetrated by previous or concurrent cognitive states of the subject. These objections adduce intuitions that seem to suggest that cognitively penetrated perceptual appearances often lack the ability to prima facie justify their contents, which clash with (PC). In response, I have suggested that the adduced intuitions cannot be vindicated by the externalist contention that cognitive penetration often makes perceptual appearances unreliable. Furthermore, I have argued at length that the same intuitions cannot be vindicated by the internalist contention that cognitive penetration often makes perceptual appearances epistemically irrational or arational. I have also defended the claim that cognitively penetrated appearances don’t lose their ability to prima facie justify their contents, and the claim that the intuitions that seem to suggest otherwise can be explained away as cases of misidentification of epistemic properties. In particular, I have suggested that the beliefs based on undetectable cognitively penetrated seemings are prima facie justified but they look defective when they are the product of cognitive faculties that appear not to function properly. Cognitive penetrability is not a problem for phenomenal conservatism.

References Bergmann M (2006) Justification without awareness. Oxford University Press, NY Boghossian P (2014) What is inference? Philos Stud 169:1–18 Boghossian P (2019) Inference, agency, and normativity. In: Jackson MB, Jackson BB (eds) Reasoning: essays on theoretical and practical thinking. Oxford University Press, Oxford, pp 101–128 Brogaard B (2013) Phenomenal seemings and sensible dogmatism. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 270–289 Broome J (2014) Normativity in reasoning. Pac Philos Q 95:622–633 Carr D (1981) Knowledge in practice. Am Philos Q 18:53–61 Cath Y (2012) Knowing how without knowing that. In: Bengson J, Moffett MA (eds) Knowing how: essays on knowledge, mind, and action. Oxford University Press, Oxford, pp 113–135 Cath Y (Forthcoming) Knowing how. Analysis Comesaña J (2002) The diagonal and the demon. Philos Stud 110:249–266 Comesaña J (2010) Reliabilist evidentialism. Nous 44:571–600 Firestone C, Scholl B (2016) Cognition does not affect perception: evaluating the evidence for ‘top-down’ effects. Behav Brain Sci 39:1–72 Frege G (1979) Logic. In: Hermes H, Kambartel F, Kaulbach F, Long P, White R (eds) Posthumous writings. University of Chicago Press, Chicago Fumerton R (2006) Epistemology. Blackwell Publishing, Malden Gatzia D, Brogaard B (2017) Pre-cueing, perceptual learning and cognitive penetration. Front Psychol. https://doi.org/10.3389/fpsyg.2017.00739

References

55

Georgakakis C, Moretti L (2019) Cognitive penetrability of perception and epistemic justification. Internet Encycl Philos. https://www.iep.utm.edu/cog-pene/. Accessed 31 May 2019 Goldman A (1979) What is justified belief? In: Pappas G (ed) Justification and knowledge. Reidel, Dordrecht, pp 1–25 Goldman A, Beddor B (2016) Reliabilist epistemology. In: Zalta E (ed) Stanford Encyclopedia of Philosophy. https://plato.stanford.edu/archives/win2016/entries/reliabilism/ Ghijsen H (2016) The real epistemic problem of cognitive penetration. Philos Stud 173:1457–1475 Huemer M (2007) Compassionate phenomenal conservatism. Philos Phenomenol Res 74:30–55 Huemer M (2013a) Phenomenal conservatism Über Alles. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 328–350 Huemer M (2013b) Epistemological asymmetries between belief and experience. Philos Stud 162:741–748 Huemer M (2016) Inferential appearances. In: Coppenger B, Bergmann M (eds) Intellectual assurance: essays on traditional epistemic internalism. Oxford University Press, Oxford, pp 144–160 Johansson P, Hall L, Sikström S, Tärning B, Lind A (2006) How something can be said about telling more than we can know: on choice blindness and introspection. Conscious Cogn 15:673–692 Locke J (1689/2008) An essay concerning human understanding. Oxford University Press, Oxford Lycan WG (2013) Phenomenal conservatism and the principle of credulity. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 293–305 Lyons J (2011) Circularity, reliability, and the cognitive penetrability of perception. Philos Issues 21:289–311 Lyons J (2016) Inferentialism and cognitive penetration of perception. Episteme 13:1–28 Macpherson F (2012) Cognitive penetration of colour experience: rethinking the issue in light of an indirect mechanism. Philos Phenomenol Res 84:24–62 Markie P (2006) Epistemically appropriate perceptual belief. Nous 40:118–142 Markie P (2013) Searching for true dogmatism. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 248–268 McGrath M (2013a) Phenomenal conservatism and cognitive penetration: the ‘Bad Basis’ counterexamples. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 225–247 McGrath M (2013b) Siegel and the impact for epistemological internalism. Philos Stud 162:723–732 McHugh C, Way J (2016) Against the taking condition. Philos Issues 26:314–331 Millikan R (1984) Naturalist reflections on knowledge. Pac Philos Q 4:315–334 Mole C (2015) Attention and cognitive penetration. In: Ziembekis J, Raftopolous A (eds) The cognitive penetrability of perception: new philosophical perspectives. Oxford University Press, Oxford, pp 218–238 Peirce CS (1905) Issues of pragmaticism. Monist 15:481–499 Plantinga A (1993a) Warrant: the current debate. Oxford University Press, Oxford Plantinga A (1993b) Warrant and proper function. Oxford University Press, Oxford Pylyshyn Z (1999) Is vision continuous with cognition? Behav Brain Sci 22:341–365 Raftopoulos A (2019) Cognitive penetrability and the epistemic role of perception. Palgrave Macmillan, London Russell B (1920) The nature of inference. In: Frohmann B, Slater J (eds) The collected papers of Bertrand Russell: essays on language, mind, and matter slater, 1919–26, Ch. 15. Unwin Hyman, London. Siegel S (2012) Cognitive penetrability and perceptual justification. Nous 46:201–222 Siegel S (2013a) Can selection effects influence the rational role of experience? In: Gendler T (ed) Oxford studies in epistemology, vol. 4. Oxford University Press, Oxford, pp 240–270 Siegel S (2013b) The epistemic impact of the etiology on experience. Philos Stud 162:697–722

56

3 Cognitive Penetrability

Siegel S (2013c) Reply to Fumerton, Huemer, and McGrath. Philos Stud 162:749–757 Siegel S (2017) The rationality of experience. Oxford University Press, Oxford Siegel S (2019) Inference without reckoning. In: Jackson BB, Jackson MB (eds) Reasoning: new essays on theoretical and practical thinking. Oxford University Press, Oxford, pp 15–23 Silins N (2016) Cognitive penetration and the epistemology of perception. Philos Compass 11:24–42 Stokes D (2012) Perceiving and desiring: a new look at the cognitive penetrability of experience. Philos Stud 158:479–492 Stroud B (1979) Inference, belief, and understanding. Mind 88:179–196 Teng L (2016) Cognitive penetration, imagining, and the downgrade thesis. Philos Top 44:405–426 Tucker C (2010) Why open-minded people should endorse dogmatism. Philos Perspect 24:529–545 Tucker C (2011) Phenomenal conservatism and evidentialism in religious epistemology. In: Clark KJ, VanArragon RJ (eds) Evidence and religious belief, Ch. 4. Oxford University Press, Oxford Tucker C (2013) Seemings and justification: an introduction. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 1–29 Tucker C (2014a). If dogmatists have a problem with cognitive penetration, you do too. Dialectica 68:35–62 Tucker C (2014b) On what inferentially justifies what: The vices of reliabilism and proper functionalism. Synthese 191:3311–3328 Tucker C (Forthcoming) Dogmatism and the epistemology of covert selection. In: Ballantyne N, Dunning D (eds) Reason, bias, and inquiry: new perspectives from the crossroads of epistemology and psychology. Oxford University Press, Oxford Vahid H (2014) Cognitive penetration, the downgrade principle, and extended cognition. Philos Issues 24:439–459 Wu W (2013) Visual spatial constancy and modularity: does intention penetrate vision? Philos Stud 165:647–669 Wu W (2017) Shaking up the mind’s ground floor: the cognitive penetration of visual attention. J Philos 114:5–32

Chapter 4

The Bayesian Objection

Abstract In this chapter I analyse an objection to phenomenal conservatism to the effect that phenomenal conservatism is unacceptable because it is incompatible with Bayesianism. I consider a few responses and dismiss them as misled or problematic. Then, I argue that the objection doesn’t go through because it rests on an implausible formalization of the notion of seeming-based justification. In the final part of the chapter, I investigate how seeming-based justification and justification based on one’s reflective belief that one has a seeming interact with each another. Keywords Phenomenal conservatism · Bayesianism · Perceptual justification · Immediate and non-immediate justification · Inferential and non-inferential justification · Reflective justification · Reflective awareness

4.1 White’s Objection According to the phenomenal conservative, (PC) If it seems to S that P, S thereby has prima facie justification for believing P.

As clarified in Sect. 2.1, the ‘thereby’ in (PC) indicates that S’s justification for P rests solely on S’s appearance that P, which need not be justified. S’s justification for P is thus immediate because it is not based on any independent justification of S. Furthermore, S’s justification for P is non-inferential because it is not available to S through some inference or possible inference of S. Schiffer (2004), Cohen (2005), White (2006) and Wright (2007) have all argued that (PC) is flawed or problematic because it is at odds with Bayesianism. These objections are very similar. I focus on White (2006) because White’s objection is the most detailed and clear, and the most discussed in the literature. All these authors crucially rely on the same implausible formalization of the intuitive notion of seeming-based justification, which I criticize and reject in this chapter. Thus, my reply to White is meant to respond—indirectly—to the other Bayesian objections as well.

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2020 L. Moretti, Seemings and Epistemic Justification, SpringerBriefs in Philosophy, https://doi.org/10.1007/978-3-030-43392-5_4

57

58

4 The Bayesian Objection

White specifically targets perceptual dogmatism—namely, phenomenal conservatism circumscribed to perceptual justification. White’s challenge could nevertheless be re-formulated to target phenomenal conservatism in other domains.1 The Bayesian assumes that our degrees of rational confidence—or credences—obey probability calculus. Furthermore, the Bayesian conceives of both hypotheses and evidence as propositions held to be true by the subject with a degree of rational confidence enclosed between 0 and 1, where 0 represents certainty in the proposition’s falsehood and 1 certainty in its truth.2 The charge raised by White against (PC) is to the effect that (PC) is inconsistent with the Bayesian account of how perceptual evidence rationally affects our credences in hypotheses. Given its intuitive appeal, ductility and an impressive list of insightful analyses, Bayesianism is enormously influential in today’s epistemology. If (PC) were actually incompatible with Bayesian reasoning, many would conclude that it is (PC) that has to go to the wall. At any rate, this clash would generate a philosophical puzzle with no easy solution. White (2006) formulates his argument synchronically. Since it actually concerns epistemic processes (i.e. acquisition of evidence and credence revision), I prefer to re-formulate it diachronically. Let Pr be the credence function of a rational subject S, and let Pr F be S’s credence function updated upon S’s learning F. I take the updating procedure to be standard conditionalization, according to which Pr F (P) = Pr(P|F) and Pr F (F) = 1. These identities state that for any propositions P and F on which Pr is defined, as S learns F, S’s credence that P equates to the conditional credence that P given F that S had before learning F, and S’s credence that F becomes the highest possible. White assumes that S’s degree of epistemic justification for believing a proposition P that S has in a given situation and S’s degree of confidence that P that S should rationally have in the same situation coincide, or—at least—that the second provides an accurate model of the first. The last assumption has a strong intuitive appeal, so I will take it on board. From this assumption, White infers that S acquires some justification for believing a proposition P from evidence F if and only if S’s rational credence that P increases upon S’s learning F.3 Let’s stipulate that: P = ‘this is a hand’, F = ‘it appears to S that P’, SH = ‘this is a fake-hand’. A fake-hand is everything that isn’t a hand but cannot be visually distinguished from a hand. ‘Examples of fake-hands include [for instance] perfect plastic replicas of hands, and moving projected holograms of hands’ (White 2006: 528), and—I would add—hallucinations of hands induced by a Cartesian demon. SH stands therefore for any error conjecture that S can conceive of, which is incompatible with P but entails 1 In

which case my response can be adjusted accordingly. good introduction to Bayesian methodology is Weisberg (2011). 3 The only if direction of this biconditional is controversial (cf. Wright 2011; Moretti and Piazza 2013). I won’t press this objection here because it would distract us from more serious difficulties of White’s arguments. 2A

4.1 White’s Objection

59

F or makes F highly probable. White imagines a scenario in which S looks at her hand for the first time—so S has a perceptual appearance that P—and he analyses the bearing of this perceptual evidence on S’s credence that P and SH. Before proceeding any further with the presentation of White’s argument, let me clarify a crucial point. White relies on a formal model that takes S’s perceptual evidence to be a reflective belief of S about her own perceptual appearance that P—that is to say, S’s reflective belief that (F) it appears to S that P, entertained by S with the highest possible confidence. The phenomenal conservative—on the other hand—would take S’s perceptual evidence to be S’s appearance that P. White doesn’t think that there is a substantive mismatch here. For he claims that S’s perceptual appearances and S’s correlated reflective beliefs have the same justifying force (at least in the cases relevant to his Bayesian proof). Unlike White, I find this claim very problematic. For the sake of discussion, I will question it only in the next section. To forestall possible confusion, let me sharply distinguish between ‘S’s having an appearance that P’ and ‘S’s being reflectively aware of an appearance that P’.4 These two expressions don’t refer to the same mental states. The first expression refers to a mental state of S—her appearance that P—which can exist in S’s mind even if she doesn’t reflect on her own mental states. The second expression refers to a more complex state of S, one that encompasses at least these three items: (i) S’s appearance that P, (iii) S’s reflective acquaintance with her appearance that P,5 and (ii) S’s reflective belief that (F) S has an appearance that P, based on her acquaintance with her appearance that P. This is White’s argument. It looks true that Pr P (F) ∼ = Pr SH (F) ∼ = 1—namely, S’s credence that F should be close to certainty when S learns P or SH. Furthermore, since S shouldn’t in general expect to have an experience of a hand, Pr(F) Pr(F) and Pr SH (F) > Pr(F). The first inequality entails via probability calculus that Pr F (P) > Pr(P). The second inequality entails via probability calculus that Pr F (SH) > Pr(SH). So, when S learns F—or, equivalently for White, when S has an appearance that P— S acquires some justification for believing both P and SH. Let’s focus on Pr F (SH) > Pr(SH). This inequality implies via probability theory that (1) Pr F (¬SH) < Pr(¬SH). (1) states that when S learns F, S’s credence that ¬SH (i.e. that SH is false) should decrease. Hence, when S has an appearance that P, S cannot acquire justification for believing ¬SH. 4 There

is a third expression that one might want to distinguish from these two expressions, that is to say: ‘S’s directing her attention to her appearance (or appearance-part) that P’. This expression refers to S’s act of selecting just one of her concurrent mental states. This act of S doesn’t require S to become reflectively aware—in the sense I clarify above—of the selected state. 5 Here ‘reflective acquaintance’ can be intended as a form direct awareness, in Fumerton (1995)’s sense, or as an introspective appearance, in Huemer (2007: 46)’s sense.

60

4 The Bayesian Objection

Now note that given that P entails ¬SH, it follows through probability calculus that (2) Pr F (P) ≤ Pr F (¬SH). That is to say, when S learn F, S’s credence that ¬SH cannot be smaller than S’s credence that P. Finally, (2) and (1) trivially entail by transitivity that (3) Pr F (P) < Pr(¬SH). (3) shows—according to White—that ‘its appearing to [S] that this is a hand can render [S rationally] confident [to any degree] that it is a hand, only if [S is] already [rationally] confident that it is not a fake-hand [to a higher degree]’ (2006: 534). We saw before that when S has an appearance that P, in White’s model, S does acquire some justification for believing P. (3) places an upper bound on the strength of this justification. For it shows—according to White—that its appearing to S that P can give S some degree of justification for believing P only if S has already, and so independently, a higher degree of justification for believing ¬SH. This entails that when S has an appearance that P, unless S has independent justification sufficient to believe ¬SH, S cannot have justification sufficient to believe P. White takes this proof to show that, against (PC), S’s appearance that P cannot give S immediate prima facie justification for believing P (cf. Tucker 2013: 18).

4.2 Responding to White’s Objection The responses to White given by supporters of (PC) fall within either of these categories: the revisionary ones, which agree with White that (PC) is incompatible with Bayesianism but propose revising Bayesianism to resolve the incompatibility, and those that retort that (PC) and Bayesianism haven’t actually been shown to be incompatible by White. Revisionary responses propose, for instance, to replace the probability function with an original superadditive function (cf. Pryor 2007); to switch to imprecise probabilities and introduce a non-standard conditionalization procedure (cf. Weatherson 2007); to re-interpret Bayesian formalism in a non-standard way (cf. Kung 2010); and to replace the probability function with an intuitionist probability function (cf. Jehle and Weatherson 2012). I won’t discuss these proposals because I think they are on the wrong track.6 In this section I will argue that White hasn’t actually proven that (PC) and Bayesianism are incompatible. Silins (2007) has made a case for the same conclusion.7 Before laying down my response to White, let me detail Silins’ response and the reasons of my discontent with it. 6 It is fair to say that these revisionary proposals aren’t ad hoc. They are typically introduced as means

to respond to White and—simultaneously—resolve independent shortcomings of Bayesianism; for instance, its inability to provide states of uncertainty (or absence of evidence) or defeaters of certain types with adequate formalizations. 7 To be accurate, Silins (2007) aims to defend, not (PC), but a position that he calls liberalism, following Pryor (2004). According to liberalism:

4.2 Responding to White’s Objection

61

The salient formal result of White’s objection is the proposition that (3) Pr F (P) < Pr(¬SH). White takes (3) to contradict (PC). Silins agrees that (3) models an important feature of perceptual justification but insists that (3) doesn’t establish that S’s having independent justification for believing ¬SH is a condition necessary and basing for S’s acquiring perceptual justification for P. (3) only establishes— according to Silins—that S’s having independent justification for believing ¬SH is a condition necessary for S’s acquiring perceptual justification for believing P. Silins concludes that since (PC) doesn’t disallow conditions for perceptual justification that are necessary but not basing, White’s inequality doesn’t contradict (PC). Silins gives a few examples of conditions for perceptual justification that are necessary but not basing, and that aren’t disallowed by (PC). Let’s consider two of them. Take the conceptually true proposition that (R) a hand is a hand. S does have (a priori) justification for believing R if S possesses the concepts required to entertain R. Presumably, S can have a seeming that (P) this is a hand, only if S has the concepts required to represent that P. But these include the concepts required to entertain P, and thus R. Consequently, if S has an appearance that P, S possesses prima facie justification for believing P but also independent justification for believing R. S’s having independent justification for R is thus a necessary condition for S’s having perceptual justification for P, but it doesn’t look like a basing condition for it. For S’s perceptual justification for believing P doesn’t appear to be constituted by, among other things, S’s independent justification for believing R. Take now the contingent proposition that (R*) S exists. Whenever S has a perceptual appearance that P, and thus prima facie justification for believing P, S also has a cogito-style independent justification for believing R*. In this case too, S’s having independent justification for R* is a necessary condition for S’s having perceptual justification for P, but this doesn’t appear to be a basing condition for it. I find Silins’ response to White problematic for at least two reasons. This is the first: suppose that S’s possessing independent justification for ¬SH were actually a necessary but not basing condition for S’s having perceptual justification for P. A consequence of this would be that if S had a seeming that P but no independent justification for ¬SH, S would not have prima facie justification for believing P, in which case (PC) would be false. Therefore, if S’s having independent justification for ¬SH were actually a necessary but not basing condition for S’s having perceptual justification for P, the phenomenal conservative would be committed to claiming that S does have some sort of default justification for ¬SH that is not constitutive of S’s perceptual justification for P. Indeed, the phenomenal conservative would be committed to making the same claim about any conceivable error conjecture relevant to the content of any perceptual seeming falling within (PC)’s domain. Note In some case C, S’s visual appearance that P gives S some justification for believing P, and what makes S’s appearance capable of giving S some justification for believing P doesn’t include S’s having some independent justification for believing other propositions. (Cf. 2007: 111–112) Silins (2007: 108) makes it clear that one C−case is the one inspected by White (2006), in which P states that there is a hand. It is quite clear that for Silins (PC) and liberalism largely overlap.

62

4 The Bayesian Objection

that showing that S has default justification of this sort8 is much more demanding than showing that S has independent justification for relatively trivial truths like R or R*. A line of response to White that did not embark the phenomenal conservative on such an arduous enterprise would be preferable. Another problem of Silins’ response is that it concedes to White that S’s having independent justification for ¬SH is a condition necessary for S’s having perceptual justification for P. Silins concedes this because he endorses White’s assumption that perceptual appearances and correlated reflective beliefs have, in the circumstances of White’s thought experiment, the same justificatory force (cf. Silins 2007: 125n22). Given that this assumption is implausible, Silins’ thesis that S’s having independent justification for ¬SH is a condition necessary for S’s having justification for P is ungrounded. This is the second reason why I take Silins’ response to White to be inadequate. I now explain why White’s assumption is implausible. I will then argue that since White’s case crucially relies on this assumption, it doesn’t get off the ground. The phenomenal conservative takes S’s perceptual evidence to be S’s perceptual appearance. White, on the other hand, relies on a formalism in which S’s perceptual evidence is S’s reflective belief that S has a perceptual appearance. The result is that when the phenomenal conservative identifies the evidential basis of S’s perceptual justification for believing a proposition X with S’s appearance that X, White’s formalism identifies the evidential basis of S’s perceptual justification for believing the same proposition X with S’s reflective belief (Y ) that S has an appearance that X. White’s formalism can accurately model the notion of justification deployed by the phenomenal conservative and describe important features of it only if the justifying force of S’s perceptual appearance that X and the justifying force of the correlated reflective belief that Y of S are identical or very similar, at least when X and Y are propositions relevant to White’s proof. This is precisely what White assumes. For example, White (2006: 535) considers whether the formal step in his objection constituted by (1) Pr F (¬SH) < Pr(¬SH) can be interpreted as describing a feature of the justifying force of S’s appearance that P (rather than S’s reflective belief that F) with respect to ¬SH. White takes it to be true that if S happens to have a perceptual appearance that P accompanied by a correlated reflective belief that F, the strength of S’s justification for ¬SH should be identical to the strength of S’s justification for ¬SH depending on S’s sole belief that F. (I agree with White on this point. I return to it in the next section.) Considering what (1) specifically states, this means that when S has a perceptual appearance that P accompanied by a correlated reflective belief that F, the strength of S’s justification for ¬SH should decrease. White then suggests that: If the rational response to its appearing that this is a hand, when [S] also believe[s] that it appears that this is a hand, is to decrease [S’s] confidence that it is not a fake-hand,9 then 8 See

Wedgwood (2013) for an ingenious attempt. See also Balcerak-Jackson (2016)’s objections.

9 White’s original passage omits ‘not’ and literally reads ‘it is a fake-hand’. This is certainly a typo.

4.2 Responding to White’s Objection

63

surely this is the rational response to the same experience when [S does] not even consider how things appear to [her]. (2006: 535)

In conclusion, according to White, S’s appearance that P alone should lessen S’s justification for ¬SH, exactly as S’s belief that F alone does. Despite White’s conviction, it appears false or very implausible that S’s appearance that P should lessen S’s justification for ¬SH. One might find it intuitive that appearances and correlated reflective beliefs must always have the same justifying force. However, our intuitions about this matter crucially depend on the conception of appearance we work with. If one adopts the conception presupposed by the phenomenal conservative—according to which appearances are non-doxastic attitudes that represent states of affairs as actualized—one will naturally conclude that appearances and correlated introspective beliefs may often have very dissimilar justifying force. In the case in point, note that the content of S’s belief that F—the proposition that it appears to S that this is a hand—and the content of S’s appearance that P—the proposition that this is a hand—are about disparate states of affairs, where the first concern S’s mind and the second the external world. Furthermore, note that these contents don’t stand in the same logical relation with SH—the proposition that this is a fake-hand. Because of all of this, S’s belief that F and S’s appearance that P presumably don’t have, when separately considered, the same evidential bearing on SH, and thus on ¬SH. Since SH entails F or makes F highly probable, it is intuitive that S’s coming to believe F should result in S’s acquiring justification for SH, and so lessening her justification for ¬SH. On the other hand, since P is logically incompatible with SH, it is very implausible that S’s appearance that P—in which it seems to S to ascertain that P—could result in S’s acquiring justification for SH.10 It is much more plausible that S’s appearance that P should result in S’s lessening her justification for SH, and so in S’s acquiring justification for ¬SH. Since S’s appearance that P and S’s correlated belief that F turn out to have divergent evidential force in the case just considered, it is very dubious that White’s formalism could accurately model the notion of seeming-based justification that (PC) aims to account for. Thus, it is very dubious that the consequences that White draws from his formal framework could capture key features of that notion of perceptual justification. To persuade the unimpressed reader that White’s objection actually goes off the track, let me re-formulate it directly in terms of seeming-based justification. If White’s objection goes through, this re-formulation of it must also go through. But it doesn’t. Since S knows that P entails ¬SH and degrees of justification are presumably closed under known entailment, when S has an appearance that P, S has some degree of justification for believing P only if S has at least the same degree of justification for believing ¬SH given the same appearance that P. Therefore, considering that S’s appearance that P lowers S’s 10 Pryor acknowledges this when he writes: ‘My view is that whenever you have an experience as of P, you thereby have immediate prima facie justification for believing P … Your experiences do not, in the same way, give you immediate prima facie justification for believing that you are dreaming, or being deceived by an evil demon, or that any of the skeptic’s other hypotheses obtain.’ (2000: 536).

64

4 The Bayesian Objection degree of justification for ¬SH, S can have some degree of justification for believing P when S has an appearance that P only if S has already—and so independently—a higher degree of justification for believing ¬SH.

Since it is false that S’s appearance that P lowers S’s degree of justification for ¬SH, this argument is unsound. To salvage the cogency of White’s argument, one might think of re-formulating it by replacing F with P itself. The hope might be that since a belief that P and an appearance that P share the same content, believing P could prove a closer surrogate than believing F for having an appearance that P. This would require replacing Pr F (P), Pr F (SH) and Pr F (¬SH) with, respectively, Pr P (P), Pr P (SH) and Pr P (¬SH), which refer to the degrees of confidence in P, SH and ¬SH that S should rationally have whenS comes to believe P (rather than F) with certainty. The goal would be proving that Pr P (P) < Pr(¬SH). It is easy to see, however, that this proposition is false. In fact, note that Pr P (P) = 1.11 Hence, given that 1 is the highest possible degree of credence, it cannot possibly be that Pr P (P) < Pr(¬SH). Perhaps one might think of proving the weaker proposition that (3*) Pr P (P) ≤ Pr(¬SH). (3*) can be interpreted as indicating that when S has an appearance that P, unless S has independent justification sufficient to believe ¬SH, S doesn’t have justification sufficient to believe P. However, (3*) is also false. Since P entails ¬SH, Pr P (¬SH) = 1. Furthermore, since S cannot be just certain that ¬SH, Pr(¬SH) < 1. Thus, Pr P (¬SH) > Pr(¬SH). The latter inequality and the proposition that Pr P (P) = 1 = Pr P (¬SH) jointly entail that Pr P (P) > Pr(¬SH). Another suggestion might be that of replacing F with P while deploying Jeffrey conditionalization rather than standard conditionalization. Suppose X is an evidential proposition of S. Jeffrey conditionalization enables S to update her credence function on uncertain evidence—i.e. on a mere change of S’s confidence in X that may not result in S’s holding X with credence 1. Let Pr old be S’s credence function before S changes her credence that X, and let Pr new be S’s credence function upon S’s changing her credence that X. Let Y be any proposition on which Pr old is defined. On Jeffrey conditionalization, (JC) Pr new (Y ) = Pr new (X) Pr old (Y|X) + Pr new (¬X) Pr old (Y|¬X) (JC) states that S’s credence that Y upon S’s changing her credence that X equates to the sum of two things: the product of S’s new credence that X by S’s conditional credence that Y given X before S changed her credence that X, and the product of S’s new credence that ¬X by S’s conditional credence that Y given ¬X before S’s changed her credence that X. Suppose now that S’s evidential proposition X is P. Jeffrey conditionalization enables S to update her credence function on a mere increase in her rational confidence that P. A mere increase in S’s confidence in P might appear to be a very close surrogate 11 I

follow White (2006) in assuming that Pr(P) > 0 so that P can actually be learned by S.

4.2 Responding to White’s Objection

65

for S’s having an appearance that P. One might hope to rehabilitate White’s argument by showing that (3+) Pr new (P) ≤ Pr old (¬SH). However, (3+) is false because one of its instances is false. To see this, consider that standard conditionalization is just a special case of Jeffrey conditionalization in which the evidential proposition is held by S with credence 1.12 Given this, in any case in which the evidential proposition P is held by S with credence 1, the formalism of Jeffery conditionalization can be replaced salva veritate with the formalism of standard conditionalization. For example, in these cases, we can replace Pr new (P) = 1 with Pr P (P) = 1, and Pr old (¬SH) = n with Pr(¬SH) = n. A consequence of this is that in any case in which the evidential proposition P is held by S with credence 1, (3+) can be re-written salva veritate as (3*) Pr P (P) ≤ Pr(¬SH). This shows that (3*) is actually an instance of (3+). Since we have seen that (3*) is false, (3+) is also false. At this point, one might still hope to obtain a working variant of White’s argument by showing that (3+) is true for some specific range of values of Pr new (P), not inclusive of 0 and 1, suitable to formalize the epistemic import of S’s entertaining an appearance that P. But this looks like something very hard to show. To wrap up, White hasn’t actually shown that (PC) is incompatible with Bayesianism. For the formal notion of justification at work in his argument doesn’t provide a sufficiently accurate model of the intuitive notion of seeming-based justification, and there is no easy way to produce a working variant of White’s argument that relies on a more accurate formalization.

4.3 Perceptual Appearances and Reflective Justification Before concluding this chapter, I would like to probe a few interesting issues that might have crossed the mind of the reader in the course of the former discussion. My investigation will pave the way for the task of appraising the actual antisceptical bite of phenomenal conservatism, carried out in the next chapter. Phenomenal conservatives can and should generalise the thesis that S’s appearance that P and S’s reflective belief that F have different evidential force with respect to SH and ¬SH to other propositions. Consider any appearance that X of a subject S capable of entertaining beliefs about her own seemings. The phenomenal conservative can legitimately claim that the justifying force of S’s appearance that X, with regard to a given proposition Z, and the justifying force of S’s reflective belief that (Y ) S has an appearance that X, with regard to the same proposition Z, need not coincide and may diverge. This claim prompts the following question: how do the justifying force of S’s seeming that X and the justifying force of S’s reflective belief that Y generally work together to determine S’s justification for Z? 12 In

fact note that when Pr new (X) = 1, (JC) reduces to Pr new (Y ) = Pr old (Y|X).

66

4 The Bayesian Objection

It is intuitively plausible that when S has a reflective belief that Y, the justifying force of S’s appearance that X becomes irrelevant to S’s epistemic standing with regard to Z. We can rephrase this epistemological thesis by stating that the evidential import that S’s sole appearance that X has for Z is overridden by the evidential import that S’s reflective belief that Y has for Z, when S also entertains this reflective belief. When the evidential import that S’s appearance that X has for Z is overridden by the evidential import that S’s belief that Y has for Z, the strength of S’s all things considered justification for Z—i.e. the strength of the justification for Z that S has when she entertain both the appearance that X and the correlated reflective belief that Y —is the same as the strength of S’s justification for Z based on S’s reflective belief that Y alone. Let’s call reflective justification S’s justification for Z based on S’s reflective belief that Y alone. To appreciate the intuitive plausibility of this thesis, consider the following thought experiment. S faces a randomly chosen table in a room. Since S’s eyes are closed, S cannot see the colour of the table. However, S is reliably told that (BK) The room has two white tables and two red tables. One of the white tables is illuminated by a hidden red light, so that it looks red, but the other three tables are illuminated by natural light.

Let’s stipulate that: R = ‘the table is red’. A = ‘it appears to S that the table is red’. SH* = ‘the table is white but looks red because it is illuminated by a hidden red light’. Given BK, it is easy to determine what the strength of S’s confidence in R and ¬SH* should be before S opens her eyes. Since S knows that the chance that R is ½ and the chance that ¬SH* is ¾, then Pr(R) = ½ and Pr(¬SH*) = ¾.13 The strength of S’s credence that R and ¬SH* can be taken to indicate the strength of S’s justification for believing R and ¬SH*. Considering that the justification threshold for plain belief is presumably very high and near to 1, the justification values ½ and ¾ are intuitively insufficient to sustain, respectively, S’s belief that R and S’s belief that ¬SH*. Accordingly, before opening her eyes, S doesn’t have justification sufficient to believe R or ¬SH*. Given BK, it is also easy to determine what the strength of S’s confidence in R and ¬SH* should be if S believed A. Since S would know that the cases in which R is true are 2/3 of those in which A is true, and the cases in which ¬SH* is true are also 2/3 of those in which A is true, Pr A (R) = Pr A (¬SH*) = 2/3. Again, this level of reflective justification would intuitively be insufficient to sustain S’s belief that R or that ¬SH*. So, if S believed A, S wouldn’t thereby have reflective justification sufficient to believe R or ¬SH*. Now imagine that S opens her eyes and has an appearance that R. Given (PC), S must acquire prima facie justification sufficient to believe R. Furthermore, since S’s 13 This

follows from the application of (some version of) Lewis’ Principal Principle. I take S’s information BK to be embedded in Pr.

4.3 Perceptual Appearances and Reflective Justification

67

seeming gives her the impression of verifying R, which is incompatible with SH*, it is plausible that S has also justification sufficient to believe ¬SH*, but not to believe SH*. Suppose that after this, S reflects on her experience and comes to believe A. It is strongly intuitive that at this point S’s appearance that R will cease to be evidentially relevant for R and ¬SH*. At this point, S’s all things considered justification for R and ¬SH* will only be determined by the evidential force of S’s reflective belief that A. Since Pr A (R) = Pr A (¬SH*) = 2/3, S will no longer have all things considered justification sufficient to believe R or ¬SH*. Let me describe a different thought experiment which might hopefully persuade readers still unconvinced that the evidential force of S’s appearance that R is overridden by the evidential force of S’s reflective belief that A. Suppose an oracle—who S knows to be virtually infallible—tells S that S will have an appearance that R at a future time t. So, now, S already believes A without hesitation.14 Imagine, the oracle instantly adds information BK about the chance that R and ¬SH* will have at t— which S also believes without hesitation. So, now, S has already a suitable amount of justification for R and ¬SH* based on A.15 Suppose that, at t, S actually comes to have an appearance that R. It is intuitive that S’s degree of justification for R and ¬SH* shouldn’t change. This suggests, again, that the evidential force of S’s reflective belief that A with respect to R and ¬SH* overrides the evidential force of S’s sole appearance that R with respect to R and ¬SH*. As we will see in Sect. 5.3, an independent reason to accept the thesis that the justification based on S’s appearances is generally overridden by the justification based on S’s correlated reflective beliefs is that this thesis plays a crucial role in resolving the problem of easy justification, which apparently afflicts phenomenal conservatism. Let me make a few final remarks. Suppose S has a seeming that X and a reflective belief that (Y ) she has a seeming that X. To begin with, note that S’s seeming-based justification for X and S’s reflective justification for X are different in type. In particular, S’s seeming-based justification for X is both non-inferential and immediate. Whereas S’s reflective justification for X is both inferential and non-immediate. S’s reflective justification for X is inferential because it is available to S through an inference from Y to X. (For instance, the inference from ‘it appears to S that the table is red’ to ‘the table is red’.) S’s reflective justification for X is non-immediate because it is constituted by the portion of S’s justification for believing Y transmissible to X through the inference. Consider now any error conjecture Z that states that although S has a seeming that X, X is false. White’s proof—generalized to any perceptual proposition X, correlated reflective belief Y and correlated error conjecture Z—shows that Pr Y (X) < Pr(¬Z). This means that the degree of S’s reflective justification for believing X has an upper 14 S’s

belief that A isn’t—rigorously speaking—a reflective one before t, but this is irrelevant. forestall possible confusion let’s make the implicit temporal references in A, R and ¬SH* explicit. Thus, A = ‘it appears to S at t that the wall is red at t’, R = ‘The wall is red at t’, and ¬SH* = ‘At t the wall is white but looks red because it is illuminated by a hidden red light’. 15 To

68

4 The Bayesian Objection

bound in the degree of independent justification that S has for believing that Z is false. However, note that we have found no reason to think that the degree of S’s seeming-based justification for believing X has such an upper bound. One might suppose that the finding about reflective justification heavily relies on interpreting justification as rational credence, and so on the Bayesian machinery. This seems false. White’s Bayesian proof has a parallel informal argument that exploits basic intuitions about epistemic justification. This is a version of the argument: since S knows that X entails ¬Z and degrees of justification are presumably closed under known entailment, when S has the reflective belief that Y , S can have some degree of justification for believing X only if S has at least the same degree of justification for believing ¬Z, given the same reflective belief. Consider now that since S’s reflective belief that Y increases S’s degree of justification for Z at least a little bit,16 the same reflective belief cannot but lower S’s degree of justification for ¬Z. Thus, S can have some degree of justification for believing P when S has the reflective belief that Z only if S has already—and so independently—a higher degree of justification for believing ¬Z. Finally, note that S’s reflective justification for X may not suffice to support S’s belief that X, though S’s seeming-based justification alone may suffice to it. (This happens, for instance, in the examples about the coloured tables examined before.) This possibility indicates that S’s reflective belief that Y (accompanied by suitable background information) can act as a defeater of S’s seeming-based justification for Y.

4.4 Conclusions White has articulated the sharpest and most detailed Bayesian objection to (PC) delivered in the recent literature. I have argued that White’s objection is nevertheless inconclusive because it relies on an implausible model of the intuitive notion of seeming-based justification. I cannot exclude that a more accurate formalization of this notion could ultimately be found. However, I have no reason to suppose that such a formalization would show that perceptual justification doesn’t have the central features that (PC) credits to it. My analysis of White’s objection lends support to the following propositions: seemings and correlated reflective beliefs can diverge in justifying power, the strength of reflective justification (but not the strength of the correlated seeming-based justification) has an upper bound in the strength of the justification for denying any relevant error conjecture, reflective justification generally overrides the correlated seeming-based justification, a reflective belief can act as a defeater of the correlated seeming-based justification.

16 Note

that Y = ‘S has an appearance that X’ and that Z entails Y and explains it to some extent.

References

69

References Balcerak-Jackson M (2016) Perceptual fundamentalism and a priori bootstrapping. Philos Stud 173:2087–2103 Cohen S (2005) Why basic knowledge is easy knowledge. Philos Phenomenol Res 70:417–430 Fumerton R (1995) Metaepistemology and skepticism. Rowman & Littlefield, Lanham, MD Huemer M (2007) Compassionate phenomenal conservatism. Philos Phenomenol Res 74(1):30–55 Jehle D, Weatherson B (2012) Dogmatism, probability, and logical uncertainty. In: Restall G, Russell G (eds) New waves in philosophical logic. Palgrave Macmillan, New York, pp 95–111 Kung P (2010) On having no reason: dogmatism and Bayesian confirmation. Synthese 177:1–17 Moretti L, Piazza T (2013) When warrant transmits and when it doesn’t. Synthese 190:2481–2503 Pryor J (2004) What’s wrong with Moore’s argument? Philos Issues 14:349–378 Pryor J (2007) Uncertainty and undermining. Unpublished manuscript downloadable at: https:// www.jimpryor.net/research/papers/Uncertainty.pdf Schiffer S (2004) Skepticism and the vagaries of justified belief. Philos Stud 119:161–184 Silins N (2007) Basic justification and the Moorean response to the skeptic. In: Gendler TS, Hawthorne J (eds) Oxford studies in epistemology, vol 2. Oxford University Press, Oxford, pp 108–140 Tucker C (2013) Seemings and justification: an introduction. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 1–29 Weatherson B (2007) The Bayesian and the dogmatist. Proc Aristot Soc 107:169–185 Wedgwood R (2013) A priori bootstrapping. In: Casullo A, Thurow J (eds) The a priori in philosophy. Oxford University Press, Oxford, pp 226–246 Weisberg J (2011) Varieties of Bayesianism. In: Gabbay D, Hartmann S, Woods J (eds) Handbook of the history of logic, vol 10. Elsevier/North-Holland, Amsterdam, pp 477–551 White R (2006) Problems for dogmatism. Philos Stud 131:525–557 Wright C (2007) Perils of Dogmatism. In: Nuccetelli S (ed) Themes from G. E. Moore: new essays in epistemology and ethics. Oxford University Press, Oxford, pp 25–48 Wright C (2011) McKinsey one more time. In: Hatzimoysis A (ed) Self-knowledge. Oxford University Press, Oxford, pp 80–104

Chapter 5

Antiscepticism and Easy Justification

Abstract In this chapter I investigate the epistemological consequences of the fact that seeming-based justification is elusive, in the sense that the subject can lose it simply by reflecting on her seemings. I argue that since seeming-based justification is elusive, the antisceptical bite of phenomenal conservatism is importantly limited. I also contend that since seeming-based justification has this feature, phenomenal conservatism isn’t actually afflicted by easy justification problems. Keywords Phenomenal conservatism · Reflective awareness · Scepticism · Responses to scepticism · Easy justification · Bootstrapping

5.1 Antiscepticism and Reflective Awareness Suppose P is the content of a possible appearance of S. A sceptical alternative SH or sceptical scenario SH alternative to P is a special error conjecture SH such that, if SH is true, (i) P is false though S has an appearance that P, and (ii) S couldn’t discover that SH is true, no matter how much or how deep S might investigate. As we have seen in Sect. 2.3, one of the asserted merits of phenomenal conservatism is that it affords us the means of a thoroughgoing response to the sceptic—in particular, the sceptic who insists that S must have independent justification for ruling out any relevant sceptical alternative in order to possess justification for believing ordinary things. For instance, suppose that S has a perceptual appearance that (P) there is a cat on the mat, and no reason to think that this appearance could be deceptive. The sceptic could argue that since it might be the case that, say, (SH) P is false but S has a hallucination that P caused by the Matrix, S has justification for believing P only if S has independent justification for ruling out SH. The sceptic will insist that since S cannot have this independent justification, S doesn’t have justification for believing P. Antisceptics can attempt various lines of reply. The phenomenal conservative would respond that since (PC) is true and S has an appearance that P, S does have prima facie justification for believing P even if S has no independent justification for believing ¬SH.1 The phenomenal conservative will conclude that 1I

take the expression ‘ruling out SH’ to mean ‘believing ¬SH’.

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2020 L. Moretti, Seemings and Epistemic Justification, SpringerBriefs in Philosophy, https://doi.org/10.1007/978-3-030-43392-5_5

71

72

5 Antiscepticism and Easy Justification

since S has no reason to think that her appearance might be deceptive in the case in point, S does have justification for believing P. Although this line of response may appear very promising or even successful, it isn’t free from problems. I will now make a case to the effect that the antisceptical bite of (PC) is importantly limited. A response to the sceptic would actually be forceful only if—I suggest—it enabled a thinker S who engaged with a sceptical argument that questioned her own possession of epistemic justification (or knowledge) to reject or seriously challenge the argument. A paradigmatic example of such a thinker is Descartes in his Meditations. It appears to me, however, that (PC) doesn’t allow for this type of response to the sceptic. As we have seen in Sect. 4.3, seeming-based justification is elusive.2 For S can lose her seeming-based justification for believing a proposition P by simply reflecting on her mental states and thereby acquiring a belief that she has a seeming that P. In these circumstances, S’s seeming-based justification for P would in fact be overridden and thus replaced by S’s reflective justification for P (the one based on S’s reflective beliefs that she has a seeming that P), where the latter justification may be very different in strength from the former. I will now show that since seeming-based justification is elusive in this sense, if S attempted to reject a sceptical argument that questioned her own possession of seeming-based justification by adducing (PC) in discussion or private reasoning, S would lose her seeming-based justification, making her appeal to (PC) ineffective. Imagine S has an appearance that (P) that there is an apple on the table, and no reason to think it might be deceptive. Suppose that at a certain point S engages with a sceptical argument that questions the claim that S has actually justification for believing P. The argument states—precisely—that since S’s appearance P can possibly be caused by the instantiation of a sceptical alternative SH to P, S could have justification for believing P only if she had independent justification for believing ¬SH. (Suppose SH states that while P is false, S has a hallucination that P caused by an evil demon.) The argument concludes that since S doesn’t have this independent justification, S doesn’t possess justification for believing P. This sceptical argument is just one of those that (PC) is claimed to enable us to defuse. The question is— therefore—whether, in the imagined circumstances, S can respond successfully to this argument by adducing (PC) and contending that she does have justification for believing P. Unfortunately for the phenomenal conservative, the answer is negative. Consider that S can actually engage with the sceptical argument under consideration only if she grasps the way in which SH puts her justification for P at risk. S can do this only if S becomes reflectively aware of her seeming that P. Furthermore, S can competently adduce (PC), to respond to the sceptic, only if she is reflectively aware of her seeming that P. For this seeming is referred to in the relevant instance of (PC) that S is committed to invoking. If S becomes reflectively aware of her seeming that P, however, even if (PC) holds true, S’s seeming-based justification for P will be overridden by S’s reflective justification for P based on S’s reflective belief that she 2I

use the term ‘elusive’ in a sense similar to one used Lewis (1996) to describe that which he thought was an important feature of knowledge. This similarity doesn’t commit me to endorsing Lewis’ views.

5.1 Antiscepticism and Reflective Awareness

73

has a seeming that P. Thus, at this point, whether or not S has justification for believing P no longer depends on S’s having seeming-based justification; so it doesn’t hinge on (PC)’s truth. Whether or not S has justification for believing P depends on S’s having reflective justification, which is not accounted for by (PC). That’s why S cannot reject the sceptical argument by appealing to (PC). In response, phenomenal conservatives might try to deny the general thesis that reflective justification overrides the correlated seeming-based justification. A basic problem of such a strategy is that the thesis appears true in the circumstances we have envisaged. Suppose that at a time t 0 , S has a seeming that P and no reason to think that it might be deceptive. In these circumstances, S’s seeming that P gives S justification for believing P. Suppose that at a successive time t 1 , S engages with the sceptical argument that questions her having justification for P. Accordingly, S becomes reflectively aware of her seeming that P. At t 0 , S didn’t have the belief that she had that seeming. S was thus unable to wonder whether her seeming was veridical or deceptive. At t 1 , S is able to pose this question and, pressed by the sceptical case, she actually does so. S finds two possible answers. A first hypothesis available to S—the perception hypothesis—states that S truly perceives that P. The sceptical argument offers S an alternative hypothesis SH, which states that P is false and S has a hallucination that P. It is intuitive that in this predicament, S’s seeming that P can no longer give S justification for believing P. Suppose in fact that, at t 1 , S’s seeming that P gave S justification for believing P. Since, at t 1 , S reflectively believes that she has that seeming, S would thereby acquire justification for believing that the seeming that she has is veridical—namely, for believing the perception hypothesis and ruling out SH. But this looks perversely circular: no seeming of S could per se provide S with justification for believing that the seeming itself is veridical. Justification for believing so requires independent evidence. Hence, it appears very plausible that, at t 1 , S’s seeming that P can no longer give S justification for believing P.3 (I return to these issues in Sect. 5.3.) At t 1 , S could have some justification for believing P if S were able to conclude that the perception hypothesis is likely to be the correct explanation of her seeming’s existence. S would get justification for this conclusion through an inference to the probable explanation of her seeming’s existence. Inferences don’t link seemings with doxastic states, they connect premises with conclusions—namely, doxastic states with other doxastic states. The inference by S would go from her reflective belief that she has a seeming that P to her belief that the perception hypothesis has a good chance to be true. At t 1 , S could possibly acquire justification for P via an inference like this. These considerations strongly suggest that, at t 1 , it is S’s reflective belief that she has a seeming that P—rather than S’s seeming itself—that provides the basis of S’s justification for P.4 At t 1 , S’s seeming that P only plays an indirect role: S’s acquaintance with this seeming forms the basis of the justification of the premise of 3 One

might alternatively attempt to stop the vicious circularity by conceding that S’s seeming that P gives S justification for P while insisting that S’s reflective acquaintance with her seeming doesn’t give S justification for believing that she has that seeming. But the last claim is very counterintuitive. 4 This discussion naturally suggests a view about epistemic justification that parallels Sosa’s celebrated thesis there are two different kinds of knowledge. According to Sosa,

74

5 Antiscepticism and Easy Justification

S’s inference—S’s reflective belief that she has a seeming that P. In conclusion, at t 1 , S’s seeming-based justification for P appears to be overridden by S’s reflective justification for P. This is why S couldn’t reject the sceptical argument by appealing to (PC). A question naturally arising at this point is whether S would actually be able to acquire enough justification to believe P once she has engaged with the sceptical argument. I cannot hope to settle this important question here. Let me make only one remark. We should concede to the sceptic that S could have justification sufficient to believe P, in this case, only if S had higher independent justification for believing ¬SH. This is so because in the imagined circumstances, S’s reflective justification for P would override S’s seeming-based justification for P and—as we have seen in Sect. 4.3—the strength of S’s reflective justification has an upper bound in the strength of S’s independent justification for believing that any relevant error conjecture is false; so, also for believing ¬SH. One might suggest that, thanks to (PC), this condition necessary for S’s having enough reflective justification for believing P is satisfied as long as SH refers to a sceptical conjecture (rather than an ordinary error conjecture). The suggestion precisely says that since SH is a terribly far-fetched, odd and unnatural proposition as only sceptical scenarios can be, if S were going to appraise SH, S would normally have a seeming that ¬SH. And this seeming would grant S enough prima facie justification to believe ¬SH. Note that if this is true, our seemings can provide us with immediate justification for ruling out sceptical conjectures. Since I have actually heard this suggestion a few times in conversation, let me briefly examine it. My view is that the suggestion faces a series of hurdles, and ultimately fails. To begin with, note that if S had a seeming that ¬SH, the conclusion licensed by (PC) would be that S has prima facie justification sufficient to believe ¬SH. However, we have seen that S could have reflective justification sufficient to believe P only if she had higher justification for believing ¬SH. Thus, even if S had a seeming that ¬SH, it is unclear that the condition necessary for S’s having enough reflective justification to believe P would be met. The argument under consideration One has animal knowledge about one’s environment, one’s past, and one’s own experience if one’s judgments and beliefs about these are direct responses to their impact—e.g., through perception or memory—with little or no benefit of reflection or understanding… [Furthermore,] one has reflective knowledge if one’s judgment or belief manifests not only such direct response to the fact known but also understanding of its place in a wider whole that includes one’s belief and knowledge of it and how these come about. (1991: 240) Animal knowledge is mere reactive knowledge, whereas reflective knowledge is mainly unifying or integrating knowledge. One could similarly distinguish between animal justification and reflective justification. S’s animal justification for believing Q is identifiable with S’s justification for Q based on her seeming that Q. Animal justification has essentially the function of certifying the rationality of S’s reactive beliefs—those directly caused by S’s appearances. S’s reflective justification for believing Q is instead justification based on S’s belief that she has an appearance that Q. Justification of this type stems from S’s ability to explain her appearance that Q by adducing Q in conjunction with other propositions about her environment and cognitive faculties. Reflective justification has the function of certifying the rationality of the beliefs of S that aim at a unified and integrated view of the world.

5.1 Antiscepticism and Reflective Awareness

75

seems to rely on a principle of seeming-based justification stronger than (PC), which should be clarified and defended. Another concern—which I have articulated in Moretti (2019)—is that S’s justification based on a seeming that ¬SH would be elusive in the same way as S’s seeming-based justification for P is elusive. S would lose her seeming-based justification for ¬SH if she engaged with a sceptical argument that questioned her possession of it. There are more serious problems: under closer scrutiny, it is quite dubious that if S inspected ¬SH, S would actually have a seeming that ¬SH, or seeming-based justification sufficient to believe ¬SH. To begin with, note that a seeming that ¬SH would not need to come with sensory or mental images of any type—it would be qualia-free. It would thus closely resemble a rational (or a priori) appearance, such as an appearance that 1 = 1. A seeming that ¬SH couldn’t however qualify as a rational seeming, for rational seemings are about necessary truths, and ¬SH doesn’t look like a necessary truth. This might raise some initial concern about the claim that S can have a seeming that ¬SH. Nevertheless, some authors think that people can entertain qualia-free seemings that are not rational seemings. This could happen, for instance, in attested cases of blindsight. According to these authors, in these cases, self-avowed partly blind subjects have accurate seemings about objects present in their environment, although they have no visual sensation (cf. Tucker 2010; Huemer 2013). Since the content of a qualia-free seeming of this type would be a contingent proposition, one can insist that there is no principled problem with the claim that S can entertain a qualia-free seeming that ¬SH, though ¬SH is contingent. Even so, the claim is questionable. A seeming that ¬SH can be free from noncognitive qualia (e.g. colours, flavours, smells, sounds, and so on) but it cannot be completely free from cognitive qualia. Phenomenal conservatives admit of at least one cognitive quale: the phenomenal force of appearances (described in Sect. 2.2). Because of its phenomenal force, a seeming that ¬SH is supposed to be an experience of verifying or ascertaining that ¬SH. However, when one inspects a proposition like ¬SH, admittedly, one normally has no experience of this sort. So, again, it looks dubious that S can have a seeming that ¬SH. To respond, one could suggest that seemings can actually be stronger or weaker because their phenomenal force is not just a matter of all or nothing but, rather, it comes in degrees (cf. Huemer 2005: 100; Koksvik 2011: 127). One could then insist that a seeming that ¬SH would just be a weak seeming. This would explain why when we inspect a proposition like ¬SH, we don’t have the feeling of verifying that ¬SH. Suppose this is correct. If S’s seeming that ¬SH were a weak seeming, presumably, it would provide S only with weak justification for ¬SH.5 Therefore, S would not have justification sufficient to believe ¬SH. So, S could not have reflective justification sufficient to believe P. In conclusion, this rejoinder doesn’t help. The only alternative reply I can think of draws on Huemer (2016)’s theory of inferential seemings, which expands phenomenal conservatism to deliver an account

5 The

weak version (pc) of (PC), presented in Sect. 2.1, can account for weak seeming-based justification of this type.

76

5 Antiscepticism and Easy Justification

of inferential justification. According to Huemer’s theory, S’s having inferential justification for Q from P requires—among other things—S’s having justification for believing P, and S’s entertaining an inferential seeming that represents Q as likely in light of P,6 where the likelihood can be relative to S’s background information B. In particular, S’s inferential seeming would include reference to B if S’s sense of likelihood constituting the seeming were generated by mental processes of S shaped by B, and if S would be disposed to acknowledge, if the issue arose, the relevance of B to the likelihood of Q. Huemer suggests that a limiting case of S’s having inferential justification for Q is one in which only B, but no premise P, is involved in the justification for Q. Here is an example: When the issue arises as to what country I occupy, it seems to me very likely (to put it mildly) that I live in the United States. This sense of likelihood is shaped by a host of past experiences and beliefs of mine that are relevant to where I live, including many experiences and beliefs that I do not now remember. That is part of why my sense of likelihood counts as referring to the probability of my living in the United States conditional on all those past experiences and beliefs. But... it is also a matter of how I would react if some of these past experiences or beliefs were raised in connection with the question of how likely it is that I live in the United States. For example, if the question arises, I will view the fact that Colorado is in the United States (which I believe) as supporting the claim that I “probably” live in the United States. (Huemer 2016: 158)

One could contend that S’s seeming that ¬SH would actually be an appearance of this type. If this is correct, the seeming would be—precisely—an appearance that ¬SH is likely. This would clarify why S would not have the feeling of verifying that ¬SH. However, also this proposal is problematic. S’s seeming that ¬SH is likely could supply S with justification sufficient to believe that ¬SH is likely. It is unclear, however, that a seeming of this sort could give S justification sufficient or even more than sufficient to believe ¬SH. One might however suppose that if ¬SH appeared extremely likely to S, S might get justification sufficient or even more than sufficient to believe ¬SH. So, the question is whether or not ¬SH would appear to S to be extremely likely. According to Huemer’s theory, how likely ¬SH would appear to S to be is determined by the content of ¬SH and background information B. Suppose for example S has a firm background belief that (ND) demons don’t exist. If SH stated that S is deceived by a demon into experiencing as if P, S might in these circumstances have a seeming that ¬SH is extremely likely. This won’t always be the case, though. Suppose S believes ND but SH states that S is hallucinated by the Matrix. Or suppose SH states that S is hallucinated by a demon but S doesn’t believe ND. In either case, S may not have a seeming that ¬SH is extremely likely. Indeed, there is no guarantee that S would have a seeming that ¬SH is extremely likely. Whether or not S would have such a seeming depends on the specific contents of SH and B. It is also important to note that if SH were a global sceptical conjecture, S’s conceiving of SH could result in altering the content of B, which could preclude 6 When

P just entails Q, the seeming represents Q as maximally likely in light of P.

5.1 Antiscepticism and Reflective Awareness

77

S from having a seeming that ¬SH is likely. Suppose for example that SH states that S is a disembodied soul in an immaterial world deceived by an evil demon into perceiving the ordinary world. Imagine that S initially believes ND (demons don’t exist) on the grounds of evidence coming from her science readings. S’s conceiving of SH could lead S to doubt that she actually has evidence for ND, thus to doubt ND. Because of this, S may become unable to entertain a seeming that ¬SH is likely. To conclude, even if Huemer’s theory of inferential seemings were correct, there would no guarantee that, upon inspecting SH, S would have seeming-based justification sufficient on more than sufficient to believe ¬SH. In summary, we have seen that the antisceptical bite of (PC) is importantly limited because S couldn’t successfully appeal to (PC) to rebut any sceptical argument that questioned her own possession of appearance-based justification for a proposition P. The problem is that if S engaged with the argument or competently invoked (PC), her seeming-based justification for P would be replaced by reflective justification for P, which may not suffice to sustain S’s belief that P and is not accounted for by (PC). We have also seen that the truth of (PC) cannot guarantee the fulfilment of a condition necessary to make S’s reflective justification robust enough to sustain S’s belief that P.

5.2 Easy Justification Objections I have argued that the fact that seeming-based justification is elusive (in the sense explained) has a negative impact on phenomenal conservatism, as it weakens its antisceptical bite. Allow me now to show that the very same fact also has a positive consequence for phenomenal conservatism, as it offers its supporters the means to reject two apparently forceful objections against it. In this section I present these objections. My cases against them will follow in the next two sections. Consider a subject S, an evidence source ES (such as perception, memory, induction, etc.) and a proposition P. Suppose ES provides S with evidence that P. Any view W (internalist or externalist) that allows S to know that P on the basis of ES regardless of S’s having independent knowledge that ES is reliable is called by Cohen (2002, 2005) a basic knowledge theory. Cohen contends that any basic knowledge theory W is affected by easy knowledge problems. In particular, on W, it will prove too easy for S to know that ES is reliable, which suggests that W is flawed. Cohen’s arguments often range from focusing on knowledge to focusing on epistemic justification. The problems of easy knowledge can indeed be reformulated in terms of easy justification. Let’s call basic justification theory any view W (internalist or externalist) that allows S to acquire justification for believing P through ES regardless of S’s having independent justification for believing that ES is reliable. According to Cohen, any basic justification theory W is plagued by easy justification problems. In particular, on W, it will prove excessively easy for S to acquire justification for believing that ES is reliable, which suggests that W is flawed. Suppose again that ES provides S with evidence that P. There are various ways to interpret the claim that ES is reliable. Let’s consider two of them. That claim

78

5 Antiscepticism and Easy Justification

can be read as stating that ES is reliable in this particular case—meaning that S’s evidence that P is not deceptive—or it can be interpreted as stating that ES is reliable in general—meaning that the propositions which ES supplies (good) evidence for are true most of the time. When Cohen speaks of the reliability of ES, he means the reliability of ES in general.7 In my discussion, I will also refer to the notion of reliability in a particular case. For it appears to me that the two principal varieties of easy justification problems described by Cohen actually arise from interpreting ‘reliability’ in these two respective senses. (I give examples below.) Hereafter, I will focus on justification. Cohen (2002, 2005) maintains that phenomenal conservatism is plagued by easy justification problems—this is indeed a recurrent criticism of phenomenal conservatism. See for example White (2006), Wright (2007), and Siegel and Silins (2015).8 In fact note that (PC) qualifies as a basic justification theory, as it allows S to have justification for P from a seeming that P regardless of S’s having independent justification for believing that the seeming is reliable, and irrespective of S’s having independent justification for believing that the seemings of the same type are accurate most of the time. Epistemologists formulate the easy justification problems as objections that target theories of basic justification—precisely, as the easy justification from closure objection and the easy justification from bootstrapping objection. Let’s see how these objections can be used to strike phenomenal conservatism. The easy justification from closure objection interprets ‘reliable’ as ‘reliable in a particular case’ and exploits the principle of closure of justification under known entailment, according to which: (JC) If S has justification sufficient to believe P and S knows that P entails Q, then S has justification sufficient to believe Q.

(JC) or a very similar principle is pervasively used in science and ordinary epistemic practices to expand the set of one’s justified beliefs via deduction. (JC) looks very plausible and most epistemologists endorse it. This is how the objection unfolds (cf. Cohen 2002). Consider a proposition R and a correlated error conjecture SH. Suppose for instance that: R = ‘the table is red’; SH = ‘the table is white but seems red because it is illuminated by a hidden red light’. Imagine S has an appearance that R and no reason to suspect that the appearance is deceptive. If (PC) is correct, S will acquire justification sufficient to believe R. Since S knows that R entails ¬SH, thanks to (JC), S will also acquire enough justification to believe ¬SH. Since SH can be replaced by any alternative error conjecture S might think of, S will get justification sufficient to believe—more generally—that her seeming is reliable. However, acquiring justification in this way appears really too easy. If one could rationally exclude all relevant error possibilities in this rather 7 See

especially Cohen (2005). authors focus on perceptual dogmatism, but their objections can be extended to hit more generally phenomenal conservatism.

8 These

5.2 Easy Justification Objections

79

simple way, one would never rely on inductive justification from past cases to dispel sceptical concerns,9 and one would never perform independent checks.10 However, as a matter of fact, people do rely on inductive justification from past cases and they often perform independent checks to exclude possible mistakes (cf. Cohen 2002: 313). There is a more fundamental reason why such a way to get justification looks counterintuitive: saying that S can rationally rule out possible cognitive errors in the way described is substantially the same as saying that a seeming of S on its own and without the help of independent evidence can give S justification for believing that the seeming itself is not deceptive. This appears hopelessly circular (cf. Siegel and Silins 2015: 787). In conclusion, since (JC) and (PC) jointly sanction this intuitively flawed way to acquire justification for rejecting error conjectures, but (JC) is intuitively very plausible, (PC) must be flawed. It is important to note that a principle alternative to (PC) which permitted S to acquire justification from a seeming only if S had independent justification for believing that the seeming is reliable would not raise the easy justification problem just described. On this alternative principle, in order to receive justification from a seeming, S should already have justification for ruling out the relevant error conjectures. Therefore, S would not acquire this justification from her seeming.11 Let’s turn to the bootstrapping objection. A version of this argument was initially raised by Fumerton (1995) and Vogel (2000) against reliabilism about justification, but it became soon clear that internalist theories of basic justification can also be targeted by it. This objection interprets ‘reliable’ as ‘generally reliable’ and makes use of both deductive and ampliative inferences. Furthermore, also this objection relies on thought experiments. Cohen (2010) for example imagines a scenario like this: having no idea whether her colour vision is reliable, S decides to test it. A friend stands above S and holds up many coloured cards, one at a time. S looks at the first card and reason: (1) (2) (3) (4)

The first card appears red to me. The first card is red. The first card is red and appears red to me. Therefore, my colour appearance matches the actual colour of the card on this occasion.

Suppose (PC) is true. Since S bases her belief (2) on her seeming that the card is red, and she has no reason to suspect that the seeming is deceptive, (2) is justified for S. Since S bases her belief (1) on her reflective acquaintance with that seeming, (1) is also justified for S. Furthermore, (3) is justified for S because she deduces it from (1) and (2); and (4) is justified for S because she deduces it from (3). Imagine now that S reasons exactly the same way for each of n cards held up by her friend. (‘The 9 For example, to rule out SH, S could consider how many times in the past she was actually tricked into believing that an object had a certain colour by a hidden light shining on it. 10 For instance, to rule out SH, S could carefully check all light sources above the table. 11 These epistemic circumstances would instantiate a case of failure of transmission of justification (see for instance Wright 2002; Moretti and Piazza 2018).

80

5 Antiscepticism and Easy Justification

second card is blue, it appears blue to me, so the second card is blue and appears blue to me’, etc.) Eventually, S will deductively infer: (5) My colour experiences have always matched the actual colour of each of the n cards that I have viewed. Since the premises of (5) are all justified for S, (5) is also justified for her. Imagine that S finally infers from (5) by enumerative induction (cf. Cohen 2005) or inference to the best explanation (cf. White 2006): (6) My colour vision is reliable. Provided that the number n of cards viewed by S is sufficiently large, S will justifiedly believe (6). The problem is—again—that acquiring justification for believing that an evidence source is reliable in this way looks in a sense too easy. It is intuitively implausible that S could arrive at justifying the claim that her colour appearances are generally reliable by relying only on her colour appearances, and without any independent verification of their accuracy. Since the deductive and inductive steps of this procedure appear straightforward, and it is very plausible that S does have justification for (1) by reflective acquaintance with her seeming, the problem must lie in (2). We should conclude, therefore, that S cannot have justification for believing (2) on the basis of her seeming in these circumstances. This is possible only if (PC) is incorrect. So (PC) must be incorrect. A principle alternative to (PC) which permitted S to acquire justification from an appearance only if S had independent justification for believing that the appearances of the same type are reliable would not raise this easy justification problem.12 For if S could acquire justification for (2) only if S had already justification for believing that (6), S wouldn’t acquire this justification for (6) through the inferential steps described above.

5.3 Answering the Easy Justification from Closure Objection As Siegel and Silins (2015) suggest, to respond to the easy justification from closure objection, phenomenal conservatives might be tempted to reject the closure principle (JC) and contend that justification need not transmit across deductive arguments like the one from R to ¬SH.13 A problem of this strategy is that even if (JC) turned out not to be true across the board, (JC) appears true when applied to ordinary propositions 12 Weisberg (2010) nevertheless contends, against Cohen, that also non-basic justification theories of

this type are affected by a variant of the bootstrapping problem, so this problem is actually a general paradox rather than an objection to specific epistemological views like phenomenal conservatism or reliabilism. See Cohen (2010) for a response. 13 An attempt in this direction is in Cohen (2002). For views that reject (JC) independent of the easy justification problem see for instance Dretske (1970) and Avnur (2012).

5.3 Answering the Easy Justification from Closure Objection

81

like R and ¬SH. (Recall that R states that table is red, and SH states that the table is white but seems red because it is illuminated by a hidden red light.) So, this line of reply doesn’t look promising. Let me discuss an alternative and influential response, worked out independently by Pryor (2004) and Markie (2005), which doesn’t involve rejecting (JC). My own response will emerge from the criticism of this response. Pryor and Markie admit that when we inspect arguments like the one from R to ¬SH, licensed (PC), we have the impression that they are in some sense defective. According to them, however, our feeling doesn’t stem from our perceiving that these arguments cannot actually transmit justification to their conclusions14 ; it rather originates from our sensing that these arguments have no dialectical power, so they are of limited use. Pryor and Markie insist that since arguments of this type are epistemically impeccable, there is no reason to think that (PC) and similar principles of immediate justification are flawed or problematic. To understand Pryor and Markie’s contention, consider the following scenario, used by Cohen (2002) to introduce the easy knowledge from closure objection: Suppose my son wants to buy a red table for his room. We go in the store and I say, ‘That table is red. I’ll buy it for you.’... [He] worries, ‘Daddy, what if it’s white with red lights shining on it?’ I reply, ‘Don’t worry—you see, it looks red, so it is red, so it’s not white but illuminated by red lights.’ Surely, he should not be satisfied with this response. (314)

According to Pryor and Markie, what actually explains the impression that the boy wouldn’t be rationally persuaded by his father to believe ¬SH in this scenario is our sensing that the argument from R to ¬SH, adduced by the father, is dialectically ineffective. Any dialectically ineffective argument is such that it is unable to persuade an interlocutor of the truth of its conclusion if the interlocutor doubts the conclusion beforehand. For the subject who does so is rationally committed to not accepting the evidence offered in support of the premise as a justification for believing it.15 The argument from R to ¬SH appears to have this problematic feature. Since the boy doubts ¬SH (he doubts whether or not the table is white but seems red as deceptively illuminated), he cannot accept his own appearance that R (that the table is red)— referred to in his father’s case—as a justification for R. So, he couldn’t rationally believe ¬SH in virtue of his father’s argument.16 Pryor and Markie think that despite this argument is dialectically powerless, the boy does acquire justification for believing ¬SH thanks to the argument. For he has a seeming that R and no defeating evidence. So, the boy must have justification for believing R,17 and he must also have justification for believing ¬SH when he appreciates that R entails ¬SH (cf. Pryor 2004: 362–368). A very similar diagnosis can be 14 Markie focuses on knowledge rather than justification, but his claims can be recast in terms of justification. 15 I follow Pryor (2004)’s account, as it is much more detailed than Markie (2006)’s. 16 As clarified in Sect. 2.3, rational commitment is a type of coherence between propositional attitudes that doesn’t require them to be justified. So, even if the boy’s doubt that ¬SH is suppositional or hypothetical and so unjustified, this attitude still rationally commits him to not accepting his seeming that R as a justification for R. 17 This is meant to be propositional justification. If the boy actually believed R, his belief would be irrational and so doxastically unjustified.

82

5 Antiscepticism and Easy Justification

produced—according to Prior and Markie—for all deductive arguments sanctioned by (PC) and similar principles of immediate justification that seem to enable us to rule our error conjecture too easily. These arguments are in fact all epistemically faultless, though dialectically ineffective. In order to assess Pryor and Markie’s claims, it is instructive to examine Cohen (2005)’s rejoinder. Cohen concedes that a diagnosis like the one detailed above can partly explain the feeling that arguments like the one from R to ¬SH are defective, but he contends that it cannot provide the full explanation. For arguments like these— according to him—are both dialectically ineffective and epistemically flawed. To support his view, Cohen produces a variant of the scenario described before. In this variant, his son no longer doubts ¬SH. He accepts that his father knows that ¬SH, and asks him why it is the case that ¬SH. The father answers in the same way as before: ‘Oh, that’s easy.[The table] looks red, so it is red, so it is not white with red lights shining on it.’ (2004: 420). Cohen emphasizes that it is still intuitive that his son would find the argument unconvincing. He observes that since the boy no longer doubts ¬SH, we couldn’t explain this intuition by claiming that the argument is dialectically ineffective. According to Cohen, this example strongly suggests that the weakness of arguments like the one from R to ¬SH, authorized by (PC), is also epistemic, not just dialectical. Although Cohen apparently scores points against Pryor and Markie in this case, I believe that these two authors are ultimately correct when they claim that the arguments like the one from R to ¬SH, authorized by (PC), are epistemically impeccable. A source of confusion that permeates through this dispute is the fact that neither Pryor and Markie nor Cohen carefully distinguish between seeming-based justification and correlated reflective justification. The deductive arguments licensed by (PC) are meant to convey—specifically—seeming-based justification, rather than reflective justification. If phenomenal conservatives drew this distinction and held these two types of justification apart, they could untangle the apparently conflicting intuitions that a cursory analysis of these arguments tends to produce. In this way, they could rebut the easy justification from closure objection. Let me show how to do this. Suppose a subject S has a seeming that R and no defeater. Thanks to (PC), S has justification for believing R based on that seeming. The thesis that if S realizes that R entails ¬SH, S’s justification based on her seeming that R can transmit to ¬SH appears to be straightforward, despite Cohen’s opinion. As said, an asserted problem of this thesis is that transmitting justification in this way seems viciously circular. For, in this way, S’s seeming that R would provide S with justification for believing that the seeming that R itself is not deceptive, which appears impossible. Under closer scrutiny, however, this claim proves false. Recall that seeming-based justification is elusive because it “vanishes” as the subject becomes reflectively aware of the relevant seeming. Accordingly, when S has justification for R based on her seeming that R, S cannot be reflectively aware of her seeming. Now, suppose that S’s justification based on her seeming that R transmits to ¬SH. Since S is not aware of having the seeming that R, this justification cannot count for S as a reason for believing that

5.3 Answering the Easy Justification from Closure Objection

83

the seeming that R itself (of which she is not aware) is not deceptive.18 On the other hand, if S became reflectively aware of her seeming, S would no longer have that seeming-based justification. Thus, S would not have a reason for believing that her seeming that R is not deceptive based on that seeming only. In conclusion, this way to get justification for ¬SH doesn’t involve vicious circularity. S can acquire seemingbased justification for believing ¬SH without circularity through the entailment from R to ¬SH and, if S actually deduces ¬SH, S can acquire a seeming-based justified belief that ¬SH without circularity. Analogous considerations apply to any other similar deductive argument authorized by (PC). One might doubt that S could actually acquire seeming-based justification for believing ¬SH via the argument from R to ¬SH. For one might suppose that S’s mere thinking of ¬SH would instantly turn S’s attention to her seeming that R, which would wipe out her seeming-based justification for the premise. But this is false. Whether S’s thinking of ¬SH would make S’s reflectively aware of her seeming that R depends on the specific circumstances. Here is an example of a situation in which S’s thinking of ¬SH would probably not make S’s reflectively aware of her seeming that R. Note first that ¬SH is equivalent to the disjunction: (A) It is not the case that the table is white or (B) it is not the case that the table seems red because it is illuminated by a hidden red light.19

Note that R entails ¬SH just because R entails A. The content of B is irrelevant; the entailment would hold true even if B were a random proposition. Now imagine that S takes an introductory course to logic. S happens to justifiedly believe R on the basis of her seeming that R. To practise logical connectives, S is asked by the teacher to deduce from R various disjunctions each of which includes one random disjunct. To do this, S keeps the disjunct A fixed and forms the other disjunct of the disjunctions by picking out in sequence all members of a list of random propositions provided by the teacher. The list happens to include B. At a certain point S deduces ¬SH. Since S justifiedly believes R, S justifiedly believes ¬SH too. Nevertheless, given the circumstances (S is doing logic rather than, say, epistemology) and S’s peculiar way to put together ¬SH, it is plausible that S’s thinking of ¬SH and deducing it from R won’t lead S to introspect herself and become aware of her seeming that R. Another asserted difficulty of arguments like the one from R to ¬SH, licensed by (PC), is that they generate a puzzle. If these deductive arguments are epistemically impeccable, why do people use inductive evidence or make independent checks, rather than simply running the arguments, to dispel concerns about the reliability of their appearances? But this puzzle can easily be unravelled. Note that when one is concerned about the accuracy of an appearance that one entertains, one is reflectively aware of it. Take then the argument from R to ¬SH enabled by (PC). When S is concerned about the accuracy of her seeming that R, and so S is reflectively aware of it, the basis of her justification for R is, not her seeming that R, but her reflective belief note that ¬SH doesn’t entail the proposition that S has a seeming that R. So, S’s believing ¬SH wouldn’t involve S’s believing that she has that seeming. 19 Since ¬(X & Y ) is equivalent to ¬X or ¬Y. 18 Also

84

5 Antiscepticism and Easy Justification

that she has that seeming. This explains why, when S is concerned about the accuracy of her seeming that R, S cannot use the argument from R to ¬SH, licensed by (PC), to justifiedly conclude that ¬SH. There is more. S can have justification for R based on her reflective belief that she has a seeming that R only if S has already justification for ¬SH—namely, inductive justification from similar past cases or justification based on an independent check of ¬SH. This clarifies why, when S is concerned about the accuracy of her seeming that R because she suspects that SH might be true, S will look for inductive justification for ¬SH or will make an independent check of it. Analogous considerations apply to any deductive argument licensed by (PC) similar to the one from R to ¬SH. Thus, it is not puzzling, after all, that people resort to inductive evidence or independent checks when they worry about the accuracy of their appearances, rather than running these arguments. We are now in a position to provide an accurate interpretation of Cohen’s thought experiments. Recall that in the first though experiment, Cohen’s son doubts ¬SH, but he is not persuaded by his father’s argument to believe ¬SH. The father argues that since it appears that R, R is true, therefore, ¬SH is also true. In the second thought experiment, the boy doesn’t doubt ¬SH. He only wants to know how his father knows that ¬SH. His father runs the same argument as before, which the son finds again unconvincing. My diagnosis of why this argument isn’t persuasive in these scenarios will complete my response to the easy justification by closure objection. Let’s start with the first scenario. Since the boy doubts the accuracy of his seeming that R at the outset, he is reflectively aware of his seeming, and thus he justifiedly believes the proposition that he has a seeming that R. This belief constitutes the basis of his justification for R. When his father claims that since it appears that R, then R is true, the boy assesses this claim. Specifically, the boy evaluates whether the proposition that he himself has an appearance that R is good reason to believe R. To boy realizes that this proposition could actually be such a good reason only if he already had a good reason to believe ¬SH. The boy then assesses the father’s second claim that since R is true, then ¬SH is true. The boy realizes that he cannot come to have a good reason to believe ¬SH in this way. For he must have a good reason to believe ¬SH in the first place to carry out the reasoning. That’s why the boy finds his father’s argument unpersuasive. This thought experiment says nothing about the argument from R to ¬SH licensed by (PC). For the argument that the boy doesn’t find persuasive is not one licensed by (PC). Pryor and Markie’s analysis of this case is partly misguided. Pryor and Markie claim that since the boy doubts ¬SH, he cannot accept his own appearance that R as a justification for R. I agree with them on this.20 This is an additional reason why the boy may find his father’s argument unconvincing. Yet Pryor and Markie also think that the boy, in this scenario, does possess justification for believing R because he has a seeming that R and no defeating evidence. They also think that the boy’s justification for R transmits to ¬SH across the entailment. Both claims are problematic. Concerning the first, my view is that the boy can have justification 20 If the boy has independent justification for ¬SH, but doubts ¬SH, he cannot take the proposition that he has an appearance that R to be a justification for R.

5.3 Answering the Easy Justification from Closure Objection

85

for believing R if he has independent (and stronger) justification for believing ¬SH. However, the basis of the boy’s justification is, in this case, his reflective belief that he has the seeming, not the seeming itself. The second claim is false: the boy’s reflective justification for R doesn’t transmit to ¬SH. Let’s turn to Cohen’s second thought experiment. In this case, the boy has thought of the possibility that his seeming that R is deceptive—so he is reflectively aware of it—but he has ruled out this possibility because he thinks that his father knows that ¬SH. When the father explains to him how one can conclude that ¬SH by adducing the argument described before, the boy finds it unconvincing. The explanation is identical to the one detailed before. To conclude, phenomenal conservatism is not endangered by the problem of easy justification from closure. Under close scrutiny, the asserted difficulties dissolve.

5.4 Answering the Easy Justification from Bootstrapping Objection Supporters of (PC) and similar principles of immediate justification almost generally acknowledge that bootstrapping arguments, such as (1)–(6) described in Sect. 5.2, are epistemically defective.21 These authors typically respond to the bootstrapping objection by insisting that if a subject S endorses (PC) or a similar principle, S is not committed by this very fact to accept the illicit step of the bootstrapping reasoning, which is almost always identified with the non-deductive one. This step would in fact be biased or defeated, as it would conflict with key methodological principles. Solutions of this type are for instance developed in Cohen (2002, 2005), Vogel (2008) and Weisberg (2010).22 (See Weisberg 2012 for a survey of internalist and externalist responses to the bootstrapping problem.) Let’s consider an example. Take again the bootstrapping argument (1)–(6). Its non-deductive step is the one carried out by S from (5) My colour experiences have always matched the actual colour of each of the n cards that I have viewed to (6) My colour vision is reliable. Vogel (2008) contends that the inference from (5) to (6) gets defeated because the whole reasoning (1)–(6) is rule-circular. Rule-circular reasoning uses a rule R to establish that R itself is reliable. In particular, the defeater is the following proposition, which Vogel thinks to be a priori true:

21 Markie (2005) nevertheless denies it and contents that bootstrapping reasoning is only dialectically

ineffective. See Cohen (2005)’s response. interprets the bootstrapping problem as a general paradox. See note 12 above.

22 Weisberg

86

5 Antiscepticism and Easy Justification (NRC) A belief that an epistemic rule R is reliable cannot be justified by the application of R. That is, neither the conclusion itself nor any belief which supports the conclusion may be justified in virtue of the application of R. (531)23

Recall the initial steps of the bootstrapping argument: (1) (2) (3) (4)

The first card appears red to me. The first card is red. The first card is red and appears red to me. Therefore, my colour appearance matches the actual colour of the card on this occasion.

We can think of (PC) as sanctioning a rule Rc , which states that if it appears to one that X has colour C, and one has no reason to doubt it, then one ought to believe that X has colour C. (6) can be read as asserting that Rc is reliable. Furthermore, note that S uses Rc to justifiedly believe (2) in the thought experiment. Since S’s reasoning (1)–(6) comes in conflict with (NRC) at its non-deductive step from (6) to (7), this step is defeated and cannot give S justification for believing (7). In short, Vogel (2008)’s response to the bootstrapping objection is the following: since (NRC) is a defeater of the non-deductive inference from (6) to (7), and (PC) is compatible with (NRC), supporters of (PC) are as such uncommitted to the nondeductive inference. So (PC) is not hit by the easy justification from bootstrapping objection. As White (2006) has emphasized, however, not all rule-circular reasoning is bad. For instance, ‘doing well in a memory game can suggest that I have a good memory, even though I can’t help but use my memory to evaluate my performance’ (530). But (NRC) would disallow this kind of reasoning. A problem of Vogel’s view is that it doesn’t distinguish between good and bad rule-circular reasoning. Weisberg (2010) has also found out that there are (impermissible) bootstrapping cases that don’t instantiate rule-circularity, so they cannot be stopped by (NRC).24 A cursory examination of the principal responses to the bootstrapping objection discussed in the literature will show that they are all problematic to some degree. I cannot exclude that new proposals might turn out to be more straightforward.25 Nonetheless, my view is that phenomenal conservatism doesn’t need any sophisticated defence of this type. The reason being that since seeming-based justification is elusive, the bootstrapping objection cannot get off the ground when directed to perceptual dogmatism. Consider again the argument (1)–(6). Note that S can justifiedly believe that (1) The first card appears red to me only if S is reflectively aware of her seeming that the card is red. Accordingly, if S also justifiedly believes that 23 (NRC)

is reminiscent of what Bergmann (2000) calls the no self-support principle. Bergman interprets Fumerton (1995: 177) as endorsing it. 24 See Cohen (2010) for further objections. 25 An interesting one is Butzer (2017).

5.4 Answering the Easy Justification from Bootstrapping Objection

87

(2) The first card is red, this justification cannot be based on S’s seeming that the card is red. S’s justification for (2), if any, must be reflective justification—that is to say, justification based on S’s reflective belief that she has that seeming. Therefore, (PC) is off the hook at the outset.

5.5 Conclusions I have investigated the epistemological consequences of the fact that seeming-based justification is elusive, in the sense that a subject S can lose this justification simply by reflecting on her seemings and becoming aware of them. I have argued that because of this feature of seeming-based justification, S couldn’t successfully appeal to (PC) to reject sceptical arguments that questioned her possessing seeming-based justification. I have also suggested that S’s seemings could not provide S with immediate justification for ruling out sceptical conjectures. These findings indicate that phenomenal conservatism doesn’t give scepticism a fully satisfactory response. On the positive side of my investigation, I have shown that since seeming-based justification is elusive, phenomenal conservatism is not afflicted by easy justification problems. When seeming-based justification and reflective justification are held apart, it becomes apparent that neither the easy justification from closure objection nor the easy justification from bootstrapping objection can strike phenomenal conservatism.

References Avnur Y (2012) Closure reconsidered. Philosophers’ Impr 12. www.philosophersimprint.org/ 012009/ Bergmann M (2000) Externalism and skepticism. Philos Rev 109:159–194 Butzer T (2017) Bootstrapping and dogmatism. Philos Stud 174:2083–2103 Cohen S (2002) Basic knowledge and the problem of easy knowledge. Philos Phenomenol Res 65:309–329 Cohen S (2005) Why basic knowledge is easy knowledge. Philos Phenomenol Res 70:417–430 Cohen S (2010) Bootstrapping, defeasible reasoning, and a priori justification. Philos Perspect 24, Epistemology:141–159 Dretske F (1970) Epistemic operators. J Philos 67:1007–1023 Fumerton R (1995) Metaepistemology and skepticism. Rowman & Littlefield, Lanham, Maryland Huemer M (2005) Ethical intuitionism. Palgrave Macmillan, Houndmills & Basingstoke Huemer M (2013) Phenomenal conservatism Über Alles. In: Tucker C (ed) Seemings and justification: new essays on dogmatism and phenomenal conservatism. Oxford University Press, New York, pp 328–350 Huemer M (2016) Inferential appearances. In: Coppenger B, Bergmann M (eds) Intellectual assurance: essays on traditional epistemic internalism. Oxford University Press, Oxford, pp 144–160

88

5 Antiscepticism and Easy Justification

Koksvik O (2011) Intuition. PhD Thesis. Australian National University. https://philpapers.org/ archive/KOKI.pdf Lewis D (1996) Elusive knowledge. Australas. J Philos 74:549–567 Markie P (2005) Easy knowledge. Philos Phenomenol Res 70:406–416 Moretti L (2019) Inferential seemings and the problem of reflective awareness. Can J Philos 49: 253–271 Moretti L, Piazza T (2018) Transmission of justification and warrant. In Zalta EN (ed) Stanford encyclopedia of philosophy (Winter Edition). https://plato.stanford.edu/archives/win2018/ entries/transmission-justification-warrant/ Pryor J (2004) What’s wrong with Moore’s argument? Philos Issue 14, Epistemology:349–378 Siegel S, Silins N (2015) The epistemology of perception. In: Matthen M (ed) Oxford handbook of philosophy of perception. Oxford University Press, Oxford, pp 781–810 Sosa E (1991) Knowledge and intellectual virtue. In: Sosa E (ed) Knowledge in perspective: selected essays in epistemology. Cambridge University Press, Cambridge, pp 225–244 Tucker C (2010) Why open-minded people should endorse dogmatism. Philos Perspect 24:529–545 Vogel J (2000) Reliabilism levelled. J Philos 97:602–623 Vogel J (2008) Epistemic Bootstrapping. J Philos 105:218–239 Weisberg J (2010) Bootstrapping in general. Philos Phenomenol Res 81:525–548 Weisberg J (2012) The bootstrapping problem. Philos Compass 7:597–610 White R (2006) Problems for dogmatism. Philos Stud 131:525–557 Wright C (2002) (Anti)-Sceptics simple and subtle: Moore and McDowell. Philos Phenomenol Res 65:330–348 Wright C (2007) The perils of dogmatism. In: Nuccetelli S, Seay G (eds) Themes from G. E. Moore: new essays in epistemology and ethics. Oxford University Press, Oxford, pp 25–48

Chapter 6

Concluding Remarks

Abstract In this chapter I draw the conclusions of my investigation into phenomenal conservatism. I argue that phenomenal conservatism isn’t actually plagued by serious problems attributed to it by its opponents, but that it neither possesses all the epistemic merits that its advocates think it has. I suggest that phenomenal conservatism could provide a more satisfactory account of everyday epistemic practices and a more robust response to the sceptic if it were integrated with a theory of inferential justification. I also identify questions and issues relevant to the assessment of phenomenal conservatism to be investigated in further research. Keywords Phenomenal conservatism · Foundationalism · Scepticism · Inferential and non-inferential justification · Reflective justification · Global ambitions of phenomenal conservatism Phenomenal conservatism is the internalist theory of immediate, non-inferential justification based on the principle (PC), stating that if a subject S has a seeming that P, in the absence of defeaters, S thereby has justification for believing P. Throughout this work I have assumed the soundness of intuitions and arguments that attest that epistemic justification is internalist. Once epistemic internalism is adopted, (PC) appears plausible: if S has a seeming that P and no reason to distrust it, the only epistemically justified attitude for S to take towards P is—intuitively—belief. Phenomenal conservatism has been celebrated for having an array of epistemological merits. The following are possibly the most significant. Phenomenal conservatism supplies ordinary cognitive practices with a rational explanation. Indeed, in everyday life we seem to take ourselves to have justification for entertaining many beliefs just because of how things appear to us to be, in accordance with (PC). Moreover, phenomenal conservatism is claimed to form the basis of fallible foundationalism. For it looks plausible that it is our appearances that put an end to the regress of inferentially-justified beliefs, when we search for an ultimate justification of our beliefs. Finally, phenomenal conservatism is said to help us overcome scepticism. For if (PC) is true, we can get many beliefs justified on the basis of our appearances even if we lack independent reasons to rule out any sceptical conjecture. Since phenomenal conservatism looks plausible and promising of a variety of epistemological

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2020 L. Moretti, Seemings and Epistemic Justification, SpringerBriefs in Philosophy, https://doi.org/10.1007/978-3-030-43392-5_6

89

90

6 Concluding Remarks

benefits, it is important to investigate whether it actually keeps its promises and withstands apparently forceful objections raised against it. In this book I have carried out such an investigation to some extent. In Chap. 3 I have scrutinized a number of views opposing phenomenal conservatism which entail that cognitively penetrated appearances would in many cases lack the power to justify their propositional contents even prima facie. I have rejected all these views as implausible, incoherent or counterintuitive. I have suggested that cognitively penetrated appearances wouldn’t lose their inherent justifying power, and that the intuitions that seem to indicate otherwise confound the property of being a belief produced by a faculty that doesn’t function properly with the property of being an unjustified belief. In Chap. 4 I have refuted an apparently forceful case made to discredit (PC), according to which (PC) is incompatible with Bayesian methodology. I have shown that the conclusion that (PC) conflicts with Bayesianism follows from a specious formalization of the intuitive notion of seeming-based justification. In Chaps. 4 and 5 I have found seeming-based justification to be elusive: when S becomes reflectively aware of a seeming that P, S’s justification based on S’s reflective belief that she has that seeming overrides S’s seeming-based justification. In Chap. 5, I have shown that since seeming-based justification is elusive in this way, (PC) is unaffected by easy justification problems. Although the above difficulties raised against phenomenal conservatism have turned out to be insubstantial, in Chap. 5 I have shown that the fact that seemingbased justification is elusive limits the antisceptical efficacy of (PC). When S engages with an argument questioning her own having immediate justification, phenomenal conservatism doesn’t give S the means to rebut that argument. The contribution of phenomenal conservatism to the fight against scepticism is thus more modest than expected. Alternative responses to the sceptic—for instance, abductive and hingeproposition replies1 —don’t seem to be subject to limitations of this sort. When S engages with an argument questioning her own having justification of the relevant type, these responses—if working at all—do enable S to rebut the argument. It is worth noticing, nevertheless, that the elusiveness of seeming−based justification curbs but doesn’t obliterate the antisceptical bite of phenomenal conservatism. Thanks to (PC), individuals who don’t reflect on their seemings can still gain seeming−based justification for their beliefs even if they have no independent justification for ruling out sceptical conjectures. These subjects are presumably the large majority. We can draw similar conclusions about the asserted capacity of phenomenal conservatism to provide ordinary epistemic practices with a rationale: phenomenal conservatism can account for quotidian justification of unreflective individuals, but it cannot account for the justification of subjects who reflect on their appearances (think for instance of microscopists, birdwatchers, pilots, witnesses in courts of law, and epistemologists). These considerations suggest that phenomenal conservatism would provide a more forceful response to the sceptic and a more comprehensive account

1 For

critical surveys of recent responses see for instance Greco (2007), Klein (2019) and Pritchard (2019).

6 Concluding Remarks

91

of ordinary cognitive practices—inclusive of those proper to reflective individuals— if it were integrated with a theory of reflective—and so inferential—justification. What theory would be suitable for this role and how phenomenal conservatism could actually be integrated with it are questions to be thoroughly investigated. Further work on phenomenal conservatism should address other issues that I have not explored in this short monograph. To begin with, my response to the Bayesian objection calls for further investigation: an important question is whether there is an adequate formal model of the notion of seeming-based justification that does justice to (PC).2 The existence of such a formalisation would bring further support to phenomenal conservatism. Let’s turn to a last issue. The fact that seeming-based justification is elusive doesn’t prevent appearances from being sources of non-inferential justification. Thus, this fact doesn’t undermine the role of (PC) as a key principle of fallible foundationalism. Indeed, (PC) has been claimed to be the sole key principle of fallible foundationalism; that is to say, one capable of accounting for all kinds of non-inferential justification—inclusive of, for instance, perceptual, introspective, rational and mnemonic justification (cf. Huemer 1999, 2001). These global ambitions of (PC), however, have been called into question (see Smithies 2019). The present work has mainly focused on perceptual justification. Future work on phenomenal conservatism should carry out a systematic investigation about which forms of non-inferential justification can actually be accounted by (PC).

References Greco J (2007) External world scepticism. Philos. Compass 2:625–649 Huemer M (1999) The problem of memory knowledge. Pac Philos Q 80:346–357 Huemer M (2001) Skepticism and the veil of perception. Rowman and Littlefield, Lanham, MD Klein P (2019) Skepticism. Stanford Encyclopedia of Philosophy. In: Zalta E (ed). https://plato. stanford.edu/archives/sum2015/entries/skepticism/ Miller B (2016) How to be a bayesian dogmatist. Australas J Philos 94:766–780 Pritchard D (2019) Contemporary skepticism. Internet Encycl Philos. https://www.iep.utm.edu/ skepcont/. Accessed 10 November 2019 Smithies D (2019) On the global ambitions of phenomenal conservatism. Anal Philos 60:206–244

2 An

interesting attempt is in Miller (2016).