Purpose and Procedure in Philosophy of Perception 2020950740, 9780198853534

Contemporary philosophy of perception is dominated by highly polarized debates. The polarization is particularly acute i

136 54 2MB

English Pages [312] Year 2021

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Purpose and Procedure in Philosophy of Perception
 2020950740, 9780198853534

Table of contents :
Cover
Purpose and Procedure in Philosophy of Perception
Copyright
Contents
Acknowledgements
List of Contributors
1: Introduction
1. Background
2. Purpose and Procedure
2.1 Destinations
2.2 Origins
2.3 On the Way: Fixed Points
2.4 On the Way: Methods
3. Philosophy and the Science of Perception
3.1 Destinations
3.2 Origins
3.3 On the Way: Fixed Points
3.4 On the Way: Methods
4. Ways Forward
5. Summary of Papers
References
2: Perceptual Paradigms
1. Introduction
2. Burge and McDowell on Disjunctivism
3. Burge v McDowell: A Clash of Paradigms?
4. Research Programmes
5. Methodological Lessons for the Philosophy of Perception
References
3: Bridging the Gap? Naïve Realism and the Problem of Consciousness
1. Introduction
2. What Is to Be Explained? Naïve Realism and the Problem of Consciousness
3. Fixed Points? Naïve Realism and Physicalism
4. Methodology? Metaphysical, Descriptive, and Transcendental
4.1 Metaphysical Naïve Realism
4.2 Descriptive Naïve Realism
4.3 ‘Default’ Metaphysical Naïve Realism
4.4 Transcendental Naïve Realism
References
4: Experiential Pluralism and Mental Kinds
1. Introduction
2. Visual Experience in the Good Case
3. The Argument from Abilities
3.1 Situation-Dependent Abilities (P1)
3.2 Explaining Possession of Situation-Dependent Abilities (P2)
3.3 The Explanatory Demand for World-Involvingness (P3)
4. World-Involving Experiences as a Mental Kind
References
5: The Tractability of the Debate on Relationalism
1. The Debate on Relationalism
2. Philosophical Disagreement and Incommensurable Paradigms
3. Perception, Hallucinations, and Bullets to Bite
4. Phenomenal Character and Introspection
5. The Superficiality Constraint
6. Advancing the Debate with the Argument from Superficiality
References
6: Neopragmatism and Philosophy of Perception
1. Introduction
2. Neopragmatism
3. A Case Study in Demystification: Naïve Realism
3.1 Naïve Realism and Disjunctivism
3.2 Demystification
4. A Neopragmatist Theory of Perception
5. Conclusion
References
7: Perceptual Experience and Physicalism
1. Introduction
2. Representationalism and Physicalism
3. Non-relationalism
4. Perceptual Experiences and Accuracy
5. Conclusion
References
8: High-Level Perception and Multimodal Perception
1. Introduction
2. Beyond Vision and the Unimodal Paradigm
3. Multimodal Perception and the Hunt for High-Level Content
3.1 The McGurk Effect
3.2 The Ventriloquist Effect
4. The High-Level Contents of the Rubber Hand Illusion
4.1 RHI on Liberalism and Conservatism
4.2 RHI and the Liberal’s Conjecture
4.3 Empirical Support for the Liberal’s Conjecture
5. Feeling What’s Not There
6. Conclusion
References
9: What Can Predictive Processing Tell Us about the Content of Perceptual Experience?
1. Introduction
2. The Content of Perceptual Experience (CPE) Debate
3. The Predictive Processing Framework
4. Counterfactual Predictions and ‘Seeing As’
5. Does the PPF Dissolve the CPE Debate?
6. The PPF as Supporting Liberalism
7. Four Illustrative Objections
7.1 Phenomenological Effects Are Attributable to Top-down Effects on the Experience of Low-Level Properties
7.2 Phenomenological Effects Are Attributable to Non-perceptual Phenomenology
7.3 The PPF Is Entirely Consistent with Conservatism
7.4 What Modality Are These So-Called Perceptual Experiences Happening In?
8. Conclusion
Funding
References
10: Wading in the Shallows
1. Introduction
2. Naturalism and Avoiding the Excess of Zettelism
3. Prediction Error Theory and Naïve Realism
3.1 Prediction Error Theory
3.2 Naïve Realism and the Asymmetric Discreditation Argument
3.3 Hallucination and Response-Dependent Properties
3.4 Perception–Cognition Divide
4. Sensory Kinds as Natural Kinds
5. Concluding Remarks
References
11: Naturalism and the Metaphysics of Perception
1. Introduction
2. Two Approaches to Theorizing about Perception
2.1 Metaphysical Theories of Perception
2.2 Psychological Theories of Perception
2.3 Comparing the Metaphysical and Psychological Debates
3. Distinguishing the Metaphysical and Psychological Debates
3.1 Is There a Difference in the Modal Strength of the Theories?
3.2 Do the Theories Result from Different Methodologies?
3.3 Is There a Difference in Whether the Explanations Are Causal or Constitutive?
3.4 Are There Different Kinds of Constitutive Explanation?
4. Conclusion
References
12: Phenomenology as Radical Reflection
1. Introduction
2. Humean, Kantian, and Husserlian Phenomenology of Perception
3. Problems with Authority
4. From Husserlian Phenomenology to Radical Reflection
5. Merleau-Pontian Phenomenology of Perception
13: Merleau-Ponty: Perception and Methodology
1. Introduction
2. Merleau-Ponty’s Phenomenology
2.1 Experience
2.2 The Transcendental-Phenomenological Reduction
2.3 Description
2.4 Analysis
3. Application of Merleau-Ponty’s Method to Perception
4. Concluding Remarks
References
14: Sensation and the Grammar of Life: Anscombe’s Procedure and Her Purpose
1. Introduction
2. The Raging Nerve Extracted
3. What Is an Object of Sight?
4. Anscombe’s Procedure
5. Two Ontological Errors in Brief
6. Anscombe’s Purpose
References
Index

Citation preview

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 16/3/2021, SPi

Purpose and Procedure in Philosophy of Perception

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 16/3/2021, SPi

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 16/3/2021, SPi

Purpose and Procedure in Philosophy of Perception Edited by

HEATHER LOGUE AND LOUISE RICHARDSON

1

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 16/3/2021, SPi

3

Great Clarendon Street, Oxford, OX2 6DP, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © OUP 2021 The moral rights of the authors have been asserted First Edition published in 2021 Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2020950740 ISBN 978–0–19–885353–4 DOI: 10.1093/oso/9780198853534.001.0001 Printed and bound by CPI Group (UK) Ltd, Croydon, CR0 4YY Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 16/3/2021, SPi

Contents Acknowledgements List of Contributors

1. Introduction Heather Logue and Louise Richardson 2. Perceptual Paradigms William Fish 3. Bridging the Gap? Naïve Realism and the Problem of Consciousness Keith Allen

vii ix

1 23

43

4. Experiential Pluralism and Mental Kinds Maja Spener

63

5. The Tractability of the Debate on Relationalism Roberta Locatelli

85

6. Neopragmatism and Philosophy of Perception Joshua Gert

107

7. Perceptual Experience and Physicalism Laura Gow

127

8. High-Level Perception and Multimodal Perception Dan Cavedon-Taylor

147

9. What Can Predictive Processing Tell Us about the Content of Perceptual Experience? Sam Wilkinson

174

10. Wading in the Shallows Paul Noordhof

191

11. Naturalism and the Metaphysics of Perception Zoe Drayson

215

12. Phenomenology as Radical Reflection Dave Ward

234

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 16/3/2021, SPi

vi



13. Merleau-Ponty: Perception and Methodology Komarine Romdenh-Romluc

258

14. Sensation and the Grammar of Life: Anscombe’s Procedure and Her Purpose Clare Mac Cumhaill and Rachael Wiseman

276

Index

295

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 16/3/2021, SPi

Acknowledgements This publication was made possible through the support of a grant from the John Templeton Foundation (via the Cambridge New Directions in the Study of the Mind project). The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the John Templeton Foundation. We also gratefully acknowledge the peer reviewing efforts of the other members of the Sense Perception in the North research network—Keith Allen, Clare Mac Cumhaill, and Paul Noordhof.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 16/3/2021, SPi

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 16/3/2021, SPi

List of Contributors Keith Allen is Senior Lecturer, University of York Dan Cavedon-Taylor is Lecturer, The Open University Zoe Drayson is Assistant Professor, University of California, Davis William Fish is Professor, Massey University Joshua Gert is Leslie and Naomi Legum Distinguished Professor, William & Mary Laura Gow is Lecturer, University of Liverpool Roberta Locatelli is Alexander von Humboldt Postdoctoral Associate Researcher, University of Tübingen Heather Logue is Associate Professor, University of Leeds Clare Mac Cumhaill is Associate Professor, Durham University Paul Noordhof is Anniversary Professor of Philosophy, University of York Louise Richardson is Senior Lecturer, University of York Komarine Romdenh-Romluc is Senior Lecturer, University of Sheffield Maja Spener is Lecturer, University of Birmingham Dave Ward is Senior Lecturer, University of Edinburgh Sam Wilkinson is Lecturer, University of Exeter Rachael Wiseman is Senior Lecturer, University of Liverpool

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 16/3/2021, SPi

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

1 Introduction Heather Logue and Louise Richardson

1. Background Consensus is hardly the norm in any area of philosophical interest, but contemporary philosophy of perception is dominated by debates in which consensus has proven particularly elusive, and discussion more heated than one might expect (given, for instance, the lack of obvious political or policy-related consequences). For example, Tyler Burge goes so far as to accuse John McDowell of holding a view that ‘ . . . exhibits ignorance of the most elementary aims, claims, and methodology of the science of perceptual psychology’ (2011, 66). One might reasonably worry that some debates in this subfield have deteriorated into irresolvable impasses. One central case is the debate over the metaphysical structure of perceptual experience. As we will see, there is no clear consensus on what exactly this debate is about, but for now, let us identify it as the many-sided debate between sensedatum theorists, adverbialists, representationalists, and naïve realists. For the sake of space, we will just focus on the two most popular sides in the contemporary debate: representationalism (sometimes called ‘intentionalism’, e.g., Tye 2000) and naïve realism (sometimes called ‘relationalism’, e.g., Campbell 2002).¹ According to naïve realism, veridical perceptual experience—an experience in which a subject perceives things in her environment as they are—simply consists in the subject perceiving things in her environment as they are. More precisely, naïve realists hold that such experiences ‘are essentially relational: [they] “extend out” into the world, and consist in the obtaining of a conscious relation of awareness or acquaintance between perceiving subjects and mind-independent objects and properties in their environment’ (Allen, this volume, 43). For example, the metaphysical structure of a veridical experience in which a subject sees a banana and its yellowness is just the subject seeing the banana and its yellowness. Since the subject of a total hallucination doesn’t perceive anything in her environment, naïve realism leads naturally if not inexorably to disjunctivism about perceptual experience: the view that there is a difference between at least ¹ Although adverbialism may be making a comeback—see Gert’s and Gow’s contributions to this volume.

Heather Logue and Louise Richardson, Introduction In: Purpose and Procedure in Philosophy of Perception. Edited by: Heather Logue and Louise Richardson, Oxford University Press (2021). © OUP. DOI: 10.1093/oso/9780198853534.003.0001

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

2

    

hallucination and veridical perception that constitutes a significant difference in mental kind.² By contrast, representationalism offers a unified account of veridical and hallucinatory experiences, in terms of the subject’s representing her environment as being a certain way (e.g., as containing a yellow, crescent-shaped object). The debate between the representationalist and the naïve realist has largely been focused on two questions. First, does perceptual experience have representational content? That is, does it involve the subject representing the world as being a certain way? An affirmative answer is entailed by representationalism, but most naïve realists answer in the negative. Second, is disjunctivism true? Most naïve realists think that their view entails it, and so defend it to the hilt, while pretty much everyone else finds it utterly implausible.³ Most of the contributions to this volume focus on the debate over the metaphysical structure of perceptual experience. This is plausibly because this is where consensus seems furthest away. However, it is not the only debate in philosophy of perception that carries a whiff of irresolvable impasse. Another that arguably falls in this category concerns the ‘admissible contents’ of perception (see, e.g., the 2009 special issue of Philosophical Quarterly and the 2013 special issue of Philosophical Studies devoted to this topic).⁴ The core issue concerns what kinds of properties we can perceive: is it just what are typically called ‘low-level’ properties (properties like colours, shapes, sizes, locations, distances, pitches, volumes, textures, smells, tastes, etc.), or do we also perceive so- called ‘high-level’ properties (e.g., natural and artifactual kind properties, semantic properties, causal properties, mental states, and evaluative properties)? Furthermore, there has been exceptionally vigorous debate amongst philosophers of perception about how to individuate the senses (see, e.g., Macpherson 2011). Some hold that the traditional taxonomy of sight, hearing, touch, smell, and taste is at least broadly correct, while others insist that empirical considerations give us reason to radically revise it. (This is not an exhaustive list of difficult debates in the philosophy of perception; we’ve just highlighted those that figure prominently in the contributions to this volume.) It is our contention that the threat of irresolvable impasse is largely down to two things. First, as we will illustrate below, there is much disagreement about how we should set about answering the questions that dominate the field (the procedure of

² Not every difference between experiences constitutes a significant difference in mental kind (see Spener, this volume, section 4). Also, note that there is a difference of opinion among naïve realists about whether illusions—roughly, perceptual experiences in which a subject perceives something in their environment but misperceives at least one of its properties—should be classified with veridical experiences or with hallucinations (see Byrne and Logue 2008, 69). ³ This is most likely because the principles discussed in section 2.3 are typically presupposed in most people’s introductions to the philosophy of perception (e.g., Descartes’ Meditations (1641/1993)). For non-disjunctivist views that are in the vicinity of naïve realism, see Johnston (2004) and Sethi (2020). ⁴ Many participants in this debate presuppose that perception has representational content—hence the framing of the issue in terms of which contents are admissible—but (at least prima facie) the core issue is independent of this presupposition.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi



3

philosophical theorizing about perception), and which of these questions are worth pursuing (the purpose of philosophical theorizing about perception). Second, much of this disagreement is unspoken: there hasn’t been enough explicit discussion of these methodological issues in philosophy of perception. We don’t want to overstate the case—there are certainly instances of methodological reflection in the literature. For example, Susanna Siegel articulates and argues for a method of phenomenal contrast (Siegel 2011, ch. 3.3), and her interlocutors have subjected this method to critical scrutiny (see Helton 2016 for an overview). But exchanges of this sort seem to be the exception rather than the rule. Even when a contributor to a debate in philosophy of perception is explicit about their purpose or their procedure, their opponents more often engage with first-order issues rather than meta-level, methodological ones. The goal of this volume is to move such methodological questions from the background to the fore, in the hope of facilitating progress where debates threaten to stagnate into impasse. We make no claim to comprehensive coverage in this volume; it does not represent all points of view. Rather, our intention is just to kick-start a trend of more explicit, systematic discussion of methodology in philosophy of perception. Note that this need not involve calling ‘time out’ on first-order debates to focus exclusively on methodological issues. It will also be fruitful to engage in first-order debates in a way that is explicitly attentive to such issues—most of the contributions to this volume are of this sort. Creating space for such methodological reflection is just a first step in making progress: not necessarily in getting all parties to agree, of course, but in identifying where progress is—and is not—to be made (e.g., new lines of enquiry, or places where ‘talking past’ has inadvertently occurred), and where—and why—differences are potentially irreconcilable, perhaps because of broader differences in approach.

2. Purpose and Procedure In order to get a sense of what the choice points are, we will frame philosophical theorizing about perception in terms of a journey: an origin, a destination, and the route we take from the former to the latter. The origin concerns the relative status of competing theories at the outset (e.g., does a particular theory have default status?). The destination has to do with the question of what the various philosophical theories of perception are trying to accomplish. The route from the origin to the destination involves at least two questions. First, we should consider which ‘routes’ are permissible: what principles should be held fixed as we theorize? Second, we should consider appropriate ‘mode(s) of transportation’: which method(s) should we employ in theorizing? The aim of this section is not to take a stand on any of these issues—rather, it is to begin cataloguing them, to help readers to think through what their own stands are in an explicit, systematic way.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

4

    

It will be helpful to consider the journey in reverse, since our conception of where we want to end up will inform our conception of how best to get there.

2.1 Destinations What is the target of philosophy of perception, exactly? The answer, of course, is perception. But that answer leaves much unsettled. After all, ‘perception’ is polysemous: it can refer, amongst other things, to the conscious experience of perceiving, or it can refer to something else, such as that which achieves the task that we take perceptual processing to accomplish. For example, Schellenberg writes that perception ‘converts informational input . . . into representations of invariant features in the environment’ (2018, 1). A philosopher of perception who adopts the latter as their target is likely to include unconscious perception within it, unlike one who is targeting the conscious experience of perceiving.⁵ One important issue is whether it is legitimate to restrict the target in this way; given that unconscious perception is a genuine phenomenon, this could affect the plausibility of what some philosophers want to say about conscious perception (Berger and Nanay 2016). In any event, it’s typically assumed that the philosopher of perception’s target at least includes conscious experiences of perceiving.⁶ So for the purposes of the rest of this introduction, we’ll focus on conscious perceptual experience (omitting the ‘conscious’ for ease of exposition from here on out).⁷ What is it about perceptual experience that philosophers of perception are trying to explain? One answer that figures prominently in the literature is perceptual phenomenal character—‘what it is like’ to have a perceptual experience (the distinctive character of seeing a yellow thing, smelling a banana, etc.). But this answer raises the further question of what it is about perceptual phenomenal character we’re trying to explain. One aim is to specify (certain kinds of) facts in virtue of which perceptual experience has phenomenal character.⁸ The various theories of the metaphysical structure of perceptual experience offer different accounts: for example, a naïve realist holds that phenomenal character consists in perceiving mind-independent entities, whereas a representationalist holds that it consists in representing one’s environment as being a certain way. Another aim ⁵ Although see Phillips forthcoming for reasons to doubt that unconscious perception is a genuine phenomenon. ⁶ Pace sceptics about consciousness (see, e.g., Frankish 2016). ⁷ A focus on perceptual experience raises the question of what kind of thing it is. That is, are perceptual experiences states, episodes, dynamic processes involving action, or something else? An important issue that falls through the cracks of this volume concerns the relationship between the ontological category perceptual experience falls into and the questions we’re asking about it (see, e.g., Soteriou 2013). ⁸ Arguably, not all the facts, since some of them won’t be accessible by typical philosophical theorizing—the kinds of facts that the metaphysician of perception seems to have in view are personal-level psychological facts (Logue 2013a).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi



5

is to explain phenomenal contrasts between experiences—e.g., the contrast in what it is like for a tree-spotting expert and a novice to see a pine tree. Susanna Siegel (2011) has argued that we can explain such contrasts only if we can perceptually experience ‘high-level’ properties like being a pine tree. (In this volume (154), Cavedon-Taylor expresses scepticism that phenomenal contrasts are fruitful explananda by which to measure competing philosophical theories of perceptual experience, on the grounds that it is difficult to rule out alternative explanations of the contrast compatible with rival theories.) More ambitiously, some hold that a theory of the metaphysical structure of perceptual experience should go some distance towards addressing the perceptual component of the ‘hard problem’ of consciousness (Chalmers 1995): that is, towards explaining why perceiving (e.g.) a yellow banana has the specific phenomenal character it does, and why it comes with any phenomenal character at all (see Fish 2009, 75–9 and Allen, this volume, 44; see Gert, this volume, 121 for a case against this proposed explanandum). Another common answer to the question of what a theory of perceptual experience is supposed to explain is its epistemic role. As with phenomenal character, this explanatory project comes in more and less ambitious versions. The more ambitious projects seek to explain how perceptual experience affords knowledge; for example, knowledge of one’s environment (McDowell 1986, 2008; Schellenberg 2018), or knowledge of the reference of demonstrative terms (Campbell 2002). By contrast, a less ambitious project is to explain the facts in virtue of which perceptual experience affords (mis)information about one’s environment and generates beliefs about it (Logue 2014). While phenomenal character and epistemic role are the explananda that loom the largest—a fact reflected by Fish’s metaphor of phenomenological and epistemological ‘hats’ that a theory of perceptual experience is supposed to fit (2010, 2)—there are other things we might want a philosophical theory of perceptual experience to explain. For instance, there are explananda related to the role perceptual experience plays in action. In this volume, Maja Spener argues that a theory of perceptual experience should explain how it grounds situationdependent abilities (e.g., the ability to pick up a cat one sees). Furthermore, the mere occurrence of a certain kind of perceptual experience could be an explanandum in its own right. Dan Cavedon-Taylor (this volume) argues that the occurrence of the Rubber Hand Illusion is explicable only if we can visually experience the ‘high-level’ property of being a hand. This is not an exhaustive list of the potential explananda; the aim here is just to give the reader a sense of the possibilities. And of course, these explananda aren’t mutually exclusive; one might think that a philosophical theory of perceptual experience is in the business of explaining all of them. However, one might think that some of these alleged explananda don’t actually help us decide between rival theories. This might be because one holds that the

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

6

    

alleged explanandum doesn’t exist—for example, a sceptic about phenomenal character is not going to set a philosophical theory of perception the task of explaining it. Alternatively, one might think that a given explanandum exists, but that it’s not the job of a philosophical theory of perceptual experience to explain it. For instance, one might insist that the epistemic role of perceptual experience is not explained by its metaphysical structure, but rather by epistemological theories (e.g., reliabilism) that are in principle compatible with any metaphysics of perceptual experience. Or one might agree that a given explanandum exists, and that it is the job of a philosophical theory of perceptual experience to explain it, but hold that each of the theories on the table do an equally good explanatory job (e.g., see Logue’s (2013b) suggestion that we might be able to explain everything that needs explaining regardless of whether we experience ‘high-level’ properties). Furthermore, the Anscombian view articulated in Mac Cumhaill and Wiseman’s contribution to this volume is one on which explaining the aforementioned explananda isn’t in the philosopher of perception’s job description. According to their interpretation of Anscombe, the job is to describe the linguistic mastery of perceptual verbs, such as ‘to see’. For example: Armed with the term ‘object of sight’ the philosopher of perception can now proceed to describe this linguistic competence by asking what patterns the verb ‘to see’ imposes on its (grammatical) objects. One line of investigation will ask which putative ‘x’ in sentences of the form ‘Mary sees x’ or ‘I see x’ are intelligible (what are the possible objects of sight). Another will investigate whether, where some ‘a’ is an object of sight, linguistic competence commits us to accepting other objects, ‘b’, ‘c’, ‘d’. For example, whether ‘I see a rabbit’ commits a speaker to, e.g. ‘I see a mammal’, ‘I see a grey fluffy shape’, or ‘I see a living thing’ (what are the relations between objects of sight). (this volume, 281)

This conception of philosophy of perception, which Mac Cumhaill and Wiseman call the grammatical approach, contrasts with what they call the ontological approach, on which ‘a theory of perception should tell us the fundamental nature of perceptual experience’ (this volume, 277). Given that the point of discerning the fundamental nature of perceptual experience is to provide metaphysical explanations of the phenomena mentioned above, the ontological approach arises out of the idea that a theory of perceptual experience should provide such explanations. The grammatical approach offers an alternative conception of what philosophers of perception should be getting up to. Joshua Gert’s contribution also involves a focus on our use of perceptual language, but his conception of philosophy of perception’s purpose is importantly different. Like Mac Cumhaill and Wiseman’s Anscombe, Gert’s neo pragmatist eschews ‘positive explanations of the fundamental metaphysical nature of the basic entities or relations’ in this domain, and holds that ‘the main type of

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi



7

question . . . the philosopher [of perception] ought to be answering is “How do I explain the talk that gives rise to [the puzzle of perception]?” ’ (this volume, 108). But a key point of contrast between the two approaches is that Gert’s neo pragmatist offers a metaphysical theory alongside his response to this question. That is, although Gert holds that it is not the point of philosophy of perception to offer metaphysical explanations of the explananda listed above, he still thinks that there is a theory of the nature of perceptual experience (viz., adverbialism) that fits better than others with the demystification of perceptual talk offered by the neo pragmatist. Another alternative purpose is suggested by Romdenh-Romluc’s characterization of Merleau-Ponty’s method. On this picture, the destination is not an explanation of any of the phenomena mentioned above, but rather the revelation of the essence of perceptual experience. For Merleau-Ponty, ‘the essence of some phenomenon is a meaning that unifies disparate experiences of that phenomenon. It is something over and above those experiences insofar as it cannot simply be reduced to a collection of such experiences. Yet it does not exist in the absence of those experiences.’ (Romdenh-Romluc, this volume, 265). This project is akin to familiar endeavours within analytic philosophy. Just as an epistemologist might aim to identify what unifies instances of knowledge, what all and only these instances have in common (e.g., justified true belief plus . . . ), a philosopher of perception might aim to identify what unifies instances of perceptual experience (see, e.g., Martin 2004, 48–52). In summary, we began this subsection by suggesting that the purpose of philosophy of perception might be to explain its phenomenal character, its epistemic role, its role in facilitating action, or the occurrence of certain kinds of experiences. Several questions arise at this point. Is it the job of philosophy of perception to explain all of these phenomena? Are any of the explananda more important than the others? Or should we instead, like Mac Cumhaill and Wiseman’s Anscombe, focus on describing our linguistic competence with perceptual verbs, or like Gert’s neo pragmatist, focus on demystification of perceptual talk in order to dissolve philosophical puzzles, or like Romdenh-Romluc’s Merleau-Ponty, focus on discerning the essence of perceptual experience? Or should we be doing all of these things?

2.2 Origins Let us now turn our attention to the starting point of the journey, which concerns the relative status of the available theories at the beginning of inquiry. On one view, there is no privileged starting point for theorizing about perception. Any comer is just as much a contender as any of the others at the outset. But this view is not compulsory. For example, on Romdenh-Romluc’s characterization of

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

8

    

Merleau-Ponty’s method (this volume), the starting point of theorizing about perceptual experience should be halfway between an ‘inner’ and ‘outer’ conception of it. Some theories of perception characterize it as ‘an inner content of consciousness that can remain just as it is, regardless of the nature or existence of the world it seemingly presents to its subject’, but Merleau-Ponty thinks this is in tension with the way experience strikes us as being, namely, as ‘an immediate openness to the world’ (Romdenh-Romluc, 260). On the other hand, we’re not justified in assuming that ‘experience presents that world as it is, in-itself ’ (RomdenhRomluc, 260). So a sensible, middle-ground starting point is the idea that ‘when we describe perceptual experience, what we are describing are the worldly objects of perception as they are experienced’ (Romdenh-Romluc, 260)—that is, both the mind-independent objects of perceptual experience and the subjective responses to them figure in its metaphysical structure. This starting point leaves open the possibility that we could argue our way to either extreme (‘inner’ or ‘outer’), but the point is that we have to earn the right to occupy one of these positions. Similarly, M.G.F. Martin has argued that disjunctivism about perceptual experience has a default status; the disjunctivist’s opponent could in principle argue her way out of the default view, but the burden is on her to make the case (2004, 47–52). In this volume, Keith Allen argues that naïve realism has a privileged status that is much stronger than just being the default. On his view, it’s not that naïve realism has an initial advantage that could be in principle outweighed. Rather, just as Strawson argued that determinism is theoretically but not practically intelligible, Allen holds that the same is true of alternatives to naïve realism. Very roughly, the idea is that just as we cannot bring ourselves to genuinely believe that determinism is true, and thus that interpersonal interactions are not as they seem (e.g., warranting reactive attitudes such as gratitude and resentment), we also cannot bring ourselves to genuinely believe that naïve realism is false, and thus that ‘perceptual experience might be other than it appears’ (59)—namely, a conscious relation of acquaintance to things around us.⁹ In short, yet another methodological question for the philosopher of perception to answer concerns whether the available theories start off on an equal footing, or whether any of them enjoy some kind of privileged status.

⁹ Although see Gert, this volume: ‘suppose for a moment that experience is—contrary to the naïve realist’s view—merely a matter of various modifications to an experiencer’s consciousness: modifications caused, but not constituted, by external objects. Even if this were true, we would still end up learning to use such phrases as “right there”, “present”, and “directly before us” in the same way we would if naïve realism were true . . . what I am arguing is that even if naïve realism is false the phenomenology is not misleading.’ (115)

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi



9

2.3 On the Way: Fixed Points As we saw in the previous subsections, the explananda and starting points of a philosophical theory of perceptual experience place constraints on our theorizing. For example, assuming that part of the point of giving a theory of perceptual experience is to account for its phenomenal character, failing to do so is a mark against a theory. And if a particular theory has a default status, its rivals have a steeper hill to climb in order to win acceptance. But the explananda and starting points are not the only sources of theoretical constraints—philosophers of perception typically appeal to other kinds of fixed points in arguing for their own theories and objecting to rivals. To lean on the journey metaphor yet again, we might think of these fixed points as ‘guardrails’ that determine permissible routes to the destination explananda. Some guardrails are rather general, in the sense of being big-picture theories or worldviews. For example, a physicalist worldview—roughly speaking, the claim that everything is physical—is regarded as a fixed point by some (see, e.g., Gert’s and Gow’s contributions to this volume), but not by all (see, e.g., Allen’s contribution). Other guardrails are very specific, in that they are claims restricted to the particular phenomenon under investigation—principles specifically about perceptual experience. And some guardrails fall somewhere in the middle—for example, principles about mental states more generally.¹⁰ Providing an extensive list of fixed points that figure in philosophy of perception is too big a task for this introduction. Instead, we will illustrate how some relatively specific fixed points function in our theorizing through a case study that concerns the debate between naïve realists and their opponents. This divide roughly tracks attitudes to the following principles: Local Supervenience: ‘the phenomenal character of a [perceptual] state locally supervenes on neural activity’ (Locatelli, this volume, 86); that is, ‘if the proximal neural conditions that occurred when one perceived an object were created in the absence of that object, the subject would have an experience with the same phenomenal character as the original perceptual experience’ (Fish 2009, 41). The Indistinguishability Principle: ‘whenever two phenomenal characters are indistinguishable, they are qualitatively identical’ (Locatelli, this volume, 85). According to this principle, a total hallucination and a veridical experience from which it is subjectively indistinguishable have the same phenomenal character.

¹⁰ To flag a couple that crop up in this volume: Berkeley’s Likeness Principle, or the claim that an idea can be like nothing but another idea (Gow, this volume, 140), and the claim that mental states are entirely internal to their subjects, in the sense that they cannot have objects outside the subject’s skull as constituents (see Gert, this volume, 116).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

10

    

These principles form the basis of arguments for: The Common Kind Assumption: ‘the idea that we could have the very same kind of experience in veridical and hallucinatory cases’ (Gert, this volume, 113, see also Martin (2004, 40)). This claim is, of course, incompatible with disjunctivism about perceptual experience.

Naïve realists (by and large) reject these principles, whereas their opponents accept them. But there isn’t much discussion in the literature of which attitude towards these principles is correct; to date, neither camp has invested much effort in persuading the other that their attitude towards these principles is the right one (although there are some exceptions—e.g., Fish 2009, ch. 5). Proponents of these principles often seem to think that the naïve realists who deny them are deeply and perhaps even hopelessly confused about the subject at hand. In this volume, Roberta Locatelli seeks to break this kind of impasse by identifying common ground. She proposes that we can find such common ground in The Superficiality Constraint: ‘It is not possible that an experience seems through introspection to have a certain type of phenomenal character, while it doesn’t actually have that phenomenal character, and there is no disabling or interfering condition in place that prevents one from introspectively realizing that the experience doesn’t, in fact, have that type of phenomenal character’ (Locatelli, this volume, 97).

The idea is that there is a very tight connection between the phenomenal character of perceptual experience and introspection of it—so tight that (in favourable conditions, at least) certain kinds of introspective errors about it are impossible. Locatelli suggests that this principle, unlike the others just mentioned, is a neutral starting point that should be accepted by all parties to the debate (this volume, 98). However, in principle, one could accept the Superficiality Constraint while denying that a total hallucination would be subjectively indistinguishable from a possible veridical perception. That is, one could deny an analog of the Local Supervenience principle concerning the subjective indistinguishability of total hallucinations, namely: if the proximal neural conditions that occurred when one perceived an object were created in the absence of that object, the subject would have an experience that is subjectively indistinguishable from the original perceptual experience. Now, it is true that almost everyone—including naïve realists—accepts this principle. This is reflected by the fact that most parties to the debate hold that a metaphysics of perceptual experience owes an account of total hallucinations that are subjectively indistinguishable from a possible veridical

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi



11

perception. However, this principle is not beyond question. After all, if one doesn’t take it for granted that the relevant sort of neural activity is sufficient for perceptual phenomenal character, then the resulting state may well be subjectively distinguishable from a possible veridical perception. Plausibly, the neural activity is sufficient for something psychological, e.g., a disposition to believe a certain proposition about one’s environment (Fish 2009, 143). But, if one has gone so far as to question whether the relevant sort of neural activity is sufficient for perceptual phenomenal character, it’s no longer clear why that ‘something’ would be subjectively indistinguishable from a possible veridical perception. Rejecting the local supervenience principles leads naturally to the approach that Romdenh-Romluc attributes to Merleau-Ponty in her contribution to this volume (261–2): focusing on experiences that we know to be possible because they’re actual, and not worrying about creatures of science fiction like total hallucinations. For if one doesn’t accept the local supervenience principles just sketched, it is an open possibility that the mental state resulting from directly stimulating the relevant parts of a subject’s brain is not a perceptual experience, and hence not a state that a metaphysics of perceptual experience ought to cover. The moral of this particular case study concerns the relationship between the fixed points one accepts and what one counts as a target for theorizing in the first place. In particular, if one doesn’t regard the local supervenience principles as fixed points for theorizing, then (arguably) one need not accept that that there are total hallucinations that are subjectively indistinguishable from veridical perceptions that one’s theory must account for. At this point, the question arises whether one should regard the local supervenience principles as fixed points. And more generally, which principles should we regard as fixed points for theorizing?

2.4 On the Way: Methods We have seen that there is much variation in destinations, origins and fixed points in philosophy of perception. Is there also variety in the methods they use to get to their chosen destinations? It might seem that the answer is: no. A common conception of philosophical method is that of cost–benefit analysis (CBA). Applied to the debate over the metaphysics of perception, this means that the choice of a theory of perception will ultimately be settled—if at all—by determining which gives the greatest benefits, at the smallest cost. Inference to the Best Explanation (IBE) is also, as Drayson argues in her contribution to this volume, a common form of argument in philosophy of perception. This fits naturally with the primacy of cost–benefit analysis: for example, that some account of perception provides the best explanation of some pertinent phenomenon will be a benefit of that account.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

12

    

Several of the papers in this volume illustrate these common argumentative strategies: as such, their innovativeness lies not in their method, but elsewhere. For example, Spener introduces the possession of situation-dependent capacities as something which a theory of perception should explain, which she argues naïve realism does best. Thus naïve realism has a previously unnoticed benefit, to be weighed against its costs. Cavedon-Taylor argues that in the debate over the admissible contents of perceptual experience, we need to provide explanations of multi- as well as unimodal experience: certain multimodal experiences, he argues, are best explained by a view on which certain high-level properties can figure in the content of perceptual experience. Locatelli introduces the Superficiality Constraint, and holds that failure to accommodate it is an additional cost to be weighed in one’s choice of a theory. This is only a near consensus on broad methodology: many, but not all philosophers of perception are engaged in CBA and IBE. As discussed in section 2.2, Allen argues that naïve realism has a special status amongst philosophical theories of perception, and thus is not to be assessed on the basis of CBA or IBE. On this view, the defence of naïve realism should be understood as part of a ‘transcendental project of explaining how it is possible that perceptual experience’ has certain characteristics [this volume, 56]. One might think that the use of transcendental arguments would be the distinctive feature of this project. Allen demurs, pointing out that such arguments are subject to difficulties and can in any case be understood as ‘limiting cases of inference to the best explanation, in which the purported explanation is the best by virtue of being the only explanation’ (this volume, 59). Beneath even this near consensus about broad methodology is an underlying plurality of more specific methods. These are not necessarily inconsistent with one another, but there is disagreement about their usefulness. This disagreement sometimes shows up in differences in how costs and benefits are weighed, and thus also in views of whether an explanation is really ‘the best’. For example, there is disagreement over what weight we should give to accommodating the phenomenology of perception in our theorizing. Introspection—reflection on experience—is the method by means of which the relevant phenomenological facts are discovered. Those who do not give much weight to accommodating phenomenology are often, also, disparaging of the method of introspection. Gert, for example, suggests that from his neo pragmatist perspective, we should be cautious about this method because what motivates philosophical interest in perception should not be our firstperson understanding of perception but the paradigmatically third-person contexts in which we talk about it. In this he is starkly at odds with the approach of phenomenologists who, as Romdenh-Romluc puts it, ‘start with experience as it is undergone from the perspective of its subject’ (this volume, 259). Gert, then, thinks great weight should be put on the capacity of a theory to accommodate how we talk about perception. More specifically, his method is to

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi



13

explain away certain potential worries by locating their source in ways in which we talk about perception, and identifying the function of these ways of talking. Mac Cumhaill and Wiseman also endorse reflection on talk about perception as a central method—in their case, in order to identify ‘complex patterns of speech and action that manifest an individual’s grasp’ of perceptual concepts (this volume, 278). These philosophers put attention to language to quite different work, and their methods (as well as their aims, as mentioned in section 2.1), though both in some sense ‘linguistic’, are importantly different. As Ward makes clear in his contribution, there is also considerable variation in methods that involve reflection on experience or introspection: methods that are broadly ‘phenomenological’. In his terminology, a ‘Humean’ phenomenologist is one who holds that all that’s required to uncover philosophically significant facts about experience is to attend to it. Phenomenologists (those working in the phenomenological tradition) have rarely been Humean in this way. Ward sets Merleau-Pontian phenomenology in context, comparing it with Humean, Kantian, and Husserlian approaches to reflection on experience. One thing that characterizes Merleau-Pontian phenomenology is a kind of critical attention to our own reflective capacities. Romdenh-Romluc drills down into the details of Merleau-Ponty’s method, especially as applied to perception. For example, MerleauPonty thinks that there are aspects of experience that are in a sense ‘hidden’ from us, and which thus require a special method to be brought into view: one such method is that of the painter, abstracting ‘from the fullness of perceptual experience’ to identify its components (this volume, 272). Another is reflecting on a range of experiences, including those associated with pathologies. A common way of understanding the role of reflection on experience in analytic philosophy of perception is as providing data: facts to be explained or fed into a cost–benefit analysis. Perhaps some of the methods of the phenomenological tradition could be employed—by analytic philosophers of perception—in this way, to provide richer and more accurate descriptions of experience against which we might test our theories. For the phenomenologists themselves, however, explanation or analysis took the form of uncovering the ‘essences’ of experience—essential features or structures. History of philosophy has its own methods—for example, exegesis and interpretation—and these are sometimes ends in themselves. Historians of philosophy working on perception may be interested primarily in providing an accurate account of a historical figure’s ideas about perception. Others attempt to relate a historical figure’s account to contemporary work; for example, by arguing that their account commits them to naïve realism, or by identifying overlooked arguments for accounts of perception in their work (see, e.g., Gomes 2017; Kalderon 2017; Allen 2020). As Mac Cumhaill and Wiseman illustrate in their contribution, the methods of history of philosophy can also be put to more radical use: they employ careful exegesis to Anscombe’s work on perception to put

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

14

    

forward a distinctive, non-metaphysical conception of the aim of philosophy of perception. In summary, whilst IBE and CBA are common ‘modes of transportation’ to our destinations in philosophy of perception, there is only near consensus about their use. And there is room for the use of—and for disagreement about—other methods too, including transcendental arguments, introspection, linguistic methods, and the methods of the history of philosophy.

3. Philosophy and the Science of Perception Differences in purpose and procedure can seem especially profound when it comes to the relationship that the philosophy of perception stands in to the science of perception—differing conceptions of this relationship are what lead Fish in his contribution to construe the subfield as divided into incommensurable research programmes. Thus, features of the journey relevant to this relationship warrant separate treatment.

3.1 Destinations A major point of disagreement is whether the philosopher of perception’s target is the same as that of the scientist. That is, are the perceptual states with which much philosophy of perception is concerned to be identified with the representational states of traditional cognitive science of perception? As Fish discusses in his contribution, some, like Tyler Burge, think so, and also that it is within the remit of science to establish that this is so. Others—most notably John McDowell—argue that whilst the states with which scientists of perception are concerned are states of perceptual systems, philosophers are concerned instead with states of perceivers: a different thing altogether (see Burge 2005, 2011; McDowell 2010, 2013). This difference in views over philosophy of perception’s target explananda, discussed by Fish, Drayson, and Noordhof in this volume, has a disruptive ripple effect on ensuing debate: for those who take perceptual states to be identical to the states discussed in the cognitive science of perception, the results of the latter have direct and obvious implications for the debate over the nature of perception. For those who don’t, the implications will never be quite so direct and obvious. Most significantly, the Burge-ian view has been thought to straightforwardly entail the falsity of disjunctivism, or to force those who accept this view to eschew consistency with science altogether. However, Campbell (2002, 120) provides a useful metaphor for how philosophers of perception whose views entail disjunctivism might best understand the results

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi



15

of vision science. On his view, we should see the representational states involved in perception not as identical to perceptual experiences, but as states of perceptual systems that function to keep a window-like means of accessing the world clear or transparent. Perception is our access to the world when these representations perform their task successfully—it is not to be identified with the representations that make it possible. Noordhof (this volume, section 3.2) argues that the naïve realist’s attempt to push representations to this explanatory level puts them in a precarious position. A different dimension of variance concerns not what is to be explained, but the kinds of explanations it is the aim of philosophy of perception to provide, and how they relate to those to be found in perceptual science. On one kind of view, the difference is fairly minimal: philosophers of perception may offer explanations that are at a different level of abstraction to those provided by scientists working on perception, but these explanations are not different in kind. On this conception of the aims of philosophy of perception, it can be seen as a kind of theoretical—or, to use Fodor’s term—speculative psychology: a subdiscipline concerned with constructing empirical theories, but which nevertheless isn’t quite scientific psychology in eschewing the use of experiments (Fodor 1975, vii). Whilst philosophers of perception are rarely explicit on this point, it seems that many eschew this Fodorian conception: talk of perception’s ‘metaphysical nature’ or what perception ‘fundamentally’ is sometimes suggests a different kind of explanatory project. For example, on one conception of the aims of philosophy of perception, the job of the philosopher is to explain how perception is necessarily, and not merely how it happens to be implemented or instantiated in us, even at a reasonably high level of abstraction. In her contribution, Drayson argues that we should not understand philosophical theories of the metaphysics of perception in this way, and that they are not different in kind to scientific theories such as the ecological or constructivist theories of perception.At least some of this variance concerning the aims of philosophy of perception is related to the fact that work in this area can emerge from quite different philosophical subdisciplines. On the one hand, much philosophy of perception is associated with philosophy of mind, which borders epistemology and metaphysics. As such, such philosophy of perception is naturally concerned with the metaphysics of perceptual states as a specific question in the metaphysics of mind, and with explaining how perception could amount to contact with and thus knowledge of the world. More recently though, some philosophers working on perception would be more comfortable describing themselves as philosophers of psychology. Much philosophy of psychology is—as Bermudez puts it—‘concerned primarily with the nature and mechanics of cognition, rather than with the metaphysics and epistemology of mind’ (2005, 15). As such, its aims overlap with those of cognitive science itself.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

16

    

3.2 Origins Related to variance in the aims of philosophy of perception is similar variance in readiness to revise our concepts of and everyday understanding of perception in light of scientific findings. This can be seen as a kind of starting point in philosophy of perception, as it seems often to be determined at the outset of investigation rather than by what is established during it. We can understand this in terms of a spectrum of views that lie between what we might call Extreme Conservativism at one end, and Eliminativism at the other. According to the Extreme Conservative in this domain, our everyday concepts and understanding of perception are in good order and should be left as they are by the philosophy of perception. In fact, the Extreme Conservative, like Wiseman and Mac Cumhaill, might take our task to be that of understanding these concepts. According to the Eliminativist in this domain our everyday concepts and understanding of perception are not in good order and should be replaced by new concepts and a new understanding provided by the science of perception. One area of philosophy of perception in which this difference shows up is the debate over the individuation and nature of the senses, discussed by Noordhof (this volume, section 4). For example, Nudds’ (2004) starting point in exploring this issue is what he calls the counting question: ‘why do we count five senses?’ He proceeds to give an account of what the senses are that is intended to be sensitive to the significance that making this distinction has to us in everyday life. Others who contribute to this debate do not start in this way. For instance, one of the questions with which Nelkin (1990) and Keeley (2002) begin, is: how can we make the concepts of the senses scientifically useful? The Eliminativist position in philosophy of perception is fairly rare, however; most philosophers of perception seem to take faithfulness to our shared everyday notions about perception to provide something of a constraint on theorizing but differ on how radical a revision to these notions is permitted.

3.3 On the Way: Fixed Points A widely accepted fixed point concerns the consistency of philosophy of perception with the science of perception. If this means merely that claims made by philosophers of perception should not contradict well-supported empirical claims, then there may be no philosophers of perception for whom this is not a fixed point: even Noordhof ’s ‘Zettelist’ accepts it (this volume, 192–3), in taking it that philosophical accounts are wholly neutral with respect to the science of the mind. Instead, apparent inconsistency will be dealt with by dispute over which empirical claims are well supported, and their implications for philosophical claims.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi



17

This is illustrated by Wilkinson’s and Noordhof ’s discussions of the significance of the Predictive Processing Framework (PPF) in their contributions. Wilkinson’s concern is with the implications of PPF for the philosophical debate over the admissible contents of experience. Some have suggested that PPF dissolves this debate, because it shows that there is no line to be drawn between perception and cognition. Wilkinson argues that PPF is consistent with drawing this distinction, but has other implications for the debate in question. In section 3 of his paper, Noordhof argues that PPF poses problems for naïve realism.

3.4 On the Way: Methods Finally, differing views of the relationship between the science and the philosophy of perception emerge in ideas about the methods appropriate to the latter. One specific issue concerns whether the explanatory strategies employed by philosophy of perception are distinct from those employed by scientists. This is of course relevant to whether philosophy of perception should be considered continuous with the science of perception in the way it is naturally thought of as being when it is carried out within philosophy of psychology. For example, Drayson notes that a dominant methodology in philosophy of perception—inference to the best explanation—is associated with scientific reasoning. This in part motivates her claim that some scientific theories of perception (ecological and constructivist theories) are of a kind with the supposedly metaphysical theories over which philosophers of perception debate. Other methods are discussed elsewhere in the volume. For example, Allen argues that in order to accommodate the ‘meta-philosophical phenomenological datum’ that philosophy of perception is in danger of reaching an impasse, we need to recognize that naïve realism has a different status to other theories of perception, and that its denial is not practically intelligible. This decidedly non-scientific argument from practical intelligibility is quite different to the kind of inference to the best explanation that Drayson takes to dominate the field.

4. Ways Forward With these sources of variation identified, how should we move forward? First, it should be borne in mind that there will be other sources of variation too: other conceptions of purposes and procedures appropriate to the field that we haven’t considered here. So one natural next step is to keep investigating varying conceptions of destinations, origins, guardrails, and modes of transportation that characterize the philosopher of perception’s journey.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

18

    

On an optimistic view, identifying these sources of variation will allow us to resolve our differences. This might take the form of providing an explicit defence of certain aims, starting points, fixed points, or methods, sometimes by stepping out of the discipline, or one’s main tradition. For instance, one might mount a defence of a certain point on the spectrum between Eliminativism and Extreme Conservativism, by making use of the literature on the status of Common Sense or ‘Folk’ Psychology. Or, those who take introspection to be an important method in philosophy of perception might defend its use in more sophisticated forms by exploring the resources provided by the phenomenological tradition. We might also reach a kind of consensus by arguing that some debates can be dissolved altogether, inspired, perhaps, by deflationary approaches in metametaphysics (e.g., Thomasson 2010). Another way to resolve our differences might be to engage in the kind of ‘genetic phenomenology’ that (according to Ward) Merleau-Ponty advocates: a genetic phenomenology ‘aims to investigate the origins of what seems reflectively evident or apodictic to us’ (this volume, 247). Identifying why certain things seem obvious to some of us may help us to make progress, either by making them obvious to all, or by undercutting our certainty about them. Alternatively, one might seek conceptions more likely to be agreeable to all from the outset. This might mean eschewing controversial fixed points or explananda, and identifying new ones, as Locatelli and Spener do. We might also notice, when we identify the sources of variation, that we disagree less than we thought. For example, if some (pursuing the Anscombian grammatical approach outlined by Mac Cumhaill and Wiseman) intend to describe the linguistic mastery that manifests our grasp of perceptual concepts, then their results may not be at odds with those of more ontologically committing projects. On a view that is less optimistic, at least about a certain kind of progress in philosophy of perception, we are better off seeing some of our sources of disagreement as irresolvable and paradigm-defining—as Fish suggests. This means that we shouldn’t expect to have to defend these paradigm-defining conceptions to those who don’t share them. Whilst this can be described as a less optimistic perspective, it would also free up researchers to make progress within their paradigm by exploring the consequences of their ideas and incorporating new data. And, as Fish points out, it is part of this perspective that diversity of paradigms is to be encouraged. If we decide to work in this way though, it is necessary that we are aware of sources and kinds of difference—without that, we’ll be unaware when we’re bumping fruitlessly against another paradigm. A third way is also available. We might, along with the optimist, seek progress and agreement where we can, and where we can’t, allow that there is value in researchers in philosophy of perception working as if in a paradigm, if only temporarily, and perhaps with an eye to more substantial progress later. Finally, we might encourage the development of the intellectual virtues required to

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi



19

imagine ourselves into another paradigm or research programme, so that we can have helpful conversations with those who do not share even our most fundamental assumptions, without lapsing into fruitlessness.

5. Summary of Papers Fish diagnoses what he sees as deadlock in philosophy of perception—especially over the role of the science of perception—in the existence of incommensurate paradigms or at least different research programmes. He offers an, in some ways, optimistic way forward (let paradigms proliferate!) which falls short of resolving this deadlock. Other papers baulk at this, offering a range of ways out. Three of these favour naïve realism. First, Allen considers whether naïve realism might have some special status which may explain the ‘metaphilosophical phenomenological datum’ that philosophy of perception is in danger of reaching an impasse. Spener’s and Locatelli’s approaches are more conservative. Both argue that there is scope for progress without reconceiving the nature of the debate as Allen does. Spener identifies shortcomings in the debate so far, specifically a lack of attention to what it takes for something to form a ‘substantive mental kind’. Arguments for naïve realism that go via claims about fundamentality are, she suggests, unsatisfying. She then offers an argument for naïve realism that is designed to meet this explanatory demand. Like Spener, Locatelli seeks to identify something missing from current debate attention to which might yield progress. Unlike Allen and Fish, she is content to view the debate in terms of standard philosophical cost– benefit analysis. The problem as she sees it is that not all of the costs and benefits have been considered, and so ‘scope for progress’ remains. Specifically, she thinks, we need to identify the naïve realist’s commitments about introspection, and then (if one seeks to defend naïve realism) show why those commitments are acceptable. Gert and Gow also consider the debate over perception’s metaphysical nature. Gert—in contrast to the three preceding papers—offers methodological considerations contrary to naïve realism. Gow examines what is often thought to be a fixed point in theorizing about perception (and which Gert also notes): physicalism. According to Gert, the neo pragmatist approach offers a way out of the subdiscipline’s deadlock and towards an adverbalist account of perception. Gow— who also discusses adverbialism at length—argues that though physicalism may seem to act as a fixed point in theorizing about perception, not even the representationalist, in their commitment to abstracta, is consistent in their commitment to physicalism. Two papers consider a different debate in philosophy of perception: that over its ‘admissible contents’. Both papers defend the liberal view that so-called ‘highlevel’ properties can figure in the content of perceptual experience, and both do so

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

20

    

by drawing on broadly empirical considerations. Cavedon-Taylor argues that attention to multimodal illusions demonstrates truth of liberalism. Wilkinson argues that an empirical picture of how the brain works (the Predictive Processing Framework) favours this same view. Noordhof and Drayson take on the relationship between the philosophy and the science of perception. Noordhof demonstrates the complexity of this relationship: whilst scientific theorizing can falsify philosophical accounts of perception, we should, he argues, ‘be cautious about allowing the fruits of scientific investigation to mould too readily our understanding of the mental’ (this volume, 192). Along the way, Noordhof returns us to the debate over the metaphysical nature of perception, providing an argument against naïve realism. Drayson argues that this philosophical debate about the metaphysical nature of perception, and the scientific debate between constructivists and ecological theories, are not as we might think orthogonal: the scientific theories also contribute to our understanding of perception’s metaphysical nature. The volume ends with three papers with a focus on historical figures. As in the rest of the volume, we make no claim to comprehensiveness in our coverage. Nevertheless the papers we include demonstrate something of what we might gain by adopting a more historical approach in philosophy of perception. In the spirit of encouraging engagement of analytic philosophers of perception with other traditions, two of these papers are on Merleau-Ponty. These papers complement each other: Ward’s contribution situates Merleau-Ponty’s phenomenological method within a larger methodological picture. In so doing, he considers the sources of authority for this method, and thus Merleau-Ponty’s approach to ‘essences’. Romdenh-Romluc’s contribution is focused on the intricacies of Merleau-Ponty’s method, and how it is applied to perception. In so doing, she offers a conception of what identifying perception’s phenomenal character might amount to. She draws our attention to Merleau-Ponty’s conception of the items that occupy the visual field, and the methods he uses to identify them. Anscombe is associated in the minds of philosophers of perception with representationalism. Mac Cumhaill and Wiseman suggest this is mistaken. They outline and propose reanimating an Anscombian grammatical approach in philosophy of perception. This is a non-ontological approach in that it involves attention to ‘direct objects of perception’ as a grammatical rather than a classificatory notion, which is to be found when paying close attention to the way in which we ask and answer questions about what is perceived. They argue that Anscombe’s approach brings into view an ethical dimension of perception—for example, the influence of prejudice on what we see—which has been overlooked in contemporary philosophy of perception.¹¹ ¹¹ Many thanks to Keith Allen, Robbie Williams, Daniel Elstein, Thomas Brouwer, Andrew Peet, Haixin Dang, and Ludovica Adamo for feedback on a draft of this introduction.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi



21

References Allen, K. 2020. ‘The value of perception’. Philosophy and Phenomenological Research 100: 633–56. Berger, J. and Nanay, B. 2016. ‘Relationalism and unconscious perception’. Analysis 76: 426–33. Bermudez, J. 2005. Philosophy of Psychology: A Contemporary Introduction. New York: Routledge. Burge, T. 2005. ‘Disjunctivism and perceptual psychology’. Philosophical Topics 33: 1–78. Burge, T. 2011. ‘Disjunctivism again’. Philosophical Explorations 14: 43–80. Byrne, A. and Logue, H. 2008. ‘Either/Or’. In A. Haddock and F. Macpherson (eds.) Disjunctivism: Perception, Action, Knowledge. Oxford: Oxford University Press. Campbell, J. 2002. Reference and Consciousness. Oxford: Oxford University Press. Chalmers, D. 1995. ‘Facing up to the problem of consciousness’. Journal of Consciousness Studies 2: 200–19. Descartes, R. 1641/1993. Meditations on First Philosophy. Third edition. Indianapolis, IN: Hackett Publishing Company. Fish, W. 2009. Perception, Hallucination, and Illusion. Oxford: Oxford University Press. Fish, W. 2010. Philosophy of Perception: A Contemporary Introduction. New York: Routledge. Fodor, J. A. 1975. The Language of Thought. Cambridge, MA: Harvard University Press. Frankish, K. 2016. ‘Illusionism as a theory of consciousness’. Journal of Consciousness Studies 23: 11–39. Gomes, A. 2017. ‘Naive realism in Kantian phrase’. Mind 126: 529–78. Helton, G. 2016. ‘Recent issues in high-level perception’. Philosophy Compass 11: 851–62. Johnston, M. 2004. ‘The obscure object of hallucination’. Philosophical Studies 120: 113–83. Kalderon, M. 2017. Sympathy in Perception. Cambridge: Cambridge University Press. Keeley, B. 2002. ‘Making sense of the senses: Individuating modalities in humans and other animals’. Journal of Philosophy 99: 5–28. Logue, H. 2013a. ‘Good news for the disjunctivist about (one of) the bad cases’. Philosophy and Phenomenological Research 86: 105–33. Logue, H. 2013b. ‘Visual experience of natural kind properties: Is there any fact of the matter?’. Philosophical Studies, 162: 1–12. Logue, H. 2014. ‘Experiential Content and Naive Realism: A Reconciliation’. In B. Brogaard (ed.) Does Perception Have Content? Oxford: Oxford University Press, 220–41.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

22

    

Macpherson, F., ed. 2011. The Senses: Classic and Contemporary Philosophical Perspectives. Oxford: Oxford University Press. Martin, M. G. F. 2004. ‘The limits of self-awareness’. Philosophical Studies 120: 37–89. McDowell, J. 1986. ‘Singular thought and the extent of inner space’. In P. Pettit and J. McDowell (eds.) Subject, Thought, and Context. Oxford: Oxford University Press. McDowell, J. 2008. ‘The Disjunctive Conception of Experience as Material for a Transcendental Argument’. In A. Haddock and F. Macpherson (eds.) Disjunctivism: Perception, Action, Knowledge. Oxford: Oxford University Press, 376–89. McDowell, J. 2010. ‘Tyler Burge on disjunctivism’. Philosophical Explorations 13: 243–55. McDowell, J. 2013. ‘Tyler Burge on disjunctivism (II)’. Philosophical Explorations 16: 259–79. Nelkin, N. 1990. ‘Categorizing the senses’. Mind and Language 5: 149–65. Nudds, M. 2004. ‘The significance of the senses’. Proceedings of the Aristotelian Society 104: 31–51. Phillips, I. forthcoming. ‘Scepticism about unconscious perception is the default hypothesis’. Journal of Consciousness Studies. Schellenberg, S. 2018. The Unity of Perception: Content, Consciousness, Evidence. Oxford: Oxford University Press. Sethi, U. 2020. ‘Sensible overdetermination’. Philosophical Quarterly 70: 588–616. Siegel, S. 2011. The Contents of Visual Experience. Oxford: Oxford University Press. Soteriou, M. 2013. The Mind’s Construction: The Ontology of Mind and Mental Action. Oxford: Oxford University Press. Thomasson, A. 2010. Ordinary Objects. Oxford: Oxford University Press. Tye, M. 2000. Consciousness, Color, and Content. Cambridge, MA: MIT Press.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

2 Perceptual Paradigms William Fish

1. Introduction When we consider issues of methodology in the philosophy of perception, one significant question concerns the nature and extent of the interaction between philosophical and empirical approaches to the subject and, in particular, the extent to which philosophical thinking about the issues can and should be constrained by empirical claims and findings. Now, when it comes to questions about the nature of perceptual experience, I don’t think that anybody doubts that visual science produces results that bear on the philosophical questions, so I take it that the view that empirical results are relevant to the philosophical discussions is now universally accepted. However, there remains some deep disagreement as to precisely how this interaction should play out, and in how much scope there is for philosophers to make claims that appear to conflict with the kind of claims that the empirical researchers themselves make and maybe, in the process, offer possible ways in which empirical findings may be reinterpreted. In this paper, I will explore this issue by looking at the recent debate between Tyler Burge and John McDowell over the status of disjunctivism. I will suggest that aspects of this debate show a number of the characteristics that led Thomas Kuhn to introduce the notion of scientific ‘paradigms’ in different domains. Against this background, I will explore whether philosophical theories in fields such as the philosophy of perception can be usefully viewed in this way and, if they can, what methodological lessons we can learn regarding how to integrate philosophical and scientific perspectives in such fields.

2. Burge and McDowell on Disjunctivism In contemporary philosophy of perception, the dominant philosophical theories of perceptual experiences are currently relationalism (or naïve realism) and representationalism. The core claim of the representationalist view is that perception is to be understood, like belief and other intentional states, as fundamentally being a way of representing the world. Representationalists therefore typically hold that perceptual states have contents that specify how the world is represented

William Fish, Perceptual Paradigms In: Purpose and Procedure in Philosophy of Perception. Edited by: Heather Logue and Louise Richardson, Oxford University Press (2021). © OUP. DOI: 10.1093/oso/9780198853534.003.0002

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

24

 

to be and, moreover, that a subject’s being in a perceptual state that represents that the world is a certain way doesn’t entail that the world is that way: just as we can have false beliefs, we can have false or misleading perceptual experiences. In this way, the representationalist theory accommodates the possibility of deceptive experiences such as illusions and hallucinations. Representationalist thinking in the philosophy of perception began to become prominent in the second half of the twentieth century—early theories that are recognizably representationalist in character can be found in Anscombe (1965)¹ and Armstrong (1968)—which corresponds quite closely to the beginnings of the cognitive revolution within psychology (e.g. Chomsky 1959; Neisser 1967), which also has a central place for notions such as representation. And although philosophical representationalism is defended in part by appeal to important philosophical considerations—such as the need to explain the conscious character of perceiving, and how experiences justify beliefs—it is no accident that it grew up alongside cognitive psychology. The development by empirical scientists of theories of perception that tried to explain how the brain computes a representation of its environment (such as Marr 1982) both influenced and supported philosophical theories that deployed notions such as representation and computation. Of course, there have always been dissenting voices, both from empirical and philosophical standpoints. Within empirical psychology, for example, the work of J.J. Gibson and his followers (e.g. Gibson 1966) has always been a significant, if still not mainstream, approach within psychology, and this influence remains present in a range of non-representational enactivist and situated/embodied cognition approaches within psychology. The primary alternative to representationalism within contemporary philosophy of perception is a view known as naïve realism or relationalism. The central plank of this alternative view is the claim that, in successful perceptual experiences, the subject is directly aware of the part of the world they perceive in such a way that the subject could not have had an experience of that kind had the world been otherwise. So rather than say, as the representationalist does, that the subject is aware of their environment through being in a mental state that represents (as it happens, truly) that the world is a certain way, the relationalist claims that the subject is aware of their environment by simply being consciously connected to (‘open to’) the world and the way it is. This claim has the consequence that the relationalist faces the tricky question of how to account for deceptive experiences such as illusions and hallucinations, particularly in cases in which they are indistinguishable from possible successful perceptual experiences. For the representationalist, such cases are accounted for as the subject’s being in the same fundamental kind of state as a successful ¹ Although see the contributions from Mac Cumhaill and Wiseman (this volume) that dispute this reading of Anscombe.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

 

25

experience—a state that represents that the world is a certain way—but in which the content is false and the world is not as the experience represents it to be.² For the relationalist, however, things are not this straightforward: if the world is not as it seems to the subject to be, then the kind of experience the subject would have had in an indistinguishable successful case is simply not available in this deceptive case. So the relationalist cannot account for indistinguishable cases of illusion and hallucination by claiming that the subject is in the same fundamental kind of state that they would have been had the experience been successful. For this reason, the majority of naïve realists/relationalists have paired their view of the nature of successful perceptual experiences with a claim known as disjunctivism. The disjunctivist rejects the inference from the claim that two mental states are indistinguishable to the claim that the states are therefore of the same fundamental kind. Instead, the disjunctivist claims that ‘we should understand statements about how things appear to a perceiver to be equivalent to a disjunction that either one is perceiving such and such or one is suffering a . . . hallucination; and that such reports are not to be viewed as introducing a report of a distinctive mental event or state common to these various disjoint situations’ (Martin 2004, 37). In the kinds of cases we are considering, then—cases of indistinguishable perceptions and hallucinations—the disjunctivist claims that, despite the indistinguishability, these may nonetheless be mental states of different fundamental kinds, which in turn opens up the logical space to give a naïve realist theory of the successful cases, alongside some other theory of the deceptive cases. In 2005, however, Tyler Burge emphatically rejected disjunctivism—and thus naïve realist or relationalist theories that rely on it—on the grounds that it is simply incompatible with the results of visual science. In a long paper entitled ‘Disjunctivism and Perceptual Psychology’, Burge begins by arguing that ‘the psychology of perception, particularly vision, has become serious science. It has well-established results and successful application of mathematical methods. There is no good reason to doubt that it provides insight not only into the mechanics of perception, but into aspects of its nature’ (2005, 9). Burge then goes on to argue that visual science is committed to two claims: first, that the theory of vision is about ‘visual states of individuals, ordinarily so-called’ (2005, 22); second, that the visual states of an individual ‘causally depend only on proximal stimulations, internal input, and antecedent psychological conditions’ (2005, 22). This means that ‘the methodology of all serious empirical theory of vision guarantees that given types of visual state can be veridical in some circumstances and non-veridical in others’ (2005, 23). As disjunctivism claims otherwise,

² Although, for representationalists who also want their theory of deceptive experiences to accommodate cases of veridical hallucination and illusion, this simple formulation will need additional refinement.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

26

 

Burge suggests, it is therefore shown to be incompatible with the science: ‘It is fairly unusual, at least since the days of Descartes and Newton, for philosophical views to be as directly at odds with scientific knowledge as disjunctivism is. Hegel’s claim that there are seven planets comes to mind’ (2005, 29). In a 2010 response to Burge, John McDowell doesn’t dispute any of Burge’s claims about the science. In particular, he accepts that ‘there are state types of visual systems in common between occasions on which animals have veridical perceptions of colour and occasions on which their colour perception is erroneous’ (2010, 248). But he goes on to say that Burge’s discussion seems remarkably insensitive to the possibility that it might matter who or what is in a state. Experiences are states of perceivers; the states that perceptual systems get into when they as it were solve the problem of moving from sensory input to representations of the environment (which is not something perceivers do even in a metaphorical sense) are states of perceptual systems . . . . I urge that a good theory of the workings of a perceptual system yields accounts of what enables a perceiver to get into perceptual states, not accounts of what it is for a perceiver to be in those states. Burge scoffs at this. But he does not understand its point . . . . I raise a conceptual question about how we should understand the language used in such theories. Burge responds, bizarrely, as if to a case of empirical ignorance. (2010, 250–1)

So whereas Burge takes the scientific claims and the philosophical claims to be about the same thing—in Burge’s terms, ‘the visual states of individuals, ordinarily so-called’—McDowell argues that this is not mandatory. Instead, he suggests, the empirical claims can be interpreted as being about states of perceptual systems, and the philosophical claims as being about states of perceivers. Any appearance of inconsistency, therefore, is mistaken, as the claims are about distinct states. In a reply to this paper, Burge rejects McDowell’s distinction, and reiterates the charge of empirical ignorance: McDowell denies that the science is about conscious perceiving or misperceiving by individuals . . . He purports to accept the science. But he takes the states centered on by the science to be entirely sub-individual states, not states that are individuals’ successful or unsuccessful perceptions. (2011, 44) McDowell believes that he can make claims about the form and nature of human perception from the armchair because he thinks that perceptual psychology does not specify the form or nature of perceptions by human beings. This remarkable view is quite clearly mistaken. The view exhibits ignorance of the most elementary aims, claims, and methodology of the science of perceptual psychology. (2011, 66)

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

 

27

I think it clear that answering conceptual questions about a science’s language depends on empirical knowledge—knowledge of what the science says. To distinguish metaphor in popular expositions from literal theory, one must have empirical knowledge of how the science works; one must know the theory. Empirical ignorance of the science does prevent conceptual claims about the science from being correct. Conceptual questions are not, in general, independent of empirical knowledge. To do philosophy of science, one has to understand the science. (2011, 70)

Essentially, Burge insists that McDowell’s alternative interpretation is itself inconsistent with the science: the science not only specifies that a particular visual state type can be veridical in some circumstances and non-veridical in others, it also specifies that the visual states it is making theoretical claims about are the very states of perceivers that McDowell contends are the subject matter of philosophy. While Burge accepts that the presentations of the science can make use of metaphors that may, strictly speaking, be eliminable (2005, 13), he insists that you can only distinguish the metaphor from the theory if you understand the theory. And when you do understand the theory, he suggests, you find that the claim that the science concerns the visual states of perceivers, not merely perceptual systems, is in fact part of the theory. In turn, McDowell replies that he merely ‘denies that perceptions and corresponding misperceptions are individuated together at the level that is fundamental to epistemology. There are two different fundamental levels, and the idea of a clash with the science gets no grip’ (2013, 272). The science explains perceptual states as manifesting responsiveness to environmental circumstances. It is not part of its aim in doing that to characterize the epistemic significance of the states; that is why Burge’s accusation that disjunctivism is inconsistent with the science misfires. Similarly, the science is not in the business of offering a conceptual explanation of content as it figures in characterizations of conscious states. My articulation brings out a ground for taking that to be a non-starter; the use of the concept of content that would be the supposed basis for the explanation is derivative from the use that would supposedly be explained. In rejecting my thesis that the explanations provided by the science are enabling, not constitutive, Burge is implicitly endorsing the claim that a conceptual explanation of content, as it figures in descriptions of conscious perceptual states, is provided by the way the science makes those states intelligible. That is a disputable philosophical claim, not one that can demand the respect due to results of empirical science. (McDowell 2013, 275)

In a nutshell, then, it appears that the crux of this disagreement turns, not on whether the science stipulates that there are visual states that can be veridical or

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

28

 

non-veridical—both parties seem to accept that it does—but on whether the science can also stipulate that these visual states are the very states that philosophers such as McDowell are talking about when they advocate disjunctivism. Burge thinks that it can—and that disjunctivism is thereby revealed to be inconsistent with vision science—whereas McDowell insists that, even if empirical scientists were to make such a claim, they would be stepping outside of science and into the realms of philosophy. This leaves the door open to reinterpret the science itself as offering a theory of the enabling conditions for visual experience as philosophers understand it, which could in turn be rendered consistent with disjunctivism and the philosophical theories that rely on it.

3. Burge v McDowell: A Clash of Paradigms? Instead of trying to take a stand on this disagreement, I want to use this debate as a springboard to explore the methodological question of how philosophical thinking about a topic should interact with, and be guided/constrained by, empirical research into that area. Should we insist that philosophy be kept on a tight rein, and that the statements made by empirical scientists can be used to refute philosophical speculations, or should we allow philosophers a little more leeway such that, in cases where philosophers make claims that appear to conflict with the science, they can at least try to offer a consistent reinterpretation of the science, even if that does conflict with how the scientists themselves conceive of what they are doing? Which of these methodological approaches will be best placed to yield philosophical progress? So, our guiding methodological question is how, in a domain of enquiry where philosophy overlaps with science, philosophical theories should be impacted by the findings of the related science. Interestingly, one of the guiding questions of traditional philosophy of science is a direct parallel to this: how are scientific theories shaped by/impacted by empirical results? One influential answer to this question echoes the first of our options above: that empirical results can falsify or refute (but not prove) scientific theories. However, Thomas Kuhn famously argued that this is not how science actually operates: if an experimental result conflicts with a (dominant) theory, then the theory is either ‘adjusted so that the anomalous has become the expected’ (1962, 53) or else the anomaly is set aside in the hope that some new development will enable it to be resolved. It is only when these anomalous results build up unresolved that theory itself is actually threatened. So Kuhn’s point is that theories are more resilient in the face of anomalous findings than we might have anticipated. In his positive account of the way science operates, Kuhn introduces us to the notion of scientific paradigms, where a paradigm is something like a prevailing worldview, incorporating beliefs about both the kinds of questions that are

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

 

29

scientifically interesting, and a broad framework for going about answering them. In periods of ‘normal science’, Kuhn suggests, there will be a single prevailing paradigm, meaning that all practising scientists will share the same background beliefs and assumptions, and it is during this phase that problematic experimental findings are not really worried about. It is only when these anomalies build up that we start to reach a time of crisis, during which competing paradigms can emerge. A central feature of Kuhn’s description of this phase of science is that crossparadigm communication and critique can be fraught. With this in mind, it is intriguing to note that a feature of the debate between Burge and McDowell is that, on a number of occasions, they seem to be talking at cross purposes. For instance, each claims to fully understand the other, but nonetheless insists that the other misunderstands something critical, yet straightforward (for Burge, McDowell misunderstands the science; for McDowell, Burge misunderstands the point of the philosophy). This reminded me of a passage from Peter Godfrey-Smith’s Theory and Reality: ‘If we look down “from above” on two people who work within different paradigms who are arguing about which is better, it will often appear that the two people are talking past each other’ (2003, 91). Kuhn’s explanation of this phenomenon is that different paradigms are ‘incommensurable’. The underlying idea of incommensurability is the idea that there are no objective measures or standards by which competing paradigms can be compared and evaluated, which leads to proponents from each paradigm believing that their approach is superior as they each accept different standards for evaluation. In his discussions of Kuhnian philosophy of science, GodfreySmith distinguishes two key dimensions of incommensurability: incommensurability of communication and incommensurability of standards (2003, 91–2). The first sense of incommensurability captures the idea that the paradigm one is in will shape the way one uses and understands language, such that ‘people in different paradigms will not be able to fully communicate with each other; they will use key terms differently and in a sense will be speaking slightly different languages’ (Godfrey-Smith 2003, 91–2). A particular example of this in the Burge–McDowell debate can be found in the different ways they talk about the notion of successful, or ‘veridical’, perceptual experiences. McDowell often describes successful experiences in terms of a subject’s ‘openness to the layout of reality’ (1994, 26). Why? Because he thinks it is an extremely natural way to express what is involved in perceptual success: ‘I think the idea that experience at its best makes aspects of objective reality present to us is completely natural and intuitive’ (2010, 245). Burge, however, doesn’t find this way of talking remotely natural: ‘Laying aside the faintly mystical associations of manifestation and revelation, these locutions, some of them metaphorical, have unobjectionable ordinary uses. They describe veridical conscious experience’ (2011, 56). So when Burge hears McDowell talking about subjects enjoying experiences in which they are open to the layout of reality, he takes him to be

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

30

 

talking about perceptual experiences that are both veridical and conscious where, by veridical, he means representationally accurate: a perceptual state is veridical if its content is true and it thereby accurately represents the world (Burge 2005, 3). Yet McDowell insists that this is not how he means to be understood: Burge . . . can get no closer to the idea . . . of things being thus and so in a way that is manifest to a subject . . . than the idea of an experience that is veridical, an experience such that things are as the subject seems to perceive them to be. Burge’s thinking has no room for the idea of an experience in which some aspect of objective reality is there for a subject, perceptually present to her. That is a more demanding condition that an experience’s being veridical. (2010, 245)

This certainly has echoes of observing two people who are, in a sense, speaking slightly different languages: each finds different ways of speaking to be ‘natural’, and attempts to translate between the two ‘languages’ are fraught with difficulty and the potential for misunderstanding. The second dimension of incommensurability discussed by Godfrey-Smith is that different paradigms will disagree about what counts as both explaining a puzzle and what counts as a puzzle that needs to be explained in the first place. Kuhn says that ‘the proponents of competing paradigms will often disagree about the list of problems that any candidate for paradigm must resolve’ (1962, 148). As an example of this, consider the case that Burge alluded to earlier: in old astronomic paradigms, they really cared about why there are the number of planets that there are. ‘When Galileo reported seeing moons circling Jupiter, he was refuted by a priori arguments that there could be no such thing, since the number of planets (i.e. objects in the solar system) was necessarily seven’ (Garfinkel 1981, 6). However, modern astronomy no longer even attempts to answer this question: ‘suppose there are nine planets. Why is this so? What explanation does modern astrophysics give us for the fact? It turns out that there is no nontrivial explanation. Modern science rejects the idea of explaining that sort of thing’ (Garfinkel 1981, 7). So it is not so much the case that modern astronomy doesn’t provide an answer to this question (although it doesn’t), it is more that modern astronomy rejects the idea that this is an appropriate question for the science to address at all. Other examples include Aristotle’s insistence that a theory of species should explain why we have the species we do—something that contemporary zoology does not even try to explain—and the early eighteenth century debates about whether Newton’s (inverse square law) explanation of gravity, which took gravity to act instantaneously and at a distance, would only be acceptable if it were supplemented with a causal mechanism for how gravitational attraction works. A similar pattern, I suggest, can be found in the Burge–McDowell debates. As the opening sections of McDowell (2010) make clear, McDowell takes epistemological considerations to be critical: ‘the point of my disjunctivism is to

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

 

31

accommodate the fact that experiences in which an appearance is a mere appearance can be indiscriminable from experiences that make knowledge of the environment available’ (2013, 260). From McDowell’s perspective, these epistemological considerations place constraints on the nature of perception, and the theoretical task is to show how this conception of perception is compatible with the science. But Burge doesn’t see that a theory of the nature of perception should be guided by epistemic constraints. As a philosopher, it’s not that Burge is uninterested by epistemology, but he certainly seems to think that this is not a question that the scientists should try to explain. He thinks that we should allow science to determine the fundamental nature of perceptual states, and then think about how this impacts on epistemology: ‘To understand the nature of warranted perceptual belief, one must understand the nature of perception and belief. Science is our best guide for perception’ (Burge 2011, 79). Similarly with consciousness. McDowell notes that it ‘is not part of [science’s] aim . . . to characterize the epistemic significance of the states . . . , the science is not in the business of offering a conceptual explanation of content as it figures in characterizations of conscious states’ (2013, 275). And Burge also seems to agree— at least with the suggestion that psychology is not in the business of explaining perceptual consciousness: Perceptual psychology as it now stands does not attempt to give a complete theory of the essence of all perceptual states. For example, it is possible that consciousness is an aspect of the essence of some perceptual states. (It is almost surely, however, not an essential feature of all perception.) The psychological theories that I have discussed do not attempt to explain consciousness. There is, currently, no scientific theory of consciousness. (2005, 46)

Yet for McDowell, an adequate theory of the nature of perceptual states should say something about consciousness. This is why he rejects the suggestion that the very idea of conscious dealings with content can be explained in terms of access by individuals to the content of states in subindividual systems . . . [because] there is no prospect of conceptual illumination of content as attributed to conscious perceptual states in terms of content as attributed to sub-individual states. (2013, 275)

To this extent, Burge and McDowell do appear to ‘disagree about the list of problems that any candidate for [an acceptable theory of perception] must resolve’. So our guiding methodological question is how, in a domain of enquiry where philosophy overlaps with science, philosophical theories should be impacted by the findings of the related science. This brief overview of the dispute between

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

32

 

Burge and McDowell suggests that it displays enough of the classic features of a clash of Kuhnian paradigms—just with philosophical theories, rather than scientific theories, playing the role of paradigms—to make it worth exploring whether we can learn something about the nature of this disagreement, as well as something methodological about how these debates might be pursued, by looking at similar discussions within more traditional philosophy of science. In particular, it suggests that the answer to our methodological question will be somewhat more complex than the simple view that empirical results can straightforwardly refute philosophical theories. So is there anything we can learn from the philosophy of science about how to understand the relationship between empirical results and philosophical theorizing? Before I go on to address this question directly, I want to begin by noting that, despite the similarities, there are some critical differences between the philosophical case and Kuhn’s description of the core scientific case. First of all, for Kuhn, ‘crises are a necessary precondition for the emergence of novel theories’ (1962, 77). A crisis occurs when there is ‘a pronounced failure in the normal problemsolving activity’ within a paradigm (1962, 74–5). In philosophy more generally, however, competing theories tend to be the norm rather than the exception, and in the particular case in question, the broadly representationalist approach to perception is far from in crisis. Second, where Kuhn has normal science (in a given field) as involving a single paradigm at any one time, this doesn’t tend to be the case in the philosophy of perception (although maybe the sense datum theory in the early twentieth century and representationalism in the 90s to early noughties come close).³ Third, whilst there may be significant areas of agreement within mainstream philosophy of perception as to the fact that perception is representational, there are just as many important differences: is content Russellian or Fregean or . . . ? Are high level properties represented? Does phenomenology determine content or content determine phenomenology, or . . . ? And so on. However, Kuhn’s ‘one paradigm at a time’ picture has always been one of the more criticized elements of his overall philosophy of science. Later developments of his core ideas explicitly attempted to allow the coexistence of paradigm-like ‘research programmes’ that were in ongoing competition.

³ Perhaps this shows that the philosophy of perception is operating more like Kuhnian pre-paradigm science, in which there is no ‘single generally accepted view’ but instead a ‘number of competing schools and sub-schools’ (1962, 12). However, as Kuhn’s distinction between pre-paradigm science and normal science is primarily about the presence or absence of consensus, rather than the presence or absence of incommensurability between approaches—Kuhn accepts that the incommensurability issues are present at this stage as well (1962, 13)—and, as we shall see, Kuhn’s claims about scientific consensus have been independently criticized, I won’t treat this as an interestingly distinct possibility in what follows.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

 

33

4. Research Programmes Imre Lakatos argued that the recognition of cotemporaneous research programmes in competition was necessary to accurately represent the history of science. He claimed that the ‘history of science suggests that . . . tests are—at least—three-cornered fights between rival theories and experiment’ (1970, 115), and that ‘there is no falsification before the emergence of a better theory’ (1970, 119). For this reason, the more competing theories the better: ‘proliferation of theories cannot wait until accepted theories are “refuted” (or until their protagonists get into a Kuhnian crisis of confidence)’ (1970, 122). If we set this important difference aside, however, we can see that Lakatos’s research programmes are otherwise quite similar to Kuhnian paradigms, albeit with a little more detail about the structure.⁴ As Lakatos describes it, a research programme is made up of a hard core surrounded by a protective belt. The hard core of a programme contains its core commitments or essential components: the claims that make the theory what it is. As these hard core commitments could not be abandoned without thereby abandoning the research programme itself, they are ‘protected’ by a ‘belt’ of potentially refutable claims, which serve to connect the hard core claims to the world.⁵ Methodologically speaking, the hard core and protective belt connect to two heuristics, which specify how research in the programme is to proceed: The negative heuristic specifies the ‘hard core’ of the programme which is ‘irrefutable’ by the methodological decisions of its protagonists; the positive heuristic consists of a partially articulated set of suggestions or hints on how to change, develop the ‘refutable variants’ of the research-programme, how to modify, sophisticate, the ‘refutable’ protective belt. (1970, 135)

So changes within the protective belt claims are not only acceptable, they are desirable. A research programme progresses—counts as progressive—by a series of such changes increasing the application of the programme to new cases, or providing a more precise treatment of existing cases.

⁴ Kuhn at least appears happy to take on Lakatos’s terminology. See, for example, 1970, 246 and particularly 256, where he says ‘Though his terminology is different, his analytic apparatus is as close to mine as needs be: hard core, work in the protective belt, and degenerative phase are close parallels for my paradigms, normal science, and crisis.’ ⁵ We might worry that philosophical theories cannot be usefully viewed on the model of research programmes, inasmuch as their core commitments are not empirically testable claims. However, Lakatos allows the hard core of a theory to be what he calls ‘metaphysical’—where ‘a contingent proposition is “metaphysical” if it has no “potential falsifiers” ’ (1970, 132)—so long as any ‘problematic instances can be explained by content-increasing changes in the auxiliary hypotheses appended to it’ (1970, 126). ‘The methodology of a research programme with a “metaphysical” core does not differ from the methodology of one with a “refutable” core’ (1970, 126–7).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

34

 

When it comes to competition between research programmes, Lakatos is a bit quiet about how to decide between them. He does, it is true, suggest that a research programme is successful if the succession of protective belt changes leads to a ‘progressive problemshift’ (1970, 133)—that is, the changes lead to novel predictions and discoveries—and unsuccessful if the changes lead to the theory explaining and predicting less. However, he also cautions patience, insisting that there is ‘sufficient rational scope for dogmatic adherence to a programme in face of prima facie “refutations” ’ (1970, 134). I will come back to this issue in due course, but first I want to explore what methodological lessons we might learn from Lakatos’s discussion of how a theory develops through modifying its protective belt. One aspect of his discussions that is particularly relevant to our current question is his detailed account of how conflicts between theoretical claims and experimental results proceed. ‘When the theoretician appeals against the verdict of the experimentalist’, he says, ‘the appeal court does not normally cross-question the basic statement directly but rather questions the interpretative theory in the light of which its truth-value had been established’ (1970, 128). To illustrate this, Lakatos explores how William Prout’s theory that all atoms are compounds of hydrogen and thus the atomic weights of all chemical elements must be expressible as whole numbers (T (Prout)) dealt with Jean Stas’s apparently falsifying experimental results (such as R (Stas)). T (Prout): all atoms are compounds of hydrogen and thus the atomic weights of all chemical elements must be expressible as whole numbers. R (Stas):

the atomic weight of Chlorine is 35.5.

Lakatos’s first move is to restate these to claims in such a way that the inconsistency is clear. He does this as follows: T’ (Prout): all atoms are compounds of hydrogen and thus the atomic weights of all pure (homogenous) chemical elements are multiples of the atomic weight of hydrogen. R’ (Stas): Chlorine X [a particular sample of Chlorine] is a pure (homogenous) chemical element and its atomic weight is 35.5. He then explores the structure of R’ in more detail, noting that it is a conjunction of two claims: R1:

Chlorine X is a pure (homogenous) chemical element.

R2:

Chlorine X has an atomic weight of 35.5.

And in turn that R1 stands for a conjunction of two longer statements:

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

 

35

T1: If seventeen chemical purifying procedures p1—p17 are applied to a gas, what remains will be pure Chlorine. T2:

X was subjected to the seventeen procedures p1—p17.

Lakatos then argues as follows: The careful ‘experimenter’ carefully applied all seventeen procedures: T2 is to be accepted. But the conclusion that therefore what remained must be pure chlorine is a ‘hard fact’ only in virtue of T1. The experimentalist, while testing T, applied T1. He interpreted what he saw in the light of T1: the result was R . . . . But what if T1, the interpretative theory, is false? Why not ‘apply’ T rather than T1 and claim that atomic weights must be whole numbers? Then this will be a ‘hard fact’ in the light of T, and T1 will be overthrown. Perhaps additional new purifying procedures must be invented and applied. (1970, 129)

In many ways, Lakatos’s point is merely a careful articulation of something we all already know—that one way of rejecting a valid argument is to reject one of its premises—alongside drawing our attention to the fact that, when viewed from the perspective of a dominant worldview or paradigm, a number of the potentially rejectable premises may actually be left hidden, or implicit, as they are simply taken for granted as part of the relevant theoretical background. In this light, consider Adam Pautz’s recent use of empirical results in his attempt to refute theories that claim that the ‘phenomenal character of an experience is fully determined by the external physical properties’ the experience relates the subject to (2013, 236). Focusing on cases of taste and smell, Pautz argues that there are two problematic sets of empirical results for views of this kind. Not only are there cases in which subjects have similar phenomenal experiences in response to dissimilar objects (such as finding that both small molecules like urea and the large organic acids found in hop oils taste bitter), there are also cases in which subjects have dissimilar phenomenal experiences in response to similar objects, such as when one chemical (gentiobiose) tastes bitter, yet an anomer of this chemical—isomaltose—tastes sweet (2013, 227). Taken together, these results can look to undermine the relationalist claim that the phenomenal character of an experience is constituted by the external object the experience relates the subject to. If we try to outline Pautz’s argument along Lakatosian lines, the first thing we would need to do is to identify the relationalist’s ‘hard core’, which—perhaps— would look something like this: T(rel): Phenomenal experiences are constituted by awareness relations to external objects.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

36

 

Pautz’s apparently falsifying experimental result is as follows: R(Pautz): Subjects can have similar phenomenal experiences in response to different external objects and different phenomenal experiences in response to similar external objects. If we then try and restate these claims in such a way that the inconsistency is clear, we would end up with something like the following: T’(rel): Phenomenal experiences are constituted by awareness relations to external objects such that similar phenomenal experiences should involve awareness relations to similar external objects, and dissimilar phenomenal experiences should involve awareness relations to dissimilar external objects. R’(Pautz): Similarity in subjects’ phenomenal experiences does not track similarity in external objects. In this case, both of these claims are really standing for conjunctions of longer statements. T’(rel) actually stands for the conjunction of the original T(rel)— phenomenal experiences are constituted by awareness relations to external objects—and T(rel)2: similar phenomenal experiences should involve awareness relations to similar external objects, and dissimilar phenomenal experiences should involve awareness relations to dissimilar external objects. This is notable primarily because T(rel)2 doesn’t look like a hard core claim of relationalism at all, but a protective belt claim at best—there seems to be no fundamental reason why a relationalist should feel compelled to accept such a claim, particularly given our everyday familiarity with scenarios, such as Austin’s bar of soap that looks like a lemon (Austin 1962, 50), in which significantly different objects look alike. To this extent, then, even if R(Pautz) were accepted, it would only seem to require a protective belt adjustment, and not abandonment of the hard core of the programme. What about R’(Pautz)? Well, there is a lot of implicit background theory going on here. I would suggest that R’(Pautz) is in fact a conjunction of at least the following four statements. T(Pautz)1: If subjects report that two phenomenal experiences are similar/ dissimilar, then their phenomenal experiences are similar/dissimilar. T(Pautz)2: Subjects report that their olfactory phenomenal experiences of chemicals A and B are similar, and that their olfactory phenomenal experiences of chemicals A and C are dissimilar. T(Pautz)3:

Chemicals A and B are dissimilar where chemicals A and C are similar.

T(Pautz)4:

Chemicals are external objects.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

 

37

To paraphrase Lakatos, let us accept T(Pautz)2—that subjects do indeed make the reports claimed. But the conclusion that therefore similarity in subjects’ phenomenal experiences does not track similarity in external objects is a ‘hard fact’ only in virtue of the other theoretical commitments. Pautz, while testing T(rel)2, applied T(Pautz)1, 3 and 4. He interpreted what he saw in the light of these interpretative theoretical claims: the result was R(Pautz) . . . . But what if one or more of the interpretative theories are false? Why not ‘apply’ T(rel)2 rather than T(Pautz)1 and claim that phenomenal reports are thereby shown to be unreliable? Or ‘apply’ T(rel)2 rather than T(Pautz)4 and claim that chemicals are thereby shown to be inappropriate candidates for the external objects of which the relationalist speaks? Or ‘apply’ T(rel)2 rather than T(Pautz)3 and claim that Pautz’s techniques for classifying chemicals as similar/dissimilar are thereby shown to be flawed? Perhaps additional new classificatory practices must be invented and applied. Yet of course, to the extent that claims such as T(Pautz)1, 3 and 4 are part of the collection of claims that Pautz takes for granted, this will look like a refutation from Pautz’s perspective. However, it is entirely legitimate for the ‘theoretician [to] demand that the experimentalist specify his “interpretative theory”, and he may then replace it—to the experimentalist’s annoyance—by a better one in the light of which his originally “refuted” theory may receive positive appraisal’ (1970, 130). Of course, as a description of the kinds of options available to the relationalist in responding to Pautz, none of this will come as any surprise. However, I do want to suggest that viewing this debate through the lens of paradigms/research programmes makes two key ideas particularly salient. The first is that, while Burge is right that philosophers should know the science, this doesn’t require accepting everything that scientists say at face value, or at least that tracking the significance of what they say is entirely straightforward. There are unlikely to be fixed, uncontroversial empirical facts. So responding in a ‘reinterpretative’ vein is not intrinsically ad hoc or irrational. Changes to protective belt claims are not just acceptable, they are welcomed, so long as they are progressive. Theories progress by developing their protective belts in such a way that the theory can accommodate new phenomena. The second lesson is that there is no requirement for the relationalist programme to develop a response to such a claim on pain of being viewed as failing. It is legitimate to shelve puzzles and anomalies in order to come back to them at a later stage. As Lakatos says, ‘Our considerations show that the positive heuristic forges ahead with almost complete disregard of “refutations” . . . The anomalies are listed but shoved aside in the hope that they will turn, in due course, into corroborations of the programme’ (1970, 137). A classic example of this is the Darwinian approach to puzzles surrounding geological time: Darwin’s theories needed Earth to be much older than geology of the time believed, but this anomaly didn’t mean that development of his theories ground to a halt. Instead, this was recognized as a puzzle, but shelved and not worried about.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

38

 

5. Methodological Lessons for the Philosophy of Perception If the debate between representationalism and naïve realism in the philosophy of perception can usefully be seen on the model of a paradigm/research programme rivalry within science, then we can answer our guiding methodological question as follows. In a domain of enquiry where philosophy overlaps with science, philosophers should first of all be familiar with relevant empirical findings and their prevailing interpretations, but they should be given some latitude to develop coherent ways of reinterpreting the science if the presented interpretations appear to conflict with the philosophical theory they endorse. Second, even if this is not presently possible, this does not mean that the philosophical approach has been refuted, and that further development of the approach should cease on pain of irrationality. Rather, it can be legitimate to shelve a problematic result in the hope that, as the theory develops further, it becomes able to be accommodated. This does, however, raise the further question of how we might ever resolve a philosophical debate and declare one theory the winner. If philosophical theories can usefully be seen along the lines of paradigm/research programme rivalries, then any particular articulation of a philosophical position will contain both core tenets of the theory that could not be abandoned without abandoning the theory itself, and a range of negotiable, amendable developments or interpretations of this core idea (indeed, in the kind of philosophical cases we are considering, I would expect the central non-negotiable hard core of a theory to contain little more than the defining statements of the philosophical position). For this reason, philosophical theses will be extremely difficult to refute as there will often be either reinterpretations or protective belt amendments available to accommodate apparently problematic arguments or findings. Moreover, given that the supporters of a paradigm/research programme will not only endorse the central hard core tenets but also a range of (at-least-potentially-revisable) protective belt claims, then standing within one camp and issuing ‘refutations’ of the other is not likely to be successful. As Kuhn himself says: The proponents of competing paradigms are always at least slightly at crosspurposes. Neither side will grant all the non-empirical assumptions that the other needs in order to make its case. Like Proust and Berthollet arguing about the composition of chemical compounds, they are bound partly to talk through each other. Though each may hope to convert the other to his way of seeing his science and its problems, neither may hope to prove his case. The competition between paradigms is not the sort of battle that can be resolved by proofs. (1962, 148)

A consequence of this is the kind of ‘talking past’ one another that is manifest in the discussions between Burge and McDowell:

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

 

39

people working within any one paradigm will have no problem saying why their paradigm is superior to the other, by citing key differences in what can be explained and what cannot. But these comparisons will be compelling only to those inside the paradigm from which the claim of superiority is being made. (Godfrey-Smith 2003, 91)

So how does the competition between paradigms or research programmes find a resolution? Well, Kuhn himself doesn’t think that there’s anything much that adherents to either paradigm can do. Instead, he explains this process as one of social psychology. In The Structure of Scientific Revolutions, for example, he quotes Max Planck: ‘a new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with it’ (1962, 151). In later work, he isn’t quite so bleak, but without really offering much more to work with. There he says, ‘take a group of the ablest available people with the most appropriate motivation; train them in some science and in the specialities relevant to the choice at hand; imbue them with the value system, the ideology, current in their discipline . . . and finally, let them make the choice’ (1970, 238). As we saw, Lakatos does give us a little more to work with, suggesting that research programmes will count as progressive/successful if the process of making theoretical revisions to the protective belt claims both leads to the theory making novel predictions and to some of those predictions being corroborated. This suggests that, when it comes to the competition between representationalism and naïve realism in the philosophy of perception, it may be possible for adherents of one paradigm/research programme to argue that the other is in a degenerating phase because it is not making novel predictions and/or the predictions it makes are not being corroborated.⁶ However, things are not quite this straightforward. This is in part because, even in areas that are also studied by empirical scientists, such as the philosophy of perception and the philosophy of consciousness, philosophers do not tend to make the kind of testable predictions that we find in scientific contexts. This is not to say that philosophical theories cannot be made to yield predictions, however. Famously, in Consciousness Explained (1991), Daniel Dennett sketched some experiments that could be used to test implications of his model of consciousness, and some of the predictions he made were indeed later corroborated. So it may not be that philosophical claims are inherently untestable and hence unscientific, but that philosophers themselves—happiest, as they are, in more conceptual

⁶ This approach has been used in philosophical rejections of competing theories in the past, including Quine’s rejection of traditional epistemology in ‘Epistemology Naturalized’ (1969) and Churchland’s rejection of folk psychology in ‘Eliminative Materialism and the Propositional Attitudes’ (1981).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

40

 

spaces—may neither be best placed nor particularly inclined to derive testable predictions from their theories. Moreover, even in situations where philosophical theories can be made to yield predictions, there may well not yet be any noncontentious way to test those predictions. For instance, suppose someone were to contend that, for a mental state to be phenomenally conscious, the subject must be aware that they are in that state. Such a philosophical theory would ‘predict’ that you will not find a phenomenally conscious state that its subject does not have cognitive access to. But how could such a prediction be tested? If we were to take reportability to be adequate evidence for phenomenal consciousness then the prediction would appear to be corroborated—as reportability implies cognitive access—yet this test is unlikely to be acceptable to people who are not antecedently inclined to accept the theory. Suppose, then, that we approach the question in a different way: perhaps we look at the neural activity that occurs in the sensory areas of the brain in clear-cut cases of reportable phenomenal consciousness, and see if similar activity is present in cases that are unreportable. If it is, then this would appear to show consciousness in the absence of access and thereby refute the prediction. Yet this test is unlikely to be acceptable to people who are antecedently inclined to accept the theory. After all, we know that the overall neural activity will be different due to the fact that the neural mechanisms that support reportability will not be active, so to assume that this is irrelevant to the question of whether phenomenal consciousness is present just looks to beg the question. So even though philosophical theories can be made to yield predictions, the nature of the issues we are grappling with may mean that we do not yet have any agreed independent way of testing those predictions. This all suggests that the appearance that one research programme is degenerating might itself be a research-programme-relative observation, and hence not compelling to those in the programme under attack. So when we do find adherents of one paradigm/research programme arguing that the other is in a degenerating phase, we should tread carefully. In cases such as these, the patience that Lakatos urges in standard scientific practice looks even more appropriate. Even in the core scientific cases, we saw that Lakatos cautions us that there is ‘sufficient rational scope for dogmatic adherence to a programme in face of prima facie “refutations” ’ (1970, 134). He also notes that it may take a long time before we can assess whether a theory does indeed predict a novel fact (1970, 155), and stresses that nascent research programmes should be afforded a period of protection in which they can be developed without threat of elimination (1970, 157). Indeed, Lakatos often emphasizes the importance of hindsight: you can only really know how successful a research programme is (or was) when we look back at it. Given this attitude towards scientific theories, one would expect Lakatos to be amenable to the idea that, in cases such as ours, it is acceptable to pursue philosophical approaches that may conflict with important aspects of the

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

 

41

currently dominant philosophical research programmes. Indeed, as Philip Kitcher argues in The Advancement of Science (1995), this is the optimal way for exploration to proceed: even in cases when there is a dominant programme that looks significantly more promising than any other, an ideal allocation of resources would still allocate researchers to other research programmes, so long as this did not significantly deplete the capacity of the dominant programme to engage in research. In her discussions of persistent disagreements within philosophy, Catherine Elgin agrees. In philosophical cases such as ours, in which each theory seems to address ‘important common problems that the other cannot’, she argues that it will be ‘better for the epistemic community as a whole that some of its members continue to accept each position’ (2010, 67). So a further methodological lesson is therefore one of tolerance: a lack of consensus is not a bad thing, and intellectual progress will be best served by allowing people to develop and defend theories, even if they may not seem appealing to us, and not in attempting to enforce homogeneity.⁷

References Anscombe, G.E.M. 1965. ‘The Intentionality of Sensation: A Grammatical Feature’. In R.J. Butler (ed.) Analytic Philosophy. Oxford: Blackwell, 158–80. Armstrong, D.M. 1968. A Materialist Theory of the Mind. New York: Routledge & Kegan Paul. Austin, J.L. 1962. Sense and Sensibilia. Oxford: Oxford University Press. Burge, T. 2005. ‘Disjunctivism and perceptual psychology’. Philosophical Topics 33 (1): 1–78. Burge, T. 2011. ‘Disjunctivism again’. Philosophical Explorations 14 (1): 43–80. Chomsky, N. 1959. ‘Review of Skinner’s Verbal Behavior’. Language 35: 26–58. Churchland, P.M. 1981. ‘Eliminative materialism and the propositional attitudes’. The Journal of Philosophy 78 (2): 67–90. Dennett, D.C. 1991. Consciousness Explained. New York: Little, Brown and Co. Elgin, C.Z. 2010. ‘Persistent Disagreement’. In R. Feldman and T.A. Warfield (eds.) Disagreement. Oxford: Oxford University Press, 53–68. Garfinkel, A. 1981. Forms of Explanation: Rethinking the Questions in Social Theory. New Haven, CT: Yale University Press.

⁷ I would like to thank Heather Logue and the (Sense) Perception in the North network for hosting the Purpose and Procedure in Philosophy of Perception conference at which this paper was delivered, as well as Zoe Drayson for her commentary on the paper, and the other attendees for a stimulating Q & A session. I would also like to thank Heather (again), as well as Louise Richardson, Paul Noordhof, Keith Allen, Yusuke Ogawa, John Matthewson, and Steve Chadwick for comments on written drafts of the paper.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

42

 

Gibson, J.J. 1966. The Senses Considered as Perceptual Systems. Boston: Houghton Mifflin. Godfrey-Smith, P. 2003. Theory and Reality: An Introduction to the Philosophy of Science. Chicago: The University of Chicago Press. Kitcher, P. 1995. The Advancement of Science: Science Without Legend, Objectivity Without Illusions. Oxford: Oxford University Press. Kuhn, T. 1962. The Structure of Scientific Revolutions. Chicago: University of Chicago Press. Lakatos, I. 1970. ‘Falsification and the Methodology of Scientific Research Programmes’. In I. Lakatos and A. Musgrave (eds.) Criticism and the Growth of Knowledge. Cambridge: Cambridge University Press, 91–196. Marr, D. 1982. Vision. San Francisco: Freeman. Martin, M.G.F. (2004). ‘The limits of self awareness’. Philosophical Studies 120: 37–89. McDowell, J. 1994. Mind and World. Cambridge, MA: Harvard University Press. McDowell, J. 2010. ‘Tyler Burge on disjunctivism’. Philosophical Explorations 13 (3): 243–55. McDowell, J. 2013. ‘Tyler Burge on disjunctivism (II)’. Philosophical Explorations 16 (3): 259–79. Neisser, U. 1967. Cognitive Psychology. East Norwalk, CT: Appleton-Century-Crofts. Pautz, A. 2013. ‘The Real Trouble for Phenomenal Externalists: The Science of Taste, Smell and Pain’. In R. Brown (ed.) Consciousness Inside and Out: Phenomenology, Neuroscience, and the Nature of Experience. Dordrecht: Springer, 237–98. Quine, W.V.O. 1969. ‘Epistemology Naturalized’. In Ontological Relativity and Other Essays. New York: Columbia University Press, 69–90.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

3 Bridging the Gap? Naïve Realism and the Problem of Consciousness Keith Allen

1. Introduction Central contemporary debates in the philosophy of perception seem to be in danger of reaching an impasse. This paper attempts to understand why this might be so, focusing on the debate between naïve realists and their opponents. Accounting for ‘what it is like’ to be a perceiving subject is one of the central phenomena that theories of perception are often expected to explain. The second section of the paper considers the naïve realist’s account of the phenomenology of perceptual experience, and the distinctive response to the Problem of Consciousness that this enables. The third section of the paper draws out one of the consequences of the naïve realist’s solution to the Problem of Consciousness, by considering the relationship of naïve realism to what many take to be a ‘fixed point’ in theorizing about perception: a commitment to physicalism. I argue there that the naïve realist’s solution to the Problem of Consciousness is inconsistent with a commitment to physicalism. In light of this, the final section of the paper considers in more detail the methodology that we should employ in deciding between philosophical theories of perception. I suggest that there is a way of thinking about naïve realism from a transcendental perspective that both helps to explain the motivation for naïve realism, and which also explains why the contemporary debate is in danger of reaching an impasse.

2. What Is to Be Explained? Naïve Realism and the Problem of Consciousness According to naïve realist or relationalist theories of perception, perceptual experiences are essentially relational: experiences ‘extend out’ into the world, and consist in the obtaining of a conscious relation of awareness or acquaintance between perceiving subjects and mind-independent objects and properties in their environment. Naïve realism differs in this respect from sense-datum theories, according to which perceptual experiences consist in a relation of acquaintance to Keith Allen, Bridging the Gap? Naïve Realism and the Problem of Consciousness In: Purpose and Procedure in Philosophy of Perception. Edited by: Heather Logue and Louise Richardson, Oxford University Press (2021). © OUP. DOI: 10.1093/oso/9780198853534.003.0003

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

44

 

something other than mind-independent objects and their properties (e.g. Price 1932; Jackson 1977), and from representationalist theories, according to which perceptual experiences represent the mind-independent environment as being a certain way (e.g. Harman 1990; Burge 2010). One of the most widely discussed motivations for naïve realism is that it provides an articulation and explanation of the qualitative, phenomenological character of perceptual experience—or in less technical terms, ‘what it is like’ to be a subject of perception (e.g. Martin 2002). In articulating and explaining the qualitative character of experience, a number of naïve realists have claimed that naïve realism thereby provides a distinctive way of approaching, and potentially resolving, the Problem of Consciousness (see e.g. Campbell 2002, 2014; Fish 2009; Kalderon 2011; Allen 2016, ch. 9). In broad terms, the Problem of Consciousness is the problem of explaining ‘what it is like’ to enjoy conscious experience.¹ The Problem of Consciousness is standardly understood to be a problem about explaining whether, and how, processes in the brain give rise to conscious experiences with their distinctive qualitative characters: for instance, seeing a red rose or smelling coffee. According to the naïve realist, however, standard attempts to explain the qualitative character look in the wrong place. For the naïve realist, the qualitative character of perceptual experience is not determined by processes in the brain; instead it is determined (at least in part) by the nature of the things in the subject’s environment that they are perceptually related to. If the naïve realist theory of perception is combined with a ‘naïve realist’ (or ‘primitivist’) theory of sensible properties like colour, according to which sensible properties like colours are qualitative mindindependent properties of things in our environment, then what it is like to perceive colour, for example, is determined (at least in part) by the nature of the colours we are consciously acquainted with. As Campbell evocatively puts the point, ‘the qualitative character of a colour experience is inherited from the qualitative character of the colour’ (1993, 189). This way of approaching the Problem of Consciousness is standardly dismissed because the Problem of Consciousness is normally understood against the background assumption that naïve realist theories of sensible qualities like colour are false: it is standardly assumed that there is nothing in our environment that could explain ‘what it is like’ to perceive colour or any other sensible qualities. If there is nothing in the mind-independent environment that could explain ‘what it is like’ to perceive colour, then it is reasonable to suppose that the problem must be one of understanding how processes in the brain, or in the physical body more generally, could give rise to conscious experiences. This background assumption

¹ It may be more accurate to speak about the Problems of Consciousness, given that there are a cluster of closely related problems about the nature of conscious experience in the vicinity, however I will set this aside here.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

̈        

45

is one that is typically shared by proponents of both reductivist and nonreductivist solutions to the Problem of Consciousness: both those who think that we can explain how processes in the brain give rise to (or are, indeed, identical to) conscious experiences, and those who think that we cannot but need instead to appeal to qualia, or conscious non-physical properties of experience. This background assumption presupposes a negative answer to what can be called the ‘Problem of Colour’: the problem of explaining whether, and how, the purely physical properties of things in our environment could give rise to qualitative properties like colours. That is, it is standardly assumed in advance that the purely physical properties of things in our environment do not give rise to qualitative properties like colours. Yet the Problem of Colour is structurally similar to Problem of Consciousness: both are problems about the relationship between the qualitative and the physical properties of things. We may therefore ask why standard ways of understanding the Problem of Consciousness presuppose an answer to the Problem of Colour? Why shouldn’t answers to the Problem of Colour instead presuppose an answer to the Problem of Consciousness—or at the very least, why shouldn’t we see both problems as manifestations of a single, underlying, problem, that need to be solved together, like the terms of a simultaneous equation?² These questions are pressing because the negative response to the Problem of Colour presupposed by standard approaches to the Problem of Consciousness problematizes the nature of consciousness and conscious experience. Consider, for example, a common non-reductivist response to the Problem of Consciousness that appeals to qualia: supposedly intrinsic, qualitative, properties of experience that are introduced to explain ‘what it is like’ to have conscious experience. There is arguably no introspective evidence for the existence of qualia so understood. Perceptual experience is transparent: when we reflect on our experiences, we seem to be aware of objects and their properties in our environment, but not any properties of our experience itself (e.g. Harman 1990, Martin 2002). Insofar as experience is transparent, the qualia theorist is guilty of bad faith: at the same time as emphasizing the importance of consciousness (in response to the reductivist), they simultaneously ignore the nature and structure of conscious experience. The qualia theorist will be guilty of bad faith in a second way if their appeal to qualia to resolve the Problem of Consciousness is in tension with their motivation for providing a negative response to the Problem of Colour in the first place. Although there are a number of reasons for thinking that there are no sensible qualities (as we perceive them) in the mind-independent environment, a common motivation for eliminating qualitative properties from the mind-independent ² Compare Shoemaker’s (2003) distinction between the ‘subjective’ and ‘objective’ explanatory gaps, and Byrne’s (2006) claim that both Problems are really versions of a more general problem about the relationship between the manifest and scientific images of the world.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

46

 

environment is broadly physicalist: it is that physics ‘tells us’ that there are no sensible qualities as we perceive them.³ But if physics ‘tells us’ that there are no sensible qualities, then it is tempting to suppose that it also ‘tells us’ that there are no qualia: qualia have no more place in modern scientific theories than sensible qualities do. At best, it looks as though qualia will need to be epiphenomenal, in which case it is difficult to see how they are able to fulfil the explanatory role for which they were introduced in the first place. Conversely, if physics doesn’t tell us that there are no qualia, then why should we accept that it tells us that there are no sensible qualities? Although it is not strictly inconsistent to suppose that physicalism might be true of the mind-independent environment but not conscious subjects, there is at least a tension here. It is important to emphasize the role of both the naïve realist theory of sensible qualities like colour and the naïve realist theory of perception in this approach to the Problem of Consciousness. According to naïve realist theories of colour, colours are mind-independent properties of things in our environment—objects, liquids, gases, light sources—that are distinct from the physical properties of things. The claim that colours are mind-independent guarantees that there are properties in the environment that can determine the qualitative character of colour experience. The claim that colours are distinct from the physical properties of things is needed because there are reasons for thinking that there are no physical properties of objects that could determine what it is like to perceive colour. So, for example, colours stand in distinctive relations of similarity and difference—for instance, orange is more similar to red than it is to blue—and admit of a distinction into unique and binary—every shade of orange, for example, is both reddish and yellowish, whereas there is a ‘pure’ instance of yellow that does not appear to be phenomenally composed in the same way. However, there are not obviously any physical properties that exhibit the same structural properties (see e.g. Allen 2016, chs. 4–6 for more detailed discussion of this and other reasons for rejecting colour physicalism). However, a naïve realist theory of colour is alone insufficient to resolve the Problem of Consciousness; the naïve realist theory of colour also needs to be combined with a naïve realist theory of perception. One way of bringing out the importance of the naïve realist theory of perception is by considering what Chalmers (2013) has called ‘the quality/awareness gap’: that properties like colours are one thing, and our awareness of them another. Given that colours are supposedly mind-independent, it is consistent with a naïve realist theory of colour that colours are never perceived—or at least, never perceived as they are. To say that colours are mind-independent is to say that their nature and existence does

³ For example, shortly before proposing the most famous version of the Knowledge Argument for the existence of qualia, Jackson argued that ‘Science forces us to acknowledge that physical or material things are not coloured’ (1977, 120).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

̈        

47

not depend upon the nature of our psychological responses; this has the consequence that colours would exist as they are even if there were no one to perceive them, or if our experiences of them were systematically misleading. So colours as the naïve realist conceives of them cannot by themselves explain what it is like to perceive colour. The relationality built into the naïve realist theory of perception provides a way of closing this gap, insofar as the act of awareness depends for its nature on that which it is awareness of. This rules out various kinds of systematic colour/ perception permutation: for instance, it rules out the possibility that yellow objects might systematically cause ‘red experiences’ (experiences with the same qualitative character as the experiences we have when we look at ripe tomatoes) (see Campbell 1993). Because perception is relational, and the qualitative character of the experience is determined by the nature of thing perceived, there is no further question about why experiences have the qualitative characters that they do. This need not be the only way of closing the quality–awareness gap. Forms of representationalist theory of perception can play a similar explanatory role— specifically, forms of representationalist theory according to which the qualitative character of experience is fully determined by the mind-independent objects and properties that the experience represents. But whereas there is no further question for the naïve realist about why experiences have the qualitative characters that they do once the object of the experience is fixed, the representationalist owes an explanation of why an experience’s representational content determines its qualitative character (e.g. Fish 2009, 78, n. 12; Logue 2012).⁴ More generally, a relational naïve realist theory of perception promises to articulate and explain what is arguably another key aspect of ‘what it is like’ to be a subject of experience. Experience is arguably not just transparent: it is not just that we are not aware of any properties of our experience, only objects and properties in our environment. Moreover, we are aware of objects and properties in perception in a distinctive way. In particular, perceptual experience appears to put us into ‘contact’ with the mind-independent world, and to ‘reach out’ to its objects. As Broad puts it, for example, visual perceptual experience is both ‘ostensibly saltatory’—it ‘seems to leap the spatial gap between the percipient’s body and a remote region of space’—and also ‘ostensibly prehensive’—it ‘seems to “bring one into direct contact with remote objects” ’ (1952, 5–6). According to the naïve realist, this aspect of the phenomenological character of experience reflects the fact that experience is essentially relational, and objects and ⁴ Some naïve realists think that fixing the object of experience does not fix the experience’s qualitative character, and that in part the qualitative character of the experience is determined by the ‘mode of perceiving’ (e.g. Logue 2012; see also Beck 2019). In this case, however, there may be a further question about why the object and the mode of perceiving fix the qualitative character of the experience.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

48

 

their properties are literally constituents of our experiences. By contrast, representationalists typically allow that experiences do not entail the existence of their objects, and so experiences of the same type could occur in the absence of the relevant mind-independent objects and properties (at least on a particular occasion). In this respect, they are seemingly unable to explain the sense in which perception puts us into ‘contact’ with the mind-independent world (see e.g. Fish 2009; Kalderon 2017; Allen 2019a and 2019b).⁵

3. Fixed Points? Naïve Realism and Physicalism Is the naïve realist’s solution to the Problem of Consciousness, encompassing a commitment both to a naïve realist theory of perception and sensible qualities, one that we should accept? One focus of the contemporary debate has been whether the naïve realist’s account of veridical perceptual experience is consistent with a credible account of hallucinations and illusions: mental events that are, at least in principle, subjectively indiscriminable from veridical perceptual experiences. This represents an important challenge to the intelligibility of the naïve realist’s account of what it is like to be a perceiving subject, and thereby their solution to the Problem of Consciousness. However, this issue has been extensively discussed elsewhere, so here I want to consider a challenge from a different angle: that naïve realists are committed to rejecting what is often considered to be a ‘fixed point’ in theorizing about perception, namely a commitment to physicalism (see e.g. Gert and Gow, this volume). The naïve realist’s solution to the Problem of Consciousness requires accepting a theory of perception according to which perceptual experiences consist in the obtaining of a conscious relation of awareness between perceiving subjects and objects in their environment (inter alia). Is this consistent with a commitment to physicalism: the thesis that everything that exists is physical? The naïve realist about sensible qualities thinks that mind-independent objects and their properties are not themselves purely physical. In holding that colours are properties of things in our environment that are distinct from their physical properties, the naïve realist accepts a non-reductive theory of colour. Depending on exactly how this view is developed, this may or may not qualify as a form of ⁵ Gert (this volume) argues from a neo pragmatist perspective that it is possible to explain the kinds of phenomenological descriptions used to motivate naïve realism without making substantive metaphysical commitments, simply by appealing to the ways in which the relevant kinds of language are used. One concern with neo pragmatist theories is whether they deliver the thoroughgoing metaphysical deflationism that they promise; for instance, Gert himself makes the seemingly metaphysically substantial claim that perceptual experiences are adverbial modifications. A more general concern is that neo pragmatist theories of this kind fail to provide metaphysical explanations where metaphysical explanations are required (see Allen 2016, 163–6 for related discussion of a neo pragmatist theory of sensible qualities).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

̈        

49

‘non-reductive physicalism’. Non-reductive physicalists hold that unreduced entities supervene on physical properties, and do so in a physicalistically explicable way: for instance, because unreduced entities stand in a relation of something like partial identity to their physical realizers, by virtue of sharing causal powers (e.g. Yablo 1995; Watkins 2005). Some naïve realists may reject the claim that colours supervene on the physical properties of objects, allowing that there can be differences in the colours of things without differences in their physical properties; and even if the naïve realist accepts that colours supervene on the physical properties of objects, they might deny that the obtaining of the supervenience relation is physicalistically explicable (e.g. Allen 2016). What about the relation of conscious awareness? Because perceptual experiences are essentially relational, and consist in the obtaining of a relation of awareness between perceiving subjects and mind-independent objects and their properties, naïve realism is inconsistent with the local supervenience of perceptual experiences on brain states: the neural state of the subject is not alone sufficient to fix the experiential facts, so there can be a difference in experience without an intra-dermal physical difference. But denying that experiences supervene locally on brain states is consistent with a global supervenience thesis, according to which a minimal physical duplicate of the world is a duplicate of the world simpliciter, and so according to which there can be no perceptual difference without a physical difference more generally. This raises two questions. First, should the naïve realist accept a global supervenience thesis? Second, is this sufficient for a commitment to physicalism? In response to the first question, there are good reasons for naïve realists to accept a global supervenience thesis. A principal reason for accepting a global supervenience thesis is that it enables the naïve realist to accept the thesis that Martin calls ‘Experiential Naturalism’. This is the thesis that experiences ‘are themselves part of the natural order, subject to broadly physical and psychological causes’ (Martin 2006, 357)—and which also have characteristic physical and psychological effects, as when we engage in perceptually guided action or form perceptually based beliefs. The naïve realist’s commitment to Experiential Naturalism is threatened by a version of the Causal Exclusion Argument, familiar from discussions of mental causation. If the physical domain is causally complete, then the physical and psychological effects of perceptual experiences, as distinct from their neural enabling conditions, will be systematically overdetermined. This overdetermination is liable to seem problematic if the causes are distinct and independent; the regular co-occurrence of causes will seem like a massive ‘cosmic coincidence’, analogous to a situation in which someone dies by being shot at precisely the moment that they hit the ground having fallen from a fourth-floor window. But this systematic overdetermination of effects by causes is far less problematic if the causes are distinct but not independent: if the perceptual facts are dependent on the physical facts, then this kind of systematic relationship

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

50

 

between the physical and the perceptual is precisely what we should expect. A global supervenience thesis allows the naïve realist to maintain Experiential Naturalism in the face of the Causal Exclusion Argument, because it allows the naïve realist to deny that perceptual experiences and their neural enabling conditions are independent causes.⁶ Second, it has been suggested that accepting a global supervenience thesis is sufficient for accepting a minimal form of physicalism (e.g. Jackson 1998), in which case a naïve realist theory that accepted a global supervenience thesis would thereby count as a form of minimal physicalism. However, it is tempting to think that supervenience theses do not themselves explain, but rather require explanation; and unless an explanation of the global supervenience thesis that is acceptable from a physicalist perspective is forthcoming, then simply accepting a global supervenience thesis will not of itself be sufficient for naïve realism to qualify as a form of physicalism (e.g. Horgan 1993; Gow this volume). Moreover, it is far from clear that a naïve realist will be in a position to meet this explanatory demand. One way of pressing this concern is by considering the apparent conceivability of ‘naïve realist zombies’: subjects with the same perceptual processing mechanisms as ordinary conscious subjects, who are embedded in the same environment, but who do not stand in conscious relations of acquaintance to things in their environment (Pautz 2013). The naïve realist who accepts a global supervenience thesis is committed to rejecting the possibility of naïve realist zombies. But whilst inferences from (apparent) conceivability to possibility are notoriously problematic, it is not entirely clear what the grounds for rejecting this possibility are. At best, it seems that the naïve realist will have to argue against the possibility of this kind of situation indirectly. Fish (2009: 75–9; 2013), for instance, argues on general grounds that functional duplicates of conscious subjects will be conscious or that various kinds of functional capacities imply consciousness. If successful, these arguments establish that naïve realist zombies are impossible, but without explaining why perceptual processing gives rise to conscious experience. In effect, the naïve realist provides an account of what it is like to perceive colour only on the assumption that there is something it is like to perceive colour in the first place. But that there is something it is like to perceive in the first place is not itself something that the naïve realist offers a substantive explanation of.⁷ At this point, we might start to wonder whether progress has been made. Although the naïve realist can accept the global supervenience of sensible qualities

⁶ Not all naïve realists (or fellow travellers) accept Experiential Naturalism (see e.g. White 1961; Hyman 1992; Stoneham 2008). But even if naïve realists reject Experiential Naturalism, the commitment to a global supervenience thesis distinguishes the naïve realist from the dualist, and represents a commitment, if not to a (broad) form of naturalism, then at least to a form of non-super-naturalism. ⁷ Thanks to Brian Cutter and John Schwenkler for comments on ‘Colour and the Problem of Consciousness’. http://philosophyofbrains.com/2017/01/13/colour-and-the-problem-of-consciousness. aspx.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

̈        

51

and relational experiences on the physical, they are not able to provide a deep explanation of why this supervenience relation holds. As such, naïve realism is not plausibly consistent with a commitment to physicalism. So is the naïve realist theory of perception, and the resolution of the Problem of Consciousness it enables, really any more intelligible than its competitors? Indeed, is it sufficiently intelligible at all?

4. Methodology? Metaphysical, Descriptive, and Transcendental The feeling that contemporary debates about the metaphysics of perception are in danger of reaching an impasse can be thought of as a ‘meta-philosophical phenomenological datum’. Just as one desideratum of philosophical theories of perception is that they articulate and explain the phenomenological data, one possible desideratum of meta-philosophical reflection on philosophical theories of perception is that it explains the meta-philosophical phenomenology. With this in mind, the final section considers in more detail the methodology employed in contemporary philosophy of perception, and how exactly we should understand the aims and nature of naïve realism. I will argue that the impasse can be explained—and perhaps avoided—if we understand naïve realism as a particular kind of theory, that is part of a particular kind of philosophical project.

4.1 Metaphysical Naïve Realism Naïve realism is often presented as one theory of the metaphysics of perception amongst others. On this way of thinking of naïve realism, it is a theory of the same kind as, and so a direct competitor to, theories like the sense-datum theory and representationalism. This way of thinking of naïve realism is naturally allied to a particular methodology of philosophical theory choice. On this approach, philosophical theories are motivated as inferences to the best explanation of some relevant range of phenomena, and determining which theory provides the best explanation of these phenomena involves engaging in a cost–benefit analysis of the competing views. So, for instance, philosophical theories of perception might be evaluated based on how faithful they are to the appearances, how well they fit with our best scientific theories, how simple or systematic they are, or more generally how well they cohere with other theoretical commitments.⁸

⁸ For further discussion of the use of inference to the best explanation in the philosophy of perception, see Drayson this volume.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

52

 

If philosophical theories of perception are motivated and assessed in this way, then they can be shown to be false. Indeed, in the case of naïve realism, this might not seem to be a merely idle possibility. On balance, we might wonder whether naïve realism really does provide the best explanation of the relevant range of phenomena. I argued in Section 2 that naïve realism provides a compelling account of ‘what it is like’ to be a perceiving subject. But as I argued in Section 3, this response to the Problem of Consciousness encounters theoretical pressure elsewhere. In addition to the challenge of accounting for illusion and hallucination, it requires us to reject what many consider to be a ‘fixed point’ in philosophical theorizing about perception: a commitment to physicalism. The naïve realist seems unable to provide a deep explanation of why colours and other sensible properties supervene on the physical properties of things in our environment; nor does it seem likely that the naïve realist will be in a position to provide a deep explanation of why processing in the brain gives rise to conscious perceptual experience. In light of this, it might not seem very promising for the naïve realist to understand naïve realism as simply one theory of the metaphysics of perception amongst others, to be assessed on the basis of a cost–benefit analysis. At the same time, this may also give the meta-philosopher of perception pause. Understanding naïve realism as one metaphysical theory of perception amongst others raises a question about whether we can explain the meta-philosophical phenomenological datum that the contemporary debate is in danger of reaching an impasse. If, on balance, it is fairly clear that naïve realism is not the best explanation of the relevant range of perceptual phenomena, then what explains the disagreement? It might be that the disagreement is merely transitory. It is possible that the current interest in naïve realist theories of perception will provide the opportunity for the philosophical community to consider in detail the view and its consequences, and the result of this will be a better appreciation of its costs and a consequent rejection of it. Whether this will transpire is difficult, if not impossible, to determine from the temporal vantage point that we occupy, within the ongoing debate (see Fish this volume). But one reason for cautious scepticism about this is the historical resilience of naïve realism. Despite having been widely rejected at various points in the history of different philosophical traditions, versions of it have had a tendency to re-emerge: compare, for instance, Austin (1962)’s response to twentieth-century sense-datum theorists, Putnam (1999) in the pragmatist tradition, Merleau-Ponty (1945) in the phenomenological tradition (Allen 2019a), as well as more recent contemporary forms of analytic naïve realism developed by Campbell (e.g. 2002) and Martin (e.g. 2002). A different possible explanation of the impasse is sociological. Contemporary naïve realism in the analytic philosophy has its roots, at least in part, in the ‘Oxford Realist’ tradition (see Travis and Kalderon 2013). It is possible to tell a story about the subsequent development and geographical spread of the view via

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

̈        

53

academic genealogies that trace back to figures in this tradition, with key figures responsible for the wider dissemination of the view including Campbell and Martin. But this cannot fully explain the uptake of the view, as not everyone who accepts some form of naïve realism in the contemporary analytic debate is a product of those institutions where naïve realism has taken root; nor does this explain the emergence of the view at other times and in other philosophical traditions. Besides, if this kind of sociological explanation is successful, it has much wider scope than explaining the acceptance of naïve realism. There will be similar lines of transmission for other philosophical theories, including sensedatum and representationalist theories. The current disagreement in the philosophy of perception might simply be an instance of a broader phenomenon. Disagreement is a pervasive feature of philosophical discourse, and many philosophical disagreements appear to exhibit the property of ‘intransigence’: that parties to the dispute maintain their views in the face of disagreement from others and are not considered irrational for doing so (Allen 2019b). It therefore might not be distinctive of contemporary philosophy of perception that it is in danger of reaching an impasse; the problem may be more general. Without trying to rule this possibility out, it is nevertheless worth considering in more detail why the disagreement in the philosophy of perception is intransigent, if it is. It might turn out that disagreements in different areas of philosophy are importantly different; alternatively, it might be that wider morals about philosophical disagreement can be drawn from considering the specific case of the philosophy of perception.

4.2 Descriptive Naïve Realism Rather than understanding naïve realism as one theory of the metaphysics of perception amongst others, a different way of understanding the theory is as a descriptive phenomenological, conceptual, or linguistic theory: a theory of how we ordinarily experience, think, or talk about perception. This would at least go some way towards explaining the impasse that the current debate is in danger of reaching. If sense-datum and representationalist theories are theories of the metaphysics of perception, but naïve realism is a descriptive theory about how we experience, think, or talk about perception, then the different theories would simply be talking past each other. However, this diagnosis is also problematic. First, it is not clear that this is how naïve realism is generally understood, at least in the contemporary analytic debate. Indeed, it isn’t obviously best understood this way even by those naïve realists who more clearly invite this kind of meta-philosophical interpretation. By way of illustration, something like a naïve realist theory is expressed in Austin (1962). Austin was an ordinary language philosopher, who describes his philosophical

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

54

 

approach as a kind of ‘linguistic phenomenology’, which examines ‘what we should say when, what words we should use in what situations’ (1957, 8). But he insists that this is not a purely descriptive linguistic project, but rather: we are looking again not merely at words (or ‘meanings,’ whatever they may be) but also at the realities we use the words to talk about: we are using a sharpened awareness of words to sharpen our perception of, though not as the final arbiter of, the phenomena.⁹

Second, and more importantly, it is tempting to think that understood as a merely descriptive thesis, naïve realism isn’t sufficiently interesting. Phenomenological, conceptual and linguistic analysis need not be easy, and so the claim that naïve realism is enshrined in the phenomenology, or in our ordinary thought and talk need not be a trivial thesis. But nevertheless, as theoreticians we are likely to want more than this; in addition, we are likely to want a theory of what perception is.

4.3 ‘Default’ Metaphysical Naïve Realism The metaphysical and descriptive approaches can be combined in various ways. One way of doing this is to treat naïve realism, like the sense-datum and representationalist theories, as a theory of metaphysics of perception, but argue that it enjoys a ‘default’ status in the debate precisely because it is the theory that best captures the phenomenological, conceptual, and/or linguistic appearances. To say that naïve realism is the ‘default theory’ is to say that it is the theory we should accept unless there are compelling reasons to reject it; this approach accords respecting the appearances a special privilege in deciding between competing philosophical views, giving them a much greater weight in the cost–benefit analysis. Thinking of naïve realism in this way has the potential to explain the metaphilosophical phenomenological datum. On the one hand, there is likely to be disagreement about what the appearances to be captured are, and exactly how to settle this. On the other hand, and more fundamentally, the naïve realist and their opponent are liable to disagree about whether the appearances should be accorded a particular privilege in adjudicating between philosophical theories of perception in the first place. If this is right, then the impasse will be generated, at least in part, by a meta-philosophical disagreement about the status of the evidence to which a

⁹ Similarly, although Merleau-Ponty says that the phenomenological approach ‘involves describing, and not explaining or analyzing’ (1945, lxxi), even in Phenomenology of Perception (and more clearly in later work) he is naturally understood as making metaphysical claims about perception and perceiving subjects. See e.g. Ward (this volume).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

̈        

55

philosophical theory of perception ought to be responsive. The sense-datum theorist or representationalist may allow that fidelity to the appearances carries some weight in deciding between philosophical theories. But they will nevertheless disagree with the naïve realist that capturing the appearances represents any particular advantage, and so ought to carry a greater weight in the cost–benefit analysis. Indeed, they may think that it represents at best a very weak, easily defeasible, consideration—or even that it provides no good reason to accept the theory at all. Is there a compelling reason for thinking that capturing the appearances ought to be accorded a particular privilege in deciding between competing philosophical theories of perception? One suggestion is that capturing the appearances stops us, as philosophers of perception, from ‘changing the subject’. In the philosophy of perception, our grip on the phenomena that we are attempting to theorize about is fixed by our first-personal experience as perceiving subjects, along with ordinary patterns of thought and talk about perception. If we end up with a theory that is at significantly at odds with the phenomenological character of perception, or with ordinary thought or talk about it, then there is a concern that we have not provided a theory of perception at all—but instead a theory of something else. The idea that what motivates the naïve realist is a concern about losing our grip on the perceptual phenomena is suggestive, and I will return to a different way of spelling this out below. The difficulty with this suggestion as it stands, however, is that from a disinterested, theoretical perspective the concern is not obviously compelling. After all, many of the competitors to naïve realism are not radically revisionary of the appearances, and share with it core theoretical commitments: for instance, the sense-datum theory shares the naïve realist’s commitment to a conscious relation of acquaintance, while representationalism shares the naïve realist’s commitment to the claim that the objects of perception are mindindependent objects in the external environment. Besides, we are in general prepared to accept that theoretical considerations will sometimes lead us to accept that appearances are deceptive: for instance, that even though whales may look like fish, they are really mammals. Indeed, we are often prepared to believe that the appearances can be systematically deceptive: for instance, that experiences as of ghosts are illusory and that beliefs and talk about ghosts and other supernatural phenomena are systemically false. So why couldn’t it turn out that the same is true of our ordinary attitudes towards our perceptual contact with the world?

4.4 Transcendental Naïve Realism Instead of thinking of naïve realism as enjoying a ‘default’ status in the debate about the metaphysics of perception, a different way of combining elements of the metaphysical and descriptive approaches is provided by a position—or perhaps

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

56

 

more accurately, family of positions—that can be called transcendental naïve realism. Transcendental naïve realism involves adopting a specific kind of metaphilosophical attitude towards naïve realism; it involves regarding naïve realist theories of colour and perception from a ‘transcendental standpoint’. There are different forms that transcendental naïve realism can take, but in broad terms, the transcendental naïve realist does not see naïve realism simply as one philosophical theory of perception amongst others, to be defended by inference to the best explanation, and to be assessed on the basis of a cost–benefit analysis that weighs performance along a number of different dimensions—even allowing that some of these dimensions might be more heavily weighted than others. Rather, viewed from the transcendental perspective, naïve realism has a special status amongst philosophical theories; indeed, part of the promise of adopting the transcendental approach is that it renders naïve realism immune to falsification. On this approach, the naïve realist theory is part of a distinctive philosophical project: the transcendental project of explaining how it is possible that perceptual experience has the distinctive characteristics that it does.¹⁰ A characteristic feature of this philosophical project is the use of transcendental arguments. Transcendental arguments start by identifying some relevant psychological fact about our experience or mental states. This provides the basis for identifying some further fact, the holding of which is claimed to be necessary for the initial psychological fact to be the case. By identifying ‘conditions for the possibility’ of experience having the distinctive characteristics it does, transcendental arguments are central to the project of explaining how it is possible that experience is the way it is. This provides one model for understanding the argument of Section 2. It is not that naïve realism is (merely) the best explanation of ‘what it is like’ to be a subject of experience. Rather, naïve realism provides an account of how it is possible that conscious perceptual experience puts us into contact with mind-independent objects and their qualitative properties: perceptual experience could only put us into contact with qualitative properties of objects if perceptual experience is essentially relational. Indeed, it is notable that other arguments for naïve realism found in the recent literature can also be (re-)formulated as transcendental arguments. So, for instance, consider the claims that perceptual experience is transparent, that we are able to refer to, and think about, objective particulars (e.g. Campbell 2002), or that we can have knowledge of the nature and existence of the external world (e.g. Logue 2012). Naïve realism could be understood as providing (merely) the best explanation of these psychological facts from amongst the available options. Alternatively, naïve realism can be understood as the conclusion of a transcendental argument: the relationality of perceptual ¹⁰ See Ward (this volume) for further discussion of the different forms that this kind of transcendental project can take.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

̈        

57

experience is a condition of the possibility of experience being transparent, enabling our ability to refer to, and think about, objective particulars, or grounding our knowledge of the nature and existence of the external world. Given that contemporary naïve realism has roots in neo-Kantian Oxford Realism, interpreting these arguments in this way may not be entirely unmotivated.¹¹ Understanding naïve realism as a transcendental theory of perception suggests an explanation of the meta-philosophical phenomenological datum. If naïve realists are at least implicitly engaged in a different type of philosophical project to proponents of other theories of perception, then this might explain why the contemporary debate appears to be in danger of reaching an impasse. The naïve realist is not doing something other than giving an account of the metaphysics of perceptual experience; in this respect, transcendental naïve realism differs from descriptive naïve realism. But unlike standard metaphysical theories of perception, the transcendental naïve realist does not take seriously the possibility that the appearances might be misleading. The explanatory project that the transcendental naïve realist is engaged in takes the description of the appearances as a fixed point in philosophical theorizing, and aims to show how it is possible that the appearances are as they are. Is this a promising way to understand naïve realism? Whether transcendental methodology is viable—and how exactly transcendental approaches differ from other philosophical approaches—has been, and continues to be, the subject of debate. I want to briefly consider two particular problems for this approach. The first problem focuses on the starting point for the transcendental argument, and can be stated in the form of a dilemma. Transcendental arguments typically attempt to use a premise that that all parties to the debate will accept, in order to derive the conclusion that a particular view is correct.¹² In general, the dilemma facing proponents of transcendental arguments is that either the fixed point is genuinely uncontroversial, but in which case it doesn’t provide a way of deciding between competing theories; or the fixed point provides a way of deciding between competing theories, but in which case it isn’t uncontroversial. Applying this to the argument of Section 2, for instance, it might be suggested that we only get an argument for a relational theory like naïve realism if we assume that perceptual experience puts us into touch-like contact with qualitative properties of the mind-independent environment. But that description of the ¹¹ This way of understanding Campbell’s version of the epistemic argument for naïve realism, for instance, is suggested by claims like ‘I will argue that if we are to acknowledge the explanatory role of experience of objects, we have to appeal to what I call a Relational View of experience’ (2002, 114, emphasis added). In the case of the transparency of experience, Martin (2002) argues that this argument is not sufficient for naïve realism, and so presents the argument from sensory imagination instead; but this could itself be interpreted as a transcendental argument that takes as its premise the Dependency Thesis, that to imagine sensorily an x is to imagine experiencing an x. ¹² Given the traditional role of transcendental arguments in combatting scepticism, the relevant interlocutor is often the sceptic.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

58

 

appearances is contentious, and is probably not one that everyone would accept. Suppose, then, that we weaken the characterization of the appearances, so that the starting point is that when we perceive, our experience is as of qualitative properties of the mind-independent environment. This assumption should be acceptable to the naïve realist’s opponents. But insofar as it is, it does not provide an argument for naïve realism, since the naïve realist’s opponents can provide alternative explanations of how this is possible: for instance, we become aware of qualitative properties of mind-independent objects by virtue of being aware of sense-data, or else by representing mind-independent objects in our environment as being a certain way. Describing the phenomenology is notoriously problematic.¹³ Indeed, difficulties describing how things appear in a way that is generally acceptable may itself go at least some way towards explaining the impasse in contemporary philosophy of perception. But this is not a problem that is necessarily unique to the transcendental approach. If naïve realism is motivated as the outcome of an inference to the best explanation, then the naïve realist who argues that naïve realism provides the best explanation of the phenomenological character of experience needs a description of the phenomenology. The same also applies to proponents of other theories of perception, assuming that they motivate their views, at least in part, by appealing to the appearances—which often they do. Setting this problem aside—at least for the sake of argument—suppose that we can agree on a description of the phenomenology. The second challenge for the transcendental naïve realist is to explain what underwrites the inference in the transcendental argument from the psychological to the non-psychological: from the fact that our experience appears to put us into touch-like contact with the world, to the fact that experience really consists in a relation of conscious awareness to qualitative properties of a mind-independent world. This is an instance of the general problem facing proponents of transcendental arguments, of justifying the inference from how we experience, think or talk about the world, to how the world really is (e.g. Stroud 1968). For Kant, the solution to this challenge involved combining an investigation into the subjectively determined forms of experience with a form of transcendental idealism, according to which the empirical world is itself constituted by transcendental subjects. It may be possible to embed a naïve realist theory of perception within this framework, and a number of commentators have argued that this was in fact Kant’s view (e.g. Allais 2015; Gomes 2017). Subsequent proponents of transcendental approaches, however, have sought to divorce the general approach from transcendental idealism, which is liable to be seen as involving obscure, unattractive, and unnecessary metaphysical claims.

¹³ An interesting case study of this is whether green is an elementary colour, instances of which are ‘unique’. See Allen (2016) for discussion.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

̈        

59

One alternative is to reconceive of the ambitions of transcendental arguments. Transcendental arguments that aim to identify some further non-psychological fact, the holding of which is necessary for an initial psychological fact to be the case, are called ‘world-directed’. But a less ambitious form of transcendental argument is ‘self-directed’. These aim only to identify some further psychological fact that is necessary for an initial psychological fact to hold. So, for instance, instead of seeking to argue from the fact that perception appears to put us into contact with qualitative properties of objects to the fact perception is relational, a more modest transcendental argument might try to argue to the conclusion that we believe that perception is relational. However, although this form of transcendental argument avoids the problem of justifying the inference from the psychology to the non-psychological, it essentially collapses back into a version of the descriptive approach considered, and dismissed, above. The second challenge to transcendental naïve realism can be understood as putting pressure on the idea that transcendental arguments alone are sufficient to establish naïve realism. Indeed, there are other reasons for suspecting that what is distinctive about the transcendental approach is not that it employs transcendental arguments, as opposed to inferences to the best explanation; after all, transcendental arguments can be understood as limiting cases of inference to the best explanation, in which the purported explanation is the best by virtue of being the only explanation. The underlying disagreement therefore may not be so much the use of a distinctive form of argument, but a refusal to base philosophical theory choice on a cost–benefit analysis, where it is in principle possible for the benefits of a view that respects the appearances to be outweighed by the costs. What distinguishes the transcendental naïve realist from the ‘default’ metaphysical naïve realist—and from other parties to the debate about the metaphysics of perception more generally—would be that the transcendental naïve realist does not take seriously the possibility that perceptual experience might be other than it appears. The key question is therefore whether this is a reasonable approach— particularly if some form of transcendental idealism is rejected, which would otherwise guarantee that experience is as it appears. I want to conclude by suggesting one answer to this question: that this is a reasonable approach insofar as the truth of naïve realism is something that—as philosophers of perception who are also perceiving human subjects—we do, or should, care about. To motivate this idea, consider a comparison to Strawson’s (1962) famous transcendental response to the apparent conflict between determinism, the thesis that all our actions are causally determined, and the reactive attitudes we experience when we engage in interpersonal relationships: for instance, gratitude, resentment, forgiveness, love, and hurt feelings.¹⁴ Strawson argues that determinism (at least as it often understood) is, in some rational sense, an intelligible

¹⁴ For a more detailed presentation of the analogy, see Allen 2020.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

60

 

philosophical doctrine. We encounter ‘bad cases’, where we regard people’s actions as merely causally determined: for instance, very young children and the severely mentally ill. Determinism is intelligible to the extent that we are able to adopt towards everyone the same detached, objective attitude that we adopt in bad cases: we understand what it is for action to be causally determined, so we understand (in some sense) what it would be for all action to be causally determined. But whilst Strawson thinks that determinism is not selfcontradictory, and indeed there may even be some general truths that could be theoretical grounds for it, he nevertheless thinks that the thesis is not practically intelligible. As he puts it, the belief in determinism: is, for us as we are, practically inconceivable. The human commitment to participation in ordinary inter-personal relationships is, I think, too thoroughgoing and deeply rooted for us to take seriously the thought that a general theoretical conviction might so change our world that, in it, there were no longer any such things as inter-personal relationships as we normally understand them. (1962, 12)

The suggestion that I want to make is that something similar is true of naïve realism. The naïve realist can concede that there is some rational sense in which alternatives to naïve realism are intelligible. We know that there are ‘bad cases’, in which experiences do not put us into contact with the world: this is what happens in cases of hallucination and perhaps illusion. The falsity of naïve realism is rationally intelligible to the extent that we can adopt towards all our experiences the same detached, objective attitude that we adopt to the bad cases: we understand what it is for experiences not to put us in contact with the world, so we understand (in some sense) what it would be for no experiences to put us into contact with the world. These views are not self-contradictory, and there may even be theoretical grounds for them. But nevertheless, they lack practical intelligibility. Naïve realism strikes a deep chord with us, at least in part because it forces itself upon us at every waking moment. Even in the study, when we are doubting the truth of naïve realism, we are thrown into the world through our perception of it. Understood in this way, the motivation for naïve realism is visceral, even almost moral. In this practical, lived sense, naïve realism is ‘intelligible’ in a way its competitors—common kind theories of perception, naturalistic theories, and traditional non-naturalistic theories—are not. If this is what really underlies the attraction of naïve realism, then it is perhaps unsurprising that the debate in contemporary philosophy of perception is in danger of reaching an impasse.¹⁵

¹⁵ Earlier versions of this paper were presented in Winnipeg, Fribourg, Budapest, and UCL. I would thank audiences on these occasions for their comments and questions, as well as Louise Richardson and Heather Logue for their comments on a written version of the paper.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

̈        

61

References Allais, L. 2015. Manifest Reality: Kant’s Idealism and His Realism. Oxford: Oxford University Press. Allen, K. 2016. A Naïve Realist Theory of Colour. Oxford: Oxford University Press. Allen, K. 2019a. ‘Merleau-Ponty and Naïve Realism’. Philosophers’ Imprint 19: 1–25. Allen, K. 2019b. ‘Should We Believe Philosophical Claims on Testimony?’ Proceedings of the Aristotelian Society 119: 105–25. Allen, K. 2020. ‘The Value of Perception’. Philosophy and Phenomenological Research 100: 633–56. Austin, J.L. 1957. ‘A Plea for Excuses: The Presidential Address’. Proceedings of the Aristotelian Society 57: 1–30. Austin, J.L. 1962. Sense and Sensibilia. Oxford: Oxford University Press. Beck, O. 2019. ‘Rethinking Naïve Realism’. Philosophical Studies 176: 607–33. Broad, C.D. 1952. ‘Some Elementary Reflexions on Sense-Perception’. Philosophy 27: 3–17. Reprinted in R.J. Swartz (ed.), Perceiving, Sensing and Knowing. New York: Anchor Books, 29–48, 1965. Burge, T. 2010. Origins of Objectivity. Oxford: Clarendon Press. Byrne, A. 2006. ‘Color and the Mind-Body Problem’. dialectica 60: 223–44. Campbell, J. 1993. ‘A Simple View of Colour’. In John J. Haldane and C. Wright (eds.) Reality: Representation and Projection. Oxford: Oxford University Press, 257–68. Campbell, J. 2002. Reference and Consciousness. Oxford: Oxford University Press. Campbell, J. 2014. ‘A Straightforward Solution to Berkeley’s Puzzle’, in J. Campbell and Q. Cassam, Berkeley’s Puzzle. Oxford: Oxford University Press. Chalmers, D. 2013. ‘Panpsychism and Panprotopsychism’. The Amherst Lecture in Philosophy 8: 1–35. http://www.amherstlecture.org/chalmers2013/. Fish, W. 2009. Perception, Hallucination, and Illusion. New York: Oxford University Press. Fish, W. 2013. ‘Perception, Hallucination, and Illusion: Reply to my Critics’. Philosophical Studies 163: 57–66. Gomes, A. 2017. ‘Naïve Realism in Kantian Phrase’. Mind 126: 529–78. Harman, G. 1990. The Intrinsic Quality of Experience, in J. Tomberlin (ed.) Philosophical Perspectives 4. Atascedero: Ridgeview. Horgan, T. 1993. ‘From Supervenience to Superdupervenience: Meeting the Demands of a Material World’. Mind 102: 555–86. Hyman, J. 1992. ‘The Causal Theory of Perception’. The Philosophical Quarterly 42: 277–96. Jackson, F. 1977. Perception: A Representative Theory. Cambridge: Cambridge University Press. Jackson, F. 1998. From Metaphysics to Ethics. Oxford: Oxford University Press.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

62

 

Kalderon, M. 2011. ‘The Multiply Qualitative’. Mind 120(478): 239–62. Kalderon, M. 2017. Sympathy in Perception. Cambridge: Cambridge University Press. Logue, H. 2012. ‘Why Naive Realism?’. Proceedings of the Aristotelian Society 112: 211–37. Martin, M.G.F. 2002. ‘The Transparency of Experience’. Mind and Language 17: 376–425. Martin, M.G.F. 2006. ‘On Being Alienated’. in T.S. Gendler and J. Hawthorne (eds.) Perceptual Experience. Oxford: Oxford University Press. Merleau-Ponty, M. 1945/2012. Phenomenology of Perception, trans. D. Landes. London: Routledge. Pautz, A. 2013. ‘Do the Benefits of Naïve Realism Outweigh the Costs? Comments on Fish, Perception, Hallucination and Illusion’. Philosophical Studies 163: 25–36. Price, H.H. 1932. Perception. London: Methuen. Putnam, H. 1999. The Threefold Cord: Mind, Body and World. New York: Columbia University Press. Shoemaker, S. 2003. ‘Content, Character, and Color’. Philosophical Issues 13: 253–78. Stoneham, T. 2008. ‘A Neglected Account of Perception’. dialectica 62: 307–22. Strawson, P.F. 1962/2008. ‘Freedom and Resentment’. Reprinted in Freedom and Resentment and Other Essays. London: Routledge. Stroud, B. 1968. ‘Transcendental Arguments’. Journal of Philosophy 65: 241–56. Travis, C. and Kalderon, M. 2013. ‘Oxford Realism’. In M. Beaney (ed.) Oxford Handbook of the History of Analytic Philosophy. Oxford: Oxford University Press, 489–517. Watkins, M. 2005. ‘Seeing Red: The Metaphysics of Colour Without the Physics’. Australasian Journal of Philosophy 83: 33–52. White, A.R. 1961. ‘The Causal Theory of Perception’. Proceedings of the Aristotelian Society 35: 153–168. Yablo, S. 1995. ‘Singling Out Properties’. Philosophical Perspectives 9: 477–502.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

4 Experiential Pluralism and Mental Kinds Maja Spener

1. Introduction We perceptually experience the world in various ways. Visual experiences are perceptual experiences which arise in exercises of one of our perceptual capacities, namely vision. These perceptual experiences essentially involve conscious visual presentation of the world: when undergoing visual experiences, it is visually for the subject as if things are presented to them. This paper is about the nature of visual experience and how to argue about it.¹ At its centre is a new argument for the view that there is kind of visual experience that is world-involving. The argument draws on the explanatory role of experience vis-à-vis certain aspects of our agency—specifically, certain types of abilities we routinely attribute to ourselves and others in ordinary situations. Before setting out the argument, let me first introduce some key notions and outline the background of the relevant debate in the philosophy of perception. It is common to recognize the following distinction. Some visual experiences are part of seeing the world, in that they figure in a cognitively successful relation to the world where you perceive the world as it really is. Some of them fail to do so. Call the first sort of case ‘good case’, the other sort ‘bad case’. My visual experience as of the carob tree in the courtyard is part of a good case if it is part of my seeing the carob tree and the courtyard, but it would be a bad case if it looks to me that way when there are no carob trees within appropriate visual range. Mixed cases are cognitively successful in some respects but not in others. Mere recognition that there are good and bad cases does not involve any commitment about whether the visual experiences at the centre of them are of different mental or psychological kinds in some substantial sense.² Both the following views are compatible with the distinction:

¹ My paper focuses exclusively on visual experience, but I will drop the qualifier ‘visual’ at times. ² The category of mental kinds is very broad, including a variety of psychological aspects. Standard candidates are phenomenal character and various types of content. Maja Spener, Experiential Pluralism and Mental Kinds In: Purpose and Procedure in Philosophy of Perception. Edited by: Heather Logue and Louise Richardson, Oxford University Press (2021). © OUP. DOI: 10.1093/oso/9780198853534.003.0004

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

64

 

Experiential monism: nature.

Visual experience has a single common psychological

Experiential pluralism: It is not the case that visual experience has a single common psychological nature. Experiential monists hold that good and bad cases involve visual experiences which are of the same basic mental kind. According to them, what makes for a given case being good or bad resides in factors that are external to the visual experiences (typically in their causal aetiology and certain counterfactual conditions); there is nothing that distinguishes the visual experience as good qua significant mental kind. Experiential pluralists hold that good and bad cases involve visual experiences which differ significantly in their mental nature. On the basis of this, they can argue that the cognitive success or failure of the good and bad cases, respectively, in which such experiences figure, is due to that difference. They hold that there is something that distinguishes the visual experience as good qua mental kind.³ These positions come in different strengths. According to extreme versions of experiential pluralism, e.g. there is nothing mental, or nothing of any significance, in common between the good case and the bad case. More precisely, an extreme version involves denial that there is anything mental in common between the good and the bad case, other than what the experiences involved have in common with coarser varieties of mental states (e.g. being a visual experience as of a tree, etc.). Less extreme versions of experiential pluralism leave room for some mental properties in common between the good and the bad case, where these common properties are specific to these cases. Such moderate versions include the claim that the good case has at least one such mental property which is not shared by the bad case. This good-case-only property is held to be significant in certain explanatory respects, and, on this basis, moderate versions judge the good case as involving an important mental kind.⁴ ³ The labels ‘experiential monism’ and ‘experiential pluralism’ are originally from (Snowdon 2008). On my use, experiential pluralism captures a great variety of positions. I am avoiding the term ‘disjunctivism’, since the latter is often used by proponents and critics of more specific positions on the nature of the experience, all of which I include on the experiential pluralist side (such as naïve realism (e.g. Martin 1997, 2002; Campbell 2002) and content-based disjunctivism (e.g. McDowell; Tye 2007). I include epistemic disjunctivism (e.g. McDowell 2010, 2013) so long as the view grounds the epistemic difference between the good and the bad case on a psychological difference between the two cases. Soteriou (2016) uses the label ‘metaphysical disjunctivism’ to cover views that postulate a psychological difference between the good and the bad case, and ‘epistemic disjunctivism’ to cover views that postulate an epistemic difference between the two cases (specifically, a difference in the kind of perceptual warrant for judgements accessible to the subject in the two cases). Epistemic disjunctivism does not automatically involve endorsing metaphysical disjunctivism. See (Soteriou 2016, esp. 197–211) for discussion of different kinds of disjunctivism, terminologies, and taxonomies and 146–51 for discussion of the metaphysical commitments of McDowell’s epistemic disjunctivism. ⁴ Byrne and Logue (2008) use the label ‘moderate disjunctivism’ to cover such views and ‘metaphysical disjunctivism’ for what I am calling extreme experiential pluralist views. Their use of the latter term thus differs significantly from Soteriou’s use.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

65

In this paper, I provide an argument for a moderate experiential pluralism. It is a novel argument because it does not lean on phenomenological or conceptual constraints on visual perception, or on epistemic constraints of the kind that extant arguments for this position do. Instead, the argument appeals to the explanatory role of experiences in the good case vis-à-vis certain aspects of our agency: possession of certain kinds of ordinary abilities.⁵ Further, I make a methodological point about how best to argue for moderate experiential pluralism. The latter takes off from the observation that claims about experiential pluralism versus experiential monism have little substance without further elucidation of the central notion of a mental kind involved. Moderate positions in particular ought to be clear on the matter. They allow for common mental properties between the good and the bad case, but nonetheless maintain that there is a good-case-only property that grounds a significant mental kind. To understand this, we need some sense of why the mental kind in question is to be theoretically prioritized. In particular, the import of a moderate view depends on the conception of the wider explanatory significance of the good-case-only property, and of the mental kind it forms in light of this explanatory role. The methodological connection between wider explanatory significance and mental kindhood does not get enough attention in the philosophical debate about perceptual experience. It is often difficult to understand the broader importance of the claims about our psychological nature that are being put forward. As a consequence, it is often difficult to relate the conception of mind put forward by proponents of experiential pluralism to conceptions put forward in other areas, especially in those that take empirical psychology as a guide. By contrast, as I will explain, my argument for experiential pluralism is sensitive to these methodological considerations. The main purpose of the paper thus goes beyond merely defending a particular position on the nature of experience according to the terms of current debate. More ambitiously, it aims to show what a thorough case for such a view requires and to propose one route to meeting this demand. In turn, providing a successful route to meet the demand counts as additional support for my argument. The plan for the paper is as follows. I briefly set out some preliminaries concerning the debate about visual experience in section 2. In section 3, I state the argument and defend it. In section 4, I explain the methodological point in ⁵ For direct appeals to phenomenal adequacy conditions, see e.g. (Martin 1997, 2002). For appeals to the distinctive warranting role of perceptual experience, see e.g. (McDowell 2010, 2013). For appeals to conceptual constraints, see (Kalderon 2018). Soteriou (2016) provides an excellent overview and detailed discussion of the variety of disjunctive positions and debate about them. My argument is perhaps closest to aspects of Campbell’s case for naïve realism, which includes appeals to the role of perceptual experience in explaining our capacities for perceptually based, demonstrative thought about mind-independent objects and their features. But at bottom, Campbell’s case crucially rests on appeals to introspectively acquired phenomenal adequacy conditions as well (for discussion, see Soteriou 2016, 83–114).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

66

 

relation to the current state of the debate, and show how my argument addresses it.

2. Visual Experience in the Good Case Experiential monists and pluralists agree on two conditions on what counts as a visual experience in a good case (‘veridical perception’). The first condition is that such experience must consciously present or convey the world the way it actually is. In a very loose sense of content, such visual experiences must have ‘veridical’ content. An experience in the bad case, of course, can also have this feature: hallucinations can convey the world as being a certain way, when the world really happens to be that way. Moreover, hallucinations can be subjectively indistinguishable from veridical perceptions, when for instance, a subject hallucinating a carob tree in the courtyard cannot tell from the inside (‘by introspection’) that they are not veridically perceiving the carob tree in the courtyard. The second condition on which both views agree is that the good case must involve a certain kind of perceptual causal contact with the world. A subject in a good case is perceptually connected with the world such that it is in virtue of this connection that the experience is veridical. The perceptual–causal connection must rule out that the visual experience in the good case satisfies the first condition accidentally or as a result of a contrived artificial manipulation. There are different ways to think about how these two conditions on visual experiences in the good case relate, i.e. how whatever satisfies the second condition, makes satisfying the first condition non-accidental and non-contrived. One suggestion—forming the focus of this paper—is to think of the two conditions as intimately related: visual experience in the good case presents the world the way it actually is, because it is itself in part constituted by the relevant aspects of the world. What is it for visual experience to be in part constituted by relevant aspects of the world? There are different options for how to spell this out, e.g. that the phenomenal character of the experience has a relational structure with worldly objects as relata, or that its representational content has object-dependent truthconditions.⁶ But the core thought in common to these options is this: a visual experience E is constituted by a given object o exactly when the occurrence of E alone guarantees the appropriate presence of o. E’s occurrence conditions include o. Let us call such a visual experience a ‘world-involving visual experience’. ⁶ There are two aspects to conscious visual presentation: content (in the loose sense) and phenomenal character, and a wide variety of views as to how these aspects relate to one another. As a result, there are different options concerning which of these aspects is the focus of world-involvingness (see Haddock and MacPherson 2008, 17ff especially). One might hold that the phenomenal character of visual experience in the good case is in part constituted by the worldly objects and properties presented in the experience, or that it is the content of such experience that is so constituted.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

67

The idea of world-involving visual experience is of central importance in the debate between experiential monism and experiential pluralism: the latter typically appeal to it in defence of their view. However, as we saw earlier, there are two separate tasks involved in an adequate defence. If there are world-involving visual experiences and these types of experience form a substantive mental kind, then experiential pluralism is true (granting that hallucinations count as genuine experiences). Thus, an adequate argument for experiential pluralism should provide reason for both the claim that visual experience in the good case is world-involving, and the claim that such experience forms a substantive mental kind. Providing reasons for merely the first claim is not sufficient.⁷ The argument in the next section serves to support both claims.

3. The Argument from Abilities Here is the argument: P1:

We have situation-dependent abilities.

P2: Possession of some of these situation-dependent abilities is explained by appeal to visual experiences in the good case. P3: If visual experiences in the good case play this explanatory role, then they are world-involving. C1:

There are some world-involving visual experiences.

C2:

Experiential pluralism is true.

Let me now clarify and defend the premises.

3.1 Situation-Dependent Abilities (P1) We are able to do many things: we are able to get to the museum, pick strawberries, play catch, etc. We attribute a wide array of abilities to ourselves and others all the time. It is an important part of how we make sense of one another, in particular, how we understand what kind of agents we are.

⁷ Tyler Burge, for instance, has long advocated a view of certain mental states as object-dependent, but is staunchly against any version of experiential pluralism (see, for example (Burge 2011; Burge 2009)). Object-dependence, on his view, is an aspect of the context in which the thinker or perceiver has a certain type of mental state—spelled out in terms of his notion of application. It is not an aspect of the psychological type of the relevant state itself. For Burge, genuine psychological kinds are those that play a role in scientific psychology.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

68

 

Abilities are properties of agents. Like other powers that agents have (e.g. competences, skills, potentialities), abilities can be had even when they are not exercised or manifested. I have the ability to walk even when I am standing still. This modal aspect of abilities admits of a distinction commonly made between what someone can do in a more general sense and what they can do with respect to more specific situations. For example, sitting at my desk I might say that right now I cannot swim but that nonetheless I can swim. The gist of this is clear: that in my current circumstances, which lack an adequate body of water nearby, I am not in a position to swim, but that, given such a body of water, the right temperature, etc., I would be in a position to swim. How these two claims about my overall swimming powers—one involving denial of ability, one involving attribution of ability—can be true at the same time is an interesting question. It is an instance of a more general issue that has been discussed extensively in debates on free will, abilities, and dispositions. For instance, Mele distinguishes between general practical abilities and specific practical abilities: Although I have not golfed for years, I am able to golf. I am not able to just now, however. I am in my office now, and it is too small to house a course. The ability to golf that I claimed I have may be termed a general practical ability. It is the kind of ability to A that we attribute to agent though we know they have no opportunity to A at the time of attribution and we have no specific occasion for their A-ing in mind. The ability to golf that I denied I have is a specific practical ability, an ability an agent has at a time to A then or to A on some specified later occasion. (Mele 2003, 447)

Mele’s distinction is routinely drawn by others using different terminology, and it is reflected in work on the semantics of abilitative senses of ‘can’.⁸ I will use ‘general abilities’ and ‘situation-dependent abilities’ to mark it. An ability is usefully individuated by appeal to an action-type (in Mele’s case, golfing, or in my example above, swimming) and a range of situations in which actions of this type are available, that determine the modal profile of the ability.⁹ Attribution of general abilities concerns what a person can do in some (suitably wide) range of situations. That is the sense in which I can swim as I sit at my desk, i.e. there is a

⁸ Whittle (2010) distinguishes between ‘local abilities’ and ‘global abilities’, Maier (2015) between ‘options’ and ‘general abilities’, Inwagen (1983) between an ability as ‘the power of an agent to act’ and ‘general ability’, Honoré (1964) between ‘can (particular)’ and ‘can (general)’, Berofsky (2005) between ‘token ability’ and ‘type ability’. See also Bird (1998) for a related distinction concerning the attribution of dispositions. ⁹ I do not assume any particular theory of abilities in this paper, only that attributions of abilities are associated with special restrictions on possibility. Relevant contemporary work on the semantics of ‘can’ predominantly relies on a framework proposed by Kratzer (1977), though see (Vetter 2013). See also (Maier 2018b) for critical development of the modal view of abilities. For an overview of theories of abilities, see (Maier 2018a).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

69

range of situations where I am in a position to swim, though I am not in one then. Attribution of situation-dependent abilities concerns what a person can do indexed to a specific situation. That’s the sense in which I cannot swim right now as I sit at my desk, i.e. in that situation I am not in a position to swim. Often, we are interested in those situation-dependent abilities that an agent has in the actual situation they are in but not always. I might consider, for instance, what I would be able to do if I were at Berlin central train station right now. The two kinds of ability are connected. In fact, it is plausible that they are ‘something like two modes of a single kind of power’ (Maier 2018a) and that one type is ontologically prior to the other. Work on free will, dispositions, and powers, and on the semantics of ‘can’, includes accounts for one type of ability (or abilitative ‘can’) in terms of the other. Both directions of order of explanation have been explored. The connection is more complex than a simple type-token relationship, since there are cases where we can have a situation-dependent ability without a corresponding general ability, and vice versa (Whittle 2010, 3–4). While it is thus uncontroversial that there is a distinction between situation-dependent abilities and general abilities, the exact nature of that distinction is not. But my paper does not depend on this controversy being resolved in any particular direction. Situation-dependent abilities constitute an important aspect of our agency and are critical to us in our everyday dealings with one another. We constantly refer to what we can and cannot do in certain situations. Assessment with respect to our general abilities is also significant in making practical sense of ourselves, but situation-dependent abilities have a central, structuring role in our daily planning and getting on with things. Having a good idea of someone’s agency relative to a specific situation is crucial to predicting and assessing what they will do, and to determining how they and others around them should act on a given occasion. Call this aspect of our agency determined by situation-dependent abilities our ‘situational agency’, noting that it is not meant to prejudge overall conceptions of rational and non-rational agency. (P1) is therefore uncontentious. There are many things that I am able to do, relative to the situation that I am in, or to a situation of interest. Given some specific situation, the class of my situation-dependent abilities determined in relation to it is typically much smaller than the class of what I can do more generally in this situation.

3.2 Explaining Possession of Situation-Dependent Abilities (P2) How do we attribute abilities to ourselves and to others? At least one routine way of doing so exploits an explanatory link we commonly take for granted between having certain mental states or episodes and having certain abilities. Here, I focus

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

70

 

on an explanatory link between having visual experiences in good cases and having certain kinds of situation-dependent abilities. Let me give some examples of the routine explanations at issue. Finding Home: Suppose you are standing outside your office building, looking around. You are visually aware of the things around you, the street, the houses, etc. This puts you in a position to find your way home now. Of course, having the visual experience on its own does not put you in this position, there are other relevant aspects of you that must be in order and your environment has to remain relatively manageable, stable, etc. But your visual experiences of your surroundings as you are standing outside the department do play a key role in explaining how it is that you are able to find your way home right now. Crossing River: Suppose you are standing at the bank of a stream full of boulders forming a natural path to hop across. You are visually aware of the things around you, the sloping river bank, the boulders, the water, the boggy patches, etc. This puts you in a position to hop across the stream. Your visual experiences of your surroundings help explain how you are able to cross the stream. Picking-up Cat: Suppose you are in my garden looking at my cat Pepperpot lounging in the sun. You are visually aware of Pepperpot sleeping on the bench. This puts you in a position to pick up Pepperpot and take him into the house. Your visual experiences of Pepperpot on the bench help explain how you are able to pick him up and carry him inside. Each case focuses on attribution of a situation-dependent ability, exercise of which involves success at doing something with respect to a specific aspect of one’s environment. The explanations concern possession of these abilities, independently of whether they are actually exercised. The general explanatory pattern is that the potential for success is explained by cognitive access to relevant facts or things in one’s environment (e.g. landmarks, boulders, Pepperpot). It takes for granted that other factors must be in place as well (adequate use of limbs, etc.), but it singles out visual experiences as a key explanatory factor. Thus, according to the explanatory pattern, the cognitive access to the relevant environmental aspects underwriting the potential for success is provided visually. It is worth emphasizing that the claim contained in the above explanatory pattern is merely that the situation-dependent abilities in question are typically, or perhaps even just sometimes, visually based. They could be had, and perhaps often are had, in virtue of other kinds of access to the relevant environmental aspects (e.g. hearing or touch).¹⁰ I will say more about this in a moment. But while the explanatory pattern I am drawing attention to does not allege exclusivity, it does assume that its application sometimes allows us to understand and explain that a

¹⁰ See (Millikan 2000) for interesting discussion of the different ways in which abilities can be had. Moreover, Finding Home also involves a longer process than the other two abilities and so requires continued cognitive access to aspects of the environment along the way.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

71

subject possesses a certain range of situation-dependent abilities. In fact, the explanatory pattern is a firm part of common sense practice: we employ these sorts of explanations—attributing a range of situation-dependent abilities by appeal to subjects’ visual experiences—in our understanding of one another’s agency and our planning all the time. Note, though, that insofar as one does appeal to visual experience in the above three examples, only visual experiences in the good case can play the required explanatory role. The subject’s visual awareness must be such that it figures in genuinely seeing their environment because we want to explain their possession of an ability which turns on adequate access to certain parts in that environment. One might object to the last point as follows. In Finding Home, suppose that instead of an experience in the good case, you are hallucinating as you come out of the department building and that your hallucinations happen to be veridical. Such veridical hallucinations would also allow us to explain plausibly how you could get home on this occasion. Similarly, in Crossing River and Picking-up Cat, appeals to veridical hallucinations of the boulders in the river, or of Pepperpot on the bench, would allow for plausible explanations of the respective situation-dependent abilities. So, according to the objection, certain visual experiences in the bad case can serve in these explanations as well. The objection is misguided and to see why is helpful to understand (P2) better. I will focus on Finding Home, though the response generalizes. It is true that visual experiences in the bad case, specifically, veridical hallucinations, can explain how it is possible for you to find your way home on that occasion. Suppose, as you exit the building, you are unaware that you are hallucinating (scenario Oblivious). Since your hallucinations are veridical, you would get home by relying on this visual information. You would be lucky, it so happens that the environment actually is the way your hallucinations present it to be. Suppose, as you exit the building, you do know that you are hallucinating and veridically so (scenario Clued-up). Then it is also possible for you to find your way home and not just by luck. Since you know that you are hallucinating veridically, you can exploit the information visually presented to get home. Thus, in both Oblivious and Cluedup, it is possible that you find your way home and your veridical hallucinations seem to play a role in this. But, as I will argue now, neither scenario can make use of the explanatory pattern to which I am drawing attention. In Oblivious, we are not explaining possession of an ability at all but merely the possibility that I perform a certain action. It is at this point that my particular choice of explanatory target—situation-dependent abilities rather than perceptually based behaviour or action—makes the important difference.¹¹ Contrast the

¹¹ This choice of explanatory target—i.e. my focus on situation-dependent abilities and not on actions—distinguishes my approach from the older discussions of the role of object-dependent or externalist mental content in the psychological explanation of action (see, e.g. Segal (1989) and Noonan

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

72

 

sense in which it is possible for you to Φ in a given situation S, with the sense in which you have it in your power to Φ in S. It being in your power to Φ in S is a modal aspect of you and having it requires more than that Φ-ing is something the circumstances that you find yourself in allow (Inwagen 1983, 9). It is possible that you pick the winning ticket from an opaque urn with thousands of tickets (you might be lucky) but it is not in your power to pick the winning ticket in this situation. You cannot choose to do so, for instance. As Millikan says ‘what one does successfully only by pure chance is not something one knows how to do. [ . . . ] Abilities [have] necessary involvement in the purposive and non-accidental order’ (Millikan 2000, 58). The point is that mere performance of Φ-ing is not sufficient for possessing the ability to Φ. On wandering through a foreign city, where all the signs are in a foreign language I do not understand, I might happen to make my way from my hotel to the museum of modern art. But I do not thereby have the ability to find my way from the hotel to the museum. I might very well not be able to get there the next morning again, since it was quite by chance that I ended up where I did today. For it not to be a chance or one-off fluke performance, it must be the case that the (potential) performance is robust in some way: that I could have gotten from the hotel to the museum again in relevantly similar conditions. More generally, for me to exercise an ability, the successful performance constituting such an exercise must be repeatable in relevantly similar circumstances. The point needs elaboration, for clearly, one’s Φ-ing entails that one can Φ and no one will deny that. Austin, for instance, when discussing his famed golfing example, says that according to ordinary English merely doing something entails that one has the ability to do it (Austin 1956). But in fact, something like the above condition on what counts as possessing an ability in the relevant sense, is standardly recognized in literatures where discussion of abilities plays a central role. Correspondingly, any notion of ability or sense of ‘can’ that applies to a oneoff performance is held to be significantly different. In this vein, Honoré talks about ‘the use of “can” (particular) in relation to particular actions’ (Honoré 1964, 464). Similarly, Mele introduces the distinction between simple abilities as opposed to intentional abilities. Simple abilities cover the sense in which one can Φ in virtue of a one-off Φ-ing (Mele’s own example is Ann’s being able to roll a six with a die in a chance game because she did so). Intentional abilities are abilities that manifest a sense of power or control that the agent has in possessing a given ability (Mele 2003, 447–8). It is the latter sense of ability that is at issue here and that admits of the distinction between situation-dependent and general (1993); also Drayson (2018)). There seems to be a certain argumentative stalemate in many of these debates, due to the fact that we can causally explain relevant behaviour and action by appeal to either object-dependent mental content, or a complex of an internal mental content and the obtaining of environmental conditions. As I argue here, situation-dependent abilities present a significantly different explanatory challenge.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

73

abilities and that are subject to the condition of repeatability of performance. Thus, when clarifying her notion of local (i.e. my situation-dependent) ability, Whittle insists: ‘It isn’t enough that I just so happen to jump over a five-foot bar on one occasion. In addition, a reliable connection between being in a particular set of circumstances and the outcome is required’ (Whittle 2010, 4). The target of explanation in Oblivious is the possibility for performance of an action—finding your way home on that occasion. When you stand outside your department and hallucinate veridically while being unaware of this, it is easy to understand how these experiences, together with the fortuitously matching circumstances, make it possible that you get home on that occasion. But these experiences do not make it the case that you have it in your power to get home because it would be merely a happy accident if you did. There is no reliable connection between having a hallucination with a given content and performance, even if that content happens to be veridical on this occasion. Were you to be in relevantly similar circumstances—having an hallucinatory experience with the same content as the veridical hallucination you are having now—it is by no means clear that you could find your way home.¹² Having it in your power to get home— having the ability to get home, on the other hand, requires more than the circumstances conspiring in this manner. I will say more about this in the next section. In Clued-up, you are in a position to find your way home, and, that is a genuine power you have rather than merely the circumstances conspiring. Being in the hallucinatory state, together with knowing that you are veridically hallucinating, does provide for a reliable connection to the environment, making the performance robust. In relevantly similar circumstances, where you have an experience with the same content, and knowledge that this experience, though hallucinatory, does present your environment accurately, you will also be successful in finding your way home. However, attributing to you the ability to find your way home in Clued-up is based primarily—or at least to a large extent—on your higher-order knowledge about your visual experiences. What makes the connection between your particular circumstances and the outcome—the successful performance of finding your way home—reliable is that you could do so again in relevantly similar circumstances: circumstances in which you have the hallucinatory experience and knowledge that it is veridical. While the visual experience you are having (the veridical hallucination) does explain the possibility of you finding your way home

¹² Could the relevantly similar circumstances be stipulated to be those in which you have a veridical hallucination of your surroundings? Veridicality is, in the case of hallucinations, a matter of luck. This would amount to stipulating the relevantly similar situations as ones in which you are always lucky in that the content of your hallucinatory experiences corresponds to the way your environment is. Any repeatability in successful performance based on such luck obtaining does not constitute a genuinely reliable connection between your visual experience and your performance: they are not anti-luck conditions.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

74

 

on this occasion, the robustness of this potential for performance needs crucial appeal to your the higher-order knowledge about the relationship between your experience and the environment. It is the latter that provides the explanatory grounds for you having the ability to find your way home, as opposed to the fact that the possibility of finding your way home is something that is open to you in your specific situation. Thus, in neither of the scenarios are we explaining possession of the target ability by appeal to visual experiences in line with the explanatory pattern I set out above. In Oblivious, the target of explanation, the explanandum, is different: we are not explaining possession of an ability in the relevant sense. In Clued-up, the explaining factor, the explanans, is different: we are not explaining the relevant ability mainly by appeal to visual experiences, but by crucial appeal to higherorder knowledge about how the experience relates to the environment. The latter again brings out that abilities can be had in different ways. The explanatory pattern to which I am drawing attention in (P2) merely requires that certain situation-dependent abilities are sometimes had via normally functioning vision, where this is consistent with them being had differently at other times. But when the pattern is at work—when we are appealing to visual experiences as the main factor to explain the possession of a given situation-dependent ability, only those in the good case can do the job.

3.3 The Explanatory Demand for World-Involvingness (P3) Turning now to (P3), in this section I argue that for visual experiences in the good case to do the explanatory job we said they do in the previous section, they are best conceived of as world-involving. According to the distinction between Φ-ings which are one-off fluke performances and Φ-ings which are exercises of the ability to Φ, the latter but not the former involve performances that satisfy some antiluck or robustness condition. Such performances must be repeatable by the subject in sufficiently similar situations. Recall, abilities are individuated in terms of an action-type Φ and a range of situations where the performance of Φ is available to the subject. In the case of situation-dependent abilities, the individuating range of situations is such that it includes (i) a particular situation with respect to which the agent’s situationdependent ability is fixed (the ‘base situation’), (ii) more situations than the base situation, (iii) those situations which are relevantly similar to the base situation. When we explain a subject’s possession of a situation-dependent ability by attributing a visual experience in the good case to her, that experience must

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

75

provide reason to pick out not only the action-type that individuates the ability, but also to fix the ability’s modal profile, such that the individuating range of situations includes (i), (ii), and (iii). The base situation is identified as the one in which the agent has a given visual experience. Which situations count as being relevantly similar to it is also fixed by reference to the agent’s experience: it is all those situations which differ minimally from the base situation, namely, from the situation in which the agent has the visual experience in question. That is, relevantly similar situations are minimally those in which the agent has a visual experience of the same type. What must be guaranteed is that these are indeed situations in which the action in question is available to the agent. World-involving visual experiences can guarantee this. The base situation is identified in terms of the agent’s having a given world-involving visual experience Ew-i. The base situation is thus one in which they have the right cognitive access to their environment for them to successfully perform Φ, where success in Φ-ing depends on such access. The range of relevantly similar situations, in which Φ is also available to the agent, comprises those situations in which the agent has experiences of type Ew-I (and nothing else relevant being amiss with the agent, i.e. their limbs are working, other cognitive faculties are in order, etc.). Since Ew-i is world-involving, that means that relevantly similar situations are those in which the required cognitive access obtains and thus things are as they are visually presented in relevant respects. If you are in a situation relevantly similar to the base one, where the latter is identified in terms of having a world-involving visual experience presenting Pepperpot as sitting on the bench, then the right cognitive access to the environment is guaranteed, such that it will put you in a position to pick him up. That is, relevant similarity requires that the situation is one in which you are having a world-involving visual experience presenting Pepperpot sitting on the bench over there. Such a situation is one in which you could easily have succeeded to pick him up if you tried. Possession of the situation-dependent ability to Φ is thus explained by appeal to having a certain type of visual experience Ew-i: by having a certain kind of access to relevant aspects of one’s environment provided by the visual experience. Importantly, this visual access explains how the potential for the performance of Φ is robust, how you could easily succeed to Φ in relevantly similar situations, viz. situations in which you have such cognitive access. The visual experience’s being world-involving thus guarantees robustness. So, conceiving of visual experience in the good case as world-involving provides a simple and elegant solution to satisfy the robustness requirement. That fact does not yet suffice for a defence of (P3), of course. Other conceptions of visual experience in the good case might also provide a decent solution. Indeed, given that the good case is characterized in part by the right kind of perceptual contact with relevant aspects of the environment, it may seem that any conception of

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

76

 

visual experience in the good case satisfies the robustness requirement. A complex state E+Cext consisting of a veridical experiential mental component together with the right kind of external hook-up to the world guarantees robustness, too. Accordingly, one might object that the fact that world-involving visual experience guarantees robustness yields no support for (P3), other than perhaps by considerations of elegance. Elegance considerations notwithstanding, the objection is wrong. There is a significant difference in how each conception of visual experience in the good case can guarantee robustness and this difference shows that there is, after all, an explanatory demand for world-involving experience providing support for (P3). If visual experience in the good case is world-involving, then conscious visual experience of the world, Ew-i does the explanatory heavy lifting with respect to both, fixing the relevant action-type individuative of the ability in question, as well as robustness. Visual experience itself—conscious visual awareness or presentation—explains possession of the situation-dependent ability under consideration because the right perceptual hook-up to the world is a constituent part of it. This account makes best sense of the explanatory pattern expressed by (P2), according to which we sometimes explain possession of situationdependent abilities by appeal to visual experiences. By contrast, if visual experience in the good case is not world-involving, then the experiential element, the conscious visual presentation of the world, loses considerable force as an explanatory factor. On the E+Cext conception of visual experience in the good case, E can be had in a bad case, too. Just like in the case of Clued-up above, conscious visual experience E per se, does fix the action-type relevant to the ability, explaining how performance of this action is possible for the subject in situations where they have E. But, the range of situations fixed by E in this way do not amount to relevantly similar situations in which such performance is robust: E on its own will not fix the relevant similarity class of situations in which one could easily have succeeded to Φ. Instead, the explanatory burden of fixing the latter range is carried first and foremost by the additional external conditions Cext which characterize cases of genuine seeing. In principle, these might obtain in conjunction with a variety of visual presentational states the agent is in, so the overall explanation needs to appeal to the complex E+Cext. Although this kind of complex state can guarantee robustness, then, it diminishes the explanatory efficacy of conscious visual awareness itself considerably. Our starting point is the ordinary explanatory pattern referred to in (P2), namely that there are occasions where we explain our possession of certain situation-dependent abilities by appeal to visual experience (accepting that other conditions, e.g. working limbs, etc. must be in place). A conception of worldinvolving visual experience does the best job of accounting for this explanatory role we accord to visual experience in ordinary practice vis-à-vis such abilities. This is because on this conception, conscious visual experience can be understood

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

77

both to provide for the selection of the relevant action-type and to underwrite the robustness of potential performance of such action, necessary to turn mere performance into an exercise of an ability. World-involving visual experience yields the most appropriately straightforward explanans and in our explanatory pattern. The other conception of visual experience in the good case, E+Cext, does not give conscious visual experience E the same comprehensive role, needing to appeal to external conditions Cext to explain robustness. The explanation via world-involving visual experiences therefore more naturally reflects common sense explanatory practice and our understanding of our agency as often grounded in visually conscious access to our environment. The explanatory pattern expressed by (P2) is reasonable: we ordinarily often do explain possession of such abilities by appeal to conscious visual experience. Furthermore, this common sense explanatory practice works: understanding our situational agency is at the core of our day-to-day planning and decision-making. It serves us well in understanding the kinds of agents we are in situ and guiding expectations and predictions about what we can and cannot do fairly well.

4. World-Involving Experiences as a Mental Kind The methodological concern raised in the introduction focuses on the gap between establishing that there is world-involving experience and establishing that such experience forms a significant mental kind. Any moderate experiential pluralism of substance must close this gap by spelling out what the mental kindhood of world-involving experience consists in. The latter requires not only an argument for a good-case-only property of experience, but also reasons for treating that property as constituting a mental kind. Proponents of different versions of experiential pluralism often rely on terms such as ‘nature’, ‘kind’, and ‘fundamental’, as in ‘fundamental level of classification’ (McDowell 2010, 252), ‘(ontologically fundamental) kind’ (Snowdon 2005, 133), to state and clarify their key claims. However, they typically do not elucidate these terms much further, thus leaving it unclear what such appeals to fundamentality, nature, basic kind, etc. amount to in this context. Consider Martin’s suggestion that the fundamental kind at issue concerns the essence of the experience. The essence of a given entity determines what kind of entity it is, according to some privileged classifications of individuals, both concrete objects and events, and that our talk of what is essential to a given individual tracks our understanding of the kinds of thing it is . . . . [E]ntities (both objects and events) can be classified by species and genus; for all such entities there is a most specific answer to the question, ‘What is it?’ In relation to the mental, and to perception in

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

78

  particular, I will assume that for mental episodes or states there is a unique answer to this question which gives its most specific kind; it tells us what essentially the event or episode is. (2006, 361)

Suggestions along these lines seem to presuppose that we have some prior and shared common sense understanding of what constitutes the fundamental level, or the most specific kinds, etc. at issue.¹³ But fundamentality is an explanatory notion and claims about basic or most specific fundamental kinds must be situated within an explanatory context which in turn is tied to a given level of description of the world. This is not to say that everything goes, of course. Without wading into the general controversies about realism about natural kinds—specifically, whether such kinds are interest-relative, or conventional in some sense that undermines their objective reality—I think it is reasonable to assume that not all classifications for any given explanatory purpose are equal with respect to the world that they describe. We do want the fundamental kinds to carve the world’s joints in some significant manner, even if they are interest-relative to some degree. This is true of mental kinds as well. Both of these aspects of mental kindhood—their individuating connection to explanatory projects and levels of description, as well as their reality—need to be articulated adequately in order to address the methodological concern about the gap between an argument for a good-case-only property of experience and reasons for treating that property as constituting a mental kind. Claims about fundamental levels of classification and similar typically do not do so in my view. The assumption of a shared common sense understanding of the relevant notion of a fundamental level at which to locate the mental kinds in question, cannot be taken for granted. There is no firm grasp of a level of reality available, even in common sense, that is both independent of explanatory projects and uncontroversial enough to yield something like Martin’s unique, most specific answers in the case of experience. So, in the background of such an assumption must be something like the thought that there is an explanatory context, a level of description, which forms the shared understanding of common sense. One likely candidate for this is ‘the manifest image’ of mentality—the idea that there is a distinctive way in which we encounter our minds, including perceptual experience, from the person-level, subjective perspective. The manifest image of mentality is meant to provide a level of description that is at the heart of our basic mental concepts, and explanatory projects hitched to it are therefore held to be revealing of the nature of mentality itself. But this background thought about the manifest image of mentality which putatively moors the shared understanding of perceptual experience providing for ¹³ Martin and McDowell, of course hold very different experiential pluralist views (see footnote 3) and this section is not meant to treat these views as the same. But given that such views typically use talk of the fundamental level, etc. this variety illustrates the need for more detail on the psychological kind invoked by such talk.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

79

unique answers to the ‘What is it?’ question, is non-trivial. What is the source of claims reflecting our shared understanding and the manifest image of perceptual experience? It is fair to say that in the debate about the nature of perceptual experience, many of them are grounded in the deliverance of introspective reflection on the phenomenal character of experience. In addition to introspection-based claims about phenomenal character, there are considerations about the epistemic role of perceptual experience in sustaining various other manifest mental capacities (e.g. demonstrative thought, empirical propositional knowledge), as well as theoretical reflection on the nature and purpose of the capacity of perception itself. It is not at all clear to me that these sources (singly or jointly) offer the bases for the required shared understanding of perceptual experience. Insofar as the manifest image is meant to reflect common sense understanding, these sources are not obviously part of ordinary thinking about the mind, if by that we mean pre-theoretic, everyday thought and talk. Considerations about the epistemic role of perceptual experience in grounding perceptual demonstratives, or about the nature and purpose of the capacity of perception are theoretically motivated considerations about the mental phenomenon in question. One has to do some work to get the ordinary person to understand the questions being asked in the first place. Even introspective reflection on the phenomenal character of experience does not genuinely yield a ‘naïve’ take on experience, where that requires isolating (say) visual experience, from any background expectations, habits, other beliefs, and non-visual experiences in the vicinity. At the very least, that kind of naïvety has to be practised some before being available to an introspecting subject. But once we accept that the shared understanding requires some theoretical sophistication and, importantly, shared explanatory interests, none of the claims expressive of it offer particularly neutral dialectical starting points that opponents will easily agree upon. Furthermore, the above suggestions typically have little to say about how exactly their explanatory projects relate to other explanatory projects concerned with the mind and perception in particular. According to Martin, the essence of experience as located at the fundamental manifest level is meant to be consistent with there being ‘a more fundamental level of reality’ on which the mental supervenes in some way (see his 2006, 360, n. 9). But it is difficult to see how exactly the ‘manifest kinds’ at issue in their debates relate to, especially, scientific kinds in perceptual psychology.¹⁴ In the wake of a thriving science of perception, philosophers of perception are (rightly, in my view) expected to ensure that their views are appropriately empirically informed. More generally, I think that the question about the reality of mental kinds demands some minimal integration of ¹⁴ The label ‘manifest kind’ is from Johnston (1997), discussed in Phillips (2018). For a related debate about the question of whether there is unconscious or subliminal perception, see, e.g. Block and Phillips (2016), Peters et al. (2017), Phillips (2018).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

80

 

the picture of the mind emerging from philosophical investigation of the manifest image, with the picture of the mind emerging from scientific investigation in psychology. In some quarters, philosophical concerns and views about perceptual experience are held to be irrelevant and unmotivated unless they are contained in our best current scientific approaches to perception (Burge 2010; Rescorla 2015). This strikes me as an unjustified restriction—based on insisting on the primacy and dominance of the explanatory project of contemporary vision science (e.g. Fish, this volume). However, to insist in return on the availability of entirely different and autonomous explanatory projects is equally unsatisfactory. It leaves the question about the reality of mental kinds—as distinct from psychological kinds—hanging, undercutting any sense that the mental kinds of one such project and the psychological kinds of the other in the end contribute to an integrated picture of the mind. My strategy—pursued in the argument from abilities—helps to refocus the issue away from insisting on the primacy or autonomy of a specific explanatory project, to the choice of appropriate explanatory targets for attribution of mental or psychological kinds more generally. According to my strategy, the explanatory targets are ordinary situation-dependent abilities. These are relatively uncontroversial candidates for counting as natural properties of the mind, from the scientific, philosophical, and common sense perspectives. They offer a genuinely neutral starting point for explanation and allow for different explanatory projects concerning them. In addition, in focusing on ordinary situation-dependent abilities as explanatory targets, my strategy sticks closely to an innocuous rule of thumb concerning the reality of natural kinds in psychology: that mental states or episodes count as real psychological properties if they are causally trackable. Moreover, if the causal efficacy of such properties can be shown to be systematic and distinctive in certain ways, there is reason to hold that these properties form a significant psychological kind.¹⁵ The way in which such causal efficacy is often established is by showing that these mental states or episodes have an important—perhaps even essential or ineliminable—role in the causal explanation of behaviour (Drayson 2018). Although my argument follows the basic outline of this approach, it does not fit the mould of a causal explanation of behaviour exactly. As we have seen, the argument turns on the explanatory role that world-involving experiences play and the explanatory targets in question are ordinary situation-dependent abilities. The latter, of course, are not instances of behaviour, if by that we understand performances of actions. But, as we also saw, such abilities are intimately connected with performances of action. Abilities encompass certain behavioural possibilities

¹⁵ This second application of the rule—underwriting natural psychological kindhood—is clearly less innocuous than the first, underwriting natural psychological properties. It can be filled out in different ways, depending on one’s views about what constitutes kindhood more generally.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

81

that characterize an agent. They determine an important part of their agency in setting a repertoire of actions that an agent has available to them in certain kinds of circumstances. So, while the explanatory targets in my argument are not instances of behaviour directly, they are properties that manifest in human behaviour. Correlatively, the explanations at the heart of my argument are themselves not causal. They are constitutive, indicating that having world-involving experience endows the subject with certain other psychological properties. But the abilities which are my explanatory target form the basis for prediction of behaviour in ordinary planning and decision-making, where we need to know what an agent can and cannot do in specific situations. Abilities themselves thus become causally trackable in terms of their exercises being causally trackable. This is reflected in the fact that common sense explanatory practice involving attribution of situation-dependent abilities is empirically informed. Attributions of these abilities have been honed by observation of success and failure of performance of relevant actions. More importantly, abilities are open to empirical study. That means not only that abilities are scientifically recognized phenomena, but that common sense explanatory practice involving them is open to revision from science.¹⁶ In light of this, we can see that my argument follows the rule of thumb for recognition of a psychological kind mentioned above in spirit, if not in practice. It follows it enough to provide a firm first step towards recognizing a certain type of mental episode as a genuine psychological kind. World-involving visual experiences explain possession of some situation-dependent abilities. In turn, these abilities are a central ingredient in behaviour-related planning and decisionmaking, and they are causally trackable via their causally efficacious behavioural manifestations. In light of their explanatory and predictive role and the fact that they are subject to empirical investigation and revision, situation-dependent abilities are genuine psychological properties of ours. In that case, mental states or episodes by appeal to which we attribute and explain possessing them, have a claim to be taken seriously as real psychological properties if not kinds as well. World-involving visual experience does have such an explanatory role and it is therefore a fitting candidate for counting as a psychological kind. A detailed account of what constitutes natural kindhood, and especially natural psychological kindhood, is beyond the scope of this paper. The positive strategy I offer is meant as a first step, charting a plausible route for making a detailed case that the type of experience in question constitutes a natural mental kind in its own ¹⁶ Much of psychology involves investigating which abilities, skills, capacities, and expertise we have and how to explain them, and, moreover, how they contrast with our ordinary conceptions of them. Two give just two examples: studies on unethical amnesia and its effect on one’s ability to regulate shame-induced social dysfunction (Kappes and Crockett 2016) and studies on our ability to provide accurate eyewitness testimony (Wells and Loftus 2003).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

82

 

right, where such kindhood is not entirely disconnected from related psychological kinds recognized by empirical psychology. There may well be other good strategies for doing so (see, e.g. Phillips 2018 for a different proposal). The particular one I have advanced here lays the groundwork for a promising approach, if the wider aim is to facilitate engagement between scientific and philosophical theorizing about perception.¹⁷

References Austin, J.L. 1956. ‘Ifs and cans’. Proceedings of the British Academy 42: 109–32. Berofsky, B. 2005. ‘Ifs, Cans, and Free Will: The Issues’. In R. Kane (ed.) The Oxford Handbook of Free Will. Oxford: Oxford University Press, 181–201. Bird, A. 1998. ‘Dispositions and antidotes’. The Philosophical Quarterly 48: 227–34. Block, N. and Phillips, I. 2016. ‘Unconscious Seeing—A Debate’. In B. Nanay (ed.) Current Controversies in Philosophy of Perception. London: Routledge. Burge, T. 2009. ‘Five Theses on De Re States and Attitudes’. In J. Almog and P. Leonardi (eds.) The Philosophy of David Kaplan. Oxford: Oxford University Press, 246–310. Burge, T. 2010. The Origins of Objectivity. New York: Oxford University Press. Burge, T. 2011. ‘Disjunctivism Again’. Philosophical Explorations 14 (1): 43–80. Byrne, A. and Logue, H. 2008. ‘Either/Or’. In Disjunctivism: Perception, Action and Knowledge. Oxford: Oxford University Press, 57–94. Campbell, J. 2002. Reference and Consciousness. Oxford: Oxford University Press. Drayson, Z. 2018. ‘Extended Minds and Prime Mental Conditions: Probing the Parallels’. In J.A. Carter, A. Clark, J. Kallestrup, S.O. Palermos, and D. Pritchard (eds.) Extended Epistemology. New York: Oxford University Press. Haddock, A. and MacPherson, F. 2008. ‘Introduction: Varieties of Disjunctivism’. In A. Haddock and F. MacPherson (eds.) Disjunctivism: Perception, Action and Knowledge. Oxford: Oxford University Press, 1–24. Honoré, A. M. 1964. ‘Can and can’t’. Mind 73 (292): 463–79. Inwagen, P. v. 1983. An Essay on Free Will. Oxford: Clarendon Press. Johnston, M. 1997. Manifest kinds. Journal of Philosophy 94 (11): 564–83. Kalderon, M. 2018. ‘Experiential Pluralism and the Power of Perception’. In John Collins and Tamara Dobler (eds.) The Philosophy of Charles Travis: Language, Thought, and Perception. Oxford: Oxford University Press, 222–36.

¹⁷ Earlier versions of this paper have been presented at Leeds, Stockholm, Cambridge, York, Durham, Hannover, and Mexico City (UAM-Cuajimalpa). I am grateful for all the excellent discussions on these occasions. Special thanks to Dan Cavedon-Taylor, Heather Logue, Louise Richardson, Alex Silk, Scott Sturgeon, Henry Taylor, Barbara Vetter, and Al Wilson.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

83

Kappes, A. and Crockett, M. 2016. ‘The benefits and costs of a rose-colored hindsight’. Trends in Cognitive Sciences 20 (9): 644–6. Kratzer, A. 1977. ‘What “must” and “can” must and can mean’. Linguistics and Philosophy 1: 337–55. Maier, J. 2015. ‘The agentive modalities’. Philosophy and Phenomenological Research 87 (3): 113–34. Maier, J. 2018a. ‘Abilities’. The Stanford Encyclopedia of Philosophy (Winter 2020 Edition), Edward N. Zalta (ed.). https://plato.stanford.edu/archives/spr2018/ entries/abilities/ Maier, J. 2018b. ‘Ability, modality, and genericity’. Philosophical Studies 175 (2): 411–28. Martin, M.G.F. 1997. ‘The Reality of Appearances’. In M. Sainsbury (ed.) Thought and Ontology. Milano, Italy: Franco Angeli, 81–106. Martin, M.G.F. 2002. ‘The transparency of experience’. Mind and Language 17: 376–425. Martin, M.G.F. 2006. ‘On Being Alienated’. In T. Gendler and J. Hawthorne (eds.) Perceptual Experience. Oxford: Oxford University Press, 354–410. McDowell, J. 2010. ‘Tyler Burge on disjunctivism’. Philosophical Explorations 13 (3): 243–55. McDowell, J. 2013. ‘Tyler Burge on disjunctivism (II)’. Philosophical Explorations 16 (3): 259–79. Mele, A.R. 2003. ‘Agents’ abilities’. Nous 37 (3): 447–70. Millikan, R. 2000. On Clear and Confused Ideas. Cambridge: Cambridge University Press. Peters, M., Kentridge, R., Phillips, I., and Block, N. 2017. ‘Does unconscious perception really exist? Continuing the ASSC20 debate’. Neuroscience of Consciousness 3 (1): 1–11. Phillips, I. 2018. ‘Unconscious perception reconsidered’. Analytic Philosophy 69 (4): 471–514. Rescorla, M. 2015. ‘Bayesian Perceptual Psychology’. In M. Matthen (ed.) Oxford Handbook of Perception. Oxford: Oxford University Press. Segal, G. 1989. ‘The return of the individual’. Mind 98 (January): 39–57. Snowdon, P. 2005. ‘The formulation of disjunctivism: A response to Fish’. Proceedings of the Aristotelian Society 105: 129–41. Snowdon, P. 2008. ‘Hinton and the Origins of Disjunctivism’. In A. Haddock and F. MacPherson (eds.) Disjunctivism: Perception, Action, Knowledge. Oxford: Oxford University Press, 35–56. Soteriou, M. 2016. Disjunctivism. London: Routledge. Vetter, B. 2013. ‘ “Can” without possible worlds: Semantics for anti-Humeans’. Philosophers’ Imprint 13 (16): 1–27.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

84

 

Tye, M. 2007. ‘Intentionalism and the argument from no common content’. Philosophical Perspectives 21 (1): 589–613. Wells, G. L. and Loftus, E. F. 2003. ‘Eyewitness Memory for People and Events’. In A. Goldstein (ed.) Handbook of Psychology: Forensic Psychology. Hoboken, New Jersey: Wiley. Whittle, A. 2010. ‘Dispositional abilities’. Philosophers’ Imprint 10 (12): 1–23.

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

5 The Tractability of the Debate on Relationalism Roberta Locatelli

1. The Debate on Relationalism One of the liveliest debates in the contemporary philosophy of perception concerns the way perception relates us to the world. The dominant alternatives in this debate are representationalism and relationalism. Representationalism, which has become something of an orthodoxy since the second half of the twentieth century, claims that perception has a representational content, capable of being veridical or illusory, which represents the way things appear to be to the subject. Relationalism denies that perception is (fundamentally) representational: instead, it is a non-representational relation of acquaintance with mind-independent objects. Much of this debate has focused on how the alternative views account for non-veridical experiences, in particular hallucinations. For representationalists hallucinations are the same kind of experience we have when we perceive veridically: the only difference from veridical perceptions is that their content typically incorrectly represents how things are in the world. But for the relationalist, the kind of experience we have when veridically perceiving (the obtaining of a relation of acquaintance with mind-independent objects) is not one you can have when hallucinating, where the relevant objects are absent. Therefore, most relationalists embrace disjunctivism, the view that perception and hallucination differ in some important way. Proponents of disjunctivism disagree as to where exactly this difference lies. For reasons I will discuss later on, here I will focus on a version of disjunctivism, often called ‘phenomenal disjunctivism’ (Haddock and Macpherson 2008), which locates the difference between perception and hallucination in their respective phenomenal character. For simplicity, I will simply use the term ‘disjunctivism’ to refer to this particular claim. Disjunctivism has been met with strenuous resistance by many philosophers, for reasons that fall under two main lines of reasoning: (1) The very notion of phenomenal character entails that whenever two phenomenal characters are indistinguishable, they are qualitatively identical, as there is no seem/is distinction when it comes to phenomenal Roberta Locatelli, The Tractability of the Debate on Relationalism In: Purpose and Procedure in Philosophy of Perception. Edited by: Heather Logue and Louise Richardson, Oxford University Press (2021). © OUP. DOI: 10.1093/oso/9780198853534.003.0005

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

86

  character. This idea goes often under the name of ‘the Indistinguishability Principle’ (Siewert 1998; Siegel 2008, 2004; Deutsch 2005; Kriegel 2013; Farkas 2006). (2) Disjunctivism is incompatible with vision science because the latter is committed to either of the following claims: (a) a visual state ‘causally depend[s] only on proximal stimulation, internal input, and antecedent psychological condition’ (Burge 2005, 22) (same cause same effect principle) (see also Smith 2008, 2002; Foster 2000; Sollberger 2007); (b) the phenomenal character of a visual state locally supervenes on the neural activity (Robinson 1994, 151–2).¹ Proponents of relationalism retort that both objections are ill-founded: (1) Nothing—either in the notion of phenomenal character or in what we know about how introspection works—forces us to accept the Indistinguishability Principle (Hinton 1967b, 1967a; Martin 1997, 2003, 2006; Fish 2008, 2009). (2) Disjunctivism is not incompatible with science because: (a) The ‘same cause same effect principle’ begs the question against relationalism, as it ‘rules out the possibility that relational states of affairs or events can form part of a causal nexus where relational states of affairs may differ purely in their distal elements’ (Martin 2006, 368); (b) Local supervenience is only a working assumption concerning the relation between neural and mental states, and is not better supported by empirical data than competing models (Fish 2009, 117–44).

Opponents of relationalism, in turn, insist on the validity of the principles they rely on or insist that, somehow, disjunctivism is ‘a bitter if not impossible pill to swallow’ (Knight 2014, 357). Each camp seems to think that the opponents grossly misunderstand their position, fail to see the inescapability of some principle (or, on the contrary, blindly accept principles that should not be taken for granted), or have an irreducibly different (possibly wrong-headed) way of thinking about the methods and purposes of philosophical reflection on perception. This fuels scepticism about the possibility of ever adjudicating on its own terms a debate which seems to rest on a disagreement over fundamental assumptions

¹ These two claims are distinct, as the local supervenience claim doesn’t entail causal dependence, but they are very similar inasmuch as they both rule out the possibility that subjects being in the same neural state can have experiences with different phenomenal characters.

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

      

87

and methodological approaches. In a paper from 2005 and in his contribution to this volume, William Fish voices this sort of scepticism and likens the debate over disjunctivism to Kuhn’s incommensurable competing paradigms. In the next section I will argue that, despite Fish’s attempts to give it a positive spin, his diagnosis of the debate at hand leaves us without any way of adjudicating it (and possibly other philosophical debates). This outcome is certainly something most philosophers would like to avoid. The good news is that we do not have to be resigned to such defeatism. Contrary to Fish, I will argue that the current lack of progress in the debate is not due to an insurmountable structural feature of this debate but to the failure to make explicit and address an implicit assumption which motivates the widespread scepticism towards disjunctivism. In order to identify this implicit assumption, in section 3 I will start by surveying how the current debate originated from the attempt to reconcile two apparently incompatible, but equally compelling, intuitions about perception. This suggests that adjudicating the debate depends on assessing the respective costs and benefits of competing views and that, most likely, no solution will come without a certain price to pay. However, not every price is equal, and one of philosophy’s tasks is to guide us in weighing up which price is worth paying. In section 4 I will argue that the failure to address an implicit assumption that operates in the background of the debate has hindered the efforts to assess the real costs of accepting relationalism. The assumption is that there is a basic constraint to any meaningful account of phenomenal character, which I dub the ‘superficiality constraint’. This is the idea that the phenomenal character of experience is tied to introspection in a way that limits the pattern of possible introspective mistakes we can make about phenomenal character. In section 5 I will argue that the superficiality constraint is effectively embedded in the way the notion of phenomenal character is used in the debate, so all parties need to accept it. In section 6 I will formulate an argument against relationalism based on this assumption, and I will briefly discuss what relationalists could say to address it. This will shed a new light on the debate and offer a roadmap to advancing it: whether the cost of accepting relationalism is one that is worth paying will depend on how satisfactory their responses to the argument from superficiality are.

2. Philosophical Disagreement and Incommensurable Paradigms In his contribution to this volume, Fish focuses on the heated dispute between Burge (2005) and McDowell (2011). He suggests that it ‘displays enough of the classic features of a clash of Kuhnian paradigms—just with philosophical theories,

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

88

 

rather than scientific theories, playing the role of paradigms’ (Fish this volume, 32). According to Kuhn, science undergoes alternating phases of normal science and revolutions. In periods of normal science, there is only one prevailing scientific paradigm—a set of beliefs, instruments, metaphysical and methodological assumptions, and models to carry out scientific enquiries. When problematic experimental findings build up and the paradigm encounters increasing difficulties in solving the new puzzles that arise, science enters a phase of revolution, where competing paradigms might emerge. The competing paradigms are, according to Kuhn, incommensurable: there is no objective way (i.e. neutral with respect to the assumptions and methods of each paradigm) of comparing and evaluating them. Fish notices elements of incommensurability in the debate between Burge and McDowell, inasmuch as ‘each finds different ways of speaking to be “natural”, and attempts to translate between the two “languages” are fraught with difficulty and the potential for misunderstanding’ (Fish this volume, 30)² and each has their own distinctive ‘list of problems that any candidate for paradigm must resolve’ (Kuhn 1962, 148, quoted in Fish this volume, 30).³ Because of this incommensurability— Fish argues—competing philosophical views, much like Kuhnian competing scientific paradigms, cannot do anything to convince the opponent. Any argument to this effect will inevitably rely on the assumptions associated with one paradigm, and thus seem compelling only to those who already accept them, but will fail to persuade the opponents who reject those assumptions. While Fish sees elements of incommensurability in the philosophical debate on perception, he is also aware of the limitations of comparing philosophical disputes to Kuhn’s scientific paradigms. The most notable difference he sees lies in the mostly diachronic nature of the Kuhnian model. For Kuhn, it is only in periods of ‘crisis’ that paradigms coexist, while in times of ‘normal science’ all research ² An example of this is McDowell’s understanding of veridical perception in terms of a subject’s ‘openness to the layout of reality’ (McDowell 1994, 26. Quoted by Fish this volume, 29), an expression that reveals his commitment to disjunctivism (if veridical perception is openness to the reality, hallucination is something else completely, which offers the mere appearance of such openness). While he finds this idea ‘completely natural and intuitive’ (2010, 245), Burge finds it rather unintuitive and confusing—nothing more than a misplaced metaphor—and is keen to translate it into his own non-disjunctivist vocabulary: an experience is ‘open to the layout of reality’ if its content is true and it thereby accurately represents the world (Burge 2005, 3. See Fish this volume, 29). ³ According to Fish, McDowell and Burge disagree on whether a philosophical account of perception should be constrained by epistemological considerations on how perception makes knowledge of the world available: McDowell’s account is primarily driven by those, while Burge doesn’t think an account of perception should be constrained by them. They also disagree, Fish claims, on what philosophy and science have to say respectively about consciousness. For McDowell a central task of a philosophical theory of perception to explain consciousness. For Burge consciousness is ‘almost surely [ . . . ] not an essential feature of all perception’ (Burge 2011, 79, quoted by Fish, this volume, 29) therefore it should not be the primary concern of the philosophical investigation of perception. As a result, McDowell and Burge have different standards for what counts as a satisfactory account of perception.

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

      

89

happens within a single prevailing paradigm. This doesn’t seem to be the case in philosophy, where competing theories are the norm rather than an exception. This leads him to turn to the work of another philosopher of science, Lakatos, whose competing research programmes bear some resemblance to Kuhn’s paradigms. For Lakatos, a research programme is successful if the successive changes to the peripheral claims of the theory made to accommodate new data yields novel predictions and some of these predictions are corroborated. Thus, for Lakatos, competing paradigms are not incommensurable, but they can be evaluated only with the benefit of hindsight. While this seems more optimistic than suggesting that competing philosophical views of perception are incommensurable, applying Lakatos’s model of scientific progress to philosophy might prove difficult. As Fish himself notices, philosophy doesn’t always yield testable predictions in the same way that science does, and while some philosophers (Fish mentions for instance Dennett) conceive of their theories as yielding predictions, many philosophers would emphatically deny that philosophy should do such a thing. Most importantly, even if we accept that at least some philosophical theories yield predictions, ‘there may well not yet be any non-contentious way to test those predictions’ (Fish, this volume, 40). Even if there was a way to derive predictions from philosophical theories and test them, ‘philosophers themselves—happiest, as they are, in more conceptual spaces—may neither be best placed nor particularly inclined’ to do so (Fish, this volume, 39). It will be up to scientists to derive predictions from philosophical theories, test them, and report back to philosophers which one of the views they have debated is more successful. Not many philosophers would like to think that deferring to science is ultimately the only way to vindicate their philosophical view. Whichever way you look at it, Fish’s diagnosis of the state of the philosophical debate about perception is disheartening. If we embrace the Kuhnian model, advancing the debate is structurally impossible. If we opt for Lakatos’s model, it will be practically impossible to do so, at least within the realm of philosophy alone. But we needn’t adopt such a defeatist diagnosis. I will argue that a more promising route forward is available. If we divide the overall debate into ‘subdebates’, we may well find that what is hindering progress is not a clash of research programmes, but a series of misunderstandings that might be avoided if philosophers were more careful about disentangling the different questions they are invested in. Some misunderstandings originate in the way the stand-off between relationalism and representationalism is often framed in terms of a disagreement about ‘the essential metaphysical structure’ (Genone 2016, 1), or ‘the most fundamental characterization’ (Brewer 2011, 94) of perception. This is problematic, because philosophers often have different views about precisely what is essential or fundamental to perceptual experiences. For some, this is whatever accounts for the conscious aspect of perception—that is, its phenomenal character. For others,

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

90

 

it’s whatever explains how perception generates knowledge of the world, or justifies beliefs about it, or motivates and prompts behaviours. For yet others, a fundamental characterization is one that explains all these phenomenological, epistemological, and behavioural explanations at once. This lack of clarity offers, as Fish remarks, ample opportunities for misunderstanding and disagreement. But we can go past the unhelpful language often used to frame the debate and avoid talking past each other by making sure we understand the restricted aim of each proposal. Interestingly, most relationalists have made it quite clear that they are primarily concerned with offering an account of the phenomenal character of perception. They think that the non-representational relation plays a fundamental role in explaining the phenomenal character of perception, which is partly constituted by the mind-independent objects one perceives.⁴ As such, they oppose representationalism to the extent that it claims that the representational content of experience can explain the phenomenal character of experience.⁵ In what follows, I will focus on this more confined debate about the phenomenal character of perception. Here, the scope for disagreement about which questions are worth asking is significantly reduced, and it becomes possible to see how we can advance the debate between relationalists and representationalists.

3. Perception, Hallucinations, and Bullets to Bite One of the most effective ways of presenting what motivates the recalcitrant disagreement between relationalists and their opponents is offered by Valberg (1992). According to Valberg, at the heart of the debate in philosophy of perception lies an antinomy between two equally reasonable attitudes towards perception. If we are ‘open to our experience’ and how it strikes us through introspection, we are led to think of experience as a relation of awareness with mind-independent objects, as per relationalism. On the other hand, however, reflection on the possibility of (certain kinds of) hallucinations compels us to deny that perception amounts to a ⁴ Martin (1997, 2002, 2003, 2006), Brewer (2008, 2011), Fish (2008, 2009), Soteriou (2013, 2016), and Campbell (2002), for instance, all frame relationalism as (primarily) offering an account of the phenomenal character of perception. ⁵ This leaves it open that relationalism (understood as an account of phenomenal character) might be compatible with allowing some other explanatory role (e.g. accounting for how perception brings about and motivates beliefs and actions) for the representational content of perception. It should also be noted that representationalism and relationalism are not the only options in this debate. Sense-data theorists claim that perception consists in a direct relation to immaterial objects called sense-data (Moore 1903; Russell 1912; Broad 1925; Price 1932; Robinson 1994; Foster 2000). Another family of theories includes views that account for the phenomenal character of perception in terms of ways of perceiving (adverbialism: see Ducasse 1942; Tye 1975, 1984; Coates 2007); for some of these views, the phenomenal character is determined by intrinsic properties of the experience (the qualia or mental paint view: see Chalmers 1996, 2004; Block 1996, 2003, 2010).

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

      

91

relation to the mind-independent objects that we ordinarily take ourselves to be perceiving. For Valberg, these two attitudes form an antinomy: they are equally compelling, but mutually incompatible, ways of thinking about perception. Once this antinomy becomes manifest, one can either ‘demonstrate that the conflict in which it consists is merely apparent’ (Valberg 1992, 42) or accept it and explain why it ‘really is a conflict’ (Valberg 1992, 42). Disjunctivists go for the first option: they stick to the relationalist intuition and argue that there is a fallacy (or several) in the argument from hallucination. Opponents of disjunctivism might well agree that relationalism would be preferable, but they contend that such an option is simply not available, because the argument from hallucination offers a reductio ad absurdum of relationalism. The argument originates from noticing that we can have hallucinations—i.e. experiences that seem to relate us to mind-independent objects in the absence of any suitable object and arguing that, because we are often unable to tell them apart from veridical perceptions, the same non-relational account that applies to hallucination must be true for perception too. From here, it has developed into many versions, roughly corresponding to subsequent refinements in response to the relationalist’s responses to it. Disjunctivism arises from the relationalist’s attempt to reconcile their view with the possibility of hallucinations: there is no need—they claim—to accept that perception doesn’t involve a relation to a mind-independent object just because hallucinations don’t, and sometimes we take one for the other (see Austin 1962; Hinton 1967a and 1967b). Proponents of the argument from hallucination bring to our attention new counterexamples and thought experiments to challenge this move. Recent discussions have focused on the following thought experiment. Suppose you see a lemon in front of you. Then the lemon is removed, while a scientist activates, through transcranial stimulation, exactly the same areas of the brain that were firing when you were seeing the lemon. You now have a hallucinatory experience as of a lemon which is indistinguishable from the veridical perception of the lemon, but in this case the experience is brought about only by the firing of certain neurons, while there is no lemon in front of you. It is more difficult for the relationalist to deny that a hallucination which is both indistinguishable from a given perception and caused by the same brain activity deserves the same account as a veridical perception. The back and forth of this debate is characteristic of what Casati (2011) calls ‘conceptual negotiation’, a practice that he deems to be at the heart of any philosophical activity. According to Casati, the main aim of conceptual negotiation is to evaluate costs and benefits of competing views, by unpacking implicit assumptions in descriptions, theories, and arguments, and exploring their unexpected consequences. The need for conceptual negotiation arises—Casati claims— when we encounter a tension between competing intuitions. Such a tension often follows the discovery of new facts, either through scientific discoveries, changes in

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

92

 

our society, or simply by new facts being brought to our attention. In our case, the tension arises from the realization that hallucinations are possible, and that this challenges the relationalist intuition. Valberg’s antinomy suggests that taking any position in this debate comes at a cost. Part of the job of conceptual negotiation is to assess how big of a bullet one has to bite when embracing a view. However, as Casati points out, the final decision as to which bullet to bite is extra-philosophical. At some point, philosophy’s job of unpacking the respective costs of competing views comes to an end, and people will still disagree on whether a certain price is worth paying. Here Casati echoes Lewis: Whether or not it would be nice to knock disagreeing philosophers down by sheer force of argument, it cannot be done. Philosophical theories are never refuted conclusively. (Or hardly ever. Gödel and Gettier may have done it.) The theory survives its refutation—at a price. Argle has said what we accomplish in philosophical argument: we measure the price. Perhaps that is something we can settle more or less conclusively. But when all is said and done, and all the tricky arguments and distinctions and counterexamples have been discovered, presumably we will still face the question which prices are worth paying, which theories are on balance credible, which are the unacceptably counterintuitive consequences and which are the acceptably counterintuitive ones. On this question we may still differ. And if all is indeed said and done, there will be no hope of discovering still further arguments to settle our differences. (Lewis 1983, x–xi)

Has the debate on relationalism reached the point where, as Lewis says, all has been said and done and there’s no hope to settle our differences? I will argue that this is not the case. There is still a lot to do to fully assess the costs and benefits of relationalism. The deep-seated scepticism with which disjunctivism is often met suggests that some hidden assumption is at play in the debate. In the next section I will briefly outline what this implicit assumption is and how it can be used to mount a further argument against relationalism, which promises to highlight the true cost of embracing it. The assumption in question—implicit in the very notion of phenomenal character—is the idea that the phenomenal character of experience is superficial with respect to introspection. For this reason, I dub the ensuing argument ‘the argument from superficiality’. Like the argument from hallucination, the argument from superficiality focuses on the possibility of hallucinations and takes the form of a reductio ad absurdum. If Lewis is right, we should not expect any argument (including reductiones) to conclusively refute a theory (despite what some proponents of the argument from hallucination purport). What reductiones show is that a theory is incompatible with certain allegedly compelling claims. However, reductiones can be more or less

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

      

93

persuasive—their persuasive force being conditional to finding the premises of the argument more convincing than the claim to be rejected, in this case relationalism. As mentioned at the beginning of this paper, relationalists contend that the premises used in the various versions of the argument from hallucination beg the question against naive realism, and are less compelling than the intuition relationalism seeks to accommodate—if not downright false. I will argue that the argument from superficiality relies on premises that relationalists cannot easily dismiss as question-begging and that are prima facie at least as compelling as the relationalist intuitions. With this argument in place, we can then examine whether relationalists can accommodate its premises and at what costs, and what would they lose if they fail or refuse to accommodate them. In other words, by considering what relationalists can say in response to the argument, we can better understand the true costs of accepting relationalism. Depending on the results, this may either reveal that the cost to be paid is higher than relationalists had so far realized, or finally appease some of the worries of the opponents. Because this paper focuses on the scope and methodology of the philosophy of perception, I won’t attempt any detailed evaluation of the argument or discussion of what relationalists could say in response to it. The aim of this paper is not to establish whether the merits of relationalism exceed its costs. Rather, it is to show that there is scope for progress in the debate, and in what direction we should move in order to achieve it.

4. Phenomenal Character and Introspection The recent debate has focused on versions of the argument from hallucination that make use of the ‘same cause, same effect’ principle or the principle of local supervenience and pivot around a thought experiment presented above, featuring hallucinations that are brought about by the same brain activity involved in perception. However, I think we can learn something important and advance the debate if we temporarily leave aside the later instantiations of the argument from hallucination and we focus on understanding what lies at the root of the original argument from hallucination and the Indistinguishability Principle (the idea that if two experiences are introspectively indistinguishable through introspection, they must have the same phenomenal character). Proponents of the Indistinguishability Principle are motivated by certain assumptions about introspection and its link to the phenomenal character of perception. Many think that, while introspection is generally fallible, it is infallible when it comes to the access it offers to the phenomenal character of experience (see Farkas 2006; Gertler 2012; Giustina and Kriegel 2017).

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

94

 

This claim of restricted infallibility is highly contentious⁶ and vehemently rejected by relationalists.⁷ While I share these reservations towards the Indistinguishability Principle and the idea of local introspective infallibility, I think that the thought that there is a special link between phenomenal character and introspection which disjunctivism makes difficult to accommodate deserves further examination. The problem is that this thought has so far been mischaracterized by the opponents of relationalism, steering the debate in an unfruitful direction. In what follows I will seek to offer a more accurate characterization of this link and examine what constraints it puts on any account of phenomenal character and how this may put pressure on disjunctivism. ‘Phenomenal character’ is a term of art introduced to focus on what it is like to have an experience, among all the other aspects of perception we may be interested in (its functional role, its object, its causal history, its representational content, if any). In turn, what it is like to have an experience is often specified in terms of what is manifest to the experiencer through introspection. Thus, the role the term ‘phenomenal character’ is designed to fulfil in the current philosophical jargon implies that it is inextricably intertwined with the notion of phenomenal character. This is not to say that the phenomenal character of experience metaphysically depends on introspection: that would be a controversial commitment about its nature that most philosophers would reject.⁸ Yet ‘direct introspective ostension’ (Kriegel 2015, 47) seems to be the only way to fix the reference for ‘phenomenal character’.⁹ This explains why it is so tempting for many to accept (more or less explicitly) the Indistinguishability Principle and dismiss disjunctivism as utterly unbelievable or confused. If an appeal to introspection is part of the descriptor that fixes the reference for ‘phenomenal character’, then introspection is the ultimate authority for knowing what the phenomenal character of an experience is: whatever the phenomenal character an experience seems to have through introspection is the phenomenal character the experience has, because there is nothing else to the phenomenal character of an experience than what appears through introspection. Or so the reasoning goes. ⁶ See Schwitzgebel 2008, 2012; Churchland 1984; Bayne and Spener 2010; Shoemaker 1996 for various arguments (coming from very different perspectives, none of which presupposes relationalism) against this restricted version of introspective infallibility. ⁷ See Hinton 1967a and 1967b; Martin 1997, 2003, 2006; Fish 2009. ⁸ Some see higher-order theories of consciousness (Lycan 1987, 2001; Rosenthal 2000, 2005a, 2005b; Carruthers 2004) as committed to this metaphysical dependence. However, most accounts construe introspection as an activity that involves attention and the deployment of higher-order concepts, they maintain that phenomenal consciousness does not require these higher-order capacities (as creatures devoid of them, or humans that are inattentive or cognitively impaired, possess phenomenally conscious experiences). ⁹ As Block (1978) famously notes, when asked to define phenomenal consciousness, a natural response is to answer as Louis Armstrong did when asked to define jazz: ‘if you have to ask, ain’t never gonna know.’

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

      

95

But this is a non sequitur. If we want to assess whether disjunctivism is a viable option, we need to tease apart which constraints are simply part of the function we have assigned to the notion of ‘phenomenal character’ from further controversial commitments about consciousness, perception, introspection, and so on, which might be surreptitiously built into it. Offering an informative yet neutral characterization of the link between introspection and phenomenal character is a delicate exercise of conceptual negotiation that I can’t hope to settle once and for all here. But I will propose an initial approximation and open up the negotiations. I find that the best way to begin is to focus on the function the notion of phenomenal character is designed to fulfil (that of singling out certain aspects of experience from others, on the basis of their privileged relation to introspection) and to compare it with another notion, which plays a similar role—that of contrasting a definite class of properties from others on the basis of their privileged relation to a source of information. The notion I have in mind is that of ‘observational properties’, used to refer to those properties (such as colour, taste, odours, sound, shape, size) that are thought to bear a special relation to perception. This is in contrast to properties which can also be seen but do not seem to be linked to perception in the same special way, such as natural kinds (being a lemon), artefactual kinds (being a chair), or causal-historical properties (being fresh).¹⁰ ¹⁰ There isn’t consensus about what properties count as observational (for instance, some would claim that colours count but shapes don’t), or even whether there is any property which is truly observational. This is irrelevant for my present purposes. All is needed for the present purposes is to accept that: (i) there is a meaningful distinction to be drawn between observational and non-observational properties and (ii) if one were to maintain that either colour, smell, shape, or size were observational, the best way to account for observationality is in terms of their being superficial with respect to perception. I also don’t need to take a stance on whether we (directly) perceive natural kinds, artefactual kinds, and so on, often referred to as higher-level properties. As I understand it, the question as to what makes certain properties observational is more fundamental than the question as to whether higher-level properties can be directly perceived. We would not be able to ask that question if we didn’t have the intuition that some properties (what I refer to here as observational properties and what are referred to as ‘low-level properties’ in the context of the debate on what can be directly perceived) are more intimately linked to perception than others—as testified by the fact that we take for granted that they are directly perceived while we consider it an open question whether higher-level properties are directly perceived. We can understand those who deny that we can perceive higher-level properties (e.g. Byrne 2009; Price 2009; Brogaard 2013) as offering a simple explanation of the intuitive distinction between observational and higher-level properties: for them what distinguishes observational properties is that they can feature in perception. However, even proponents of the claim that we can perceive higher-level properties (e.g. Bayne 2009; Siegel 2006, 2009; Nanay 2011a, 2011b) can maintain that there is a distinction: what they will need is an alternative explanation for this distinction which is not simply in terms of what can and cannot be directly perceived. What I offer here can be seen as an attempt to clarify what is special about observational properties which remains neutral with respect to the debate on what type of properties feature in perception. Thanks to Heather Logue for pointing out the need to clarify this.

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

96

 

Similarly to what happens with the notion of phenomenal character, often philosophers either fail to specify the nature of observational properties’ relation to perception or give clearly inadequate characterizations, yet assume as obvious the existence of a special connection of some sort with perception.¹¹ However, the notion of ‘observational properties’ is less elusive than that of ‘phenomenal character’, and we can rely on the fact that the relation between perception and observational properties has been more broadly examined. Thus, we can hope to learn some lessons about the phenomenal character of perception by first looking at the relatively more manageable literature on observational properties. Peacocke (1983) notices that observational and non-observational concepts (such as, for instance, concepts of natural kind properties) have a specific pattern of epistemic possibilities. It is possible for something to look like a tomato from all the different angles it may be seen, by a subject whose perceptual mechanism works properly and has the relevant concept of being a tomato, and yet it not be a tomato, but, say, a fake plastic tomato or even a synthetic replica of tomato produced by Kraft laboratories. On the other hand, something cannot seem to a subject with good sight and no relevant cognitive deficiency to be consistently yellow across time, under changing lighting conditions which include optimal ones, without actually being yellow (see Peacocke 1983, 99).¹² If we fail to see that an object is a fake tomato rather than a real tomato, it might be our perception’s fault (maybe our vision is blurry and we can’t tell that the plastic tomato out of a child’s grocery shop set is a fake tomato). But it’s also possible that the fault doesn’t lie with perception: in the case of the synthetic fake tomato, for instance, there is no perceivable difference between the real and the fake tomato. The difference is hidden to perception: it lies in their respective chemical composition. It can only be discovered by non-perceptual means of investigation. In the case of redness, if something appears red but is not red, this can only be because something is wrong with either the conditions of observation or the perceptual mechanism, not because the difference between real red and ‘fake red’ is hidden to perception. An effective metaphor often used to characterize this distinction is in terms of superficiality. A property like redness is superficial in the sense that we can know that it’s instantiated by something on the basis of perception alone. ¹¹ Here is an example of inadequate characterization: ‘Properties are observational in so far as they are presented or represented in perceptual experience. Properties are represented in experience by means of appearances. Thus we can say that observational properties are those properties that objects can appear to have when we perceive them’ (Langsam 2000, 69). This characterization is inadequate because it is too inclusive: objects can appear to have all sorts of properties, not just observational properties such as colours and tastes. An object can appear to be a lemon, stale, or lonely. We can see his characterization of observationality as assuming that we can’t perceive higher-level properties. Or else we lose the ability to distinguish between observational and non-observational properties, because any property that we can perceive would count as observational. ¹² Peacocke focuses on observational concepts, but it is plausible to assume that observational concepts refer to classes of properties to which the epistemic constraint he identifies applies.

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

      

97

5. The Superficiality Constraint This metaphor of superficiality and the related notion of epistemic possibilities are useful for understanding what constraint applies to an account of the relation between phenomenal character and introspection. Remember that the task at hand is to characterize the intuitive link between introspection and phenomenal character in a way which is informative, yet neutral with respect to controversial commitments. Philosophers have tried to characterize it with various notions, such as infallibility, certainty, incorrigibility, or luminosity. But these are all controversial. The idea that the phenomenal character is superficial with respect to introspection in the same way observational properties are superficial with respect to perception is weaker than any of these alternatives, yet it is informative, inasmuch as it grasps the intuitive distinction between, on one hand, phenomenal character and, on the other hand, other aspects of our mental life that may be introspected, but do not bear the same deep link to introspection. We often know through introspection alone what mental kinds we instantiate: we can reflect on our experience and judge that we are undergoing a perceptual experience, or an episode of recollection, or that we feel certain emotions. However, in some cases, whether we are in one kind of mental state as opposed to another is something that goes beyond the scope of introspection. For example, introspection doesn’t offer any clue as to whether the experience is a veridical perception or a hallucination that is subjectively indistinguishable from it. The only way to adjudicate that is to assess how the experience is brought about. On the other hand, it doesn’t seem possible for a difference in phenomenal character to go beyond the scope of what is accessible through introspection and only be discoverable through non-introspective enquiries. To put it in terms similar to those used by Peacocke with respect to observational properties, while introspective mistakes are possible (for instance because one doesn’t pay enough attention or because one’s introspective capacities don’t work properly), the pattern of possible mistakes is constrained by the following rule, which I call ‘the Superficiality Constraint’: (SC): It is not possible that an experience seems through introspection to have a certain type of phenomenal character, while it doesn’t actually have that phenomenal character, and there is no disabling or interfering condition in place that prevents one from introspectively realizing that the experience doesn’t, in fact, have that type of phenomenal character. The Superficiality Constraint is a very minimal claim. It is an epistemic constraint that merely restricts the pattern of possible mistakes about the phenomenal

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

98

 

character of experience and remains neutral with respect to various commitments about the nature of introspection, phenomenal character, perception, and consciousness. As such, it is prima facie compatible with disjunctivism and cannot be said to beg the question against it. To appreciate how the Superficiality Constraint is indeed very modest and neutral, it is useful to contrast it with two other claims in its vicinity: (a) revelation and (b) luminosity. (a) Revelation was introduced by Johnston as an alleged feature of our common-sense understanding of colour: (R): ‘the essential nature of (for example) canary yellow is fully revealed by visual experiences as of canary yellow things.’ (Johnston 1992, 138)

It might be tempting to think that the Superficiality Constraint is akin to Revelation for the phenomenal character of experience. But this would be a mistake. While revelation is about the essential nature of a colour, that which makes a colour the property it is and not something else, the Superficiality Constraint is not about essential properties: it remains silent with respect to the nature of phenomenal characters. It is not about what makes the phenomenal character what it is rather than something else: it more modestly dictates the conditions under which an experience may seem to have a phenomenal character it does not have (whatever makes the phenomenal character what it is). (b) Luminosity is a thesis about mental occurrences that Williamson (2000) defines as follows: ‘(L) For every case α, if in α [the mental occurrence] C obtains , then in α one is in a position to know that C obtains’ (Williamson 2000: 95). While mental occurrences such as pain and bodily sensations are often thought to be luminous, Williamson famously argues that no mental state is luminous. By allowing that there might be interfering conditions that impair one’s introspective capacities, the Superficiality Constraint is weaker than luminosity. If disabling or interfering conditions are in place, one may not be in a position to know that an experience with a certain phenomenal character obtains, although it remains true that the phenomenal character is superficial, because the fact that a certain phenomenal character is present could in principle be detected (where the interfering or disabling conditions are not in place) through introspection alone. Once we have distinguished the Superficiality Constraint from Revelation and Luminosity, we can more easily see in what respects the Superficiality Constraint is minimally committal and, as such, ‘relationalist-friendly’: (i) Unlike Luminosity—and unlike the Indistinguishability Principle, the principle of Same Cause, Same Effect and Local Supervenience used in

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

      

99

the arguments from hallucination—the Superficiality Constraint is compatible with the disjunctivist commitment: due to the possibility of interfering and disabling conditions, it allows for hallucinations to seem to have a (relationalist) phenomenal character that they in fact lack. (ii) Unlike Revelation, the Superficiality Constraint doesn’t commit to the idea that introspection fully reveals the nature of a phenomenal character. This fits very well with what relationalists are likely to say about the phenomenal character of both perception and hallucination. They clearly want to avoid the idea that the nature of the phenomenal character of hallucination is fully revealed by introspection. And while they are likely to think that there is some introspective support for their claims about the phenomenal character of veridical perception, they are likely to think that additional extra-introspective considerations are required to support their view.¹³ As we have seen in the previous section, the idea of a special link between introspection and phenomenal character is part and parcel of the notion of phenomenal character. Denying such a link would be tantamount to refusing to use ‘phenomenal character’ in the way it is arguably designed to be used when introduced—i.e. to refer to that aspect of experience that bears a special connection to introspection. If one were to deny the existence of such a connection, one would not simply embrace a controversial explanation of what the phenomenal character of perception is, one would use the locution ‘phenomenal character’ in a way that changes the explanandum, and fails to refer to what everybody else in the debate means by it. The idea that some special link exists between introspection and the phenomenal character of perception is what has more or less implicitly motivated the widespread resistance towards relationalism. However, opponents so far have failed to identify the specific nature of this link, which hindered any progress in assessing whether or not there is some tension between how we think of phenomenal character and disjunctivism. With the notion of the Superficiality Constraint on hand, we can now hope to make some progress in this direction.

6. Advancing the Debate with the Argument from Superficiality The Superficiality Constraint is something that relationalists can prima facie accept, because, contrary to the Indistinguishability Principle, it allows for introspective mistakes about phenomenal character. ¹³ An example of argument which heavily relies on extra-introspective considerations is the argument from transparency in Martin (2002).

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

100

 

Disjunctivists can explain indistinguishable hallucinations without violating the Superficiality Constraint. They can say that something interferes with or disables the subject’s introspective capacities. For instance, the subject might not be attentive enough, or their brain might not work properly.¹⁴ However, it becomes problematic when it comes to accommodating the type of hallucinations introduced in the thought experiment presented in section 3—hallucinations that are indistinguishable from a perception and brought about by the same neural activity. In this thought experiment, ex hypothesis, there is no interfering or disabling condition to which disjunctivists can appeal to explain why hallucination seems to have a relationalist phenomenal character (one that relationally puts us in contact with objects and is constituted by them). It is part of the set-up of the thought experiment that the brain activity is identical in the two situations, that the subject is employing exactly the same cognitive capacities, is exerting the same degree of attention, and no other occurrence in their stream of consciousness interferes with one’s introspective capacities. The only difference between a perception and the corresponding perfect hallucination is that the object is present in the former case and absent in the latter. But, in order for a condition to interfere with introspection, it must be something that is potentially relevant to introspection. It is not clear how the mere absence of an object can be relevant to one’s introspective capacities. In this respect, introspection is different from perception, and the analogy between phenomenal character and observational properties breaks down. Perception is sensitive to objects in the world. For this reason, facts about the world (e.g. the presence of an occluding object, or of certain lighting conditions) count as disabling or interfering conditions for seeing that an object has a certain property. But introspection tracks mental occurrences, so it is natural to think that only facts about one’s psychological state and the neurological mechanisms that underlie them can count as disabling and interfering conditions. Thus, disjunctivism seems committed to denying that the phenomenal character of these types of hallucination is subject to the Superficiality Constraint. This makes the position untenable, as the way we use the notion of phenomenal character in the philosophical jargon requires us to accept the Superficiality Constraint. Or so the argument goes.

¹⁴ What counts as an interfering or disabling condition for introspection depends in part on how we conceive of introspection. The more pluralist one’s account of introspection is (where multiple processes and capacities are involved, rather than a single mechanism), the more opportunity for disabling and interfering conditions there are. Without committing to any specific account of introspection, we can say that, since attention is certainly a crucial component of introspection, any factor that may interfere with or disable one’s introspective attention counts as a potential disabling or interfering condition. This might be the presence of competing stimuli that divert attention (such as overwhelming emotions), altered brain states due to illness or drug use, or impaired cognitive capacities.

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

      

101

This argument, which I call ‘argument from superficiality’, only relies on unpacking implicit assumptions that are part and parcel of the way the notion of ‘phenomenal character’ is employed in the philosophical debate. Therefore, it is more compelling than the various versions of argument from hallucination, whose key assumptions are deemed question-begging by relationalists. Of course there are a number of things disjunctivists can say in response to this argument. I don’t claim to have offered a conclusive argument against disjunctivism: as we have seen, theories survive their refutations. On the contrary, having identified and spelt out the source of the (as of yet implicit) resistance towards the view can give disjunctivists the tools to address these worries and win over some opponents, especially those who might appreciate the benefits of relationalism but are worried by its implications. The first option open to relationalists is of course to reject the Superficiality Constraint. They might grant that the Superficiality Constraint is constitutive of the way most people use the notion of phenomenal character, but they use it differently. What they mean by ‘phenomenal character’ is the property of relating the subject to certain objects in the environment. If this commits them to violating the Superficiality Constraint, then tant pis for superficiality: for them the phenomenal character of an experience goes beyond what is accessible through introspection. This, however, is not an easy solution. I have argued that the Superficiality Constraint doesn’t follow from a particular view of phenomenal character, but rather simply from the way we assign the reference of ‘phenomenal character’. There can’t be a disagreement about the nature of phenomenal character if we don’t agree on what we talk about when we talk about phenomenal character. And if my analysis in the previous section was correct, they couldn’t engage in any debate with their opponents about the nature of phenomenal character if they were to deny the Superficiality Constraint: they would just use the same word to refer to something else entirely. If relationalists seriously want to pursue this strategy, they cannot simply reject the superficiality constraint, but they have to offer an alternative understanding of the relation between phenomenal character and introspection, one that accommodates our intuitions equally well as the Superficiality Constraint. Another option is to accept the Superficiality Constraint but avoid the problem it poses for hallucinations by arguing that hallucinations don’t have any phenomenal character. This is a strategy already adopted by some relationalists (Fish 2009; Dokic and Martin 2012; Logue 2012) in response to the traditional argument from hallucination. Eliminativism about the phenomenal character of hallucination faces several objections, the most compelling of which, in my opinion, is the following. In this view, the apparent sensory conscious phenomenology is due to cognitive phenomenology (see Fish 2009, 98–9, n. 19). But it is not clear that one can accurately account for the (apparent, according to eliminativists) sensory

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

102

 

phenomenology of hallucinations in terms of cognitive phenomenology (see Vega-Encabo 2010, 190).¹⁵ Additionally, this strategy won’t work, because as long as we concede that it falsely seems to a hallucinating subject that they are in a state with phenomenal character, one would still violate the superficiality constraint.¹⁶ What seems to me the most promising option for relationalists is to show that their view is compatible with the Superficiality Constraint. They could do so by insisting that in the hallucinatory case there is something that interferes with one’s introspective capacities. The relationalist could insist that, since the object constitutes the phenomenal character of one’s veridical perception, and this is something one can introspect, the lack of the object that seems to constitute the phenomenal character of hallucination is relevant to introspection. Interestingly, some relationalists have proposed theories of introspection that could support such a claim. The most developed theory of introspection of this kind I am aware of is proposed by Soteriou (2013, ch. 8). He argues that introspection works by focusing one’s attention on the object of perception by focusing on how they subjectively strike us: introspection is then structurally parasitical on perception. So the mere fact that in introspection one fails to be in the perceptual relation to the world one takes herself to be is enough to impair one’s ability to introspect properly and access the real phenomenal character of perfect hallucination.¹⁷ There are certainly more things relationalists could say to try and accommodate the Superficiality Constraint or to replace it with a less problematic specification of the relation between phenomenal character and introspection. Exploring their options will in any case lead to a better understanding of their commitments and will offer new elements to help us decide whether the costs of accepting disjunctivism exceed its benefits. What’s noteworthy is that trying to deny the Superficiality Constraint and trying to accommodate it seem to both depend on a clarification of what relationalists take introspection of the phenomenal character to be. This suggests that the main disagreement between relationalists and their opponents may lie in how they respectively understand introspection and that the focus of the debate should shift in that direction. Fish identifies a lack of progress in the debate, and attributes it to a structural (and, as such, insurmountable) problem: the incommensurability of the paradigms adopted by proponents of alternative views. Instead, I have argued that the ¹⁵ As Pautz (2013) notices, the viability of this proposal depends on how satisfactorily one can counter traditional arguments against Rosenthal’s ‘higher-order thought theory’ of consciousness, on which Fish’s idea that consciousness of hallucination is a by-product of cognitive attitudes heavily relies. ¹⁶ Thanks to Heather Logue for pointing this out to me. ¹⁷ It seems to me that a view of introspection along these lines plays a central role in much of the arguments for disjunctivism in Martin (2006). A similar theory has been proposed by Logue (2012), who pairs it with a commitment to eliminativism about the phenomenal character of hallucination.

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

      

103

current stall is due to specific shortcomings—in particular, failing to acknowledge, from both sides of the debate, the deep-seated intuition that underlies the rejection of disjunctivism (i.e. the Superficiality Constraint). This diagnosis of the debate is preferable because it makes space for the possibility of adjudicating the debate and offers some suggestion as to what could tip the scale one way or another.¹⁸

References Austin, J. L. 1962. Sense and Sensibilia. Oxford: Oxford University Press. Bayne, Tim. 2009. ‘Perception and the Reach of Phenomenal Content’. Philosophical Quarterly 59 (236): 385–404. Bayne, Tim and Maja Spener. 2010. ‘Introspective Humility’. Philosophical Issues 20 (1): 1–22. Block, Ned. 1978. ‘Troubles with Functionalism’. In W. Savage (ed.) Perception and Cognition. Minneapolis: University of Minnesota Press, 261–325. Block, Ned. 1996. ‘Mental Paint and Mental Latex’. Philosophical Issues 7: 19–49. Block, Ned. 2003. ‘Mental Paint’. In Martin Hahn and B. Ramberg (eds.) Reflections and Replies: Essays on the Philosophy of Tyler Burge. Cambridge, MA: MIT Press, 165–200. Block, Ned. 2010. ‘Attention and Mental Paint’. Philosophical Issues 20 (1): 23–63. Brewer, Bill. 2008. ‘How to Account for Illusion’. In Adrian Haddock and Fiona Macpherson (eds.) Disjunctivism: Perception, Action, Knowledge. Oxford: Oxford University Press, 168–80. Brewer, Bill. 2011. Perception and Its Objects. Oxford: Oxford University Press. Broad, Charlie D. 1925. The Mind and Its Place in Nature. London: Routledge and Kegan Paul. Brogaard, Berit. 2013. ‘Do We Perceive Natural Kind Properties?’ Philosophical Studies 162 (1): 35–42. Byrne, Alex. 2009. ‘Experience and Content’. Philosophical Quarterly 59 (236): 429–51. Burge, Tyler. 2005. ‘Disjunctivism and Perceptual Psychology’. Philosophical Topics 33 (1): 1–78. Burge, Tyler. 2011. ‘Disjunctivism Again’. Philosophical Explorations 14 (1): 43–80. Campbell, John. 2002. Reference and Consciousness. Oxford: Oxford University Press.

¹⁸ I would like to thank Hemdat Lerman, Matthew Soteriou, Maarten Steenhagen, Daniel Vanello, and the editors Heather Logue and Louise Richardson for their insightful and helpful comments on drafts of this paper. I also benefited greatly from feedback by many people, when I presented some of the material discussed here in various occasions. Among those, I would like to especially thank Bill Brewer, Chiara Brozzo, Dan Cavedon-Taylor, Mark Kalderon, Keith Wilson, and Hong Yu Wong.

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

104

 

Carruthers, Peter 2004. ‘HOP over FOR, HOT theory’. In R. J. Gennaro (ed.) HigherOrder Theories of Consciousness: An Anthology. Amsterdam: John Benjamins, 115–36. Casati, Roberto. 2011. Prima Lezione Di Filosofia. Laterza. Chalmers, David J. 1996. The Conscious Mind: In Search of a Fundamental Theory. Oxford: Oxford University Press. Chalmers, David J. 2004. ‘The Representational Character of Experience’. In Brian Leiter (ed.) The Future for Philosophy. Oxford: Oxford University Press, 153–81. Churchland, Paul M. 1984. Matter and Consciousness. Cambridge, MA: MIT Press. Coates, P. 2007. The Metaphysics of Perception: Wilfrid Sellars, Perceptual Consciousness and Critical Realism. London: Routledge. Deutsch, Max. 2005. ‘Intentionalism and Intransitivity’. Synthese 144 (1): 1–22. Dokic, Jérôme and Jean-Rémy Martin. 2012. ‘Disjunctivism, Hallucination and Metacognition’. WIREs Cognitive Science 3: 533–43. Ducasse, C.J. 1942. ‘Moore’s Refutation of Idealism’. In P.A. Schlipp (ed.) The Philosophy of G.E. Moore, Vol. I, 225–51. Cambridge: Cambridge University Press. Farkas, Katalin. 2006. ‘Indiscriminability and the Sameness of Appearance’. Proceedings of the Aristotelian Society 106 (2): 39–59. Fish, William C. 2008. ‘Disjunctivism, Indistinguishability, and the Nature of Hallucination’. In Adrian Haddock and Fiona Macpherson (eds.) Disjunctivism: Perception, Action, Knowledge. Oxford: Oxford University Press. Fish, William C. 2009. Perception, Hallucination, and Illusion. New York: Oxford University Press. Fish, William C. this volume. ‘Perceptual Paradigms’. In Heather Logue & Louise Richardson (eds.), Purpose and Procedure in Philosophy of Perception. Oxford: Oxford University Press. Foster, John A. 2000. The Nature of Perception. Oxford: Oxford University Press. Genone, James. 2016. ‘Recent Work on Naive Realism’. American Philosophical Quarterly 53: 1–26. Gertler, Brie. 2012. ‘Renewed Acquaintance’. In Declan Smithies and Daniel Stoljar (eds.) Introspection and Consciousness. New York: Oxford University Press, 89–123. Giustina, Anna and Uriah Kriegel. 2017. ‘Fact-Introspection, Thing-Introspection, and Inner Awareness’. Review of Philosophy and Psychology 7: 1–22. Haddock, Adrian and Fiona Macpherson. 2008. ‘Introduction: Varieties of Disjunctivism’. In Adrian Haddock and Fiona Macpherson (eds.) Disjunctivism: Perception, Action, Knowledge. Oxford: Oxford University Press. Hinton, J. M. 1967a. ‘Experiences’ Philosophical Quarterly 17 (66): 1–13. Hinton, J. M. 1967b. ‘Visual Experiences’. Mind 76 (April): 217–27. Johnston, Mark. 1992. ‘How to Speak of the Colors’. Philosophical Studies 68 (3): 221–263. Knight, Gordon. 2014. ‘Disjunctivism Unmotivated’. Phenomenology and the Cognitive Sciences 13 (2): 355–372.

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

      

105

Kriegel, Uriah. 2013. ‘The Phenomenal Intentionality Research Program’. In U. Kriegel (ed.) Phenomenal Intentionality. Oxford: Oxford University Press. Kriegel. Uriah. 2015. The Varieties of Consciousness. New York: Oxford University Press. Kuhn, T. S. 1962. The Structure of Scientific Revolutions. Chicago: University of Chicago Press. Langsam, Harold. 2000. ‘Why Colours Do Look Like Dispositions’. Philosophical Quarterly 50 (198): 68–75. Lewis, David. 1983. Philosophical Papers, Vol. 1. New York: Oxford University Press. Logue, Heather. 2012. ‘What Should the Naïve Realist Say About Total Hallucinations?’ Philosophical Perspectives 26 (1): 173–99. Lycan, William G. 1987. Consciousness. Cambridge, MA: MIT Press. Lycan, William G. 2001. ‘A simple argument for a higher-order representation theory of consciousness’. Analysis 61 (1): 3–4. Martin, Michael G. F. 1997. ‘The Reality of Appearances’. In M. Sainsbury (ed.) Thought and Ontology. Milan: Franco Angeli. Martin, Michael G. F. 2002. ‘The Transparency of Experience’. Mind and Language 4 (4): 376–425. Martin, Michael G. F. 2003. ‘Sensible Appearances’. In T. Baldwin (ed.) The Cambridge History of Philosophy. Cambridge: Cambridge University Press. Martin, Michael G. F. 2006. ‘On Being Alienated’. In Tamar S. Gendler and John Hawthorne (eds.) Perceptual Experience. Oxford: Oxford University Press. McDowell, John. 1994. Mind and World. Cambridge, MA: Harvard University Press. McDowell, John. 2011. ‘Tyler Burge on Disjunctivism’. Philosophical Explorations 13 (3): 243–55. Moore, G. E. 1903. ‘The Refutation of Idealism’. Mind 12 (48): 433–53. Nanay, Bence. 2011a. ‘Do we Sense Modalities with our Sense Modalities?’ Ratio 24 (3): 299–310. Nanay, Bence. 2011b. ‘Do we See Apples as Edible?’ Pacific Philosophical Quarterly 92 (3): 305–22. Pautz, Adam. 2013. ‘Do the Benefits of Naïve Realism Outweigh the Costs? Comments on Fish, Perception, Hallucination and Illusion’. Philosophical Studies 163 (1): 25–36. Peacocke, Christopher. 1983. Sense and Content: Experience, Thought, and Their Relations. Oxford: Oxford University Press. Price, Henry H. 1932. Perception. Second edition. London: Methuen. Price, Richard. 2009. ‘Aspect-Switching and Visual Phenomenal Character’. Philosophical Quarterly 59 (236): 508–518. Robinson, Howard. 1994. Perception. London: Routledge. Rosenthal, David. 2000. ‘Metacognition and Higher-Order Thoughts’. Consciousness and Cognition 9: 231–42.

OUP CORRECTED AUTOPAGE PROOFs – FINAL, 18/3/2021, SPi

106

 

Rosenthal, David. 2005a. ‘The Higher-Order Model of Consciousness’. In Rita Carter (ed.) Consciousness. London: Weidenfeld & Nicolson. Rosenthal, David. 2005b. Consciousness and Mind. Oxford: Oxford University Press. Russell, Bertrand, 1912/1997. The Problems of Philosophy. Oxford: Oxford University Press. Schwitzgebel, Eric. 2008. ‘The Unreliability of Naive Introspection’. Philosophical Review 117 (2): 245–73. Schwitzgebel, Eric. 2012. ‘Introspection, What?’ In Declan Smithies and Daniel Stoljar (eds.) Introspection and Consciousness. Oxford: Oxford University Press, 29–48. Shoemaker, Sydney. 1996. The First-Person Perspective and Other Essays. Cambridge: Cambridge University Press. Siegel, Susanna. 2004. ‘Indiscriminability and the Phenomenal’. Philosophical Studies 120: 91–112. Siegel, Susanna. 2006. ‘Which Properties are Represented in Perception?’ In T. Gendler and J. Hawthorne (eds.) Perceptual Experience. Oxford: Oxford University Press. Siegel, Susanna. 2008. ‘The Epistemic Conception of Hallucination’. In Adrian Haddock and Fiona Macpherson (eds.) Disjunctivism: Perception, Action and Knowledge. Oxford: Oxford University Press. Siegel, Susanna. 2009. ‘The Visual Experience of Causation’. Philosophical Quarterly 59 (236): 519–40. Siewert, Charles. 1998. The Significance of Consciousness. Princeton: Princeton University Press. Smith, A. D. 2002. The Problem of Perception. Harvard: Harvard University Press. Smith, A. D. 2008. ‘Disjunctivism and Discriminability’. In Adrian Haddock and Fiona Macpherson (eds.) Disjunctivism: Perception, Action and Knowledge. Oxford: Oxford University Press. Sollberger, Michael. 2007. ‘The Causal Argument Against Disjunctiivsm’. Facta Philosophica 9 (1): 245–67. Soteriou, Matthew. 2013. The Mind’s Construction: The Ontology of Mind and Mental Action. Oxford: Oxford University Press. Soteriou, Matthew. 2016. Disjunctivism. London: Routledge. Tye, Michael. 1975. ‘The Adverbial Theory: A Defence of Sellars against Jackson’. Metaphilosophy 6 (2): 136–43. Tye, Michael. 1984. ‘The Adverbial Approach to Visual Experience’. Philosophical Review 93 (2): 195–225. Valberg, Jerome J. 1992. ‘The Puzzle of Experience’. In Tim Crane (ed.) The Contents of Experience: Essays on Perception. Oxford: Oxford University Press, 18–47. Vega-Encabo, Jesús. 2010. ‘Hallucinations for Disjunctivists’. Phenomenology and the Cognitive Sciences 9 (2): 281–93. Williamson, Timothy. 2000. Knowledge and its Limits. Oxford: Oxford University Press.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

6 Neopragmatism and Philosophy of Perception Joshua Gert

1. Introduction It is a truism that, other things being equal, philosophical arguments do well to avoid controversial premises. But it is equally obvious that they have to start somewhere. Neopragmatism starts by self-consciously appealing only to mundane empirical facts about human beings that are almost completely philosophically uncontroversial: at least as uncontroversial as anything in the empirical sciences. Here are examples of such facts: we live in a physical world, are biological creatures that have evolved over time, and have senses that are responsive to the physical properties of things in our environment. Also, we have an innate capacity to learn language, both by being explicitly trained and by observing the linguistic behaviour of adults. With these facts firmly in mind, the neopragmatist notes that we talk about such things as possibilities, numbers, social groups, colours, rightness and wrongness, and—of course—sensations and perceptual experiences. This kind of talk can give rise to a worry about how we come to know anything about such things, especially in cases in which they seem to have no causal powers, or to be inaccessible to our senses, or to be essentially private and incommunicable in nature. Traditional philosophical approaches to these worries address them by seeking either to reduce these problematic things to less worrisome entities, or to show that they are merely useful fictions, or to make their odd metaphysical status somehow more palatable. The neopragmatist, in contrast, focuses on language as a social practice, and seeks to explain the function of the talk that gave rise to the worries. If this talk can be fully explained without mystery, the worries disappear. There is an additional assumption that it is useful to make when applying neopragmatism to the domain of perception. This is that while it is (of course) perfectly in order to talk about the referent of a word or concept, and the truth or falsity of an utterance or proposition, we should not think of reference as a relation or of truth as a property. Rather, we should adopt deflationary views of truth and reference. This assumption is certainly controversial. But it is not a basic assumption of neopragmatism itself. Rather, as discussed below, it is the result of applying

Joshua Gert, Neopragmatism and Philosophy of Perception In: Purpose and Procedure in Philosophy of Perception. Edited by: Heather Logue and Louise Richardson, Oxford University Press (2021). © OUP. DOI: 10.1093/oso/9780198853534.003.0006

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

108

 

the neopragmatist method to truth-talk and reference-talk themselves.¹ That is, neopragmatists argue for these deflationary views, given the much less controversial starting points I have already mentioned. Given the focus of this chapter on philosophy of perception, I will be taking deflationism about truth and reference for granted. Given the above starting points, the neopragmatist then sets out to demystify mystifying phenomena on a case-by-case basis. The relevant sort of demystification is not achieved by offering positive explanations of the fundamental metaphysical nature of the basic entities or relations of whatever domain puzzles us. Of course, the neopragmatist strategy is often usefully supplemented by pointing out the implausible metaphysical consequences of rival views—and I myself will be doing this. Still, the main type of question that the neopragmatist thinks the philosopher ought to be answering is ‘How do I explain the talk that gives rise to this puzzle?’ In the domain of philosophy of perception, a primary puzzle is— unsurprisingly—the puzzle of perception. This puzzle begins by registering the intuition that veridical perceptual experience presents us, in a direct, nonmediated way, with mind-independent objects. But, it then notes, we can have the same perceptual experience whether we are actually seeing a goat or are only hallucinating one. This seems to entail that even in cases of veridical perception, what we are aware of is not actually the goat. The puzzle of perception can push a philosopher in various directions. Historically, some were led to endorse the sense-data theory.² More recently, naïve realists about perception have been led to disjunctive theories that deny the puzzle’s presupposition: that we have the same fundamental kind of experience in hallucinations and veridical perceptions.³ Contrary to this claim, disjunctivists hold that indistinguishable veridical and hallucinatory experiences are of fundamentally different natures. It is now almost unanimously held that the sense-data theory is hopeless. And even those who defend disjunctive views often allow that they are not really attractive on their face. Rather, disjunctivism is primarily to be defended as necessary for the coherence of naïve realism, which is supposed to be attractive for independent reasons. The neopragmatist will try to explain why it is appropriate for us to talk in the ways that lead philosophers to endorse sense-data theory or disjunctivism, without committing us to either of those controversial theories. The rest of this paper is divided into three sections. Section 2 explains neopragmatism in more detail. Section 3 illustrates the way in which neopragmatist explanations of puzzling phenomena can undermine the motivation for contentious and metaphysically heavyweight accounts of perceptual experience. Since neopragmatist techniques have already been deployed against sense-data theories, ¹ In this connection, see Price (1988, 2003) and Båve (2009). ² Russell (1912); Price (1932); Jackson (1977). ³ Martin (2006); Logue (2012).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

109

and since I myself have deployed them elsewhere against representationalist views, the target for demystification will be the phenomena that lead theorists to endorse naïve realism and its accompanying disjunctivism.⁴ In section 4 I sketch a positive view of perception to which neopragmatism might lead us: a form of adverbialism that relies on the idea that our sensory states are information-bearing, but not, in any robust sense, representational.

2. Neopragmatism Neopragmatism takes very seriously the idea that human linguistic practice is a natural phenomenon: a social practice of making noises, marks, or gestures of various sorts, that allows us to do a wide variety of things. The neopragmatist stresses that the assertoric form is just one of many forms that utterances can take. There are also orders, requests, greetings, thankings, expressions of pain, pleasure, and panic, and many others. By avoiding thinking of language as, in the first instance, a means for expressing thoughts—thoughts that we might have had quite independently of language—the neopragmatist avoids the need to explain how we can think about, say, numbers, prior to learning number-talk. Relatedly, it allows the neopragmatist to explain how we can say, and mean, such things as ‘three plus eight is eleven’ without having had any causal or mental contact with metaphysically weird entities called ‘the numbers’.⁵ Similarly, it allows us to explain how it is that we can say such things as ‘I’m not sure what colour it actually is, but it has a green look to me because I’ve just been looking at a red light’ without having the capacity to notice a special sort of thing—a look—that is quite different from a visible property of an object.⁶ The neopragmatist’s focus on the idea that language is a tool for getting things done means that, for her, the important philosophical question will typically be ‘What is this word or set of words doing for us?’ rather than ‘What does this word refer to?’ Call the first of these questions the pragmatic question, and the second the correspondence question. To think that the correspondence question is the right one to ask whenever one is philosophizing about a puzzling philosophical notion is to presuppose the following three things: (1)

There is a univocal and substantive relation of reference.

(2)

It is the job of object and property words to refer to objects and properties.

⁴ For neopragmatist arguments against sense data, see Sellars (1963). For my own arguments against representationalism, see Gert (2017, ch. 8). ⁵ An explanation of this sort is given by Paul Benacerraf in Benacerraf (1983). ⁶ An explanation of this sort is given by Wilfrid Sellars in Sellars (1963).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

110

 

(3)

Once one figures out which objects and properties these are, one’s task is basically finished, since this information will tell you everything philosophically interesting that you need to know about the proper use of the term.

But it is notable that the pragmatic question can be asked of the word ‘refer’ itself, and of related words such as ‘true’ and ‘represent’. And it is not only notable, but plausible, that the answer to the pragmatic question, in these cases, is something that does not require us to find a relation ‘out there in the world’ to which the term ‘reference’ refers. Rather, as has already been mentioned, one might adopt any of a variety of deflationary accounts of truth-talk and reference-talk. Huw Price is a neopragmatist who offers a distinctive deflationary explanation truth-talk.⁷ Put most simply, his view is that the assertoric form, and words such as ‘true’ and ‘false’, help members of a linguistic community push each other into a uniformity of psychological states, in cases in which such uniformity is useful. To see how this works, let us step back for a moment from neopragmatism, and consider expressivism. Expressivist views have been advanced to account for our use of assertoric language in the domains of morality, causality, and probability, among others. Expressivists about these domains hold that simple assertions of the relevant sort express psychological states that are not, strictly speaking, true or false, since they do not express beliefs that can ‘match the world’. But they nevertheless hold that it is not surprising that we use the assertoric form to express our moral views, or our probability estimates; it is the assertoric form that signals the appropriateness of truth-assessments and the reason-based arguments that back up such assessments. The happy result of applying truthassessments to utterances that are not, strictly speaking, true or false, is that people end up sharing, say, expectations about what sort of behaviour gets punished (morality), or about what will happen as the result of certain physical interventions in the world (causality), or confidence levels regarding whether or not something is the case (probability). I’ve just noted that, for the expressivist, when an utterance takes the assertoric form, it can express a ‘preferred’ psychological state—one that will be assessed as true by someone who is also in that state—or it can express a state that conflicts with this, and that will be assessed as false by such a person. This forms the basis of an explanation of the fact that ‘quasi-assertions’ in the domains of morality, causality, or probability behave just like ‘real’ assertions, in the sense of obeying the law of non-contradiction. And that is the beginning of an explanation of the fact that standard first-order logic applies even to these quasi-real domains. At this point, a natural question is how to distinguish genuinely realistic discourse from merely quasi-realistic discourse, if we grant the success of the expressivist’s

⁷ Price (1988, 2003, 2013).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

111

quasi-realist project. The neopragmatist position can be seen as a result of challenging the main presupposition of this question: that there is in fact an important metaphysical distinction to draw here. That is, the neopragmatist holds that the explanation of truth-talk in domains such as morality, causality, probability, and so on, is the very same explanation of truth-talk in the domains of middle-sized dry goods and physics. We do not need a property of truth to explain truth-talk: not even the old-style pragmatist’s idea that the true is the useful. Rather, what is useful is (a) the whole practice of making assertions and assessing them as true or false and (b) the whole practice of talking about, say, morality, modality, and so on. An isolated assertion need not be useful in order to be true. As goes truth, so goes reference. That is, the neopragmatist holds that it is a mistake to think that reference-talk concerns a relation between a word and its referent. Rather, we call certain expressions ‘referring expressions’ in virtue of the way they are used in assertions. For referring expressions, it always makes sense to ask ‘What does this expression refer to?’, just as, for assertions, it always makes sense to ask ‘Is this assertion true?’ Paralleling the case of truth, the question about reference is apt even if ‘refers’ is understood in a deflationary way. On one such view (admittedly, an overly simple one) to say that ‘X’ refers to Y is simply a way of saying that X is Y. And since the latter claim does not assert a semantic relation between a linguistic item and a thing, neither should the former be so understood.⁸ Deflationary accounts of reference save us the trouble of finding naturalistic entities—or non-natural or supernatural entities—to which number words, or colour words, or psychological words bear the reference relation, since there is no such relation. That is, it relieves us of the obligation to answer the correspondence question. Of course, it also sets us the task of answering the pragmatic question. Because the neopragmatist recognizes that language is a public practice, and one that is taught by each generation to the next, she will take our ground-level practices to be ones that are teachable by means of public criteria. That is, a language teacher will have to rely on publicly observable phenomena in order to correct that language-learner’s utterances. So the first property words taught will include those for observable properties, like colours and shapes, and kind and stuff words. Mastery of the use of words for these sorts of properties and things is mastery of their meanings. Nothing else is required to know their meanings. It is important to note that in explaining how language is taught, it is possible— indeed, unavoidable—that we make use of some vocabulary that we have not yet explained the teaching of. So, for example, in explaining how a child learns the word ‘red’, we will say such things as ‘If the child calls something “red” when it isn’t actually red, but only looks red because of the lighting, the teacher will correct ⁸ See Båve (2009) for a more adequate deflationary account that is somewhat too complex to present here.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

112

 

the child, perhaps saying that it only looks red.’ This explanation makes use of the word ‘red’. To those who are opposed to neopragmatism, this can wrongly seem to involve an objectionable sort of circularity. But in fact it is no more problematic than it is to use a typewriter in order to type out instructions for assembling a typewriter. Knowing the meaning of a word such as ‘red’ is much like knowing how to type; it is a skill, and is not particularly hard to master, given instruction. Knowing what it is for someone to know the meaning of the word ‘red’, on the other hand, is like knowing how a typewriter is put together; it is a matter of theoretical knowledge. There is nothing suspicious in granting ourselves or our audience the first sort of knowledge as we go about pursuing or imparting the second. Whenever neopragmatism is the right strategy for demystifying a particular philosophical issue, it is important to keep in mind the fact that language is a social practice. It is especially important to keep this in mind when theorizing about such things as sensations and perceptual experience, since if one loses sight of it then the temptation becomes very great to think that the best way of investigating such phenomena is by introspection. How else is one supposed to get knowledge of the referents of such items? This is why, as Bill Fish observes, the main theses that underwrite much contemporary philosophy of perception ‘are motivated, in part, by our first-person understanding of what is involved in being a perceiver’.⁹ Neopragmatism thinks these motivations should—at best—be treated with great caution, since it is not from the first-personal perspective that we see the paradigmatic uses of talk of perception, sensation, and experience.

3. A Case Study in Demystification: Naïve Realism Now that neopragmatism has been sketched, I turn to some criticisms of naïve realism, and its accompanying disjunctivism. In the first subsection, 3.1, I’ll explain what these views amount to, and highlight some questions to which they give rise. Then, in subsection 3.2, I’ll use neopragmatism to provide a nonmysterious explanation of some of the phenomena that have led theorists to endorse naïve realism. One is what has been called the openness of visual experience.¹⁰ And another is the fact that perceptual experience can give us insight into what external, mind-independent perceptible properties are like. The neopragmatist therefore is offering a characteristic two-pronged attack; first, the metaphysical commitments of a target view are shown to lead to mysteries. And, second, it is shown how attention to language deprives the target view’s starting points of the force they were taken to have. ⁹ Fish (2010, 8). ¹⁰ Crane and French (2017). It is also often called ‘presentational character’. See Fish (2009), 11–16.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

113

3.1 Naïve Realism and Disjunctivism According to naïve realism, the fundamental metaphysical nature of the perceptual experience that one has when one actually sees a kitten is in a certain way like the fundamental nature of kicking a stone. Both the experience and the kicking are essentially relational. When one kicks a stone, one is in direct physical contact with the stone. And when one has a veridical perceptual experience of a kitten, one is in direct perceptual contact with the kitten. Neither the experience, nor the kicking, could be what it is, fundamentally, without its object actually being part of it. Kicking a stone is not simply making certain motions with one’s foot—not even when there happens to be a stone in the right place. Rather, to kick a stone is to be in a certain kind of direct physical contact with the stone. Similarly, to have a veridical perceptual experience of a kitten is—according to the naïve realist—to be in direct perceptual contact with it; it is not factorable into the experience and the object in such a way that one could possibly have the very same experience without the object. One of the consequences of naïve realism has taken on a life of its own: the thesis of metaphysical disjunctivism. According to this view, it is a mistake to think that the fundamental nature of a veridical experience of a kitten could be the same as the fundamental nature of even a perfect hallucination of a kitten. After all, in the latter case, there is no kitten there, so there is no perceptual relation to a kitten. Disjunctivism is inconsistent with the sense-data view, according to which the difference between seeing and hallucinating a kitten is not to be found in the fundamental nature of the experiences—which in both cases is an awareness of certain sense data—but in the contingent causes of the experiences. Similarly, disjunctivism is inconsistent with many representational views of experience, according to which the fundamental nature of a perceptual experience is explained by reference to its representational content. It will be inconsistent with versions of this view—which include mainstream ones—if they allow that a hallucinatory experience might have the very same representational content as a veridical one. And it is inconsistent with adverbialism too. This last is unsurprising, since adverbialists hold that perceptual experiences are not relational at all. Rather, just as a string might be vibrating in a certain way, the adverbialist holds that a conscious subject can be experiencing in a certain way. And to have a certain sort of perceptual experience is just to be experiencing in the relevant way. One reason disjunctivism is inconsistent with so many theories is that its denial is not just an incidental entailment of those rival theories. Rather, those theories are in part motivated by the great intuitive plausibility of what we might call ‘the common kind assumption’: the idea that we could have the very same kind of experience in veridical and hallucinatory cases. Even proponents of disjunctivism often admit that it is defended more as a way of preserving naïve realism than as a

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

114

 

thesis with its own independent appeal.¹¹ One possible source of the plausibility of the common kind assumption is the idea that the nature of a conscious perceptual experience is exhausted by what it is like to have it, and that what is it like for us, at any moment, supervenes on what is going on in certain areas of our brains. Taken together with the idea that it is at least metaphysically possible to suffer a hallucination by having the same activity in the relevant areas of one’s brain as one has in a veridical experience, the conclusion is that a veridical experience can be of the same kind as a hallucination. This is not to say that naïve realists have no moves to make here to defend their denial of the common kind assumption.¹² Still, given the plausibility of that assumption, and of the considerations that support it, it would be nice have a theory that respected the motivations for naïve realism in a way that avoided disjunctivism.¹³ As I’ve noted is typical of much theorizing about perception, one significant motivation for naïve realism stems from reflection on experience from a firstpersonal point of view. In particular, it has seemed to many theorists that the phenomenology of normal visual experience suggests that we are in direct perceptual contact with the objects in our environment. Here are some representative claims: seeing seems to ‘bring one into direct contact with remote objects.’ Mature sensible experience (in general) presents itself as [ . . . ] an immediate consciousness of the existence of things outside of us. Veridical perception, illusion, and hallucination seem to place objects and their features directly before the mind. The ripe tomato seems immediately present to me in experience [ . . . ] The world is just there.¹⁴ These passages express the openness of perceptual experience. A central motivation for naïve realism is that it vindicates openness in a straightforward way.¹⁵ This is the first of two motivations that I will address in the following subsection.

¹¹ See Martin (2004, 37); Logue (2015, 199). ¹² See Martin (2004) and Fish (2009, esp. ch. 5). ¹³ Stoneham (2008), Conduct (2012), and Sethi (2020) defend the possibility of versions of naïve realism without metaphysical disjunctivism. Still, the disjunctive ‘bump in the rug’ unavoidably shows up in other places. Stoneham and Conduct end up defending a disjunctive account of the objects of experience, while Sethi ends up holding that the very same instance of redness can be fully grounded, simultaneously and independently, in (a) the perceiver’s neurological state and (b) the surface of a tomato. ¹⁴ These are taken from a collection presented in Hellie (2007, 266). The claims are made by, respectively, Broad (1952, 33), Strawson (1979, 47), Sturgeon (2000, 9), and Levine (2006, 179). ¹⁵ There are other motivations for naïve realism that are beyond the scope of the present discussion—though not, I think, beyond the scope of neopragmatist treatment. Martin (2002), for example, offers an argument based on sensory imagination, and McDowell (2013) argues that naïve realism helps address external-world scepticism.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

115

3.2 Demystification The openness of visual experience can seem quite mysterious. How is it that the mind ‘leaps the gap’ between one’s eyes and the tomato, so that we end up in direct, immediate mental contact with the tomato? If naïve realism provides a good explanation of this mysterious phenomenon, this would provide some reason to endorse it—in the absence of a better explanation. But, I will argue, neopragmatism can explain the openness of experience in a way that involves fewer controversial metaphysical commitments than does naïve realism, and that in fact renders it comparatively unmysterious. On the neopragmatist view, the openness of experience is to be explained by highlighting the role experience plays in teaching the distinctions between external physical reality and such things as illusion, pretend, and so on. Briefly put: of course we rightly take what we see to be what is right there, or manifest, because we learn how to use phrases like ‘right there’ and ‘manifest’ in a way that makes essential use of the senses, and that presupposes that whatever one senses is right there, manifesting itself to one’s senses. To be manifest or right there is to be available for ostension, and ostension involves sensing. To get a clearer view of the neopragmatist explanation, suppose for a moment that experience is—contrary to the naïve realist’s view—merely a matter of various modifications to an experiencer’s consciousness: modifications caused, but not constituted, by external objects. Even if this were true, we would still end up learning to use such phrases as ‘right there’, ‘present’, and ‘directly before us’ in the same way we would if naïve realism were true. As a result, perceptual experiences—as opposed to beliefs or memories—would still be described as experiences in which it seems as if objects are right there, present, directly before us. So we would end up making the very same claims that seem to some to motivate naïve realism. Conclusion: phenomenological appeals to the openness of sense experience cannot support naïve realism over other sorts of views. It is worth stressing that the point I am making is not that there is a metaphysical possibility that we might be misled by the phenomenology of experience, and end up making systematically false claims about what experience reveals to us as ‘right there’. Naïve realists are already willing to grant that experience could seem to be as of an objective world even if their view were false. Their complaint is that if we adopt a rival view, we have to hold that we actually are misled by the way things appear, not that it is a mere possibility. And they hold that it is better to have an account of perception that avoids attributing massive error to experience. I agree. But what I am arguing is that even if naïve realism is false, the phenomenology is not misleading. That is, even if naïve realism is false, the right thing to say is that perceptual experience typically reveals what is right there in front of us, just as it seems to do. After all, when we make such claims, we will be using these

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

116

 

phrases correctly and consistently, in accord with the meanings we mastered when we learned to use them. As against neopragmatism about perceptual talk, naïve realism takes itself to be making important metaphysical claims. One of these is the assertion of a certain kind of unmediated relation between mind and world—often called ‘acquaintance’—the obtaining of which constitutes a perceptual experience.¹⁶ I think this metaphysical claim is part of what gets its opponents up in arms. They find it hard to see how a conscious mental state—as opposed, perhaps, to the content of that state—could possibly include, in its very nature, an external object. This is perhaps part of the reason why Russell, who made early use of the notion of acquaintance in connection with sense data, later described it dismissively as ‘something like a mystic union of knower and known’.¹⁷ And sense-data theory at least made the objects of acquaintance sufficiently ‘local’ to make the notion of direct mental contact with them hardly more mysterious than sense data themselves already were. But it is certainly questionable whether the relation of acquaintance, born in sense-data theory, is viable when transplanted into a theory on which the objects of acquaintance are mind-independent, and are separated from us in space. Let me try to put a finer point on the worry about acquaintance with external objects. It is certainly plausible that some mental states, such as beliefs, are objectinvolving: that is the moral of Twin-Earth thought experiments. But naïve realists distinguish beliefs from perceptual experiences precisely by saying that, in the case of perceptual experiences, the objects are present. There is no problem in believing something about a physical object that has ceased to exist. That is because the ‘aboutness’ here can at least plausibly be explained in causal/historical terms. But if perceiving an object requires it to be present so that one’s experience can be constituted by a relation to it, then veridical perception requires the past to be present. After all, both light and sound require time to reach us from their sources. Do we ever see stars? To say we do not see them seems odd. But if we do see them, then our minds evidently reach back millions of years into the past to make them present to us.¹⁸ We avoid all such issues if we regard the experience itself as local, partially caused but not constituted by its object. Neopragmatism allows us to respect the openness of experience without reliance on problematic metaphysical postulates about relations between minds and things. That is, it explains why it is correct to say that perceptual experience seems to present us with features of our environment, and, in good cases, manifests to us what is in fact right there. That claim is not to be explained by doing any ¹⁶ Logue (2012, 211); Fish (2009, 14). ¹⁷ Russell (1921, 234). ¹⁸ For a surprisingly relevant discussion, see Carr (1911). Johnston (2004, 139) addresses the timelag issue by holding that seeing an object ‘is an event materially constituted by the long physical process connecting the object seen to the final state of the visual system’. In my view, this seems plausible only when we restrict attention to relatively local objects of vision. It seems very awkward to say that the event of an astronomer seeing the star Icarus began billions of years before life emerged on Earth.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

117

metaphysics, however. Rather, it is to be explained by looking at the rules for using the words ‘reality’, ‘right there’, ‘seems’, and related expressions. To get a better view of how the neopragmatist explanation goes, first imagine someone—call her Ellen—who has lost her vision in an accident as an adult. After the accident, Ellen is fitted with a prosthetic system that allows her to get around in the world with the aid of grid of pressure points on her back, relaying information from a pair of cameras mounted on her head. When we try to imagine how this might work, and how Ellen might use it to know about objects, reach for them successfully, describe their features, and so on, the sighted among us naturally imagine translating tactile sensations into a visual image, or even into propositional beliefs about the locations and properties of objects. So our inclination might be to say that in such a case, the tactile experiences do not put Ellen in direct perceptual contact with the reality beyond her skin. Rather, we might say that they only provide her with a tactile model of that reality, or tactile information about that reality, which requires interpretation before it can be acted upon. Indeed, even Ellen might say this. And she might be right. Now imagine a community of people, born without functioning eyes, but fitted at birth with the same sort of device that Ellen makes use of. These people, as they learn language, have their attention drawn to objects in the range of their cameras by other people pointing at those objects. They also learn how to use the phrase ‘right there in front of them’ in the same way we do. One such person—call her Michelle—might become a philosopher, and might come to defend a naïve realist view of her camera-aided perception, pointing out that the information coming in through the array of dots on her back puts her in contact with objective reality. And she might clarify this claim by saying that those sensations make those objects ‘present’ to her, or ‘there’ for her in a way that is totally different from the way beliefs about those objects do. After all, when Michelle learned the phrases ‘present’ and ‘right there’, it was in the context of objects being pointed out to her that were in fact right there, in front of her cameras. What is the difference between Michelle and Ellen? Of course, there is likely to be a difference in the ease with which they make use of the information coming in through the arrays of dots on their backs. This difference in ease might be quite substantial, resembling the difference between someone speaking a language as her mother tongue, and someone who has learned it as a second language.¹⁹ But it is very plausibly not a difference in the fundamental metaphysical nature of their perceptual experiences themselves. If they accept this, mainstream naïve realists should hold that Michelle is mistaken in endorsing naïve realism about her sensorysubstituted camera-and-electrode-mediated visual perceptual experience—at least if such a view is false of Ellen’s ‘interpreted’ experience of the objects in front of her.

¹⁹ Compare Austin (1962, 113).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

118

 

There are, of course, ways of arguing that the difference between Ellen and Michelle’s perceptual experiences is a radical difference in metaphysical kind. My point does not really depend on denying this, though I do deny it. Rather, my goal is to explain the naturalness of talk of manifestation of reality, or of objects being right there, open to view, and seeming to be so, without making any appeal to the phenomenal nature of experience from the first-personal point of view. The sensory-substitution thought experiment does this, since it only makes stipulations about the kind of information a person gets via her senses, and about how a person learns to speak about the objects and properties about which this information treats. Given these stipulations, it is not surprising that members of the community I’ve just described will explain what it is like for them by making the sorts of claims that the naïve realist takes—mistakenly—to support naïve realism over other views. The phenomenal report ‘it seems to be right there, open to view’ will be a matter of things seeming the way they do when they are, in fact, right there, and open to view. A second phenomenologically based argument for naïve realism begins by contrasting it with views on which there is only contingent connection between a property of a perceived external object and the phenomenal character of an experience of that object. Heather Logue complains about such views that If the connection between the phenomenal character and the property is contingent, then the phenomenal character is a mere sign of the instantiation of that property. There is no deep connection between what it is like to perceive instances of that property and what things with that property are like independently of experience.²⁰

Here we have another target for demystification. It looks like there is an important contingency in any non-relational view of the phenomenal nature of experience. So how can we explain how sense experience informs us about what things in the external world are like? It is not difficult to see how the above problem motivates naïve realism. If we think of perceived objects and their properties as part of the very nature of veridical experiences, we eliminate the apparently troubling contingency. But the neopragmatist has a less metaphysical way of addressing the mystery. That way involves looking for Logue’s ‘deep connection’ between the character of perceptual experiences and the nature of experienced objects by examining our use of the phrase ‘what it’s like’. That phrase is used in two distinct but related ways. We use it in the first way when talking about what things with a certain property are like. We use it in the second way when talking about what it’s like to

²⁰ Logue (2012, 229).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

119

perceive an instance of a certain property. The second use is salient to philosophers discussing phenomenal character. In this second sense it is by definition the case that ‘what it’s like’ only characterizes experiences—conscious experiences, in fact—and in the absence of conscious beings there would be no such thing. This is a perfectly acceptable sense. But of course we also use the phrase ‘what it’s like’ to describe objects in the external environment. For example, we describe what a gothic arch is like (it is curved in such-and-such way), what memory foam is like (it is soft, but dense), what the paint colour ‘autumn sky’ is like (it is a dark pastel blue that is quite pure), and so on. It is no surprise that the two forms of what-it’s-like talk are intimately related. The practice of talking about and describing phenomenal character (what it’s like to have an experience) is introduced by using a prior practice of talking about external objects and their properties (that is, what those objects are like). As young children are first learning to engage in this prior practice, they also learn that what they are tempted to say is not always right. Instances of disagreement and error allow them to be taught to reserve commitment by saying that an object looks or seems red, rather than that it is red. They also learn how to explicitly deny what they might earlier have been tempted to say, by saying that an object merely looks red. The process of learning to talk in these ways is illustrated in Sellars’ wellknown story of John the necktie salesman.²¹ This story explains the use of such claims as ‘It’s like I’m looking at a green necktie, even though I know what the necktie is really like; it’s blue.’ That is, much of our talk of what it’s like for us is parasitic on talk of what objects are like. At this point some may complain that I have changed the subject: that the phenomenal character of our experience is distinct from things looking (or smelling or tasting) a certain way to us when we have those experiences. They will claim that even if two people correctly report that something looks red to them, it may be that the phenomenal character of their experiences is totally different—inverted, perhaps. And they might say that it is this third sense of ‘what it’s like’ that Logue is discussing. But it is not clear whether this particular complaint is available in defence of naïve realism. If two people are genuinely perceiving the same shade of red, then, according to naïve realism, the phenomenal character of their experiences will be the same. Nor is neopragmatism friendly to the possibility of spectrum inversion. There cannot be a public linguistic practice that gives meaning (that is, use) to a word that describes something essentially private and incommunicable. The apparent possibility of spectrum inversion seems to me to stem from a mistaken picture of what ‘what it’s like’ talk is describing, as if it were describing something very much like an

²¹ Sellars (1963, §§14–15).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

120

 

inner duplication of the outside world, with corresponding observable properties to which each of us has private access.²² What the neopragmatist provides is a language-based explanation for the fact that there is a deep connection between what it’s like to see a red object, and what red objects are like; the latter objects have a property (redness) that we must appeal to in explaining the former. This will not satisfy everyone; some philosophers want what it’s like to experience a red thing to reveal ‘what red things are like independently of experience’.²³ In fact, Logue herself does not want this in the case of colour.²⁴ She thinks colour experience is largely determined by features of the observer, rather than by features of the colour observed. So it is a consequence of her view—one that is arguably in tension with common sense—that no one knows what redness is like (at least, on the basis of visual experience). But she thinks the opposite claim is true of shapes. Even if she is right about shape, however, this does not require us to say that a square object partially constitutes our perception of it. All it requires is that the phenomenal character of experiences of a square thing contains—not necessarily, but actually, given our sensory apparatus—certain information about that square thing. In containing that information, it tells us something about what the square is like. For example, it might contain the information that it has four special parts—the corners—that behave differently than the other parts, and that any given corner behaves in the same way as any other, unlike in the case of a rhombus. It is no surprise that what it is like to experience external objects contains a lot of useful information about their shapes, colours, distances, temperatures, and so on. Our sense organs receive informationbearing signals from those objects, and evolutionary forces explain how it is that subsequent signal processing extracts some of that information.

4. A Neopragmatist Theory of Perception So far, I’ve tried to show how neopragmatism can do some positive work in philosophy of perception. It can explain the openness of perceptual experience without making controversial and problematic metaphysical assumptions. And it can explain why what it’s like to perceive instances of a property can often tell us a lot about what things with that property are like, independently of experience. In this final substantive section I want to suggest, very briefly, the form that a positive neopragmatist theory of perception might take. Recall, the starting points for neopragmatist theorizing include the following. First, we human beings live in a

²² See Gert (2017, 194–208). ²³ This is the thesis of Revelation, popularized by Mark Johnston. I have argued that it is too strong elsewhere. See Johnston (1992, 223); Gert (2017, 23–5). ²⁴ Logue (2012, 215).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

121

physical world and have senses that are responsive to the physical properties of things in our environment. And, second, we have an innate capacity to learn language in certain ways, both by being explicitly trained and by observing the linguistic behaviour of adults. Now I will add a third starting point: that we humans are conscious beings in the sense that there is something it is like to be one of us—at least when we are conscious. It is certainly a mystery as to how and why there are conscious beings, just as it is a mystery as to how and why there is a material world. But, despite its being mysterious in this sense, it is almost completely uncontroversial. Every major theory of perceptual experience takes it as a starting point. They differ in their explanations of what makes it the case that we have the particular conscious perceptual experiences we do. Sense-data theorists add sense data, and some sort of acquaintance relation. Representationalists add a robust form of intentionality to the nature of the mental. And naïve realists add, in the first instance, some means by which the external world can comingle with the mental. The neopragmatist, on the other hand, will try to make no controversial ontologically committing moves of this sort at the explanatory level. But it is common ground for all these theories that, at any given waking moment, there is something it is like to be me: that there is a way it is for me. Some may take the task of a theory of perceptual experience to be to identify the facts in virtue of which we are perceptually conscious. If this amounts to the claim that one of the tasks of such a theory is to solve the hard problem of consciousness, then I think the right response is to deny it. And even if this were one of the goals of a theory of perceptual experience, no current theory so much as takes a real step in its direction. To postulate an acquaintance relation that results in perceptual consciousness when it connects our minds with sense data is to cram all of the mystery of the hard problem into the notion of sense data, along with an unexplained notion of acquaintance and an unexplained mechanism by which its instantiation results in conscious experiences of a certain sort. Naïve realism, it is true, avoids a commitment to sense data. But the acquaintance relation and its unexplained role in generating conscious experience remain. And similar complaints can be made about the thesis that perceptual consciousness is the result of a special kind of intentional relation holding between our minds and certain sorts of content. If we are looking to deal with tractable mysteries, it would be good if we could do our philosophy of perception without initially postulating the existence and explanatory power of such things as sense data, acquaintance relations, or a substantive notion of intentionality. That doesn’t mean that talk of such things is, ex ante, ruled out. Perhaps neopragmatist theorizing will lead us to see that some of it is perfectly in order. But even if this should happen, the result would not be an inference to the best explanation of the sort that naïve realists, sense-data theorists, and representationalists offer. Rather it would be a linguistic

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

122

 

demystification, providing naturalistic answers to pragmatic questions about the relevant sort of talk. Because of this, I suspect that even if neopragmatist theorizing ultimately led to a version of naïve realism, the overall theory—the naïve realism taken together with the neopragmatist methodology—would not satisfy many current proponents of naïve realism. That is because they see their view in metaphysical terms. In any case, my own view is that neopragmatism most plausibly leads to a version of adverbialism. The remainder of this section is the beginning of a sketch of an argument for such a view. What it’s like for us is—uncontroversially—in large part the upshot of the world impinging on our sensory systems. When this happens, the ultimate result, in terms of the neural correlates of consciousness, is that the matter in certain parts of our brains is behaving in a certain way, and the way it is behaving—the way the neurons are firing—carries information. By this I do not mean that it represents anything. I only mean that it is not a completely random signal. It is, in contrast, like a blood splash pattern or a set of tire tracks. It is no surprise that much of the information in the inputs to our senses is preserved in the more central neurological states to which they give rise, since it is also uncontroversial that our sensory systems and brains have been shaped by evolution so that they allow us to get around in the world successfully: reaching for things in the locations they actually are, for example. Although blood splash patterns and tire tracks contain information that we might describe in terms of probabilities relative to certain relevant background data, we do not describe them as true or false, or even as accurate or inaccurate. Rather, they can be misleading, or not, to someone—or some system—that has the capacity to ‘decode’ them. In the case of sensory-information-bearing neural states, we can start out saying the same thing. That is, we can think of them as information-bearing, but not as representational in any robust sense. On the view I’ve just described, the neural states associated with perception are not relational, and they contain information in virtue of the way they are. One might therefore call it adverbialism about sensory-information-bearing states or ‘informationtheoretic adverbialism’ for (somewhat) short.²⁵ If we put information-theoretic adverbialism together with the view that the relevant neurological states are tightly correlated with phenomenal states—an admittedly empirical hypothesis, though one that has a great deal of plausibility—the result is that phenomenal states contain pretty much the same information as the neurological ones. If this is right, then, when we are talking about phenomenal states, we can make use of the very same notion of information as applies to blood-splash patterns. That is, we can say that sometimes the way it is for us is misleading, so that we end up reaching for things that are not there, or

²⁵ See Gert (forthcoming) for a fuller presentation of the view.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

123

responding with ‘blue’ when asked the colour of something that is green. But, in contrast with blood splash patterns, when our phenomenal states are misleading in this way, it is understandable that the language of accuracy and inaccuracy— and the related notions of veridicality, illusion, and hallucination—might develop. The explanation here will appeal to the same sorts of practical reasons that the neopragmatist uses to explain the emergence of truth-assessments and the assertoric form. The view I have just described is very close to traditional adverbialism. One important innovation is that its reference-fixing descriptions of ways of experiencing appeal to the information those ways contain. These information-theoretic descriptions are not strained by the complexity of visual experience. We can simply use ‘that’-clauses such as ‘that contains the information that there are two squares about a foot in front of me: a green one on the right and a red one on the left.’ In this way, it avoids earlier awkward—and inadequate—constructions such as ‘redly and squarely’ or ‘redly-separated-from-redly-ly’.²⁶ It might seem that these earlier descriptions were more ‘direct’ than the information-theoretic ones. But in fact, they also got their meanings by having their references fixed by descriptions. Moreover, those descriptions were limited to mentioning the standard causes of the relevant ways of experiencing, which was itself the source of many difficulties. To conclude this section, consider the following dismissal of adverbialism: There are familiar reasons why the adverbial theory is indefensible [ . . . ] and their source can be traced back to the theory’s failure to accommodate even the apparent relationality of perception.²⁷

If what I argued in section 3 is correct, this criticism is misguided. The apparent relationality of perception—its openness—is a matter of the naturalness of certain ways of describing perceptual experience, including when we describe it to ourselves. And the neopragmatist can explain why those ways are natural, and apt, even if we understand the experiences themselves as completely local phenomena.

5. Conclusion In a sense, the neopragmatist position involves a linguistic and conceptual analog of the naïve realist’s metaphysical view. For example, the neopragmatist holds that

²⁶ Chisholm (1957); Tye (1984). The latter monstrosity was part of Tye’s attempt to describe experiences of at least two red patches in a way that avoided object-talk. ²⁷ Crane (2006, 142–3).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

124

 

there is no understanding talk of hallucinatory experience without a prior understanding of talk of veridical experience. This mirrors some naïve realists’ claims that there is no understanding the nature of hallucinatory experience except in relation to veridical experience.²⁸ Or, to take another example, the neopragmatist holds that we teach talk of veridical experience in ways that make essential use of the very objects of experience. This mirrors the naïve realist’s claim that veridical experience is partially constituted by the very objects of experience. Because the neopragmatist mirrors, at the linguistic or conceptual level, what the naïve realist holds at the level of referent, it is no surprise that if the neopragmatist is right, the errors in naïve realism are the result of a confusion of these levels. That is, they are a result of confusing meaning with referent. One source of this confusion is the view—rejected by neopragmatists—that in order to teach or learn a referential concept, one need only draw attention to the referent somehow, in order to establish a referential relation, and that awareness of the referential connection is enough to impart linguistic competence. But it is central to the neopragmatist view to deny this and indeed to deny that the notions of direct acquaintance or reference bear any weight in explaining the acquisition of conceptual competence.

References Austin, J. L. 1962. Sense & Sensibilia. G. Warnock (ed.). Oxford: Clarendon Press. Båve, Arvid. 2009. ‘A Deflationary Theory of Reference’. Synthese 169: 51–73. Benacerraf, Paul. 1983. ‘What Numbers Could Not Be’. In Hilary Putnam and Paul Benacerraf (eds.) Philosophy of Mathematics: Selected Readings. New York: Cambridge University Press, 272–93. Broad, Charlie Dunbar. 1952. ‘Some Elementary Reflexions on Sense-Perception’. Philosophy 27: 3–17. Carr, H. Wildon. 1911. ‘The Time Difficulty in Realist Theories of Perception’. Proceedings of the Aristotelian Society 12: 124–36, 183–7. Chisholm, Roderick. 1957. Perceiving. Ithaca, NY: Cornell University Press. Conduct, Matthew. 2012. ‘Naïve Realism Without Disjunctivism About Experience’. Consciousness and Cognition 21: 727–36. Crane, Tim. 2006. ‘Is There a Perceptual Relation?’ In Gendler and Hawthorne (2006), 126–46. Crane, Tim and French, Craig. 2017. ‘The Problem of Perception’. The Stanford Encyclopedia of Philosophy (Spring 2017 Edition), Edward N. Zalta (ed.). https:// plato.stanford.edu/archives/spr2017/entries/perception-problem/

²⁸ See Martin (2004, 2006).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

    

125

Fish, William. 2009. Perception, Hallucination, and Illusion. New York: Oxford University Press. Fish, William. 2010. Philosophy of Perception: A Contemporary Introduction. New York: Routledge. Gendler, Tamar and Hawthorne, John (eds.). 2006. Perceptual Experience. Oxford: Oxford University Press. Gert, Joshua. 2017. Primitive Colors: A Case Study in Neopragmatist Metaphysics and Philosophy of Perception. Oxford: Oxford University Press. Gert, Joshua. forthcoming. ‘Information-theoretic Adverbialism’. Australasian Journal of Philosophy. Hellie, Benj. 2007. ‘Factive Phenomenal Characters’. Philosophical Perspectives 21: Philosophy of Mind: 259–306. Johnston, Mark. 1992. ‘How to Speak of the Colors’. Philosophical Studies 68: 221–63. Johnston, Mark. 2004. ‘The Obscure Object of Hallucination’. Philosophical Studies 120: 113–83. Jackson, Frank. 1977. Perception: A Representative Theory. Cambridge: Cambridge University Press. Levine, Joseph. 2006. ‘Conscious Awareness and (Self-)Representation’. In Uriah Kriegel and Kenneth Williford (eds.) Self-Representational Approaches to Consciousness. Cambridge, MA: The MIT Press, 173–198. Logue, Heather. 2012. ‘Why Naive Realism?’ Proceedings of the Aristotelian Society 112: 211–37. Logue, Heather. 2015. ‘Disjunctivism’. In Matthen (2015), 198–216. Martin, M. G. F. 2002. ‘The Transparency of Experience’. Mind and Language 4: 376–425. Martin, M. G. F. 2004. ‘The Limits of Self-Awareness’. Philosophical Studies 120: 37–89. Martin, M. G. F. 2006. ‘On Being Alienated’. In Gendler and Hawthorne (2006), 354–410. Matthen, Mohan. 2015. The Oxford Handbook of Philosophy of Perception. Oxford: Oxford University Press. McDowell, John. 2013. ‘Perceptual Experience: Both Relational and Contentful’. European Journal of Philosophy 21: 144–57. Price, Henry. 1932. Perception. London: Methuen. Price, Huw. 1988. Facts and the Function of Truth. New York: Blackwell. Price, Huw. 2003. ‘Truth as Convenient Friction’. The Journal of Philosophy 100: 167–90. Price, Huw. 2013. Expressivism, Pragmatism and Representationalism. New York: Cambridge University Press. Russell, Bertrand. 1912. The Problems of Philosophy. London: Oxford University Press.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

126

 

Russell, Bertrand. 1921. The Analysis of Mind. London: Allen and Unwin. Sellars, Wilfrid. 1963. ‘Empiricism and the Philosophy of Mind’. In Science, Perception and Reality. London: Routledge & Kegan Paul, 127–96. Sethi, Umrao. 2020. ‘Sensible Over-Determination’. The Philosophical Quarterly 70: 588–616. Stoneham, Tom. 2008. ‘A Neglected Account of Perception’. dialectica 62: 307–22. Strawson, Peter. 1979. ‘Perception and its Objects’. In Graham Macdonald (ed.) Perception and Identity: Essays Presented to A. J. Ayer with His Replies. London: Palgrave, 41–60. Sturgeon, Scott. 2000. Matters of Mind. London: Routledge. Tye, Michael. 1984. ‘Adverbialism’. The Philosophical Review 93: 195–225.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

7 Perceptual Experience and Physicalism Laura Gow

1. Introduction Although there seems to be widespread disagreement regarding how we should go about answering questions about perception—whether we should look to science to provide us with all the answers, or whether first person reflection can reveal the nature of perceptual experience—there is one thing that many philosophers do agree on: our philosophical account of perceptual experience should be compatible with physicalism (although see Allen, this volume). In other words, we haven’t provided a successful account of perceptual experience if that account essentially involves non-physical entities. This general commitment to physicalism has had a significant impact on the direction the philosophical debate about perception has taken in recent years. It used to be the case that questions about the nature of our perceptual experiences were approached from within an indirect realist framework. Indirect realists hold that we perceive the world around us indirectly, in virtue of directly perceiving some other kind of entity. The best known version of this view is the sense-datum theory, according to which all of our perceptual experiences are essentially relations to sense-data (Russell 1912; Broad 1925; Price 1950; Ayer 1956; Jackson 1977; Lowe 1992; and Robinson 1994). One of the biggest challenges facing any theory of perception is to account for hallucinations—experiences where we seem to be aware of entities that are not in fact present in our local environment. The sense-datum theorist has a ready response to this challenge; hallucinations are like veridical perceptual experiences in that they too consist simply in the awareness of sense-data. Sense-datum theory therefore provides an explanation of what we are aware of during a hallucination as well as providing a unified account of all of our perceptual experiences— hallucinations, illusions, and veridical experiences are all essentially relations to sense-data. These experiences differ only with respect to how they are related to the world, and veridical experiences are classified as such because they are related to the world in the right kind of way. (What counts as the right kind of way for our sense-data-involving perceptual experiences to be related to the mindindependent world varies between different versions of the theory. Causal history is important—veridical perceptual experiences are those which have been caused

Laura Gow, Perceptual Experience and Physicalism In: Purpose and Procedure in Philosophy of Perception. Edited by: Heather Logue and Louise Richardson, Oxford University Press (2021). © OUP. DOI: 10.1093/oso/9780198853534.003.0007

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

128

 

by our interacting with the world through our sense organs. A veridical visual experience will have been caused by photons reflected from objects in our local environment impacting on our retinas, and so on. In addition, according to some versions of the view, the sense-data involved during veridical perceptual experiences will resemble the objects in the world which caused the experience. The relational analysis provided by the sense-datum theory is motivated by the assumption that seeming to be aware of an object instantiating various properties must involve a genuine awareness of something which instantiates those properties. In other words, if we seem to be aware of a purple unicorn (during a hallucination, say), then there must be something purple and unicorn-shaped for us to be aware of, even if it isn’t a mind-independent physical object, like an actual unicorn. This assumption is called the phenomenal principle, and is one of the driving forces behind the sense-datum theory. Here is Howard Robinson’s version: If there sensibly appears to a subject to be something which possesses a particular sensible quality then there is something of which the subject is aware which does possess that quality. (Robinson 1994, 32)

To endorse the phenomenal principle is to adopt a non-physicalist metaphysics. During hallucinations, there aren’t any (relevant) physical things in one’s local environment for one to be aware of, and yet the phenomenal principle insists that in such cases we must be aware of something. This must therefore be a non-physical something. Indeed, sense-data are typically understood to be minddependent, non-physical entities.¹ And so we can see that although the sensedatum theory provides an answer to the question of what we are aware of during hallucinations, the explanation is not a physicalist one. The sense-data theory is now a minority view in the contemporary debate, and the reason given for rejecting it is almost always based on its failure to conform to a physicalist metaphysics.² This shift is interesting; contemporary philosophy of perception has abandoned the phenomenal principle as a fixed point for theorizing, and has adopted physicalism instead. I suspect that this alteration to our starting point when thinking about perception is not in fact due to a deep change of opinion regarding our preferred metaphysical theory, but is instead a result of circumstances. In ¹ At least, this is the dominant view. ² There are other problems with sense-data besides their non-physical nature. Epistemological concerns include the so-called ‘veil of perception’ problem—the worry that the mind-independent world is entirely unknowable since we are only ever directly aware of mind-dependent sense-data. Other problems concern the location of sense-data. These entities are said to have the properties they are perceived as having, such as shape and size. They must therefore occupy a location. But it is difficult to come up with a plausible account of the location of sense-data since they are mind-dependent entities.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

129

other words, I expect that we have always seen the attractions of physicalism— certainly since Descartes’ interaction problem (the problem of trying to explain how physical and non-physical entities can causally interact) (Elisabeth of Bohemia and Descartes, 2007). Although not all philosophers are physicalists, non-physical entities tend to enter the picture only when there are no physical entities available to do the explanatory work required. Hallucinations present exactly this kind of challenge. There aren’t any physical objects to explain such experiences, and so we posit non-physical objects, as a last resort, as it were. I would like to tentatively suggest that the shift to physicalism does not therefore represent a radical shift in ideology, but rather, has been brought about by the development of a view which makes it possible and plausible to deny the phenomenal principle, and, by so doing, avoid the need of introducing non-physical entities. Representationalism (or intentionalism) is now the dominant position in contemporary philosophy of perception, and it has achieved this status through its claim to provide a successful and comprehensive physicalist account of the mind. Indeed, it was developed for this very purpose (See Dretske 1995, 2000 and Tye 1995, 2000. Note that Dretske’s 1995 is entitled ‘Naturalizing the Mind’.) Like the sense-datum theory, representationalism provides a unified account of perceptual experience.³ Veridical, illusory, and hallucinatory perceptual experiences are fundamentally the same kind of mental state—they are all representational states with representational content. All of our perceptual experiences say something about the way the world is—if they get the world right then the experiences are veridical; if they get the world wrong, they are illusory or hallucinatory.⁴ On this view, the phenomenal character of our perceptual experiences is grounded in (or constituted by, or identical to) their representational content, and so if the representational content of two experiences is the same, then their phenomenal character will also be the same, whether they are veridical, illusory, or hallucinatory.⁵

³ This distinguishes these views from naïve realism (sometimes called ‘the relational view’) which analyses veridical perceptual experiences as consisting in relations to the ordinary physical objects in our local environment. Naïve realists adopt disjunctivism when faced with the problem of hallucination, and claim that hallucinations are entirely different kinds of mental state from veridical perceptual experiences. Since the debate I am focusing on in this paper takes place between representationalists and their non-relationalist opponents, I will assume, in line with these views, that a theory which can provide a unified account of perceptual experience is to be preferred. For an excellent introduction to disjunctivism, see Byrne and Logue 2009. ⁴ The existence of veridical hallucinations (hallucinations which happen to match the way the world is) makes the above definition merely a first-pass characterization of the veridical versus non-veridical distinction. ⁵ The precise nature of the relationship between phenomenal character and representational content need not concern us here.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

130

 

The idea that our perceptual experiences have content would seem to allow for a new explanation of hallucinations. More importantly for contemporary philosophy of mind, it allows for a physicalist explanation of hallucination. This is because representationalism provides a way of denying the Phenomenal Principle. As Tim Crane points out: [P]erception, like belief, is a form of representation of the world, and it is not true of representations in general that when a representation represents that something has a property, there is something which does have that property. (Crane 2006, 132)

Representationalists therefore seem able to give a unified account of all our perceptual experiences without positing anything non-physical to explain hallucinations. However, recently, a number of philosophers have questioned representationalism’s compatibility with physicalism, and have pointed out that mainstream representationalists typically seem to rely on relations to non-physical entities. The main aim of this chapter is to reinforce and expand on this argument. I will suggest that positing non-physical entities is not just a dispensable feature of some forms of representationalism, in fact there is something essentially non-physicalist about the way representationalism is standardly formulated. This is interesting for two reasons: first, it brings to light the possibility that the leading physicalist account of perception is in fact incapable, as it stands, of meeting the criteria laid down by a genuinely physicalist metaphysics. Second, it shows that a commitment to the popular idea that the philosophy of perception should proceed in accordance with physicalism leads to the rejection of the dominant view in the philosophy of mind. Those philosophers who have criticized representationalism for positing relations to non-physical entities have provided an alternative account, which is advertised as being the position we should endorse if we want to hold on to our commitment to physicalism (particularly if we want our account to include illusions and hallucinations as well as veridical perceptual experiences). They argue that perceptual experience must be wholly non-relational: perceivers enter into causal relations with their environment, but our experiences do not involve any essential relations. In the last part of this paper I will point out one of the difficulties which arises for this view: by making perceptual experience wholly non-relational, it is difficult to do justice to the idea that perceptual experiences can be assessed for accuracy. I should point out that my aim in this chapter is not to defend physicalism, nor to promote a particular position in the philosophy of perception, but rather to point out some of the interesting consequences of holding on to the idea that our investigations should proceed along physicalist lines.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

131

2. Representationalism and Physicalism Representationalism ostensibly meets the demands of physicalism rather easily, since representing the world to be a certain way does not require an ontological commitment to the entities which are represented. A veridical experience of a horse is a representational state which ‘says’ that a horse is present, and a hallucination of a unicorn is a representational state which ‘says’ that a unicorn is present. In the first case, a horse is present (say) and the experience is veridical. In the second case, there is no unicorn present, and the experience is a nonveridical hallucination. Importantly, neither the veridical nor the hallucinatory experience involve an essential relation to the particular object which is represented. As we saw above, representationalists sometimes draw analogies between perception and other representational states, like beliefs, and point out that just as beliefs do not require an ontological commitment to the entities our beliefs are about, neither do perceptual experiences. In the same way that we can have beliefs about unicorns, so we can have perceptual experiences (hallucinations) about them too. However, the representationalist’s claim to provide a genuinely physicalist account of perceptual experience has recently come under attack, most notably by Susanna Schellenberg (2011), Uriah Kriegel (2011), and David Papineau (2014). They point out that once we look into the details of mainstream representationalism, there seems to be a noticeable reliance on relations to non-physical entities. The ontologies of mainstream representationalist accounts contain what Schellenberg calls ‘peculiar entities’, that is, entities that do not fit comfortably within a genuinely physicalist ontology. The peculiar entities we find in representationalist ontologies are not the non-physical, minddependent objects posited by the sense-datum theorist; instead they tend to be non-physical, mind-independent abstract objects. To begin with an obvious example, many representationalists hold that perceptual experiences are propositional states and so involve relations to propositions, which are generally considered to be abstract objects. (see Byrne 2005; Glüer-Pagin 2014; Pautz 2007; Stoljar 2004; Thau 2002; Tye 1995). It has also become popular to posit uninstantiated properties to explain the phenomenal character of hallucinations. (See Bealer 1982; Bengson et al. 2011; Dretske 2000; Forrest 2005; Horgan, Tienson, and Graham 2004; Johnston 2004; Lycan 2001; McGinn 1999; Pautz 2007; Sosa 2007; and Tye 2002, 2014a.) Interestingly, not only does this seem to constitute a move away from physicalism, it also seems to involve a return to the phenomenal principle, at least in spirit. The original phenomenal principle claims that even hallucinations involve the awareness of some sort of object, whereas here, we are talking about the awareness of properties which aren’t instantiated by objects. Still, the central motivation behind the principle is that even hallucinations must involve an awareness relation to

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

132

 

something, and this idea gets preserved in this more recent version. This return to the spirit of the phenomenal principle is surprisingly popular. The following quotes illustrate the widespread commitment to uninstantiated properties amongst both philosophers who endorse externalist versions of representationalism, and those who are internalists.⁶ Dretske writes: Hallucinations are experiences in which one is aware of properties (shapes, colors, movements, etc.) without being o-conscious [object-conscious] of objects having these properties . . . . Hallucinating about pumpkins is not to be understood as an awareness of orange pumpkin-shaped objects. It is rather to be understood as p-awareness [property-awareness] of the kind of properties that o-awareness of pumpkins is usually accompanied by . . . . Awareness (i.e. p-awareness) of properties without awareness (o-awareness) of objects having these properties may still strike some readers as bizarre. Can we really be aware of (uninstantiated) universals? Yes, we can, and, yes, we sometimes are. (Dretske 2000, 163)

And Tye says: [Y]ou cannot attend to what is not there. But on my view there is an uninstantiated quality there in the bad cases . . . an un-instantiated quality is present in hallucination. (Tye 2014b, 51)

Horgan, Tienson, and Graham also seem to hold that perceiving subjects are acquainted with uninstantiated properties during hallucinations: When experience presents various apparent objects as apparently instantiating properties and relations such as shape-properties and relative-position relations, experience thereby acquaints the experiencing subject with such properties and relations . . . . Experientially presented apparent instantiation of the properties and relations suffices to acquaint the experiencing subject with them . . . whether or not the experiencing subject is ever experientially presented with actual instantiations of them. (Horgan, Tienson, and Graham 2004, 304)

Adam Pautz holds a similar view, although he replaces the acquaintance relation with a relation he calls ‘sensorily entertaining’. He says: ⁶ The distinction between internalism and externalism within representationalism is not important for present purposes. Briefly, externalists think that the representational content of perceptual experiences ultimately depends upon relations between the subject and their environment—causal, historical, or teleological, for example. (Dretske 1995, 2000; Tye 1995, 2000; Millikan 1984; Papineau 1984). Internalists deny this, and hold that representational content and phenomenal character are internally determined (Horgan and Tienson 2002; Loar 2003; Crane 2006).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

133

Let E be the property of having an experience with the phenomenal character of the experience one in fact has on viewing a red ellipse, an orange circle and a green square . . . Intentionalism [or representationalism] holds that having E is identical with the property of sensorily entertaining a general proposition or complex property that may be represented as follows: . (Pautz 2007, 524)

I have said that this commitment to uninstantiated properties constitutes a move away from physicalism. Since it is possible to be a realist about universals without being a realist about abstract objects (if one holds the Aristotelian view that universals are always instantiated, for example), and it is only a commitment to abstracta which is a problem for physicalism, I need to explain this in more detail. After all, the physicalist could say that during hallucinations we are aware of properties which are only locally uninstantiated; they are instantiated somewhere. For the proponent of the uninstantiated property account to be charged with contravening the principles of physicalism, it is necessary to show that they are committed to the metaphysically problematic, Platonic conception of universals. On this view, universals exist independently of their instantiations—they are abstract objects which exist outside the spatiotemporal realm. If representationalism involves uninstantiated properties understood in the Platonic sense, then questions about its compatibility with physicalism naturally arise. Brad Thompson argues that representationalists who subscribe to the uninstantiated property analysis of the phenomenal character of hallucinations are indeed committed to the questionable Platonic interpretation. He describes a situation where we can have hallucinations of properties which are not instantiated anywhere. If the representationalist maintains that we can indeed have experiences of properties which are not instantiated anywhere, then they must sign up to the Platonic view of abstracta. Thompson asks us to imagine that every red object in the world has been painted another colour. Since it seems plausible to suppose that it would still be possible to hallucinate red even though there would no longer be any actual instantiations of red, the uninstantiated property theorist must endorse the Platonic conception of universals to be able to deal with the example (Thompson 2008). The situation Thompson describes is, of course, rather far-fetched. What is more, it is always open to the representationalist simply to deny the supposition that we would still be able to hallucinate red in the world Thompson describes. So let me provide a more persuasive example of a situation in which we seem to be aware of properties which are not instantiated anywhere. It is more persuasive because it involves experiences we can all enjoy right now. With a little help, we can all have colour experiences as of colour properties that are not instantiated anywhere. In his 2007, Paul Churchland provides the means by which we can experience a number of ‘impossible colours’—impossible because no objects could

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

134

 

genuinely possess such colour properties. For example, by staring at a yellow circle on a grey background and then looking at a maximally black stimulus one will experience an impossibly dark blue. The experienced blue will be as dark as the maximally black stimulus, which is impossible for any objective blue. (See Churchland 2007, ch. 9 for other examples.) Of course, the important point to take away from this is that analysing such experiences in terms of uninstantiated properties is only possible if the Platonic conception of universals is endorsed. Many, if not most physicalist philosophers seem to be guilty of a kind of ‘doublethink’ when it comes to abstract objects, and rely on such entities freely. This is surprising; although it is notoriously difficult to give a satisfactory definition of ‘abstracta’, everyone should agree that (whatever they are) abstracta will not meet the criteria (whatever they are) for being physical. For instance, we can be reasonably sure that one of the necessary conditions for qualifying as ‘physical’ is to be spatiotemporal, and abstract objects are not, of course, spatiotemporal entities. (See Rosen 2018 for a comprehensive discussion of possible ways of defining abstracta.) Admittedly, I am working with quite a simple and basic definition of physicalism. I understand physicalism to be the view that we should only allow physical entities into our ontology. (Of course, this raises the question of what counts as a physical entity!) One might think that abstracta can be accommodated within a physicalist ontology so long as they can be said to supervene on the physical; so long as once the physical facts are fixed, so are the facts about abstracta.⁷ (See Allen, this volume, for further discussion.) It is outside the scope of the present paper to delve too deeply into this issue, however, it seems to me that such a definition of physicalism is far too lenient. Some versions of dualism would be able to qualify as physical on this view, namely, any version which has the non-physical substance supervene on the physical. This suggests that physicalism in any meaningful sense will need to be more restrictive.⁸ (Daniel Stoljar discusses these issues in his 2017.) It is telling that proponents of the uninstantiated property view do seem, on the one hand, to want a strict version of physicalism since they tend to reject sensedatum theory out of hand for its use of mental intermediaries in perception. Of course, if it would be permissible to allow non-physical abstracta into our physicalist ontologies so long as they supervene on the physical, it would be entirely open to the sense-datum theorist to make the same move and claim that sensedata supervene on the physical too. The fact that this potential manoeuvre isn’t considered before the sense-datum theory is rejected suggests that proponents of ⁷ I would like to thank Heather Logue for suggesting I address this possibility. ⁸ Uriah Kriegel (2011) and David Papineau (2014) also take issue with the idea that our experiences, which are concrete, spatiotemporal entities, essentially involve abstract entities. Kriegel offers a principle he calls ‘the explanatory closure of the realm of concreta’, which expresses the plausible demand that everything in the realm of concreta should be explainable solely by appeal to what goes on in the realm of concreta (Kriegel 2011, 146).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

135

the uninstantiated property view do not consider abstracta to be acceptable because they supervene on the physical (and are, to that extent, compatible with a physicalist metaphysics). Instead, it seems to be a case of ‘they’re problematic, but needs must’. Although abstracta tend to appear in philosophical theories more frequently than sense-data, it seems to me that the former are, in fact, more problematic, metaphysically speaking, than the latter. After all, sense-data are concreta—they are real, concrete entities, albeit non-physical entities. Surprisingly, Tye comes very close to noticing this himself. As part of his discussion of the sense-datum theory, Tye considers an alternative way of defining sense-data—as non-physical, non-mental entities. He says: And sense-data, conceived of as non-mental entities, face other problems. For example, where are they located? In the same space as physical objects? How is this possible? Further, how can their qualities make a causal difference? (Tye 2014b, 50)

It is interesting to note that the very same concerns arise for his uninstantiated property theory. Indeed, it is hard to miss the similarity between the sense-datum theory and the representationalist’s attempt to explain hallucinations in terms of relations to uninstantiated properties. Schellenberg takes particular issue with views which posit an acquaintance or awareness relation to peculiar entities (see footnote 1 of her 2011). Although I agree that there is certainly something mysterious about the idea that we can be acquainted with or aware of non-physical entities, if the goal is to produce a thoroughly physicalist account of all perceptual experiences (including hallucinations), then it is the peculiar entities themselves that are the problem.⁹ This point is significant for two reasons. First, it means that representationalist views which do not posit uninstantiated properties but do make use of propositional contents are equally problematic from a (genuinely) physicalist perspective. And second, it reveals that the dependence on peculiar entities is not an incidental feature of only some representationalist views, rather, it seems to be an essential part of mainstream representationalism itself. Let me explain this in more detail. It is uncontroversial that the defining feature of representational states is that they have content, and there are good reasons for thinking that all versions of mainstream representationalism are committed to content being abstract. Of course, if representationalists are committed to abstract contents, we do not need to consider every version of mainstream representationalism individually ⁹ Pautz (2007) also believes it is the awareness relation that is problematic with the externalist’s uninstantiated property idea. Again, I agree that postulating an awareness relation to abstract entities is particularly counter-intuitive, but the metaphysical worries arise whatever the relation.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

136

 

to assess whether it relies on peculiar entities; all versions rely on peculiar entities. It is important to note that endorsing the idea that perceptual states have content does not automatically force one to accept peculiar entities. As we will see later on, it is possible to create a version of representationalism/intentionalism upon which perceptual states do have content, but this content isn’t abstract. There is a particular and rather interesting reason why mainstream representationalists find themselves committed to abstract contents. To bring out this reason it will help to introduce a distinction between what I’ll call ‘objectrelationalism’ and ‘content-relationalism’. Both sense-datum theory and naïve realism are versions of ‘object-relationalism’ since they hold that the perceiving subject stands in an essential (and not just causal) relation to the objects of perception—for sense-datum theories the direct objects of perception are sensedata, and for naïve realists, they are ordinary objects in the local environment. Now, it turns out that mainstream representationalism is a version of ‘contentrelationalism’. Although proponents of this view deny that essential relations hold between perceiving subjects and the objects of experience, they maintain that subjects stand in some sort of two-place relation to the contents of their perceptual experiences. In other words, the content of a particular perceptual experience is or involves something ontologically distinct from the subject’s experience. Consequently, although representationalists set themselves up as direct opponents of naïve realism they haven’t in fact moved very far away from a relational analysis of perceptual experience after all.¹⁰ It is easy to overlook the essential relationality of representationalism since it has become popular to refer to naïve realism as ‘relationalism’ or ‘the relational view’. This is unfortunate inasmuch as it masks the fact that relationalism is just as much a theoretical commitment of representationalism as it is of naïve realism—it’s just that representationalists are content-relationalists whereas naïve realists are objectrelationalists. Crane makes the relational element of representationalism explicit in the following: Every intentional state, then, consists of an intentional content related to the subject by an intentional mode. The structure of intentionality is therefore relational, and may be displayed as follows: Subject. Intentional mode. Intentional content. (Crane 2003, 39)

¹⁰ A point of clarification is in order. My claim that representationalists advocate what is ultimately a relationalist account of perceptual experience does not commit them to an ‘act-object’ view. Representationalists posit a relation to a content, but they need not hold that this abstract content is the object of the perceptual experience in the same way that sense-data or ordinary objects are the objects of perceptual experience for the sense-data theorist and the naïve realist (respectively). For an argument that some representationalist views collapse into disjunctivism, see Gow 2018.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

137

And Pautz says: Standard intentional states must be taken to be relations to intentional contents or properties or other abstract objects. (Pautz 2013, 205)

Pautz is explicit about the abstract nature of contents in the quote above, and since many philosophers think that perceptual experiences have a propositional content, it is easy to see why this version of mainstream representationalism is committed to abstracta. But it is the fact that mainstream representationalists endorse content-relationalism that explains why they must accept peculiar entities within their ontology. This point is most easily brought out by considering hallucinations. The contents of hallucinations, whether understood to involve propositions, properties, or property complexes, will have to be abstract if the subject is to stand in a two-place relation to them. Let me explain this in more detail. Representationalists can (and sometimes do) try to analyse the contents of veridical perceptual experiences in terms of the actual physical objects and properties which are represented during the experience. This manoeuvre is often motivated by appealing to the idea that perceptual experience is transparent—we are never aware of features of our experiences, but only of the objects and properties our experiences represent.¹¹ If the content of a veridical experience can be analysed solely with reference to the physical objects and properties in the subject’s local environment, then the content of these experiences need not involve abstracta. However, it is not possible to account for the content of hallucinations only by appealing to physical objects and properties in the subject’s local environment. This is simply because the subject is not relevantly related to any physical objects and properties during her hallucinatory experience. Indeed, this is the reason why many representationalists appeal to uninstantiated properties to explain hallucinations in the first place. To summarize, according to mainstream representationalism, all perceptual experiences involve relations, not to objects, but rather to contents which are (at least in the hallucinatory case) nonphysical, abstract entities.

3. Non-relationalism At this juncture, one might wonder why mainstream representationalists commit themselves to content-relationalism. After all, it is this commitment which

¹¹ See Harman 1990 for the contemporary origins of this idea and Tye 2000, 2014b for a detailed development. See Gow 2016 for an argument that the transparency claim is based on a confusion and cannot serve the purpose to which it is assigned.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

138

 

explains their failure to provide a genuinely physicalist account of perceptual experience. And in fact, the solution offered by mainstream representationalism’s critics is to analyse perceptual experiences as non-relational representational/ intentional states (Kriegel 2011; Mendelovici 2018; Nida-Rümelin 2011). Along with mainstream representationalists, non-relationalist representationalists deny object-relationalism. That is, they deny that perceptual experiences involve any essential relations to objects (there will, at least in the good cases, be causal relations to objects). However, they also deny content-relationalism; they deny that perceptual experiences involve relations to contents. Adverbialism, the original non-relational view developed by Chisholm (1957) and Ducasse (1942) is thought to be unable (or unwilling) to accommodate the intentionality of perceptual experience (see Gert this volume.) In other words, perceptual experiences are thought not to have contents at all on this view. Examples of this assumption are easy to find. Tye says: ‘we could even deny that perceptual experiences have contents (as, for example, adverbial theorists do)’ (Tye 2007, 610). And Crane describes adverbialism as the view which explains ‘all features of what it is like to have an experience in terms of intrinsic, nonintentional qualities of experience’ (Crane 2006, 142). However, contemporary advocates of non-relational views do not share this commitment, and instead offer a non-relational analysis of intentionality itself (Kriegel 2011; Mendelovici 2018; Nida Rümelin 2011). So contemporary non-relationalism is a version of intentionalism/ representationalism: perceptual experiences have contents. It is the denial of content-relationalism which distinguishes non-relationalist representationalists from mainstream representationalists. Instead of construing content as something abstract to which the perceiver stands in a two-place relation, according to the non-relational representationalist framework, contents characterize experiences; they are ways of experiencing. Adverbialists notoriously advocated a change in how we talk about our experiences to reflect the essentially non-relational nature of perceptual experience. They suggested that we exchange our existing ‘relational’ terminology with terminology that exposes the true metaphysical nature of perceptual experience. ‘I see red’ becomes ‘I see redly’. This terminology reflects the fact that seeing something red is not essentially a matter of standing in a relation to something red; instead, it is a matter of seeing in a certain way—‘redly’ refers to a way of experiencing. All non-relational accounts of perceptual experience are committed to denying that perceptual experience involves any essential two-place relations. We stand in causal relations to our environment, and when things go well we perceive our environment, but our perceptual experiences involve neither essential relations to objects nor to contents. Denying both object-relationalism and contentrelationalism allows non-relational theories to be genuinely physicalist. Perceptual experiences are non-relational states or processes of perceivers, and we can

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

139

identify these states or processes with neural states or processes.¹² Although nonrelational views are able to meet the demands of physicalism, there are further consequences of denying that perception involves essential two-place relations. In the remainder of this chapter I will outline one potential difficulty relating to whether we can still assess perceptual experiences for accuracy on a non-relational representationalist framework.

4. Perceptual Experiences and Accuracy It is natural to think that our perceptual experiences are assessable for accuracy. The terms ‘veridical’ and ‘illusory’ imply a conception of perception according to which our experiences can get the world right, or get the world wrong. What is more, the idea that our perceptual experiences are representational or intentional would usually be regarded as entailing a commitment to accuracy conditions. So let’s get clear on what assessing our perceptual experiences for accuracy would involve. Consider the following quotation: Attributing accuracy to something thus involves assessing it with respect to something else . . . (Siegel 2010, 31)

This seems exactly right. Assessing for accuracy involves, at the very least, making a comparison between two things. When it comes to perceptual experience, assessing for accuracy would require comparing the way the world actually is with the way the world is presented in the experience. If the way the world is presented in the experience matches the way the world is, then the perceptual experience is accurate. Michelle Montague has offered a well-developed theory about the process of making accuracy assessments which she calls the ‘matching view’ (Montague 2013). She claims that unless there is a sufficiently close match between how the world is and how it is represented in the perceptual experience, then true perceiving (when ‘perceiving’ is understood as a success verb) will not have occurred. In other words, if our experience gets things wrong to a significantly large degree, then we cannot call the experience one of perceiving the world, even if it has been caused in a similar way to cases where perceiving does take place (such as stimulation by photons in vision and by sound waves in audition).¹³ ¹² Of course, non-relationalism isn’t essentially physicalist—one could be a substance dualist yet still believe that perceptual experiences are non-relational states or processes of perceivers. The important point is that, unlike mainstream representationalism which relies on relations to abstract entities, nonrelationalism enables us to identify a perceptual experience entirely with a neural process. ¹³ Incidentally, Montague’s position on this is a very strong one. She says she will assume, for ease of explication, the falsity of the eliminativist view of secondary qualities like colours and sounds. This is perhaps misguided since, if eliminativism about colours (and so on) is correct—and indeed, it seems to be the prevailing view amongst vision scientists—it seems that Montague will have to claim that we

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

140

 

It is easy to understand how the process of assessing for accuracy works by using sense-datum theory as an example. Assessing our perceptual experiences for accuracy on the sense-datum theorist’s framework is a matter of comparing the sense-data the perceiving subject are directly aware of with the local, mindindependent environment. If there is a sufficient degree of resemblance between the sense-data and the objects and properties in the world which were causally involved in the experience, then the matching condition has been met, and we have a case of accurate, or veridical, perception. Of course, there is a well-known objection to the indirect realist’s proposal, originating with George Berkeley. Aiming his objection towards the views of his contemporaries, who appealed to ideas rather than sense-data, Berkeley stated that ‘an idea can be like nothing but an idea’ (Berkeley 1948–57, PHK 8). It isn’t entirely clear what his reasons were for this claim. Perhaps the thought is that it doesn’t make sense to say that two things resemble each other unless they can be compared, and because we only have direct access to our own ideas (or sensedata), we can never be in a position to make the required comparison. Or Berkeley could be claiming that entities from different metaphysical categories—mental entities like ideas and sense-data on the one hand, and physical entities in the mind-independent world on the other—are too different to be compared. (See Downing 2013 and Winkler 1989 for further discussion.) Indirect realists can respond to these objections relatively easily. In response to the first point, they can concede that we can’t make the comparison ourselves, but hold that there is still a fact of the matter about whether our ideas or sense-data resemble physical objects. Indirect realists can operate with a relatively weak notion of resemblance, and this enables them to respond to the second point; they can argue that so long as the ideas or sense-data possess some of the same properties as objects in the world, it makes sense to think that a comparison can take place, and that an adequate degree of resemblance can obtain. Further problems may arise since the red of the sense-datum tends not to be thought of as exactly the same property as the red in the world, but some sort of relation exists between them and so we can see what direction the sense-datum theorist can move in. The fact that sense-datum theory posits something (sense-data) which stand in a relation to the perceiver’s experience at least provides the theory with something that can be compared with the world. It is clear that a comparison can, at least in principle, take place on this account, and so the idea that our

never see the world. (See Gow 2014 for discussion.) After all, not only is it a central feature of the phenomenology of our visual perceptual experiences that we experience objects in the world as being coloured, for the most part, seeing objects’ colours constitutes our seeing the object. In other words, in typical circumstances, we see objects in virtue of seeing their colours. If it turns out that objects are not in fact coloured in the way our experience takes them to be, then according to Montague’s analysis, none of our visual perceptual experiences will ‘match’ the world to a degree sufficient to warrant ‘seeing’.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

141

perceptual experiences can be assessed for accuracy on an indirect realist framework is, therefore, perfectly coherent. It should also be clear how our perceptual experiences can be assessed for accuracy on the view offered by mainstream representationalism. Recall that on this view, our perceptual experiences involve a relation to a content, such as a proposition. We can therefore assess our perceptual experiences for accuracy by comparing the content of our experience—the way our experience says that the world is—with the way the world actually is. If the way the world is matches the content of the experience, then it is accurate or veridical, if not, it is inaccurate or non-veridical. Again, the reason the comparison required for assessing for accuracy can take place is because mainstream representationalists posit something (a content) to which the perceiver is related, and this content component can be compared with the world. However, when it comes to trying to assess perceptual experiences for accuracy on a non-relationalist framework, problems start to surface. To begin with, it is important to reiterate that the reason we are able to assess our perceptual experiences for accuracy on both mainstream representationalism and sense-datum theory is because our experiences involve something which we are able to compare with the world. On mainstream representationalism, our experiences involve a relation to a content, and on sense-datum theory, our experiences involve relations to sense-data. Putting aside issues regarding whether we are in fact able to carry out the comparison ourselves, the fact that our perceptual experiences involve something with which a comparison can be made ensures that there is a fact of the matter about whether or not our experiences match the world. This allows our experiences to qualify as either accurate/veridical or inaccurate/non-veridical. Now, on a non-relationalist framework, our perceptual experiences do not involve essential relations to either objects or contents. This means that there isn’t anything—neither sense-data, nor propositional content, nor any other kind of content component—which can be compared with the world to see whether it matches. In other words, it is difficult to see how a comparison could even begin to take place—what exactly are we to compare with the world? It is quite clear that making a comparison necessarily involves (at least) two things. When we are wondering whether our perceptual experiences match the way the world is (and are therefore veridical) the experience must provide us with something which can be compared with the world. On the sense-data view, our perceptual experience involves sense-data, which can be compared with the world to see if the perceptual experience deserves to be classified as veridical or not. So our comparison involves the world on the one hand, and sense-data on the other. On mainstream representationalism, our perceptual experience involves a distinct content which says something about the way the world is, and can therefore be compared with how the world actually is. So in this case, our comparison involves

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

142

 

the world on the one hand, and a representational content on the one other. But on non-relationalism, we just have a single, non-relational mental state/process. There is nothing (no thing) with which to carry out the comparison; for the content of the state is not something distinct and separable from the subject’s experience, the content is simply a characterization of the non-relational experience. Since making a comparison involves two things, and non-relational perceptual experiences do not provide us with anything distinct from the experience with which to make our comparison with the world, the matching condition cannot be met. Perhaps the non-relationalist will respond to the challenge by proposing that we compare the whole non-relational perceptual experience with the world. This is problematic. After all, perceptual experiences are just ways of experiencing on this view, and so the proposal would be that on the one hand we have the world and on the other, a way of experiencing. Not much sense can be made of the claim that we can compare the world with a way of experiencing. How would the world have to be to match a way of experiencing? To paraphrase Berkeley; surely a way of experiencing can only be like another way of experiencing. The difficulty that non-relationalists have with meeting the matching condition is most clearly brought out by employing the adverbialist’s terminology, which, for all its shortcomings, reflects very clearly the non-relational nature of perceptual experience. Consider a perceptual experience as of a purple unicorn. In order to reflect the non-relational nature of this experience, the adverbialist will characterize it as a subject ‘seeing purpley and unicornly’. Again, this reflects the fact that perceptual experiences are a matter of experiencing a certain way rather than a matter of standing in a relation to something. In order for this experience to be assessed for accuracy, we need to carry out a comparison with the way the world is, but what precisely are we to compare with the world? The experience does not involve a relation to an object (like a sense-datum) which can be compared with the world to see if it matches. Neither does it involve a relation to a content which ‘says’ something about the way the world is, and can render the experience accurate so long as the way the world actually is matches with this content. Contents on non-relationalism do not make claims about the way the world is; instead, they are simply ways of experiencing, and we cannot compare a way of experiencing with the world. ‘Anna sees purpley’ is on a par with ‘Anna walks slowly.’ It is clear that it makes little sense to ask whether Anna’s walking slowly is accurate or not, and it is equally unclear what sense can be made of asking whether Anna’s seeing purpley is accurate or not.

5. Conclusion A commitment to physicalism characterizes contemporary philosophy of mind. Indeed, the idea that our account of perceptual experience should be compatible

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

143

with physicalism seems to be one of the few assumptions that are shared by the majority of philosophers. My aim in this paper has not been to defend this assumption, nor to defend the theory of perception which seems best placed to provide a genuinely physicalist account of our perceptual experiences. Instead, my aim has been to explore some of the consequences that arise from this widespread commitment to physicalism. I have argued that mainstream representationalism, which advertises its physicalist credentials as its major selling point, is not a genuinely physicalist view. Versions of this theory typically make essential use of relations to abstract, non-physical entities in their accounts. What is more, I have argued that so long as content is construed as something to which perceivers stand in some sort of two-place relation, all versions of mainstream representationalism will fail to conform to a physicalist metaphysics. This is because content must (at least in some cases) be abstract (and therefore nonphysical) if the perceiver is to stand in a two-place relation to it. The physicalist solution to these problems is to adopt non-relationalism. On non-relationalist views, perceptual experiences should not be analysed as relations to objects nor to contents. Perceptual experiences are non-relational mental states, and the content of the experience is a way of characterizing the experience— contents are ways of experiencing. Although non-relational views are able to meet the physicalist’s criteria, I have argued that they encounter difficulties when trying to explain how our perceptual experiences can be assessed for accuracy. This is because (at least according to a popular view) assessing for accuracy requires comparing an aspect of the experience with the mind-independent world. On the sense-datum theory we can compare sense-data with the world, and on mainstream representationalism we can compare the content of the experience with the world, but if perceptual experiences are non-relational then they cannot provide anything with which a comparison to the world can be made. Perhaps there are other legitimate ways of assessing perceptual experiences for accuracy besides the matching view. However, as things stand, it looks as if the generally held belief that we should begin our investigations into the nature of perception from a physicalist standpoint may result in our having to give up on the idea that our perceptual experiences can be assessed for accuracy.

References Ayer, A. J. 1956. The Problem of Knowledge. London: Macmillan. Bealer, G. 1982. Quality and Concept. Oxford: Clarendon Press. Bengson, J., Grube, E., and Korman, D. 2011. ‘A New Framework for Conceptualism’. Noûs 45 (1): 167–89. Berkeley, G. 1948–57. The Works of George Berkeley, Bishop of Cloyne. A.A. Luce and T.E. Jessop (eds.). London: Thomas Nelson and Sons. 9 vols.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

144

 

Broad, C.D. 1925. Mind and its Place in Nature. London: Routledge and Kegan Paul. Byrne, A. 2005. ‘Perception and Conceptual Content’. In M. Steup and E. Sosa (eds.) Contemporary Debates in Epistemology. Oxford: Blackwell, 231–50. Byrne, A. and Logue, H. (eds.). 2009. Disjunctivism: Contemporary Readings. Cambridge, MA: MIT Press. Chisholm, R. 1957. Perceiving: A Philosophical Study. Cornell University Press. Churchland, P. 2007. Neurophilosophy at Work. New York: Cambridge University Press. Crane, T. 2003. ‘The Structure of Intentionality’. In Q. Smith and A. Jokic (eds.) Consciousness: New Philosophical Perspectives. Oxford: Oxford University Press. Crane, T. 2006. ‘Is there a Perceptual Relation?’ In T. Gendler and J. Hawthorne (eds.) Perceptual Experience. Oxford: Oxford University Press. Downing, L. 2013. ‘George Berkeley’. The Stanford Encyclopedia of Philosophy. Edward N. Zalta. https://plato.stanford.edu/archives/spr2013/entries/berkeley Dretske, F. 1995. Naturalizing the Mind. Cambridge, MA: MIT Press. Dretske, F. 2000. Perception, Knowledge and Belief. Cambridge: Cambridge University Press. Ducasse, C. J. 1942. ‘Moore’s ‘The Refutation of Idealism’. In P. Schlipp (ed.) The Philosophy of G.E. Moore. Evanston, IL: Northwestern Press. Forrest, P. 2005. ‘Universals as Sense-Data’. Philosophy and Phenomenological Research 71 (3): 622–631. Glüer-Pagin, K. 2014. ‘Looks, Reasons, and Experiences’ in B. Brogaard (ed.) Does Perception Have Content? Oxford: Oxford University Press. Gow, L. 2014. ‘Colour’. Philosophy Compass 9 (11): 803–13. Gow, L. 2016. ‘The Limitations of Perceptual Transparency’. The Philosophical Quarterly 66 (265): 723–44. Gow, L. 2018. ‘Why Externalist Representationalism is a Form of Disjunctivism’. Ratio 31 (S1): 35–50. Harman, G. 1990. ‘The Intrinsic Quality of Experience’. In J. Tomberlin (ed.) Philosophical Perspectives 4/Action Theory and Philosophy of Mind. Atascadero, CA: Ridgeview, 31–52. Horgan, T. and Tienson, J. 2002. ‘The Intentionality of Phenomenology and the Phenomenology of Intentionality’. In D. Chalmers (ed.) Philosophy of Mind: Classical and Contemporary Readings. Oxford: Oxford University Press. Horgan, T, Tienson, J. and Graham, G. 2004. ‘Phenomenal intentionality and the brain in a vat’. In R. Schantz (ed.) The Externalist Challenge. New York: Walter De Gruyter. Jackson, F. 1977. Perception: A Representative Theory. Cambridge: Cambridge University Press. Johnston, M. 2004. ‘The Obscure Object of Hallucination’. Philosophical Studies 120: 113–83.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

145

Kriegel, U. 2011. Sources of Intentionality. Oxford: Oxford University Press. Loar, B. 2003. ‘Phenomenal Intentionality as the Basis of Mental Content’. In M. Hahn and B. Ramberg (eds.) Reflections and Replies: Essays on the Philosophy of Tyler Burge. Cambridge, MA: MIT Press. Lowe, E. 1992. ‘Experience and its Objects’. In T. Crane (ed.) The Contents of Experience. Cambridge: Cambridge University Press. Lycan, W. 2001. ‘The Case for Phenomenal Externalism’. Philosophical Perspectives 15: 17–35. McGinn, C. 1999. ‘The appearance of colour’. In Knowledge and Reality: Selected Essays. Oxford: Clarendon Press. Mendelovici, A. 2018. The Phenomenal Basis of Intentionality. Oxford: Oxford University Press. Millikan, R. 1984. Language, Thought and Other Biological Categories. Cambridge, MA: MIT Press. Montague, M. 2013. ‘The Access Problem’. In U. Kriegel (ed.) Phenomenal Intentionality. New York: Oxford University Press. Nida-Rümelin, M. 2011. ‘Perceptual Presence and Perceptual Awareness: A Subjectivist Account of Openness to the World’. Philosophical Issues 21 (1): 352–83. Papineau, D. 1984. ‘Representation and Explanation’. Philosophy of Science 51: 550–72. Papineau, D. 2014. ‘Sensory Experience and Representational Properties’. Proceedings of the Aristotelian Society 114 (1): 1–33. Pautz, A. 2007. ‘Intentionalism and Perceptual Presence’. Philosophical Perspectives 21 (1): 495–541. Pautz, A. 2013. ‘Does Phenomenology Ground Mental Content?’. In U. Kriegel (ed.) Phenomenal Intentionality. New York: Oxford University Press. Price, H. H. 1950. Perception. Second edition. London: Methuen. Elisabeth of Bohemia and Descartes, René. 2007. The Correspondence between Princess Elisabeth of Bohemia and René Descartes, edited and translated by Lisa Shapiro. Chicago: University of Chicago Press. Robinson, H. 1994. Perception. New York: Routledge. Rosen, G. 2018. ‘Abstract Objects’. The Stanford Encyclopedia of Philosophy. Edward N. Zalta (ed.). https://plato.stanford.edu/archives/win2018/entries/abstract-objects/ Russell, B. 1912. The Problems of Philosophy. New York: Oxford University Press. Schellenberg, S. 2011. ‘Ontological Minimalism about Phenomenology’. Philosophy and Phenomenological Research 83: 1. Siegel, S. 2010. The Contents of Visual Experience. Oxford: Oxford University Press. Sosa, D. 2007. ‘Perceptual Friction’. Philosophical Issues 17 (1): 245–61. Stoljar, D. 2004. ‘The Argument from Diaphanousness’. Canadian Journal of Philosophy Supp. 30: 341–90.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

146

 

Stoljar, D. 2017. ‘Physicalism’. The Stanford Encyclopedia of Philosophy, Edward N. Zalta (ed.). https://plato.stanford.edu/archives/win2017/entries/physicalism/ Thau, M. 2002. Consciousness and Cognition. Oxford: Oxford University Press. Thompson, B. 2008. ‘Representationalism and the Argument from Hallucination’. Pacific Philosophical Quarterly 89 (3): 384–412. Tye, M. 1995. Ten Problems of Consciousness. Cambridge, MA: MIT Press. Tye, M. 2000. Consciousness, Color and Content. Cambridge, MA: MIT Press. Tye, M. 2002. ‘Visual Qualia and Visual Content Revisited’. In D. Chalmers (ed.) Philosophy of Mind: Classical and Contemporary Readings. Oxford: Oxford University Press. Tye, M. 2007. ‘Intentionalism and the Argument from No Common Content’. Philosophical Perspectives 21 (1): 589–613. Tye, M. 2014a. ‘What is the Content of a Hallucinatory Experience?’ in B. Brogaard (ed.) Does Perception Have Content? Oxford: Oxford University Press. Tye, M. 2014b. ‘Transparency, Qualia Realism and Representationalism’. Philosophical Studies 170: 39–57. Winkler, K. P. 1989. Berkeley: An Interpretation. Oxford: Clarendon Press.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

8 High-Level Perception and Multimodal Perception Dan Cavedon-Taylor

1. Introduction Two central topics in contemporary philosophy of perception are the admissible contents of perception debate and increasing recognition of perception’s multimodality. Puzzlingly, each is typically addressed in isolation from the other. Notably, the question of which properties are admissible in perceptual content is largely debated within a unimodal, visual framework. This is surprising, since perceiving the world via only one sense tends to be a rare exception. Multimodal perception is the norm, and so philosophers interested in which properties are admissible in perceptual content owe it to themselves to explore the issue from a multimodal perspective. This paper has two goals: the first is to motivate such a reorientation of the admissible contents debate, charting its potential significances. The second is to explore the possibility that at least one empirical study of multimodal perception, the rubber hand illusion, supports a so-called Liberal (or “high-level” or “rich”) account of perception’s admissible contents. One reason that Liberals should welcome the argument developed here is that it constitutes a reply to the following objection to their view: including high-level properties, like natural and artefactual kinds, in perceptual content intolerably broadens the scope of what is to count as a perceptual illusion. Alex Byrne, for instance, writes: Visual illusions, as the object of study in the visual sciences, concern properties like shape, motion, colour, shading, orientation and the like, not properties like being tired, belonging to Smith or being a lemon. There is thus no immediate reason to take (visual) perceptual content to include the proposition that o is a lemon, and the like. (2009, 449; see also Byrne and Siegel 2016)

If what I say here is right, then we have reason to think that perceptual content can include such high-level properties precisely on the basis that there is a perceptual illusion, the rubber hand illusion, that, as an object of study by empirical psychology, does not concern low-level properties alone. Dan Cavedon-Taylor, High-Level Perception and Multimodal Perception In: Purpose and Procedure in Philosophy of Perception. Edited by: Heather Logue and Louise Richardson, Oxford University Press (2021). © OUP. DOI: 10.1093/oso/9780198853534.003.0008

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

148

 -

First, what is the admissible contents debate about? Suppose one thinks that perceptual experience has content, that it attributes properties to objects and so represents the world as being some way. Someone who held this view must take a stand on what properties can figure in perceptual content; that is, which ways can perception represent the world as being? Two positions can be distinguished. According to Conservatives (Byrne 2009; Price 2009; Brogaard 2013a; Tye 1995), perceptual experience is limited to representing so-called “low-level” sensibles: Vision:

colour, shape, motion, spatial location, and spatial orientation.

Audition: loudness, pitch, timbre, and duration. Touch: pressure, vibration, temperature, weight, texture, shape, and spatial location. Taste:

saltiness, sweetness, sourness, bitterness, and umami.

Olfaction: camphorousness, pungency, floralness, muskiness, and earthiness.¹ According to Liberals (Bayne 2009; Nanay 2011a and 2011b; Siegel 2006 and 2011), perceptual content can also include what have come to be called “highlevel” properties. A tomato may not simply be represented as red and round; it may be represented as being the kind of thing it is: a tomato. Similarly, tables and chairs may be represented as tables and chairs, not just 3-D expanses of colour and shape at various locations and orientations. According to the Liberal, the representation of such natural and artefactual kinds, i.e. “being a tomato”, “being a table”, etc. is not always a matter of post-perceptual inference, but can be a matter of how one literally sees objects to be.² What about those who deny the representational account of perception? Relationalists hold that perception acquaints subjects with particular objects and/or property-instances, rather than representing such object-property complexes (Martin 2002; Campbell 2002; Brewer 2011; Fish 2009). For relationalists, the debate cannot be framed in terms of what properties perception represents, but should be framed in terms of what properties perception can acquaint one with. Here I will treat the Liberal/Conservative debate as one concerning the contents of ¹ Insofar as low-level properties are typically identified with a conjunction of both the special and common sensibles of the perceptual modalities, there is some debate to be had about precisely which properties are low-level, since there is debate about whether all such properties are really represented in each modality. Consider: should we think of thermal perception as part of touch? What of nociception? The sensibles of olfaction are even less well-defined and typically fail to reflect language used by ordinary people (Kaeppler and Mueller 2013). Here, I simply list some odour terms from biochemist John Amoore’s (1977) influential classification. ² Although much of the debate between Conservatives and Liberals centres on the admissibility in perception of natural and artefactual kinds, the issue may be debated for a host of other properties too, including causal properties (Siegel 2009) as well as action-based and dispositional properties (Nanay 2011a and 2011b).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-    

149

perception. However, in assuming such representationalist terminology, I am simply adopting the terminology in which this issue has, for better or worse, so far been debated. Although some have expressed scepticism about the possibility of framing the debate in terms friendly to the relationalist (Brogaard and Chomanski 2015; Cavedon-Taylor 2015), for the purposes of this paper I will assume that “represents” and its cognates can be treated by relationalists as “acquaints one with” and its cognates (see Siegel 2006; Silins 2013; Reiland 2014; Logue 2013; Bengson 2013; Nanay 2013; Toribio 2018; Di Bona 2017). Indeed, a number of relationalists have weighed in on the debate, some in favour of Liberalism (Fish 2009, 2013; Johnston 2006), others in favour of Conservatism (Martin 2010) and some arguing that there is no fact of the matter (Logue 2013). Why does this debate matter? First, consider the low-level/high-level distinction. It is notoriously difficult to precisely distinguish low-level properties from high-level ones. It might be that the best we can do is say that low-level properties are the ones that we all agree are perceived, while high-level properties are the ones that we debate about (Logue 2013, 1–2; see also Siegel 2006 and Macpherson 2012; but see Martin 2010). But if the low-level/high-level distinction is this ambiguous, one might wonder why anyone should care about the debate. Some recent arguments conclude that it matters for the epistemology (Siegel 2006) and metaphysics (Cavedon-Taylor 2015) of perception, along with the question of perception’s penetrability by cognition (Macpherson 2012) and the possibility of perceptually based action (Nanay 2013). But the debate impacts issues further afield as well. For instance, some hold that we can know others’ minds via perception (McDowell 1982; McNeill 2012; Cassam 2007) and some claim that we can know what is the morally right thing to do via perception (Audi 2013; Cuneo 2013; McBrayer 2010). It is not always clear whether such philosophers think of mental and moral properties as represented in perception, rather than in perceptual belief (though see Werner 2016; Brogaard 2016; Toribio 2018; and Newen 2017). But insofar as any philosopher would wish to articulate a fully perceptual epistemology of others’ minds and moral knowledge, they would thereby require the truth of Liberalism. Getting the right answer to the question of perception’s admissible contents has the potential to impact a number of issues in philosophy of mind, epistemology, and moral philosophy. I proceed as follows: in section 2, motivate the importance of transposing the admissible contents debate into the multimodal framework. In section 3, I explore two widely studied multimodal illusions, the McGurk effect and the ventriloquist effect, and conclude that neither support Liberalism. In section 4, I explore a multimodal illusion that, on balance, does: the rubber hand illusion. In section 5, I consider objections. Section 6 concludes.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

150

 -

2. Beyond Vision and the Unimodal Paradigm Philosophers of perception are increasingly aware of the distorting effect that “visuocentric” biases have had on their field. In theorizing about perception in general, it is becoming clear that we shouldn’t restrict our focus to vision in particular. To take just one example, consider the relation between our senses and space. Vision is a spatial sense, but not in the same way that touch is (Martin 1992). What we see, we see to be located in space at a certain distance and direction from our own body; not so when we touch those things and matters are similarly complex in the case of audition (Nudds 2009). Olfaction, by further contrast, has only very minimal spatial content. What we smell, we typically smell to be located at an undifferentiated location of “here” rather than smelling those things to be distally located (Batty 2011; Matthen 2005). Concerning the relationship between our senses and space, we will have the wrong view of that relationship if we generalize from vision alone. Visuocentrism, while distorting, is sometimes said to be understandable (O’Callaghan 2011a). We often don’t pay attention to how things in our immediate environment feel (e.g., your feet in your socks), how they sound (e.g., the tick of the clock), how they taste (e.g., the lunch you wolfed down before a meeting), and so on. Whereas with vision, things are constantly there, immediately salient to one. All it takes is being awake! But visuocentrism, while understandable, remains problematic. The senses are hugely diverse, both in number and nature (see the papers in Macpherson 2011 and Stokes et al. 2014). We will have a skewed account of the senses if we try to shoehorn all of them into a visual mould. One way that we can break with the visuocentric perspective is to simply ask questions about non-visual modalities that we have historically asked about vision (Di Bona 2017). But this procedure fails to correct the chief wrong of visuocentrism, and a more radical break from it is available. For what is objectionable about visuocentrism is not simply that there are senses it overlooks, but that it is symptomatic of an outlook which misconstrues the nature of perception, both at the experiential and subpersonal levels. In our typical, everyday interactions with the world, it is almost inevitable that more than one sense will be stimulated at a time. And, very often, more than one sense is stimulated by one and the same object. For instance, the things we see are often the things we smell, as when we savour the aroma of coffee before a first sip; the things we touch are sometimes the things we hear, as when we drum our fingers on a table; and the sound of another’s voice is something one both sees and hears to originate from their lips. Experientially, this is not always a matter of co-conscious awareness; that is, mere simultaneous representation of the world via one sense modality and also via another. Typical episodes of multimodal perception involve a stronger degree of unification that mere temporal simultaneity of distinct unimodal experiences

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-    

151

(O’Callaghan 2015; Bayne 2014). Although some so-called “minimally” multimodal experiences may be characterized in this way, e.g., when hearing voices in the café and also feeling the hardness of one’s coffee cup, other multimodal experiences cannot. When we talk face-to-face, my experience of your lips is a single audiovisual one rather than an overall experience in which I perceive your lips twice over, once via vision and again via audition. All this is mirrored in sensory processing, where the binding of information from distinct senses, and which creates a single ‘multisensory event’ (Vatakis and Spence 2007, 744), is exceptionally well-studied by cognitive psychology (Bedford 2001; Calvert et al. 2004; Driver and Noesselt 2008; Welch and Warren 1980). So the true nature of perception will not be respected if we simply ask questions of non-visual senses traditionally asked of vision. (see also O’Callaghan 2016). In following such a procedure, we risk simply reorienting existing debates around one of those other senses, all the while continuing to ignore how perception actually functions for us, i.e. multimodally. Indeed, the Liberal/Conservative debate is increasingly audition-centred (see Brogaard 2018; Di Bona 2017; Nes 2016; and Reiland 2015). Multimodal perception remains overlooked, never mind olfaction, touch, and gustation. Is the neglect of multimodal perception justifiable in the case of the Liberal/ Conservative debate? Perhaps it is. Indeed, there are reasons for thinking that unimodalism is the best strategy to pursue here. First, unimodal perception is, by its nature, simpler than multimodal perception. By sticking with a unimodal approach, we simplify the Liberal/ Conservative debate, making the issue more tractable than it would be from a multimodal perspective. The question of the admissible contents of multimodal perception, being more complex and challenging than its unimodal cousin, can then be usefully informed by what we have discovered in the simpler, unimodal cases. Second, addressing the Liberal/Conservative debate from a multimodal perspective appears to be a redundant procedure. For once we’ve examined this issue from the perspective of the individual senses in isolation, we then seem to have our answer as to whether there is high-level content in multimodal perception. No separate multimodal investigation is required. For instance, suppose that the outcome of a completed unimodal investigation is that vision and audition have high-level content, while olfaction, touch, and gustation do not. It then follows that for any multimodal experience in the traditional five senses (of a suitably mature perceiver) we have an answer as to whether high-level content is present. A multimodal visual-olfactory experience of a spraying skunk would have highlevel content due to vision’s having high-level content: one would see the skunk to be a skunk, while smelling only a pungent odour. By contrast, a multimodal tactile-olfactory experience of touching and smelling a rose in the dark would have only low-level content: one would feel there to be an object which is soft in

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

152

 -

some places, spikey in others, and which is vaguely floral-smelling, but the experience would be incapable of representing the object to be a rose per se. Thus, framing the admissible contents of perception debate in multimodal terms looks unable to tell us anything that we won’t simply glean from the results of a completed unimodal procedure. The difficulty with these motivations for unimodalism is that they presuppose multimodal perception to be a mere conjunctive blend of the contents of visual, auditory, tactual, etc. experiences. Granted, this may sometimes be the case. For instance, Frederique de Vignemont (2014) usefully distinguishes between additive and integrative multimodality. Multimodal perception is additive when it combines experiences in two individual modalities and when those experiences concern different properties of the perceived object. For instance, I may see a car’s wheel spin on the tarmac and also hear its high-pitched chirp as it drives away. Multimodal perception is integrative when, by contrast, at least one instance of the same common sensible is represented by each modality. For instance, in grasping a ball, one sees and feels its sphericality. In additive and integrative cases, multimodal perception is a conjunctive blending of the properties represented by the component experiences, either non-redundantly or redundantly. But one may further distinguish what Robert Briscoe (2016) calls “generative” multimodal perception and O’Callaghan (2016) calls “constitutive” multisensory perception. In these cases, the properties represented multimodally do not appear in one or other of the multimodal experience’s component experiences. A key example is flavour perception, which combines gustation, olfaction, as well as tactual and trigeminal sensations to produce experiences of novel and emergent properties, e.g., “cherryness”, “peatyness”, “astringency”, etc. and which are not themselves found in any of flavour’s component experiences, including taste (Smith 2015). Generative/constitutive multimodal experiences are highly significant for the admissible contents debate. For suppose that a completed unimodal investigation yielded the result that high-level properties are not represented in, e.g., unimodal visual experience, nor unimodal auditory experience, nor unimodal tactual experience (and so on for the other senses). From such a unimodal perspective, this is a situation in which Conservatism has been proved true and Liberalism proved false. But from a multimodal perspective, this conclusion does not follow; given the existence of generative/constitutive multimodal experiences, it would remain open that high-level content appears exclusively in multimodal perception in the way that flavour properties appear exclusively in flavour perception. Thus, the above motivations for unimodalism not only wrongly assume that multimodal perceptual experience is simply the sum of its modality-specific parts, but they stack the deck, albeit modestly, in favour of Conservatism.³ Generative/ ³ Is flavour perception just high-level multimodal perception? Not obviously. Due to the ambiguous nature of the high-/low-level distinction, Conservatives may plausibly deny that flavour properties are

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-    

153

constitutive multimodal perception should motivate Liberals in particular to take very seriously the idea of pursuing the contents of perception debate in a multimodal framework. Importantly, the above is not to say that multimodal perception, by its nature, favours Liberalism. It would be a mistake, given the existence of generative/ constitutive multimodality, to reason from the putative absence of high-level content in unimodal perception to its absence in multimodal perception. But it would similarly be a mistake to reason that the additional complexity of multimodal perception supplies reason to think that high-level content will be found there. Yet this latter reasoning seems implicit in some influential work on highlevel content. Consider the follow remarks of Susanna Siegel’s: [The debate] is less interesting when applied to certain multimodal experiences, compared with purely visual experiences. For instance, the idea that . . . a visual and kinesthetic experience represents that your hand is pulling open a zipper . . . is somewhat less surprising than corresponding theses about what visual experience represents. (2011, 25–6)

Siegel goes on to say that focusing on unimodal cases, vision in particular, ‘will maximize the punch’ (p. 26) of Liberalism. If what I have said above is correct, then the reverse is true, since this procedure characterizes perception in an artificial manner that has little bearing on how we actually perceive. In addition, consider Siegel’s claim that unimodalism is the more interesting perspective from which to examine the admissible contents debate, on the implied basis that it is already intuitive that high-level content is present in multimodal perception. Does the multimodality of perception a priori favour Liberalism in this way? Liberals such as Siegel should refrain from being so cavalier. Conservatives deny that there is high-level content in perception when looking at a zipper; it looks to be a small, grey object with a certain shape and which bears visually perceptible spatial relations to other objects. Will Conservatives find it intuitive that any highlevel property is perceptually attributed to the zipper simply by virtue of the subject’s also having touched the zipper? Surely not.⁴ So while a unimodal perspective on the debate may favour Conservativism over Liberalism, it is not the case that the situation is reversed once one adopts a multimodal perspective. While Liberals owe it to themselves to explore the possibility that there is

high-level and may claim that such properties ought to be considered low-level for the flavour modality. After all, such properties pass the test of being ones that we all agree are represented by flavour. ⁴ The high-level property allegedly represented here, according to Siegel, is “pulling”, a causal property. For simplicity I focus on the kind property “being a zipper”. Conservatives are no less apt to deny that any causal property is perceptually represented merely by virtue of touching the zipper simultaneous with seeing it.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

154

 -

high-level content in multimodal perception, thereby thinking beyond a unimodal perspective, they should not simply assume that high-level content will be found there. To sum up so far, we have good reasons to distrust the results of any unimodal investigation into the admissible contents of perception. At best, that approach will be artificial. At worst, it will be necessarily incomplete in ways that potentially favour Conservatism. On the other hand, we should not assume before investigation that there is high-level content in multimodal perception; this is a further, substantive issue. Finally, it is worth noting the role of empirical findings in the debate. The initial arguments developed in favour of Liberalism took an armchair form insofar as they relied upon intuitions about so-called phenomenal contrast cases, particularly the contrast between expert and novice experiences. Simplifying, there is, intuitively, a contrast in phenomenal character between a perceiver’s perceptual experience of a type of object before and after acquiring a disposition to recognize objects of that type. For instance, seeing pine trees in a forest for the first time seems experientially different from seeing them once one learns to recognize such trees. The best explanation for this experiential difference was thought to be that acquiring a disposition to recognize pines is a matter of the property “being a pine” coming to figure in the content of one’s perceptual (visual) experience of them. Gaining a disposition to recognize pines, ‘you can spot the pine trees immediately: they become visually salient to you’ (Siegel 2011, 100). The difficulty with phenomenal contrasts is now well-recognized: there are alternative explanations for the phenomena that appear no less plausible. Maybe the recognitional disposition manifests itself in perception by modulating one’s attention to previously experienced low-level properties rather than producing the perceptual representation of an additional, high-level property (Block 2014; Nanay 2011b; Briscoe 2015; Price 2009; Fish 2013; Prinz 2013). Alternatively, the phenomenal change may be at the level of cognitive, not perceptual phenomenology (Price 2009; Tye 1995). Another option is that the change takes place at a level between perception and cognition, one involving a state called a “seeming” (Brogaard 2013b; Lyons 2005; Reiland 2014). Phenomenal contrast cases were introduced in order to counter the stalemate that would likely arise between Liberals and Conservatives were they to attempt to settle the debate via a direct appeal to introspection (Siegel 2007). Hence, contrasting mature, “expert” perceptual experience with “novice” perceptual experiences, i.e. where no high-level properties are represented, seemed necessary. But as things have panned out, this method too seems to have produced an impasse. One of the most significant developments in this debate is the turn by Liberals to clinical disorders and experimentally observed effects in order to support their view, including: visual associative agnosia (Bayne 2009), unilateral neglect (Nanay 2012), adaptation after-effects (Block 2014; Fish 2013; Di Bona 2017), and rapid

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-    

155

categorization or gist-perception (Fish 2013; Bayne 2016). The empirically grounded arguments for Liberalism that have resulted, some of which retain aspects of the method of phenomenal contrast, have been thought to represent an improvement over purely armchair ones.⁵ Still, their success remains limited insofar as they assume a unimodal perspective. To better support their view, Liberals should look to results from not merely the psychology of perception, but to the psychology of multimodal perception in particular.

3. Multimodal Perception and the Hunt for High-Level Content The standard way in which psychology has sought to study multimodal perception is via perceptual illusions.⁶ I will concentrate on three of the most widely studied examples: (i) the McGurk effect; (ii) the ventriloquist effect; and (iii) the rubber hand illusion.⁷ In this section, I will argue that (i) and (ii) fail to support Liberalism. Then, in section 4, I will present considerations that favour high-level perceptual content being a necessary background condition of (iii), the rubber hand illusion.

3.1 The McGurk Effect The McGurk effect (McGurk and MacDonald 1976) is an example of an auditoryvisual multimodal illusion. In one version of the illusion, an audio recording of a tokening of the /ba/ phoneme, when paired with a visual recording of a person moving their mouth to produce the /ga/ phoneme, results in an overall multimodal, audiovisual perception of a tokening of the /da/ phoneme.

⁵ These arguments still leave room for dissent (see Briscoe 2015; Brogaard 2013a; Raftopoulos 2015; Helton 2016). Fish (this volume) discusses how scientific claims are susceptible to philosophical reinterpretation. ⁶ The definition, and existence, of perceptual illusions (in the philosopher’s sense) is increasingly debated (Travis 2004; Kalderon 2011; Fish 2009; Brewer 2011). I do not wish to prejudice those issues. If it turns out that we should say that the subject’s perceptual responses to stimuli in these experiments are veridical, nothing I say here should be affected. ⁷ This is just a small sample of multimodal illusions. I ignore here some particularly well-known examples which, like the stream-bounce illusion (Sekuler et al. 1997) and double-flash illusion (Shams et al. 2002), seem explicable entirely in terms of the perception of low-level properties, e.g., shape and motion, along with beeping tones and other simple auditory stimuli.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

156

 -

How might Liberalism figure in the explanation of this illusion? The effect is commonly described in the psychological literature as a matter of speech perception. However, one should not infer from this that the effect necessitates hearing high-level properties, i.e. meanings and related semantic features. Indeed, the effect involves no such thing. Phonological features of a language, like phonemes, are to be regarded as the audible “building blocks” of language. Phonemes like /ba/, /ga/ and /da/ are semantically relevant, but are not semantically significant by themselves. Perceptual experiences of them are not high-level (O’Callaghan 2011b, 801). However, the Liberal may nonetheless argue that the effect depends upon vision’s having high-level content. In particular, they may claim that the effect requires vision to represent the object making the /ga/ movements to be lips or a mouth. The McGurk effect is an instance of facial perception no less than speech perception. So perhaps it is here that one can locate high-level content.⁸ This strategy claims that the representation of high-level properties is a necessary background condition for the illusion, granted that the property illusorily represented in multimodal perception, the /da/ phoneme, is low-level. Experimental results tell against this account of the effect. Rosenblum and Saldana (1996) induced the McGurk effect in subjects who were entirely unaware that they were looking at faces and facial features. In their experiment, the visual stimuli were dot-lights, e.g., of the sort used in motion-capture and CGI, placed on a darkened subject’s lips, teeth, tongue, chin, and cheeks. In the video display shown to participants, only the dots and their motions were visible against a black background; the face and its features could not be seen. When the image of the dot-points was static, none of the participants correctly guessed that they were looking at a mouth. After the experiment was concluded, most participants correctly identified the stimulus, but nearly a quarter guessed incorrectly, showing themselves to be unaware that they were looking at a face. (They guessed the darkened stimulus to be “an owl”, “a butterfly”, and “sheep”.) Crucially, questionnaires filled out by these subjects indicated that the McGurk effect had nonetheless occurred: such subjects had a multimodal, auditory-visual experience of the /da/ phoneme, but no representation in vision (or belief) of a face—only moving points of light. It would thus seem that the McGurk effect does not depend on one’s representing the visual stimulus in a high-level way, e.g., as a face. Rather, it simply depends on the visual representation of spatial-motion properties. Such properties are canonically low-level.

⁸ There is room for debate about whether seeing faces as such is a high-level perceptual phenomenon (Brogaard and Chomanski 2015) or a low-level one (Lyons 2009).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-    

157

3.2 The Ventriloquist Effect A second multimodal illusion which features prominently in the experimental literature is the ventriloquist effect (Howard and Templeton 1966), a familiar one in which the sound of a ventriloquist’s voice at one location, x, is heard to come from a distinct location, y, by the mouth of their “dummy”. Contrary to popular wisdom, this effect does not rely on the ventriloquist “throwing” their voice. Rather, it involves what psychologists call the “visual capture of sound”. When the illusion is successful, the dummy’s mouth “captures” the sound of the ventriloquist’s voice due, in part, to temporal synchrony between movements of the dummy’s mouth and the production of speech by the ventriloquist. The two become crossmodally “feature-bound” or “phenomenally fused”, at the location of the dummy. Does this illusion favour Liberalism? On the face of it, one might think the chances are no better than for the McGurk effect. After all, this is an illusion regarding a spatial property, i.e. the location of a sound, and spatial properties are paradigm examples of low-level properties. But recall the second route to Liberalism explored in relation to the McGurk effect: Liberals may claim that the perceptual representation of high-level properties is a necessary, background condition for the illusion and its phenomenal fusion to occur, granted that the property illusory represented at the multimodal level, the sound’s spatial location, is itself low-level. How might this idea be developed? The reasoning might go as follows: the explanation for why visually experiencing a dummy’s mouth to move in time with a ventriloquist’s speech results in phenomenal fusion is that this is a case of (visual) facial perception and (auditory) speech perception. That is, the ventriloquist’s voice over here is heard to be located over there with the dummy’s mouth only because the voice is auditorily represented to have the high-level property “being speech” (i.e. it is not simply heard as a collection of meaningless tones), while the dummy’s mouth is visually represented to have the high-level property “being a mouth” (i.e. it is not merely a moving object having a certain colour and shape). The two stimuli become grouped in perception only in virtue of them first being visually and auditorily represented as being the kinds of things they are, kinds which naturally go together: a mouth and speech. And so perception represents such high-level properties as “being speech” and “being a mouth” as a background condition for the effect. As promising as the above idea might seem, the empirical literature suggests it to be on the wrong track. For one, the visual capture of sound occurs when auditory stimuli are mere tones and pulses (rather than speech) and are paired with visual stimuli that are simple LED lights (rather than mouths) (Radeau and Bertelson 1976; Bertelson and Radeau 1981). Indeed, psychologists Argiro Vatakis

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

158

 -

and Charles Spence (2007, 745) lament the fact that the ventriloquist effect tends to be studied using ‘arbitrary, or nonmeaningful, combinations of stimuli, such as flashing lights and brief tones’ rather than ‘informationally rich’ everyday objects. But the fact that psychologists can induce the ventriloquist effect with nonmeaningful audiovisual stimuli is good news for Conservatives. It casts significant doubt on high-level content being a background condition for the illusion. A potential reply on behalf of the Liberal would be to grant that high-level content isn’t necessary for the ventriloquist effect in general, but that in conditions when the auditory stimulus is speech, a mouth-like visual stimulus is necessary to capture it. This would be to claim that representing a visual stimulus as a mouth is required for the visual capture of speech. However, the empirical literature doesn’t bear this out either. Radeau and Bertelson (1977) found visual capture of speech occurred in situations in which the visual stimulus was lights that flashed in time with speech. So far we have seen reason to doubt both that the ventriloquist effect involves either the multimodal representation of a high-level property or that high-level perception in unimodal perception is a background condition for the effect. But there is another potential explanatory role for high-level content here, which is as an enhancer of multimodal, audiovisual fusion. On this line of thought, high-level content is not necessary for the ventriloquist effect; rather, when present, it aids or strengthens the effect. This conjecture is empirically supported. Consider two experiments in one of the earliest studies of visual capture of sound (Jackson 1953). In one experiment, the visual stimulus was a kettle, an artefact kind, and the audio stimulus was a kettle’s whistle, a sound associated with an artefact kind. In another, the stimuli were bells and flashes, non-meaningful stimuli whose pairing is more arbitrary. In both, the visual stimuli successfully captured the auditory stimuli; asked to point to the location of the sound, subjects reliably pointed in the direction of the visual stimulus as opposed to the actual origin of the sound. However, in comparing the results of the pointing tasks across the two experiments, participants appeared to bias audiovisual fusion when the stimuli were kettles and whistles. Asked to point to where the sound (whistle) was coming from, subjects pointed closer to the visual stimulus in the kettle condition than in the flash/bell condition. This enhanced strength of response towards familiar, everyday objects versus non-meaningful stimuli in the case of the ventriloquist effect is robust (see Vatakis and Spence 2007). The Liberal line of argument under consideration says that the perceptual systems of subjects in these experiments were biased towards fusing the whistle sound with the kettle because they saw and heard the stimuli in that very way, i.e. as kettles and as whistles. Granted, had they failed to represent the stimuli in a high-level way, there still would have been a degree of multimodal, audiovisual fusion, as illustrated by the flash/bell experiment.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-    

159

Conservatives, however, can give equally plausible explanations of the bias. Everyday stimuli such as kettles and whistles can be considered ‘informationally rich’ (Vatakis and Spence 2007), relative to non-meaningful stimuli like flashes and tones. This relative richness can be understood as a greater degree of complexity in terms of low-level properties; it need not be a matter of the everyday objects in Jackson’s experiments being perceptually represented in a high-level way. Kettles, whistles, mouths, and speech tend to be more intricate and complex with respect to their shapes, colours, pitches, etc. than are mere flashing lights and beeping tones. Audiovisual fusion may not bias everyday stimuli by virtue of perceptually representing them as having high-level properties, but simply by virtue of such stimuli having a greater number of low-level properties (see also Deroy 2015, section 4.2). How is this conjecture on behalf of the Conservative explanatory? For one, it fits with discussion in the psychological literature of the role of bottom-up factors in effecting audiovisual fusion. For as well as spatial and temporal coincidence playing a key role, another important factor is dimensions of what Spence (2011, 972) calls synaesthetic congruency between stimuli, where this is precisely a matter of ‘correspondences between more basic stimulus features (e.g., pitch, lightness, brightness, size) in different modalities’. Indeed, there are well-known audiovisual synaesthetic correspondences between such low-level properties as, e.g., hue and pitch, brightness and loudness, and brightness and high pitches. So, when stimuli used to test audiovisual fusion are relatively numerous in terms of their low-level properties, as everyday objects are, one should naturally expect a greater degree of fusion than when simpler stimuli like tones and flashes are used. Everyday objects, by virtue of having a greater number of low-level properties, present greater opportunities for synaesthetic congruency. And where there is greater congruency between audiovisual stimuli, it is well-known that one’s perceptual system is thereby more likely to fuse together such stimuli (Welch and Warren 1980). In addition, since it is plausible that subjects have background beliefs linking whistles to kettles, but not beeps to flashes, Conservatives may argue that the bias subjects exhibit in such experiments simply reflects a post-perceptual decision bias as to the whistle’s spatial location (see Bertelson and de Gelder 2004, 151–6). In sum: we have reason to think that the ventriloquist effect is an instance of a broader audiovisual fusion effect that neither produces, nor relies upon, nor is definitively enhanced by, the perceptual representation of high-level properties.

4. The High-Level Contents of the Rubber Hand Illusion The rubber hand illusion (RHI) (Botvinick and Cohen 1998) is a multimodal illusion in which tactual sensation is referred to a false, rubber hand. While in the ventriloquist effect we find audiovisual fusion, RHI involves a fusion of vision and

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

160

 -

touch. Subjects have one of their hands placed out of view, and a fake appendage placed before them where their hand would ordinarily be. Subjects are then instructed to visually fixate on the rubber hand while it is stroked with a paintbrush. At the same time, their out-of-sight hand receives matching cutaneous stimulation. Introspective reports and behavioural indicators of proprioception having “drifted” towards the rubber hand suggest that subjects begin to experience (non-veridically) the stimulating of the rubber hand; the sensations actually occurring on their hand become spatially referred to the fake one. As the title of Botvinick and Cohen’s study aptly puts it: the rubber hand feels the touch that the eye sees. As in the case of both the McGurk and ventriloquist effects, the properties represented here multimodally do not seem candidate high-level ones, as again the illusion concerns a low-level spatial property. It remains to be seen whether high-level perceptual contents are, however, a necessary background condition for the illusion to occur.

4.1 RHI on Liberalism and Conservatism RHI can be regarded as an illusion in two interrelated ways. First, RHI involves an illusion of bodily ownership: one proprioceptively senses (non-veridically) the rubber hand from the inside. Second, RHI involves a tactual illusion: the cutaneous stimulation of one’s actual hand is felt (non-veridically) to be located on the rubber hand. Indeed, Botvinick and Cohen (1998, 756) describe the illusion as revealing ‘a three-way interaction between vision, touch and proprioception’. But one might simplify Botvinick and Cohen’s description of RHI, referring to it as revealing mere visual-tactual interaction, since proprioception is plausibly necessary for touch to begin with, if not partially constitutive of it (Martin 1992).⁹ From this perspective, it is no surprise that RHI is an illusion of both bodily ownership and tactual sensation. For, if proprioception is partly constitutive of touch, then we can picture the proprioceptive illusion of bodily ownership as the factor that causally explains the (non-veridical) referral of tactual-cutaneous sensation to the rubber hand. In approaching the question of whether there is high-level perception in RHI in terms of such content being a background condition for the illusion, let us begin by focusing on the idea that RHI involves an illusion of bodily ownership. A natural way to unpack this idea is to say that RHI involves a non-veridical, proprioceptive representation of the rubber hand as having the property “mine”. ⁹ RHI might be regarded as multimodal along several dimensions on the basis that touch itself is not a single sense but is multimodal in its own right. Touch combines experiences of pressure, shape, and vibration, as well as temperature, and relies not only on receptors in the skin but also those located in muscles, joints, and tendons (see Fulkerson 2016, section 2).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-    

161

At first glance, this might seem like a candidate high-level property. But that impression should immediately be dispelled. The property “being mine” is best thought of as a low-level property of proprioception, given that it is ubiquitous, if not definitive, of the modality. It seems precisely the kind of property that, when proprioception is construed as a sense modality (Macpherson 2011; Schwenkler 2013), all would agree upon proprioception as representing. Let us switch focus now from the content of proprioception in RHI to the content of touch. It is difficult to see on what grounds a Liberal could argue that high-level tactual content is a necessary background condition for RHI. This would be to claim that RHI requires subjects to tactually represent the object stimulating their out of sight hand to be a brush or a paintbrush, rather than simply representing pressure, textures, vibration, etc. on their skin. This is a fairly implausible suggestion, and so reflection on the content of touch in RHI also fails to motivate the existence of high-level content. What of the visual component of RHI? Here the case for Liberalism begins to look more plausible. To see why, let us contrast how Conservatives and Liberals are likely to construe the visual content of subjects undergoing RHI. On a Conservative account of the content of vision in RHI, subjects visually represent the rubber hand as having a certain colour, shape, size, spatial location, and spatial orientation. In short: it looks to be a wavy-shaped object of a certain colour. This content is likely to be accurate, barring abnormal experimental conditions. Thus, according to Conservatives, the content of vision in RHI is veridical: although vision is a constitutive element of RHI’s multimodality, RHI is an illusion in proprioceptive and tactual content alone. Liberals, by contrast, are apt to claim that the stimulus rubber object is represented by vision, erroneously, as a hand. That is, Liberals are likely to think that the rubber hand is visually represented as not merely an object with a certain size, shape, and colour, but as an object which has the high-level property “being a hand”. The rubber appendage does not have this property—it’s not a hand. So, RHI, on Liberalism, is a threefold illusion with misrepresentation in (high-level) visual content as well as in (low-level) proprioceptive content and in (low-level) tactual content. Putting this together, the crucial matter is that the Liberal thinks of RHI as a multimodal perceptual illusion in which an object is non-veridically seen/proprioceived/tactually felt to have the property “my hand”. Vision supplies the “hand” part of this complex, erroneous content, while proprioception and touch supply the “mineness”. But for conservatives, RHI is at most a twofold illusion in proprioceptive and tactual content alone. In RHI, on this latter view, nothing is amiss from vision’s perspective. In giving conflicting accounts of the extent of misperception in RHI, Conservatives and Liberals cannot both be right. In the remainder of section 4, I shall show why the Liberal account is to be preferred.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

162

 -

4.2 RHI and the Liberal’s Conjecture If proprioception drives the referral of tactual sensation from one’s actual hand to the rubber hand, then what, in turn, drives that initial, proprioceptive referral? The answer, everyone agrees, is vision: seeing (not feeling) the rubber hand to be stroked over there, while at the same time feeling (not seeing) one’s own hand to be stroked over here, constitutes a conflict of spatial information, similar to the audiovisual conflict in the ventriloquist effect. As often occurs with conflicting spatial information across two or more senses, if one of those senses is vision, then that is what tends to dominate. Notice that the above sketch explains the tactual-proprioceptive referral wholly in terms of perceptually low-level properties, i.e. conflicting spatial information generated by temporally synchronous visuo-cutaneous stimulation. But the Liberal should insist that there is an additional explanatory factor at work. Given that they are apt to think of RHI as involving visual misrepresentation of the rubber object as being an actual hand, they surely ought to exploit this as a partial explanation of the tactual-proprioceptive referral. That is, the Liberal should be inclined to say that the tactual-proprioceptive referral occurs, in part, because the rubber appendage is visually represented as being a hand (in addition to its being represented as having the low-level properties just mentioned). Why would that be explanatory? The answer is that if the property “being a hand” is visually attributed to the rubber appendage, then the visual system would be classifying the stimulus as a body part, something that is apt to be proprioceptively perceived and which is a suitable object of tactual-proprioceptive experience. Thus, the candidacy of the rubber appendage for tactual-proprioceptive experience is a product of it being seen in a high-level way. Call this “the Liberal conjecture”. It is important at this point to note that a number of bottom-up factors, which involve perceptually low-level properties, are necessary for RHI. These don’t simply include the synchronous tactual stimulation of one’s actual hand and visual experience of the fake appendage receiving spatially competing stimulation. There must also be congruency between the subject’s hand and the rubber appendage in terms of rotation and orientation (Farnè et al. 2000) and there must be left-hand/leftrubber-hand congruency (or right-hand/right-rubber-hand congruency, as the case may be) (Tsakiris and Haggard 2005). Unless these perceptually low-level conditions are in place, RHI fails. By contrast, colour is a low-level property that matters little, since RHI is not constrained by whether the rubber hand matches subjects’ skin tone (Farmer et al. 2012). Thus, the Liberal conjecture is not that visually representing the rubber stimulus (erroneously) to be a hand, even while receiving synchronous visuo-cutaneous stimulation from the experimenter, is sufficient for RHI. The conjecture is only that it is a necessary one, along with the perceptual representation of the above-mentioned low-level properties.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-    

163

4.3 Empirical Support for the Liberal’s Conjecture The Liberal’s account of the extent of misperception in RHI is that there is misperception not just in tactual-proprioceptive content, but in visual content as well—the rubber stimulus looks to be a hand—something the Conservative will deny. The Liberal conjecture being explored here is that this is necessary to explain why the stimuli object is then felt (non-veridically) to be one’s own hand. On this view, high-level visual content is a necessary background condition for the illusion. Naturally, Conservatives will deny the Liberal’s conjecture and will insist that only the perception of low-level properties is fully sufficient for RHI. Worryingly for the Liberal, early studies of RHI suggest that Conservatives may get their wish. Indeed, one of Botvinick and Cohen’s (1998, 756) initial claims about RHI was that ‘intermodal matching’ is a ‘sufficient’ condition for the tactual-proprioceptive referral; what they likely have in mind is synchrony in terms of visuo-cutaneous stimulation, a perceptually low-level phenomenon which, if sufficient for RHI, would leave no room for the Liberal conjecture. But the idea that RHI can be explained solely in terms of bottom-up, low-level factors, is now widely rejected. More recent, experimental results suggest that RHI requires top-down influences and that bottom-up, stimulus-driven cues, while necessary, are insufficient (De Vignemont et al. 2005; Haans et al. 2008; Tsakiris and Haggard 2005; Tsakiris et al. 2008; Tsakiris et al. 2010). A recent review is emphatic on this matter: Converging evidence from RHI studies, studies on visuo-tactile extinction on neuropsychological patients, and neurophysiological studies on monkeys suggests that correlated multisensory stimulation and spatial proximity are necessary but not sufficient for the integration of a visual stimulus to peripersonal space or for the experience of ownership during the RHI. (Tsakiris 2010, 706, emphasis my own)

This appears to be good news for the Liberal, since this is exactly what they are likely to claim. They will likely say that RHI requires something extra, over and above the representation of low-level properties, and they say precisely what that something is: visually representing the instantiation of the natural kind “being a hand”.¹⁰

¹⁰ Liberals need not, and probably should not, say that this is the only top-down factor. The aforementioned psychologists who claim that top-down factors are necessary for RHI typically have in mind the subject’s body schema. It should be stressed that this need not be at odds with the idea that RHI involves seeing the rubber stimulus to be a hand. Indeed, the two claims may well be complementary insofar as seeing the rubber stimulus in this high-level way could be responsible for schemaincorporation, or vice versa. I leave exploration of this idea for another occasion.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

164

 -

This is only to suggest that the Liberal may be on the right track. Can one find further support for the Liberal conjecture? One way to settle the matter more decisively would be to consider a prediction that the conjecture generates: when the physical stimulus used to induce the illusion is not apt to be seen to be a hand, there will be no RHI. Strikingly, vindication of this prediction can be found in three recent studies of RHI. First, in Tsakiris and Haggard (2005), RHI failed when the visual stimulus was a wooden stick. Similarly, Kalckert et al. (2019) failed to induce RHI using a balloon. Clearly, neither stimulus is a good candidate for being visually represented as a hand. So, from the perspective of the Liberal, it is entirely unsurprising that such studies would fail to induce RHI: sticks and balloons, unlike hands, aren’t represented by vision as falling under any kind whose instances are suitable objects for being proprioceptively sensed. The fact that tactual-proprioceptive referral failed in such conditions is as exactly as the Liberal conjecture predicts. Third, in the experiments conducted by Tsakiris et al. (2010), a number of stimuli were used, beginning with a wooden plank, and progressively modifying the stimulus’ shape across experiments so that it became more hand-like (see Figure 8.1). Crucially, of the stimuli depicted in Figure 8.1, only stimulus 5 induced RHI. Not even stimulus 4 was sufficient. Again, the Liberal has an explanation: only stimulus 5 was seen by subjects to be a hand. This is plausible since, on the face of it, only stimulus 5 is uncontroversially visually passable, without further contextual information, e.g., jewellery, as being a hand.¹¹ No

Stimulus 1

Stimulus 2

Stimulus 3

Stimulus 4

Stimulus 5

Figure 8.1. Stimuli used in Tsakiris et al. (2010) Credit: Reprinted by permission from Springer-Verlag, Experimental Brain Research, ‘Hands Only Illusion: Multisensory Integration Elicits Sense of Ownership for body parts but not for non-corporeal objects’ Manos Tsakiris et al. © 2009.

¹¹ It is plausible that with further contextual information, stimulus 4 would be seen to be a hand, e.g., if jewellery and/or a watch were appropriately placed on it. Also, one is perhaps more likely to think that stimuli 3 and 4 are visually passable as hands when the stimuli are lined up together and seen concurrently with stimulus 5, as in Figure 8.1. But this was not the viewing condition of subjects in these experiments.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-    

165

surprise, the Liberal will think: only stimulus 5 could induce RHI, since only that stimulus, in being apt to be seen to be a hand, could confer vision with content sufficient to fool tactual and proprioceptive awareness. Conservatives, by contrast, ought to be deeply puzzled by the results of Tsakiris and Haggard (2005), Kalckert et al. (2019), and Tsakiris et al. (2010). According to Conservatives, nothing can look to be a hand. Here’s the rub: then why is it that only visual stimuli exceptionally similar to hands trigger RHI? We ought to be able to deduce from Conservativism some minimally explanatory claim about why RHI failed in these experiments, in the way that we have deduced from Liberalism a very clear answer on this matter. But the Conservative comes up short. Indeed, the Conservative is at risk of taking it to be a brute fact that only an exceptionally narrow range of groupings of low-level properties, “gestalts”, can refer proprioceptive and tactual content in RHI. However, this is far from satisfying, since it is unlikely to be by fluke that these suitable gestalts are identical to the visual gestalts typical of hands. This would amount to simply taking it for granted that visual experience of the rubber hand’s size and shape properties does, whereas visual experience of other stimuli’s size and shape properties does not, refer the tactualproprioceptive sensation. Moreover, in pursuing this line of reply, the Conservative would risk ignoring the previously mentioned fact that RHI is partially effected in a top-down manner, since the low-level gestalts that they will appeal to are perceptually represented in a purely bottom-up way. Neither this, nor the explanatory paucity of the Conservative view regarding why RHI fails in the above three experiments, are acceptable. Hence RHI lends support to Liberalism. Another potential reply would be for the Conservative to concede that the rubber hand has to be represented as a hand at some level for RHI to occur, but insist that this representation is not at the level of conscious perceptual experience. Read one way, e.g., as claiming that the rubber hand is represented as a hand in belief, the reply has no credibility; subjects are not fooled into thinking the stimulus is an actual hand. Read another way, e.g., as claiming that the rubber hand is represented as a hand in unconscious perceptual experience, the reply is more credible, but is a significant concession to Liberalism, if not an outright abandonment of Conservatism; Liberals, for consistency’s sake, may take their thesis to be about both conscious and unconscious perception and not merely the former. Setting the matter of consistency aside, Conservatives would still owe a compelling story about why the property “being a hand” can only be represented by perception unconsciously and not also consciously. A third potential reply would be for Conservatives to claim that the narrowness of the range of gestalts capable of inducing RHI is not a brute fact, but is explained by the further fact that tactual-proprioceptive sensation can only be referred from one’s actual hand to an object very much visually like it in terms of low-level properties. But this does nothing to help remove the mystery surrounding the

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

166

 -

Conservative account of RHI failure in the above three experiments. Without some representation in the subject’s psychology of the rubber hand as a hand, it is still inexplicably brute why such referral should occur simply on the basis of visually experiencing a gestalt that happens to be similar to those of actual hands, but which is not, in addition, represented as being in any way hand-related from the subject’s point of view. As discussed in the paragraph above, there seem no plausible options for the Conservative to identify as the bearer of the relevant “hand” content. The Liberal’s account of RHI gains a measure of further support insofar as it predicts that somatoparaphrenic subjects will be strongly susceptible to the illusion. Such subjects have reduced feelings of ownership for their bodies and typically have weakened proprioception (Vallar and Ronchi 2009). As such, they seem all the more likely to have their tactual-proprioceptive sensations dominated by the conflicting visual information that occurs in RHI. This prediction appears vindicated by at least one study of RHI in a somatoparaphrenic subject (Van Stralen et al. 2013). Strikingly, the subject reported feelings of ownership for the rubber hand simply when looking at it, prior to the onset of tactual-cutaneous stimulation. The Liberal has a partial explanation for this surprising result: the stimulus was being seen to be a hand, i.e. a body part, something apt for ownership, hence why a somatoparaphrenic subject reported feelings of ownership towards it, even before tactual stimulation.¹² Unless they saw the rubber appendage as a hand, their having feelings of bodily ownership for it would appear unexplained. This is precisely how matters stand on Conservativism.

5. Feeling What’s Not There Frederique De Vignemont (2014) has recently developed an argument that seems to show that RHI cannot support Liberalism. De Vignemont considers whether the possession of a sortal concept, in perception or cognition, is necessary for RHI. (Sortal concepts are typically thought to be constituents of high-level perceptual content.) They write: [I]n the Rubber Hand Illusion, it seems unlikely that the participants feel that their hand is F, see that the rubber hand is F, erroneously judge that their rubber hand is their own hand, and then only integrate what they feel with what they see. It may rather be the reverse. Participants do not identify the rubber hand as their hand and then experience the illusion; rather, they experience the illusion and

¹² Again, one shouldn’t forget about various bottom-up factors needing to be in place as well. After all, somatoparaphrenic subjects do not have feelings of bodily ownership for each and every hand they observe in everyday life.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-    

167

only then do they judge as if the rubber hand were their hand. The identification of the rubber hand as one’s hand is not a prerequisite of visuo-somatosensory binding; it is a consequence of it. (2014, 144)

But it is important not to misconstrue the Liberal conjecture. Liberals are likely to think that subjects in RHI perceptually represent the rubber stimulus as having the multimodal, erroneous content “my hand”, with proprioception and touch supplying the “mineness” and vision supplying the “handness”. This is to agree with De Vignemont that the illusion of ownership occurs prior to the identification of the hand as “my hand”. But it is to disagree that what follows is that sortal concepts are not involved in such multimodal fusion. What De Vignemont does not consider is whether participants first erroneously identify via perception the stimulus to be “a hand” (granted they do not first identify it as “my hand”, at least not prior to the illusion having occurred and the integration of vision with proprioception and touch). Moreover, De Vignemont mentions some RHI experiments that seem to pose a problem for views of RHI along the lines of Liberal conjecture. Notably, Guterstam et al. (2013) elicited the illusion in the absence of any stimuli, referring tactual sensation into empty, peripersonal space. The effect was that participants felt as if they had an invisible hand, yet in the absence of stimuli that could be visually represented as a hand. What can the Liberal say about this? Worryingly for them, Guterstam et al. (2013) are not alone in inducing RHI via non-bodily stimuli. For instance, Armel and Ramachandran (2003) referred tactual-proprioceptive feedback to the corner of a table, and Ramachandran and Hirstein (1998) referred it to a shoe. However, the Liberal may take heart from the fact that these results are not widely replicated, indeed the former never so (Ma and Hommel 2015, 76). Indeed, it is striking that Guterstam et al. (2013) failed to induce RHI when using a wooden box as stimulus. This is entirely in line with the results of Tsakiris and Haggard (2005), Kalckert et al. (2019), and Tsakiris et al. (2010) and which were marshalled above as having the potential to vindicate the Liberal conjecture. Insofar as the replicability of Guterstam et al.’s (2013) own results remain to be seen, we don’t have here a knock-down objection to the Liberal conjecture. There is also a further question about whether to type this illusion as a variety of RHI or whether to type it as its own sui generis “invisible hand” illusion. Less defensively, suppose we grant that Guterstam et al. (2013) have shown that RHI can be induced in a purely bottom-up manner, via the perception of low-level properties alone. The fact that RHI can be so induced when no physical object is present does not show that it can also be induced in a purely bottom-up, low-level manner when a physical object is present, without any need of high-level content. Again, the results of experiments by Tsakiris and Haggard (2005), Kalckert et al. (2019), and Tsakiris et al. (2010) seem to confirm precisely this: low-level, bottom

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

168

 -

up, stimulus-driven cues are insufficient for RHI when the visual stimulus is a physical object. The fact that they may be sufficient when no physical object is present is therefore not to the point.

6. Conclusion Reflection on RHI suggests two conclusions. First, that high-level properties figure in perceptual content. Second, that Conservatives who claim that a consequence of the Liberal view is an intolerable broadening of the scope of perceptual illusions, particularly from the perspective of perceptual psychology, should pursue other arguments against the view.

References Amoore, J. 1977. ‘Specific Anosmia and the Concept of Primary Odors’. Chemical Senses 2 (3): 267–81. Armel, K., & Ramachandran, V. S. 2003. ‘Projecting Sensations to External Objects: Evidence from Skin Conductance Response.’ Proceedings of the Royal Society B: Biological Sciences 270 (1523): 1499–506. Audi, R. 2013. Moral Perception. Princeton, NJ: Princeton University Press. Batty, C. 2011. ‘Smelling Lessons’. Philosophical Studies 153: 161–74. Bayne, T. 2009. ‘Perception and the Reach of Phenomenal Content’. Philosophical Quarterly 59 (236): 385–404. Bayne, T. 2014. ‘The Multisensory Nature of Perceptual Consciousness’. In. D. Bennett and C. Hill (eds.) Sensory Integration and the Unity of Consciousness. Cambridge, MA: MIT Press. Bayne, T. 2016. ‘Gist!’ Proceedings of the Aristotelian Society 16 (2): 107–26. Bengson, J. 2013. ‘Presentation and Content A Critical Study of Susanna Siegel. The Contents of Visual Experience’. Noûs 47 (4): 795–807. Bedford, F. 2001. ‘Towards a General Law of Numerical/Object Identity’. Cahiers de Psychologie Cognitive 20 (3–4): 113–75. Bertelson, P. and Radeau, M. 1981. ‘Cross-modal Bias and Perceptual Fusion with Auditory-Visual Spatial Discordance’. Perception and Psychophysics 29: 578–84. Bertelson, P. and de Gelder, B. 2004. ‘The Psychology of Multimodal Perception’. In C. Spence and J. Driver (eds.) Crossmodal Space and Crossmodal Attention. Oxford: Oxford University Press. Block, N. 2014. ‘Seeing-As in the Light of Vision Science’. Philosophy and Phenomenological Research 89 (1): 560–72.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-    

169

Botvinick, M. and Cohen, J. 1998. ‘Rubber Hands “Feel” Touch that Eyes See’. Nature 391: 756. Byrne, A. 2009. ‘Experience and Content’. Philosophical Quarterly 59 (236): 429–51. Byrne, A. and Siegel, S. 2016. ‘Rich or Thin?’ In B. Nanay (ed.), Current Controversies in Philosophy of Perception. New York: Routledge. Brewer, B. 2011. Perception and its Objects. Oxford: Oxford University Press. Briscoe, R. 2015. ‘Cognitive Penetration and the Reach of Phenomenal Content’. In A. Raftopoulos and J. Zeimbekis (eds.), The Cognitive Penetrability of Perception. Oxford: Oxford University Press. Briscoe, R. 2016. ‘Multisensory Processing and Perceptual Consciousness: Part I’. Philosophy Compass 11 (2): 121–33. Brogaard, B. 2013a. ‘Do We Perceive Natural Kind Properties?’ Philosophical Studies 162 (1): 35–42. Brogaard, B. 2013b. ‘Phenomenal Seemings and Sensible Dogmatism’. In C. Tucker, (ed.) Seemings and Justification. Oxford: Oxford University Press. Brogaard, B. 2016. ‘Perceptual Appearances of Personality’. Philosophical Topics 44 (2): 83–103. Brogaard, B. 2018. ‘In Defense of Hearing Meanings’. Synthese 195 (7): 2967–83. Brogaard, B. and Chomanski, B. 2015. ‘Cognitive Penetrability and High-Level Perception’. Pacific Philosophical Quarterly 96 (4): 469–86. Calvert, G., Spence, C. and Stein, B. (eds.) 2004. The Handbook of Multisensory Processes. Cambridge, MA: MIT Press. Campbell, J. 2002. Reference and Consciousness. Oxford: Oxford University Press. Cassam, Q. 2007. The Possibility of Knowledge. Oxford: Oxford University Press. Cavedon-Taylor. 2015. ‘Kind Properties and the Metaphysics of Perception: Towards Impure Relationalism’. Pacific Philosophical Quarterly 96 (4): 487–509. Cuneo, T. 2013. ‘Reidian Moral Perception’. Canadian Journal of Philosophy 33 (2): 229–58. de Vignemont, F. 2014. ‘Multimodal Unity and Multimodal Binding’. In. D. Bennett and C. Hill (eds.) Sensory Integration and the Unity of Consciousness. Cambridge, MA: MIT Press. de Vignemont, F., Tsakiris, M., and Haggard, P. 2005. ‘Body Mereology’. In G. Knoblich, I.M. Thornton, M. Grosjean, and M. Shiffrar (eds.) Human Body Perception from the Inside Out. Oxford: Oxford University Press. Deroy, O. 2015. ‘Multisensory Perception and Cognitive Penetration’. In J. Zeimbekis and A. Raftopoulos (eds.) The Cognitive Penetrability of Perception. Oxford: Oxford University Press. Di Bona, E. 2017. ‘Towards a Rich View of Auditory Experience’. Philosophical Studies 174 (11): 2629–43.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

170

 -

Driver, J. and Noesselt, T. 2008. ‘Multisensory Interplay Reveals Crossmodal Influences on “Sensory-Specific” Brain Regions, Neural Responses, and Judgments’. Neuron 57 (1): 11–23. Farmer, H., Tajadura-Jimenez, A., and Tsakiris, M. 2012. ‘Beyond the Colour of my Skin: How Skin Colour Affects the Sense of Body-Ownership’. Consciousness and Cognition 21 (3): 1242–56. Farnè, A. Pavani, A. Meneghello, and F. Làdavas, E. 2000. ‘Left Tactile Extinction Following Visual Stimulation of a Rubber Hand’. Brain 123 (11): 2350–60. Fish, W. 2009. Perception, Hallucination and Illusion. Oxford: Oxford University Press. Fish, W. 2013. ‘High-Level Properties and Visual Experience’. Philosophical Studies 162 (1): 43–55. Fulkerson, M. 2016. ‘Touch’. E. Zalta (ed.). The Stanford Encyclopedia of Philosophy (Spring 2016 Edition). https://plato.stanford.edu/archives/spr2016/entries/touch/ Guterstam, A. Gentile, G., and Ehrsson, H. 2013. ‘The Invisible Hand Illusion: Multisensory Integration Leads to the Embodiment of a Discrete Volume of Empty Space’. Journal of Cognitive Neuroscience 25 (7): 1078–99. Haans, A., Ijsselsteijn, W., and de Kort, Y. 2008. ‘The Effect of Similarities in Skin Texture and Hand Shape on Perceived Ownership of a Fake Limb’. Body Image 5 (4): 389–94. Helton, G. 2016. ‘Recent Issues in High-Level Perception’. Philosophy Compass 11 (12): 851–62. Howard, I. and Templeton, W. 1966. Human Spatial Orientation. London: Wiley. Jackson, C. 1953. ‘Visual Factors in Auditory Localization’. Quarterly Journal of Experimental Psychology 5: 52–65. Johnston, M. 2006. ‘Better Than Mere Knowledge? The Function of Sensory Awareness’. In T. Gendler and J. Hawthorne (eds.) Perceptual Experience. Oxford: Oxford University Press. Kalckert, A., Bico, I., and Xi Fong, J. 2019. ‘Illusions with Hands, but Not with Balloons—Comparing Ownership and Referral of Touch for a Corporal and Noncorporal Object After Visuotactile Stimulation’. Perception 48 (5): 447–55. Kalderon, M. 2011. ‘Color Illusion’. Noûs 45 (4): 751–75. Kaeppler, K. and Mueller, F. 2013. ‘Odor Classification: A Review of Factors Influencing Perception-Based Odor Arrangements’. Chemical Senses 38 (3): 189–209. Logue, H. 2013. ‘Visual Experience of Natural Kind Properties: Is There Any Fact of the Matter?’ Philosophical Studies 162: 1–12. Lyons, J. 2005. ‘Clades, Capgras and Perceptual Kinds’. Philosophical Topics 33 (1): 185–206. Lyons, J. 2009. Perception and Basic Beliefs. Oxford: Oxford University Press.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-    

171

Ma, K. and Hommel, B. 2015. ‘Body-Ownership for Actively Operated Non-Corporeal Objects’. Consciousness & Cognition 36: 75–86. McBrayer, J. 2010. ‘A Limited Defense of Moral Perception’. Philosophical Studies 149 (3): 305–20. McDowell, J. 1982. ‘Criteria, Defeasibility and Knowledge’. Proceedings of the British Academy 68: 455–79. McGurk, H. and MacDonald, J. 1976. ‘Hearing Lips and Seeing Voices’. Nature 264: 746–8. McNeill, W. 2012. ‘On Seeing that Someone is Angry’. European Journal of Philosophy 20 (4): 575–97. Macpherson, F. (ed.) 2011. The Senses: Classic and Contemporary Philosophical Perspectives. Oxford: Oxford University Press. Macpherson, F. 2012. ‘Cognitive Penetration of Colour Experience: Rethinking the Issue in Light of an Indirect Mechanism’. Philosophy and Phenomenological Research 84 (1): 24–62. Martin, M.G.F. 1992. ‘Sight and Touch’. In T. Crane (ed.) The Contents of Experience. Cambridge: Cambridge University Press. Martin, M.G.F. 2002. ‘The Transparency of Experience’. Mind and Language 4 (4): 376–425. Martin, M.G.F. 2010. ‘What’s in a Look?’ In B. Nanay (ed.) Perceiving the World. Oxford: Oxford University Press. Matthen, M. 2005. Seeing, Doing, and Knowing: A Philosophical Theory of Sense Perception. Oxford: Oxford University Press. Nanay, B. 2011a. ‘Do we Sense Modalities with our Sense Modalities?’ Ratio 24 (3): 299–310. Nanay, B. 2011b. ‘Do we See Apples as Edible?’ Pacific Philosophical Quarterly 92 (3): 305–22. Nanay, B. 2012. ‘Perceptual Phenomenology’. Philosophical Perspectives 26 (1): 235–46. Nanay, B. 2013. Between Perception and Action. Oxford: Oxford University Press. Nes, A. 2016. ‘On What we Experience when we Hear People Speak’. Phenomenology and Mind 10: 58–85. Newen, A. 2017. ‘Defending the Liberal-Content View of Perceptual Experience: Direct Social Perception of Emotions and Person Impressions’. Synthese 194 (3): 761–85. Nudds, M. 2009. ‘Sounds and Space’. In M. Nudds and C. O’Callaghan (eds.) Sounds and Perception: New Philosophical Essays. Oxford: Oxford University Press. O’Callaghan, C. 2011a. ‘Lessons from Beyond Vision (Sounds and Audition)’. Philosophical Studies 153 (1): 143–60. O’Callaghan, C. 2011b. ‘Against Hearing Meanings’. Philosophical Quarterly 61 (245): 783–807.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

172

 -

O’Callaghan, C. 2015. ‘The Multisensory Character of Perception’. Journal of Philosophy 112 (10): 551–69. O’Callaghan, C. 2016. ‘Enhancement Through Coordination’. In B. Nanay (ed.) Current Controversies in Philosophy of Perception. New York: Routledge. Price, R. 2009. ‘Aspect-Switching and Visual Phenomenal Character’. Philosophical Quarterly 59 (236): 508–18. Prinz, J. 2013. ‘Siegel’s Get Rich Quick Scheme’. Philosophical Studies 163 (3): 827–35. Radeau, M. and Bertelson, P. 1976. ‘The Effect of a Textured Visual Field on Modality Dominance in a Ventriloquism Situation’. Perception and Psychophysics 20: 227–35. Radeau, M. and Bertelson, P. 1977. ‘Adaptation to Auditory-Visual Discordance and Ventriloquism in Semirealistic Situations’. Perception and Psychophysics (2): 22. Raftopoulos, A. 2015. ‘What Unilateral Visual Neglect Teaches Us About Perceptual Phenomenology’. Erkenntnis 80 (2): 339–58. Ramachandran, V. & Hirstein, W. 1998. ‘The Perception of Phantom Limbs’. Brain: A Journal of Neurology 121 (9): 1603–1630. Reiland, I. 2014. ‘On Experiencing High-Level Properties’. American Philosophical Quarterly 51 (3): 177–87. Reiland, I. 2015. ‘On Experiencing Meanings’. Southern Journal of Philosophy 53 (4): 481–92. Rosenblum, L. and Saldana, H. 1996. ‘An Audiovisual Test of Kinematic Primitives for Visual Speech Perception’. Journal of Experimental Psychology: Human Perception and Performance 22 (2): 318–31. Schwenkler, J. 2013. ‘The Objects of Bodily Awareness’. Philosophical Studies 162 (2): 465–72. Sekuler, R., Sekuler, A., and Lau, R. 1997. ‘Sound Alters Visual Motion Perception’. Nature 385: 308. Shams, L., Kamitani, Y., and Shimojo, S. 2002. ‘Visual Illusion Induced by Sound’. Cognitive Brain Research 14: 147–52. Siegel, S. 2006. ‘Which Properties are Represented in Perception?’ In T. Gendler and J. Hawthorne (eds.), Perceptual Experience. Oxford: Oxford University Press. Siegel, S. 2007. ‘How Can We Discover the Contents of Experience?’ Southern Journal of Philosophy 45 (1): 127–42. Siegel, S. 2009. ‘The Visual Experience of Causation’. Philosophical Quarterly 59 (236): 519–40. Siegel, S. 2011. The Contents of Visual Experience. Oxford: Oxford University Press. Silins. N. 2013. ‘The Significance of High-Level Content’. Philosophical Studies 162 (1): 13–33. Smith, B. 2015. ‘The Chemical Senses’. In M. Matthen (ed.) The Oxford Handbook of Philosophy of Perception. Oxford: Oxford University Press.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-    

173

Spence, C. 2011. ‘Crossmodal Correspondences: A Tutorial Review’. Attention, Perception, & Psychophysics 73 (4): 971–95. Stokes, D., Matthen, M., and Biggs, S. (eds.) 2014. Perception and Its Modalities. Oxford: Oxford University Press. Toribio, J. 2018. ‘Visual Experience: Rich but Impenetrable’. Synthese 195 (8): 3389–406. Travis, C. 2004. ‘The Silence of the Senses’. Mind 113 (449): 57–94. Tsakiris, M. 2010. ‘My Body in the Brain: A Neurocognitive Model of BodyOwnership’. Neuropsychologia 48 (3): 703–12. Tsakiris, M., Carpenter, L., James, D., and Fotopoulou, A. 2010. ‘Hands Only Illusion: Multisensory Integration Elicits Sense of Ownership for Body Parts but not for NonCorporeal Objects’. Experimental Brain Research 204 (3): 343–52. Tsakiris, M., Costantini, M., and Haggard, P. 2008. ‘The Role of the Right TemporoParietal Junction in Maintaining a Coherent Sense of one’s Body’. Neuropsychologia 46: 3014–18. Tsakiris, M. and Haggard, P. 2005. ‘The Rubber Hand Illusion Revisited: Visuotactile Integration and Self-Attribution’. Journal of Experimental Psychology: Human Perception and Performance 31 (1): 80–91. Tye, M. 1995. Ten Problems of Consciousness. Cambridge, MA: MIT Press. Vallar, G. and Ronchi, R. 2009. ‘Somatoparaphrenia: A Body Delusion. A Review of the Neuropsychological Literature’. Experimental Brain Research 192 (3): 533–51. van Stralen, H., van Zandvoort, M., Kappelle, L., and Dijkerman, H. 2013. ‘The Rubber Hand Illusion in a Patient with Hand Disownership’. Perception 42 (9): 991–3. Vatakis, A. and Spence, C. 2007. ‘Crossmodal Binding: Evaluating the “Unity Assumption” Using Audiovisual Speech Stimuli’. Perception & Psychophysics 69: 744–56. Welch, R. and Warren, D. 1980. ‘Immediate Perceptual Response to Intersensory Discrepancy’. Psychological Bulletin 88 (3): 638–67. Werner, P. 2016. ‘Moral Perception and the Contents of Experience’. Journal of Moral Philosophy 13 (3): 294–317.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

9 What Can Predictive Processing Tell Us about the Content of Perceptual Experience? Sam Wilkinson

1. Introduction A central debate in philosophy of perception concerns the content of perceptual experience, which is, roughly, what is conveyed to the subject by, and only by, her perceptual experience. There are those who are ‘conservatives’ about this, claiming that only ‘low-level’ properties enter into this content (e.g. Tye 1995; Dretske 1995; Byrne 2016), and there are those who are more ‘liberal’, and allow ‘higherlevel’ properties (e.g. Siewert 1998; Siegel 2006; Bayne 2009; Nanay 2011). In this chapter, I present the predictive processing framework (PPF) and explore the consequences that it may have for this debate. One view about this (e.g. Lupyan and Clark 2015) is that the PPF dissolves this debate in the following way. The debate is about where a particular line (viz. between perception and cognition) should be drawn, but there is no such line according to the PPF. I argue against this and claim instead that the PPF is best understood as supporting liberalism. I end by defending this position against four objections.

2. The Content of Perceptual Experience (CPE) Debate The debate we are interested in, which I am calling the ‘Content of Perceptual Experience Debate’, or CPE debate for short, concerns the types of properties that can enter into the content of perceptual experience. What is the significance of saying that properties of a certain type enter into the content of perceptual experience? A concrete case (loosely borrowed from Nanay 2011) will be helpful here. Suppose you are looking at a green apple, and suppose that you know that it’s a Granny Smith apple, and, furthermore, that it was grown in Chile. This apple has:

Sam Wilkinson, What Can Predictive Processing Tell Us about the Content of Perceptual Experience? In: Purpose and Procedure in Philosophy of Perception. Edited by: Heather Logue and Louise Richardson, Oxford University Press (2021). © OUP. DOI: 10.1093/oso/9780198853534.003.0009

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     ? (i) (ii) (iii) (iv)

175

a certain shape and colour. the property of being an apple. the property of being a Granny Smith apple. the property of having been grown in Chile.

Neither side of the debate will want to say that all of those properties of the apple enter into the content of your perceptual experience. In particular, even the most liberal of liberals will likely accept that (iv) just isn’t the right kind of property for your perceptual experience to convey. You may come to know that the apple has that property (viz. was grown in Chile), but you can’t have known that solely on the basis of your perceptual experience. It can enter into the content of, say, a judgement or supposition, but not of a perceptual experience. And that is a large part of what this debate is supposed to be about: drawing a principled line between the sorts of properties that can enter into the contents of perceptual experiences on the one hand and judgements (or other less committal propositional attitudes) on the other. What this amounts to, more practically speaking, is where you would attribute the ‘blame’ for any inaccuracy. Suppose, now, that you believe that the apple has all of (i–iv), and this time you are wrong about (iv): it wasn’t grown in Chile, it was grown in France. It would be odd to say that, in such a situation, your perceptual experience alone had misled you. It was an error of judgement, either based on someone telling you the wrong thing, or you drawing a false inference (perhaps based on the time of year, you wrongly assumed a southern hemispheric provenance). Now, the conservative wants to say something similar about (iii), and also about (ii). If you are wrong about it being a Granny Smith, or about it being an apple, but you are not wrong about it being a certain colour or shape (suppose it’s a very realistic waxwork), this is not (not ever) a case of your perceptual experience alone leading you astray. Rather, it is the result of an incorrect inference based on an accurate perceptual experience. One reason why the conservative may think this is because your perceptual apparatus is working just fine and you’re not subject to some kind of optical illusion. Moreover, the conservative may point to a more basic motivation for her position: the sorts of properties to which our eyes are sensitive are limited (and indeed we could even define ‘low-level’ properties as those properties). You can’t have the property of being an apple enter into the content of a perceptual experience, because our eyes just aren’t sensitive to appleness per se. To put this another way, our visual systems don’t respond differentially to apples and to very realistic apple waxworks. The liberal can counter this by pointing to an important distinction between being misled by your perceptual experience and errors in perceptual processing.¹ ¹ Arguably most optical illusions (e.g. the Muller-Lyer illusion) are cases of your visual system working fine, but your visual experience misleading you.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

176

 

Just because there aren’t errors in perceptual processing, it doesn’t mean that your full-blooded perceptual experience isn’t leading you astray. The liberal could appeal to the idea that even when you know it is a waxwork, if it is realistic enough, you have to override the experience of it as a real apple, an experience that wouldn’t be had by a subject who had never encountered an apple before and had no conception of apples. The classic arguments for liberalism exploit this idea that experiences had by those who have relevant knowledge or expertise differ phenomenologically from those had by those who lack it. In these ‘contrast cases’ you are invited to compare two cases where one would intuitively say that there is a phenomenological difference, in spite of the fact that the low-level properties represented (e.g. colour and shape) remain constant. It is then suggested that the phenomenological difference is best explained by the fact that in the two cases the ‘highlevel’ properties represented are different because, in one of the cases, the highlevel properties cannot be represented due to lack of knowledge, exposure or expertise. Contrast cases abound, but perhaps the classic example is from Siegel (2006): Suppose you have never seen a pine tree before, and are hired to cut down all the pine trees in a grove containing trees of many different sorts. Someone points out to you which trees are pine trees. Some weeks pass, and your disposition to distinguish the pine trees from the others improves. Eventually, you can spot the pine trees immediately: they become visually salient to you. Like the recognitional disposition you gain, the salience of the trees emerges gradually. Gaining this recognitional disposition is reflected in a phenomenological difference between the visual experiences had before and after the recognitional disposition was fully developed. (Siegel 2006, 491)

Siegel goes on to argue that this phenomenological difference is attributable to the fact that the kind property ‘pine tree’ enters, after recognitional expertise is gained, into the contents of the perceptual experience. Responses to contrast cases can adopt two broad tactics in support of conservatism (as Siegel 2010 acknowledges): 1) Although the phenomenology has changed between the two cases, it is not perceptual phenomenology that has changed (it is, for example, emotional, affective or even cognitive phenomenology). 2) The phenomenology has changed, and it is indeed perceptual phenomenology, but that is because the low-level properties represented are different (for example, because it affects low-level processing, or how attention is directed). This amounts to denying the founding assumption

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     ?

177

of the contrast: the low-level properties are not being kept constant after all.² In short, conservatives usually acknowledge the phenomenological differences (although see Tye 1995) between the contrast case cases, but tend to give them an alternative explanation. At this point it is worth mentioning that the desire to seek an alternative explanation that is consistent with conservatism must come from an independent motivation to adopt conservatism. One such motivation for conservatism, which I have already mentioned and which arguably makes it the default position, is that our eyes, ears, etc. seem to only respond to certain low-level features: our eyes respond to light, our ears to sound waves, etc. The framework that I am about to present goes some way towards undermining this motivation.

3. The Predictive Processing Framework The predictive processing framework (PPF) makes us rethink both what the brain does in general, and the nature of perceptual experience in particular. Contrary to standard accounts, the brain does not take inputs from the outside world, process them, and pass them on (such a framework is exemplified by the use of ‘box-andarrow’ diagrams in cognitive psychology). Instead, the brain is to be viewed as a prediction machine. Whenever information impacts on your sensory surfaces, it is already, even at the earliest stages, greeted by a prediction on the part of your nervous system. Your perceptual encounters with the world never occur in a vacuum, free of temporal context. Furthermore, what determines your conscious percept at a given time is not the inputs that causally and temporally precede the experience, but rather the hypotheses that your brain has adopted in order to best predict future inputs. In other words, your brain is always staying one step ahead of the sensory manifold, and your experience is determined by your brain’s best hypotheses. This framework has received empirical and theoretical support from a wide variety of sources and disciplines, and has far-reaching implications for how we are to think about all aspects of cognition (see Clark 2013, 2016; Hohwy 2013), including atypical cognition (for a treatment of psychosis, see Fletcher and Frith 2009; for a treatment of auditory verbal hallucinations, see Wilkinson 2014; for a treatment of autism, see Pellicano and Burr 2012). Perhaps the simplest way to present the thinking behind the PPF is to reflect on two things: ambiguity and efficiency. The brain’s main task is to figure out what is ‘out there’ based on what is impacting on the peripheries of the nervous system, ² O’Callaghan (2011) uses this tactic for the auditory experience of speech, in particular to argue against the view that we can hear meanings.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

178

 

and it has to do so as efficiently (namely, as quickly, accurately, and with as little energy expenditure) as possible. Since the incoming information is ambiguous (namely, compatible with different hypotheses) the brain has to use more than just the ‘fit’ to select one hypothesis over another. In particular, it has to use its ‘priors’, namely, its expectations (usually based on past statistical regularities). Thus with every stimulus, your brain is already greeting it with an expectation that helps to disambiguate it. This tactic is not only a good way of disambiguating a stimulus, and of doing so quickly: it is also highly energy efficient. To see why, consider an analogy with data compression. Data compression is about minimizing ‘information load’, namely, that which is explicitly represented in a signal. A standard method for doing this between a given sender/receiver pair, is for the sender to only pass in the signal what is newsworthy, namely, what the receiver hasn’t already predicted and hence can construct for itself. The same is said to occur between different regions of the nervous system: all that gets passed up is what the relevant part (the ‘receiver’) hasn’t already predicted, which in the PPF is called ‘prediction error’. In turn, your brain is constantly trying to improve its predictions. Thus you get the dictum: all the brain ever does is minimize prediction error. The extent to which this very strong and universal claim is true is something that I will put to one side. However, there is evidence converging on the idea that the brain is in the business of doing something like this (see Clark 2013 for a presentation of some of this evidence from a wide variety of different disciplines), and what is important for my purposes is that this has far-reaching consequences for how we are to think about perceptual experience. What is of vital importance here is that things don’t impact on our nervous systems out of context, against a static background: our nervous systems are in a constant fluctuating state of anticipation. Your experience is not constructed out of sensory inputs, but is rather the result of a dynamic, predictive hypothesisbuilding process. One important aspect of the PPF is that the predictive hypotheses that the brain selects are hierarchically organized, with the hypotheses of one level providing the data for the next. ‘Higher’ parts of the hierarchy are, roughly, those parts that are further away from the sensory surfaces. These tend to be at lower temporal frequencies (longer timescales), and at higher levels of abstraction. ‘Lower’ parts of the hierarchy are closer to the sensory apparatus. These tend to be at higher temporal frequencies (shorter timescales), and at low levels of abstraction. They correspond, for example, to early stages of visual processing: your brain’s early statistically driven attempts to make sense of noisy inputs. Of course, in order to express these neurally encoded predictions we need to use descriptions in natural language (in this case English), but there is nothing linguistic about the priors/ hypotheses (‘Light tends to come from above’/‘This is a face’) themselves (indeed, it is a matter of some dispute whether they themselves are representational at all

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     ?

179

(see Gladziejewski 2016 for arguments for, and Orlandi 2016 for arguments against)).

4. Counterfactual Predictions and ‘Seeing As’ According to the PPF, perceptual experience consists in your brain selecting the hypotheses that generate the best predictions of imminent sensory data. However, the perceptual conditions at any given time hugely underdetermine the hypotheses that could be appropriately selected. The nature and complexity of the hypotheses is left open, as is, in a related manner, the nature and reach of the predictions that such hypotheses may generate. If you put an apple in front of a fully sighted human being in good lighting conditions, then the hypotheses that their brain selects could (over and above the low-level hypotheses that settle questions about colour and shape) vary enormously. Perhaps the person has never seen an apple before, so they experience it as ‘That thing, which has that shape, appears to have solidity etc . . . .’ Perhaps the person has never seen an apple, but recognizes it as a fruit (‘That strange fruit I’ve never seen before’). There is no limit to the type and degree of variation we could dream up. For example, other variation may not come from lack of exposure or expertise (which would instil Siegel’s ‘recognitional disposition’) but from background information about that very specific perceptual encounter. Perhaps the person has good reason to think (e.g. were told by a reliable source) that, in spite of appearances, it’s a waxwork apple. Yet other, relatively short-lived variations may arise from the person’s practical concerns. Perhaps the person is an impressionist painter and momentarily brackets the apple-ness simply to focus on the arrangement of colour in her visual field. In mainstream analytic philosophy of perception these variations would be chalked up to differences in ‘seeing as’, which amounts to the application of different concepts to, or different conceptualizations of, the same sensory data. That is not strictly incompatible with the PPF. Perhaps, for example, one can think of ‘concepts’ as the very same thing as the ‘hypotheses’ that the PPF posits. However, two caveats should be flagged. First, according to the PPF, such ‘concepts’ aren’t to be seen as applied to perceptual experience but rather as generating (or determining) the experience.³ Second, the PPF will inevitably explain the differences in concepts or conceptualizations in terms of different predictions. This could play out in the following way. Everything has to boil down to predictions, and the reach of these predictions can—and does—vary enormously. To borrow an insight from Seth (2014), our ³ This may be in keeping with the views of ‘conceptualists’ about perceptual experience (e.g. Brewer 1999; McDowell 1994).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

180

 

hypotheses are counterfactually rich, and can vary in their counterfactual richness. What he means by this is that my brain doesn’t simply have predictions about what is going to happen, namely, how the sensory manifold will change given what is likely to happen in the near future. There is also (at least in the human brain) a wealth of prediction about how the sensory manifold would change given relevant counterfactual circumstances—circumstances that needn’t happen and may even be unlikely to happen. Another way of putting this is that my nervous system isn’t simply content with superficially predicting what will happen next, but seeks to comprehend more fully the statistical structure of the world, which involves, in part, the tacit positing of the underlying natures and dispositions of things, even if those natures and dispositions are never explicitly revealed, since their being revealed might involve an excessively complicated and/ or unlikely set of counterfactual circumstances. This deeper and more costly enterprise clearly amounts to a wise long-term investment, since it prepares you for a causally complex and hard-to-predict world.⁴ My experience of an apple as an apple is partly down to my brain’s predictions about what would happen if I were to do any number of things (walk around it, pick it up, eat it etc.) or if any number of things were to be done to it (e.g. if it were smashed with a sledgehammer), even if those possibilities aren’t likely to be actualized (e.g. I have every intention of staying still, of continuing to view the apple from this angle, and there is no threat of the apple being smashed with a sledgehammer). My nervous system’s ‘appreciation’ of these unactualized possibilities structures my experience and enables it to delve below the surfaces of objects to the postulated nature of those objects. Notice that the apple could be in a glass cabinet, such that I never get to confirm or falsify predictions that determine my experience of it as a real apple. Suppose that, unbeknownst to me, the object in the cabinet is a highly realistic waxwork apple. What it is for me to see it, wrongly in this case, as a real apple is the plethora of counterfactual predictions pertaining to real apples (e.g. that I can eat it; that it would smash in a certain way etc.). Some readers may have noticed a similarity between these counterfactually rich predictions and so-called sensorimotor contingencies (O’Regan and Noe 2001). Indeed, this is a similarity that Seth (2014) notes, and uses to great effect (as Noe himself does (2005) within an enactivist framework) in making sense of perceptual presence (and variations therein).⁵ However, it is crucial to see that not all counterfactual predictions are sensorimotor contingencies. They are not necessarily latent expectations about what would happen if I were to do something. Many of the relevant expectations are to do with what will happen to the object in

⁴ Indeed, nowhere is causal complexity greater, and predictability more important, than in the social realm (as it involves the perception of other agents). ⁵ ‘Perceptual presence’ refers to the experience of something as being present to you, as part of your world.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     ?

181

certain circumstances that do not involve my acting at all. This is not to contradict Seth (2014) since I think it is likely that the predictions that are relevant specifically to perceptual presence are indeed action-centred counterfactual predictions (viz. sensorimotor contingencies) (see Wilkinson 2020). There is, however, rather more to perceptual experience than perceptual presence.

5. Does the PPF Dissolve the CPE Debate? What does the PPF have to contribute to the CPE debate? One view, expressed by Lupyan and Clark (2015), but also easily derivable from what other PPF theorists say (e.g. Fletcher and Frith 2009; Hohwy 2013) is that the PPF dissolves the CPE debate altogether.⁶ The debate is based on a misguided picture of the human mind. The debate is about where you draw the line between the perceptual and the cognitive, and what the PPF tells us is that there is no such line to be drawn. What you have all the way ‘up’, and indeed all the way ‘down’, is a delicate dance between perception and cognition. Such theorists will want to deny, for example, that there is a principled distinction to be drawn between perception and belief on the grounds that there is no fundamental difference between my visual system’s ‘beliefs’ about what is out there, and my belief that my train leaves at 10:30. It is simply a difference to do with where in the hierarchy these are located: one is ‘high up’, abstract, and operates over longer timescales; the other is ‘low down’, concrete, and operates over shorter timescales. But these differences simply fall out of the fundamental nature of the hierarchy: they are, at bottom, the same kind of phenomena. If we apply this line of thinking to Siegel’s contrast cases, and to liberalism more generally, the question of whether, e.g., natural kinds enter in the content of perceptual experience is to miss the point. My nervous system has a certain hierarchical predictive structure that reflects my understanding of the world, and it both shapes and is shaped by my engagement with the world. Is the phenomenology of my experience of a Granny Smith apple, as an apple expert, different from the experience of a naïve observer because of something perceptual, or because of a judgement made on the basis of a percept? For the PPF theorist, it is based on something judgement-like in my nervous system. But this is no victory for the conservative, because, then again, the same applies to my appreciation of the object’s colour and shape. It is judgement all the way down. I want to argue against this view. In spite of similarities (e.g. they have mind-toworld direction of fit) there is a major functional difference between these two kinds of ‘judgements/beliefs’, namely, those that I make/have and those that my ⁶ Fletcher and Frith (2009) in particular argue that hallucinations and delusions in psychosis are fundamentally the same phenomenon, simply operating at different levels in the hierarchy.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

182

 

nervous system makes/has. That difference is to do with the way in which they are formed or updated. I just need to hear an announcement that my train has been delayed by ten minutes in order for me to update my belief about my train’s departure. To get my visual system to update its ‘belief ’, e.g., that those two lines in the Muller-Lyer illusion are in fact the same length, is rather more difficult. It would, if it turned out to be possible at all, require a lengthy process of gradual perceptual relearning.⁷ And isn’t that where we should naturally draw the line? Belief is about what I believe as a rational agent, and, as a result, can be responsive to one-off pieces of evidence, whereas perception is about what a part of my nervous system ‘believes’ as a result of statistically driven priors, and it may continue to ‘believe’ this whatever I myself come to discover (for example, by measuring the lines of the Muller-Lyer illusion).

6. The PPF as Supporting Liberalism I now want to take things one step further. Not only does the PPF not dissolve the CPE debate, it can also be interpreted as offering support to liberalism. The most important lesson from the PPF is that our nervous system constructs hypotheses that delve beneath the surfaces of what is explicitly, sensorially presented, in order to best predict what will or might (or could) be sensorially presented in the future. These hypotheses will be tweaked and updated if they fail to do this. However, the crucial move for my purposes lies in realizing that it is the hypotheses themselves that ultimately determine the content of experience, and not the sensory inputs that primarily serve to keep them accurate and anchored to the world.⁸ In light of this theoretical perspective, the kinds of properties that can enter into the contents of experience are not restricted to the sorts of properties that my sensory apparatus can detect. To think that such a restriction applies would be to conflate the sort of data that can falsify or confirm a hypothesis, with the content of the hypothesis itself. Typically, a hypothesis is rich and theory-laden, whereas the data against which the predictions it generates are compared are in and of themselves sparse, ambiguous, and open to innumerable interpretations. However, as we are about to ⁷ Indeed it is believed that those who have developed in non-rectilinear environments are less susceptible (or not at all) to the Muller-Lyer illusion (Berry 1968). ⁸ Of course, the importance of this ‘anchoring’ shouldn’t be downplayed. (After all, when it goes wrong in different ways, you get different kinds of disconnection from reality.) In a case of perception, the experience is causally dependent on signals from the outside world (you wouldn’t get that experience without them) but it is not (psychologically, proximally) determined by them (or, perhaps better, constituted by them), in that in principle (though not in practice) you could get that experience without them. In any case, we are not interested here in perceptions vs. hallucinations; we are interested in the richness of perceptual experiences, and that richness comes from the hypotheses selected. We can imagine super-advanced human beings who, with the same sensory apparatus, when engaged in the very same perceptual encounter with the world, have an experience of unfathomable richness compared to our own (or, conversely, others whose experience is impoverished).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     ?

183

see, this is not to say that there are no restrictions on the kinds of things that can enter into the contents of perceptual hypotheses (as opposed to more ‘cognitive’ hypotheses). Indeed, the reason why we experience a rich, structured and meaningful world is precisely because the hypothesizing of our nervous systems far outstrips, delves far beyond, the sensory data at their disposal. Indeed, one might even go as far as to say that, not only does the PPF remove the restriction of sensory detectability, it also makes nonsense of the very idea of raw sensation at the experiential level. To the extent that any sensation is consciously experienced, it has already been accounted for to some extent by a top-down hypothesis. The closest approximation in experience to the notion of raw sensation is sensory data that has only been accounted for by a relatively low-level hypothesis. In light of all of this, one might ask: What kinds of properties aren’t allowed in experience? As we’ve seen, the relevant restrictions are no longer about the kinds of things that your sensory surfaces are sensitive to. But this is not to say that there are no restrictions at all. These restrictions concern the sorts of hypotheses that your nervous system can generate at perceptual timescales. Two crucial points are embedded in this: (i) the hypotheses that your nervous system can generate, and (ii) at perceptual timescales. To clarify (i), it helps to rehearse a line of criticism. Are we not, the critic might ask, saying that these are the contents of judgements (albeit perceptual ones)? The reason why this is not the case is because it is crucial to distinguish judgements that I make, from the ‘judgements’ that my nervous system (or a part of my nervous system) makes.⁹ The latter ‘judgements’ are what determine the content of my experience, and, crucially, can be at odds with the judgements that I make. Nowhere is this more clearly seen than in cases where we knowingly experience illusions. When, for example, I experience the Muller-Lyer illusion, my nervous system ‘judges’ that the lines are different lengths, whereas I judge them to be the same length (for example, because I have measured them). What we typically think of as a perceptual judgement is when I endorse the ‘judgement’ that my nervous system has made, namely, the lower-level ‘inference’ that generates the perceptual content. Regarding (ii), hypotheses generated at ‘perceptual timescales’ are hypotheses that are relevant to my online engagement with a given stimulus. Thus there are all sorts of judgements, ways that we take the world to be, that are too abstract, or general, or temporally distant or distributed to be perceptual hypotheses. For example, my belief that Paris is the capital of France, or that my train leaves at noon tomorrow, or, indeed, that that apple was grown in Chile, these hypotheses (and they are in an important sense hypotheses) are not directly relevant to my ⁹ Note the deliberate use of so-called scare quotes where I’m using personal-level vocabulary to capture something subpersonal but functionally similar.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

184

 

online engagement with the object in question. Historical and political properties (like having been grown somewhere, or being the capital of somewhere) can certainly be of relevance, in the right contexts, to my engagement with something, but they are not the sorts of things that make a difference to my there-and-then, stimulus-dependent engagement with the object perceived. In contrast, the hypothesis that the thing in question is a real apple is very relevant to my thereand-then engagement with it. What about the property of being a Granny Smith? Such a hypothesis could be at a perceptual timescale (in fact, there’s no reason to think that it wouldn’t be at the same timescale as the real apple hypothesis). However, what is less clear is whether the Granny Smith content can be part of my nervous system’s hypothesizing, rather than my own. It seems that an ability to recognize Granny Smiths is not sufficient for this. This could still be the content of an inferential judgement (as in the early stages of Siegel’s example). But I see no reason why a high degree of expertise couldn’t enable the Granny Smith hypothesis to be something that my nervous system automatically selects in appropriate circumstances. The property of having been grown in Chile, however, is inappropriate in a number of ways: it is too abstract, too distant in both space and time, and so on.

7. Four Illustrative Objections I’d like to present four illustrative objections to what I have argued. Each of them can be addressed in ways that usefully clarify the position I’m presenting. Since my claim is reliant on the viability of the PPF, the objections don’t question that. They rather cast doubt on my suggestion that the PPF supports liberalism. The first two will be familiar, since they are effectively the same as the two standard responses to contrast cases. The last two are new.

7.1 Phenomenological Effects Are Attributable to Top-down Effects on the Experience of Low-Level Properties When my brain adopts one multilevel hypothesis instead of another competing one that would be equally acceptable in light of current perceptual conditions (e.g. the real apple vs. the waxwork hypotheses) what we get is indeed a change in phenomenology. However, this change in phenomenology is not because highlevel properties enter into the contents of perceptual experiences but rather because, as the PPF so strongly emphasizes, there are pervasive top-down influences on how the low-level properties are experienced. In other words, top-down effects change the properly perceptual content, but simply by changing which lowlevel properties are experienced. That, in itself, the objection goes, is perfectly

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     ?

185

consistent with conservatism, given that it is only these now-changed low-level properties that enter into perceptual experience. That is true. However, this objection underestimates both the nature and the reach of top-down influences within the PPF. It is the top-down predictions (the hypotheses) that determine your experience, not the bottom-up sensory prediction error, even at the front-line sensory surfaces. In other words, even the ‘lowest level’ of perceptual experience is only possible thanks to top-down processing. As we’ve seen, within the PPF there is no such thing as ‘raw sensory experience’. To the extent that there is any sensory data at all, it is already, even at the earliest stages, being greeted by expectations on the part of your nervous system, and the ‘rawest’ sensory experience, if it is to be considered an experience at all, is simply the result of a low-level hypothesis. So why (and indeed where) should you draw the line between the low-level and high-level aspects of experiences if both are ultimately generated top-down? Short of answering this, you should either say that neither the lower-level nor the higher-level aspects can feature in perceptual content, in which case nothing can, or you allow that both can. In short, if the PPF is correct, either you do away with the very notion of perceptual content or you should be liberal about it. As I have already argued, in the context of discussing illusions that remain in the absence of judgements to the contrary, I think there are good reasons to retain a notion of contents that are perceptual as opposed to more robustly cognitive (i.e. associated with judgement).

7.2 Phenomenological Effects Are Attributable to Non-perceptual Phenomenology The second objection is that, when our experience is of these higher-level things in virtue of its predictive structure that might be part of the overall experience, but not of the perceptual experience. Two obvious candidates for such non-perceptual phenomenological contributions to the experience are, first, affective phenomenology and cognitive phenomenology. I examine these in turn. So, can the relevant phenomenological contribution be seen as a matter of accompanying affect rather than perceptual content per se? An initial response to this is that, indeed, perceptual experiences, and especially anticipatory aspects of those experiences, are often affectively charged, but that doesn’t make them any less perceptual. Not only are even the most mundane forms of perceptual experience subtly affective, affect is also constantly modulating perceptual processing (e.g. Phelps, Ling and Carrasco 2006; Villeumier and Driver 2007). Affect doesn’t merely accompany perceptual processing. This phrasing would suggest that you could remove it and leave the perceptual processing intact. In truth, affect and perception are profoundly intertwined. This is evidenced, for example, by rare cases of brain damage where affective disruptions correspond to profound

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

186

 

changes in what is experienced. These changes can lead the subject to claim that loved ones have been replaced by identical-looking impostors, as in the Capgras delusion, or even, as in the case of the Cotard delusion, that one is dead and that the world isn’t real. What about cognitive phenomenology? Here it depends on what you mean by this. Sometimes ‘cognitive phenomenology’ is taken to refer to the phenomenology of ‘cognitive’ (as opposed to ‘conative’) propositional attitude states and events. Thus we would be talking here about a phenomenology of belief or judgement. This doesn’t seem like a good candidate. Not only is the very existence of cognitive phenomenology in this sense highly contentious (see Bayne and Montague 2011), but the relevant aspects of perceptual content are supposed to be independent of what I judge. I can continue to experience something as edible or fragile or a real apple, to the extent that my nervous system continues to select that hypothesis, even though I myself judge this to be inaccurate. Another option is to think of ‘cognitive phenomenology’ as meaning the ‘phenomenology of cognition’ where cognition is broadly construed so as to capture the subject matter of the cognitive sciences. This would not be about beliefs and judgements, but may include, for example, conceptualization and categorization. In this case, the answer is that, yes, the phenomenological difference between seeing something as a real and a fake apple is a matter of cognition, but cognition in this sense would in any case be an integral part of perception, and the generation of perceptual content.

7.3 The PPF Is Entirely Consistent with Conservatism My argument rests on the idea that the PPF supports liberalism because, according to the PPF, it is not the sensory input that determines perceptual content, but the hypotheses used to adequately predict the dynamics of future input. However, the objection goes, nothing prevents someone from adhering to the PPF whilst claiming that it is only perceptual hypotheses that determine perceptual content and these concern only low-level properties. In other words, only a liberal version of the PPF (rather obviously) supports liberalism, whereas there is an equally coherent version that supports conservatism. In short, the PPF does not in and of itself support liberalism. This objection is extremely helpful in allowing me to clarify the precise nature of my central claim, on which I’d like to say two things. First, I don’t want to say that the PPF is inconsistent with conservatism. One could imagine a hierarchically arranged predictive processing architecture of a conservative sort. Indeed, perhaps some actual organisms have nervous systems that approximate this. My claim is simply that human beings aren’t like this. My reasons for making this claim are partly due to the kinds of organisms that we are: we are, for example, deeply social and linguistic animals that develop

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     ?

187

rapid-timescale and inflexible expertise (as reflected within the PPF by the sorts of hypotheses that our nervous systems generate) about domains of reality that are underdetermined by, and far outstrip, the merely sensory. In a related vein, the claim that we have a perceptually conservative nervous system strikes me as phenomenologically implausible. More mundane liberal intuitions are onto something: we don’t just see colours and shapes, but apples, chairs, faces, etc. The second thing I’d like to say is that, while not strictly incompatible with conservatism, in focusing on hypothesis generation rather than sensory input for the determination of perceptual content, the PPF undermines some of the motivation for conservatism (namely that our sensory surfaces are restricted in the kinds of things that they can respond to). And with this motivation undermined, why would you fight the various arguments for liberalism? What is more, conservatism is further undermined when you reflect on whether it is plausible that the human nervous system would restrict itself to generating predictive hypotheses about the here and now, at rapid, action-guiding timescales, that are only about (what gets called in the CPE debate) low-level properties. Clearly delving beyond these superficial sensory features is an integral part of how we experience the world and is a vital part of our predictive success, success which concerns what is right here in front of me, to be responded to now. I see no reason why this online, here-and-now informational state should not be called a perceptual state.

7.4 What Modality Are These So-Called Perceptual Experiences Happening In? Perceptual experiences happen in a given modality. I see colours, hear sounds, etc. But when I experience edibility, or fragility is this something that I specifically see? And how is it that my eyes are supposed to respond differentially to these properties? Something that’s edible or fragile can impact upon my visual system in exactly the same way as something that is not those things. A potential response to this, again, involves a clarification of the relationship between sensation and perception. Sensation is necessary for perception, but certainly not sufficient.¹⁰ I become perceptually informed about the world in virtue of my sensory apparatus. But much of this is cross-modal and multimodal.¹¹ The natural way to talk about these cases is not to say that I literally see something’s edibility, but rather that I experience that something is edible in virtue of seeing it. But it could be in virtue of touching it, or smelling it, or through a

¹⁰ Although this necessity is probably only physical rather than metaphysical. ¹¹ And in any case, there is very little consensus about how to individuate modalities in the first place (see Macpherson 2011).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

188

 

combination of senses. Perhaps, in light of the PPF, one should not really talk about perception of high-level properties in a different way to how we talk about perception of low-level properties. That is to say, perhaps we don’t literally see shapes, or even colours, but rather perceive that things have certain shapes or colours in virtue of sight. We don’t hear sounds, but rather perceive that certain sounds were produced (or indeed that there was a distant car crash) in virtue of audition. Remember, after all, that it is the perceptual hypotheses that your nervous system selects in order to best predict sensory input that determines the content of your perceptual experience, and not the sensory input itself.

8. Conclusion I explored the consequences that predictive processing has for how we are to think about the contents of perceptual experience. Those who are sympathetic to the PPF have tended to think that the debate between liberals and conservatives goes away. In contrast, I have argued that the PPF should be understood as supporting liberalism. This support comes in two stages. First, motivations for conservatism that come from constraints about the kinds of properties that our sensory organs are sensitive to are undermined since the emphasis in the PPF is on the hypotheses that best predict sensory input, rather than the input themselves (indeed, the idea of a ‘raw sensory input’ makes no sense within the PPF, at least not as a component of experience). Second, the challenge then becomes how one should draw the line between the hypotheses that count as perceptual and those that are more cognitive (belief-like). At this point I have appealed to two things. First, the fact that my beliefs and the ‘beliefs’ of my nervous system are rather different, and the way they function (e.g. in formation and update) are very different. It is only the latter that should be thought of as perceptual.¹² The second thing is that perceptual hypotheses are about what is around me in the here and now, and serve to guide potential actions at relatively short timescales.

Funding European Research Council (XSPECT—DLV-692739)

¹² Or at least a subset should be. There are hypotheses that my nervous system adopts to glean information not directly about the word, but to flag bodily change, some of which might be indirectly informative about the world. This is where the PPF accounts for emotion (Seth 2013). This may not count as perception, but there are clearly aspects of emotional experience that are very perception-like.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     ?

189

References Bayne, T. 2009. ‘Perception and the reach of phenomenal content’. Philosophical Quarterly 59 (236): 385–404. Bayne, T. and Montague, M. (eds.). 2011. Cognitive Phenomenology. Oxford and New York: Oxford University Press. Berry, J. W. 1968. ‘Ecology, perceptual development and the Müller-Lyer illusion’. British Journal of Psychology 59 (3): 205–10. Brewer, B. 1999. Perception and Reason. Oxford: Oxford University Press. Byrne, A. 2016. ‘The epistemic significance of experience’. Philosophical Studies 173 (4): 947–67. Clark, A. 2013. ‘Whatever next? Predictive brains, situated agents, and the future of cognitive science’. Behavioral and Brain Sciences 36 (3): 181–204. Clark, A. 2016. Surfing Uncertainty. Oxford: Oxford University Press. Dretske, F. 1995. Naturalizing the Mind, Cambridge, MA: MIT Press. Fletcher, P. C. and Frith, C. 2009. ‘Perceiving is believing: A Bayesian approach to explaining the positive symptoms of schizophrenia’. Nature Reviews Neuroscience 10(1): 48–58. Gładziejewski, P. 2016. ‘Predictive coding and representationalism’. Synthese 193: 559–82. Hohwy, J. 2013. The Predictive Mind. Oxford: Oxford University Press. Lupyan, G. and Clark, A. 2015. ‘Words and the world: predictive coding and the language-perception-cognition interface’. Current Directions in Psychological Science 24: 279–84. Macpherson, F. 2011. ‘Individuating the Senses’. In Fiona Macpherson (ed.) The Senses: Classic and Contemporary Philosophical Perspectives. Oxford: Oxford University Press. McDowell, J. 1994. ‘The Content of Perceptual Experience’. Philosophical Quarterly 44: 190–205. Nanay, B. 2011. ‘Do we see apples as edible?’ Pacific Philosophical Quarterly 92 (3): 305–22. Noë, A. 2005. ‘Real Presence’. Philosophical Topics 33 (1): 235–64. O’Callaghan, C. 2011. ‘Against hearing meanings’. Philosophical Quarterly 61 (245): 783–807. O’Regan, J. K. and Noë, A. 2001. ‘A sensorimotor account of vision and visual consciousness’. Behavioral and Brain Sciences 24 (5): 883–917. Orlandi, N. 2016. ‘Bayesian perception is ecological perception’. Philosophical Topics 44 (2): 327–51. Pellicano, E. and Burr, D. 2012. ‘When the world becomes “too real”: a Bayesian explanation of autistic perception’. Trends in Cognitive Sciences 16 (10): 504–10.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

190

 

Phelps, E.A., Ling, S., and Carrasco, M. 2006. ‘Emotion facilitates perception and potentiates the perceptual benefits of attention’. Psychological Science 17: 292–9. Siegel, S. 2006. ‘Which Properties are Represented in Perception?’ In T. Gendler Szabo and J. Hawthorne (eds.) Perceptual Experience. Oxford: Oxford University Press, 481–503. Siewert, C. 1998. The Significance of Consciousness. Princeton: Princeton University Press. Seth, A. K. 2013. ‘Interoceptive inference, emotion, and the embodied self ’. Trends in Cognitive Sciences 17 (11): 565–73. Seth, A. 2014. ‘A predictive processing theory of sensorimotor contingencies: Explaining the puzzle of perceptual presence and its absence in synaesthesia’. Cognitive Neuroscience 5 (2): 97–118. Tye, M. 1995. Ten Problems of Consciousness. Cambridge, MA: MIT Press. Vuilleumier, P. and Driver, J. 2007. ‘Modulation of visual processing by attention and emotion: windows on causal interactions between human brain regions’. Philosophical Transactions of the Royal Society B: Biological Sciences 362: 837–55. Wilkinson, S. 2014. ‘Accounting for the varieties and phenomenology of auditory verbal hallucinations within a predictive processing framework’. Consciousness and Cognition 30: 142–55. Wilkinson, S. 2020. ‘Distinguishing Volumetric Content from Perceptual Presence within a Predictive Processing Framework’. Phenomenology and the Cognitive Sciences 19 (4):791–800.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

10 Wading in the Shallows Paul Noordhof

1. Introduction Many philosophers characterize their position on perceptual experience as naturalistic. The issue I want to address under this heading is more specific. In brief, it concerns the extent to which developments in the sciences related to the study of perception, and the senses more generally, should have an influence upon philosophical theories of perceptual experience (hereafter the sense-related sciences). The issue receives a natural formulation within the context of the truth of physicalism. According to Frank Jackson and Kim Sterelny, Moderate Naturalists are physicalists (Jackson 1994, 155; Sterelny 1990, xi). To fix ideas, we may characterize this in the way that they do. Any world which is a minimal physical duplicate of our world is a duplicate simpliciter of our world. (Jackson 1998, 12)

This kind of moderate naturalism leaves open a variety of positions concerning the nature of perceptual experience and, in particular, various attitudes that one might have to the bearing of the sense-related sciences on it. In the second section, I’m going to identify two ways in which attempts have been made to isolate the influence of sense-related sciences to the philosophical theories of perceptual experience falling short of outright denial of the relevance of sense-related sciences. The first says that philosophical theories of perceptual experience don’t concern themselves with the level of detail required to distinguish between different sense-related scientific theories. The second says that important discriminations concerning perceptual experience are conventional and don’t correspond to real discriminations in nature to which sense-related scientific theories are addressed. In brief, the discriminations don’t correspond to natural kinds, that is, kinds the classification of which reflects their explanatory value (LaPorte 2004, 19). I explain why these claims fail in crucial cases. The third section focuses on the potential clash between prediction error theories of perception and cognition and naïve realist approaches to the nature of perception. The fourth focuses upon whether results from the sense-related sciences show that sensory kinds are not natural kinds. In which case, there is said

Paul Noordhof, Wading in the Shallows In: Purpose and Procedure in Philosophy of Perception. Edited by: Heather Logue and Louise Richardson, Oxford University Press (2021). © OUP. DOI: 10.1093/oso/9780198853534.003.0010

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

192

 

to be a choice between abandoning the traditional distinctions or reconfiguring them as conventional. I argue otherwise. I focus on these two cases because the first shows how psychological work may have an impact on at least one currently fashionable philosophical theory of perception whereas the second shows the extent to which we should be cautious about allowing the fruits of scientific investigation to mould too readily our understanding of the mental.

2. Naturalism and Avoiding the Excess of Zettelism Conscious perceptual experience is, at least partly, characterized in terms of its phenomenal character. My perceptual experience of the trees and grassland has a certain phenomenal character determined by the phenomenal properties of the experience. There is a familiar challenge to naturalism characterized in the way that I did above, namely that there is an explanatory gap between the properties identified by the sense-related sciences to characterize the brain, and its environment, and phenomenal properties (Levine 1983, 357–58). From our point of view this is not the most interesting challenge. The concern is not whether something is left out of the naturalistic picture of perceptual experience. The concern is rather the extent to which the sense-related sciences may inform us about perceptual experience at all, as the philosopher characterizes it (Martin 1997a, 75). At the extreme end of such concerns is the position that is expressed by Ludwig Wittgenstein in Zettel (1967). He writes No supposition seems to me more natural than that there is no process in the brain correlated with associating or with thinking; so it is impossible to read off thought-processes from brain processes. I mean this: if I talk or write there is, I assume, a system of impulses going out from my brain and correlated with my spoken or written thoughts. But why should the system continue further in the direction of the centre? Why should this order not proceed, so to speak, out of chaos? (1967, 608)

He concludes that it is possible that psychological phenomena cannot be investigated physiologically (1967, 609). His remarks are not just a standard observation, which an externalist about mental events and states might make, that our mental life cannot be read off brain processes independent of reference to the environment in which the subject is located. Equally, Wittgenstein’s remarks are not a plea for psychology over neuroscience. His comments would apply as much to psychology. It is rather the claim that no property of the mental life we recognize every day— beliefs, desires, thoughts, perceptions, sensations, and what it is like to undergo them—are explicable by properties of the brain even if physicalism is true.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

193

After its Wittgensteinian inspiration, let’s call such a position Zettelism. Where proponents of the explanatory gap take its existence to indicate that there is a potential problem with physicalism, Zettelists deny that the explanatory gap indicates anything about physicalism. Neurophysiological and psychological properties aren’t the right kind of things to explain phenomenal properties. However, Zettelism has a significant downside. Apart from any prima facie plausibility it may have to some people sceptical about ontological theorizing about the mind, there’s a dearth of positive arguments in favour of it. The challenge that Zettelism presents to those who hope to preserve philosophical theorizing on perception from the impact of the sense-related sciences is identifying considerations that justify moderation of the impact while avoiding simply being Zettelists. It is not enough to accuse your opponents of scientism, an intellectual position with no adherents but which may be characterized in the present context as simply taking across the results of the sciences to resolve some target philosophical subject matter while dismissing the philosophical standards that were in play to assess the success of the resolution (compare Noordhof 1995, 814). Two necessitation theses will help to refine the discussion for what follows: Capacity Necessitation and Representationalism. Capacity necessitation is a particular characterization of psychological explanation. It holds that S has capacity C in environment E if (i) S has constituents s₁, s₂, s₃, . . . sn organized in a certain way O (i.e. O(s₁, s₂, s₃, . . . sn) in E; (ii) S’s constituents have capacities C₁, C₂, C₃, . . Cn; (iii) Metaphysically necessarily, if C₁, C₂, C₃, . . Cn stand in a certain way R to each other, then C is instantiated (i.e. C₁, C₂, C₃, . . Cn are sub-capacities responsible for the instantiation of C) and (iv) Nomically necessarily, if O(s₁, s₂, s₃, . . . sn), then R(C₁, C₂, C₃, . . Cn). (compare Cummins 1983, 17, 29–31)

If C is S’s capacity to visually perceive, psychology tries to understand this capacity in terms of particular sub-capacities of constituents of S. Attempts to limit the impact of the sense-related sciences on philosophical theories of perception deny that this strategy is appropriate for the capacities attributed to persons (i.e. S ≠ (a person) P) although do apply to the proper understanding of constituents of persons in terms of their capacities and the sub-capacities that capacity necessitate them (for the strategy Fodor 1968, 109–20; Cummins 1983, 28–43; for its potential limitations Dennett 1969, 93–6; Dennett 1978, 153–4). Representationalism holds that Any subject that is a minimal representational duplicate of another subject, is a duplicate in the phenomenal character of their mental life.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

194

 

A subject is minimal representational duplicate of another subject if and only if the representational properties either instantiated by them, or in them by their constituents, are the same. The role of representationalism will be clearer in subsequent sections. Physicalism, representationalism, and capacity necessitation don’t have to be interpreted as claiming that the physical, the representational, and the sub-capacities are more fundamental, but it is reasonable to take them to imply that they are not less fundamental than what they realize. Reasons for rejecting the application of capacity necessitation to the personal are sketchy. Some claim that this is because the capacities identified at a personal level are those that make a person intelligible (e.g. Hornsby 1997, 161). Others claim that while perception involves the collection of information and should be understood as a ‘mode of sensitivity or openness’ the sub-personal processes, which psychologists and neuroscientists focus on to understand perception, involve the processing of information where the attribution of content is merely metaphorical (McDowell 1994, 197). But it is hard to deny that even if appeal to capacity necessitation were correctly limited to the sub-personal, information from the sense-related sciences has an impact on the personal level, or the whole organism. For example, we might attribute to a person the capacity to detect the taste of cherries on one’s tongue but discover that only the sweetness of the cherries is detectable by the tongue. The cherry flavour is detected retronasally by receptors involved in olfaction and is referred to our tongues (Richardson 2013, 222–23). The fact that frogs’ internal processing is only sensitive to small moving black specks suggests that frogs are not informed of the presence of bugs by their visual apparatus. Their response is to something less specific (McDowell 1994, 196). Where does it end? More generally, if a philosophical account of perception has developed with reference to a framework taken from the sense-related sciences—as McDowell suggests with regard to the Gibsonian perspective informing his own position— then the empirical inadequacy of the framework upon which one is drawing has implications for the motivation of one’s own position (McDowell 1994, 202–3, see Burge 2005, 47, for the point that Gibson’s approach is subject to empirical defeat). Nevertheless, the gestured support might not be essential to one’s position. There may be additional considerations in favour of a position that inoculate it more effectively against empirical undermining. I mention two that will be salient in what follows. The first is that philosophical theories about perception are at a level of generality that it is rare for a substantial theory in the sense-related sciences successfully to discriminate between them. They are capacity necessitation neutral. As Martin puts it, we are stuck in ‘the shallows of the mind’ when philosophically reflecting on the nature of our experience (Martin 1997a, 89–91, 97). The second is that sensory kinds are conventional rather than natural kinds, so the

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

195

capacities they characterize aren’t to be understood in terms of distinctive psychologically identified sub-capacities. Against the first point, I will explain how the sense-related sciences successfully discriminate between two philosophical theories of perception because of the commitments of the debate in play. Against the second point, I will argue that the claim that the kinds are conventional is unmotivated. I will close by offering a general characterization of the circumstances in which the application of the sense-related sciences is appropriate that serve to define a moderate naturalism I recommend.

3. Prediction Error Theory and Naïve Realism I will offer three arguments in favour of prediction error theory and naïve realism being incompatible: the Asymmetric Discreditation argument, the Argument from Response Dependent properties, and the Argument from No PerceptualCognitive Divide. Before I set out the arguments, I outline the key features of prediction error theory. The conclusion is not quite that if prediction error theory is true, then naïve realism is implausible (e.g. Rescorla 2015, 703–5). It is rather that the only way for naïve realists to avoid the conflict would involve a reconfiguration of the debate between naïve realist theories of perception and representationalist theories, so that the key claims of naïve realism simply place constraints upon the proper form a representationalist theory should take. For space reasons, and because I’m not arguing for one or the other, I bracket discussion of whether talk about sensory appearances supports representationalism or naïve realism. I suspect it favours neither (for the resources to resist Charles Travis’s (2004) argument for naïve realism, see Noordhof 2018, 84–95).

3.1 Prediction Error Theory Proponents of Prediction Error Theory take there to be one organizing principle of perceptual-cognitive processing (they don’t recognize a substantial distinction between these). The basic idea is that the brain has a perceptual model of the world, makes predictions about the sensory input it will receive, and adjusts the model on the basis of prediction error. Prediction error is a clash between the sensory input in fact received and the sensory input predicted. The sum of prediction errors made has been called free energy. In which case, the one principle of brain activity is the minimization of free energy. This has been modelled in Bayesian terms. PPosterior ðDistal Perceptual HypothesisjSensory inputÞ ¼ Pprior ðDistal Perceptual HypothesisÞ  PðSensory inputjDistal Perceptual HypothesisÞ ðHohwy 2013; 159Þ

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

196

 

The posterior probability of the distal perceptual hypothesis is given by the prior probability of the distal hypothesis weighted by the probability that hypothesis gives the sensory input received normalized by the probability of the sensory input. Prediction error is minimized when P(Sensory input/Distal Perceptual Hypothesis) is maximized. A subject’s perceptual system will, if functioning appropriately, select the perceptual hypothesis that is most probable given the sensory input the subject receives. There are complications that do not touch the basic issues of concern to us but there is one exception. Although I formulated prediction error theory in terms of distal perceptual hypotheses, prediction error theorists see perceptual-cognitive processing as a hierarchy of increasing generality, the hypotheses generating the predictions at one level are the input predicted by a higher level. In ascending the hierarchy, the hypotheses become less and less sensitive to the changes at the lower end of the hierarchy. These levels involve a shift from fast timescale causal regularities due to the postulation of edges and lines to more slow timescale hypotheses concerning objects and environments (Hohwy 2013, 27–32; Fletcher and Frith 2009, 54–5). At each level in the hierarchy, the focus is on identifying the prediction errors of level n + 1 with regard to the input at level n, and passing the required adjustments up the hierarchy as an input to level n + 2. Such a focus involves an efficient way to use cognitive resources. The system does not have to process all of the input but only departures from the predicted input. This is the predictive coding element of the approach taken from the data compression techniques of computer scientists (e.g. the use of the simplest set of rules to generate an image together with noted departures from the rules). Proponents of prediction error theory have differed over their interpretation of this approach regarding philosophical theories of perception. Some have held that the theory tells us that perception is indirect and what we perceive is the brain’s best hypothesis about the world causing our internal states (Hohwy 2007, 323). The suggestion is that the means of representation in the brain responsible for the representation of the best hypothesis are themselves the objects of awareness. Others hold that the content of our perceptions is the content of the brain’s updated best hypothesis (Clark 2013, 492). This is prima facie evidence that prediction error theory is neutral between (at least some) philosophical theories of perception. It can either be interpreted in a way congenial to sense datum theory where we mediately perceive objects and properties in the world by mediately perceiving, in this case, physical properties of the brain representing its best hypothesis or in a way congenial to representationalism with the content of perception being the content of the best hypothesis (for examples of sense datum theories see Jackson 1977, 15–29, for examples of representationalist theories see Dretske 1995; Tye 1995). Matters are different in the case of naïve realism.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

197

3.2 Naïve Realism and the Asymmetric Discreditation Argument Naïve realists contrast their position with representationalism about perception and yet, if their position is to be compatible with prediction error theory, deny that the latter’s appeal to representations in perceptual hypothesis formulation and error checking conflicts with their position. At root, the argument of this section is either that they cannot motivate this asymmetric attitude to the involvement of representations or they must concede that the distinctive features of perception to which they draw attention fail to set up a contrast between their own position and representationalism. Naïve realists hold that perceiving subjects are non-representationally aware of objects and properties in the world. The non-representational awareness is a relation between subjects and the objects and properties they perceive (Fish 2009, 14–15). A standard reason for emphasizing the non-representational character of awareness is that if perception is understood representationally, then having a perceptual state with a certain phenomenal character is compatible with the objects and properties not being present (Fish 2009, 14–15; Martin 1997b, 84–5; Travis 2018, 345). Naïve realists insist that these objects and properties constitute the phenomenal character of perceptual states (Martin 1997b, 84). It must be acknowledged that some proponents of representationalist accounts of perception say that perceptual states having a certain phenomenal character is compatible with the objects and properties they concern not being present (e.g. Dretske 2003, 70–1). However, other proponents of representationalism do not. Some propose a disjunctive representationalist account of perceptual experience in which demonstrative elements give an essentially context-dependent content to perception requiring the presence of the objects or properties they concern (McDowell 2009, 267–8; Soteriou 2000, 184–7). Others take the contextdependent content of perceptions to be a contingent feature of perceptual states, possessed when the conditions are met for successful demonstrative reference, although the perceptual state can be characterized independently, partly in terms of these demonstrative elements, whether the conditions are met or not (Burge 1991, 200, 208–10). Thus, it is important to naïve realists to emphasize the nonrepresentational nature of the awareness to characterize the distinctive thesis of their position. The representations to which prediction error theorists appeal seem to conflict with this commitment. The conflict can only be avoided if naïve realists insist that, when they claim that perceptual awareness is non-representational, that is compatible with allowing that representations may play a role in the perceptual processing identified by psychologists. To this end, they may say that naïve realism is committed to the relation of perceptual awareness to objects and properties in the world being psychologically fundamental (e.g. Brewer 2011, 48, 92; Logue 2015, 208–9). One difficulty with such a claim is that, as we noted

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

198

 

earlier, capacity necessitation, physicalism, and representationalism support the view that capacities, representational properties, and narrowly physical properties are no less fundamental than that which they necessitate. A distinction between psychological fundamentality and non-science specific fundamentality is unlikely to help unless the capacity necessitation explanations offered by psychologists are taken to be about something other than the psychological. If naïve realists claim the awareness is psychologically fundamental at the personal level, then they either need an explanation of why explanatory stories at the sub-personal level don’t throw this into question, on pain of collapsing into Zettelism, or defend a form of emergentism about relations of non-representational awareness in which, somehow, non-representational awareness explains why the psychological, although potentially sub-personal, story holds (Noordhof 2003; Noordhof 2010). To avoid this, they need a discreditation thesis: an account of why prediction error theorists’ representational properties, of the representations to which they appeal, don’t make perceptual awareness representational in the relevant sense. My argument is that, although they can supply some preliminary motivation for such a thesis, when we look at naïve realists’ rejection of representationalist accounts of the features of perception, which naïve realists emphasize, naïve realists undermine their attempt to isolate their theory from considerations drawn from the sense-related sciences. Naïve realists point to the following features of perception. (Confrontation) Perception involves confrontation with objects and their properties in the environment that, itself, cannot be mistaken. (Travis 2013, 29–31) (Conveyance) Perception is a relationship between something to which the laws of truth don’t apply (objects and properties in the world) to something which they might (representations). (Travis 2013, 8) (Unmediated Awareness) 2013, 31)

Perception involves unmediated awareness. (Travis

(Constitution) Objects and properties in the perceivers’ environment constitute what the perceiver’s perception of the environment is like. (Martin 1997b, 84) (Ontological Priority) Visual perception is a capacity for seeing whose manifestation, seeing, is ontologically prior to the capacity. (Kalderon 2018, 230) These claims are not immediately incompatible with representationalist approaches to perception. The first claim superficially falls out from perception being factive alone: if I visually perceive that p, then p is true. This is not what Travis has in mind. His observation concerns the proper way to characterize the object of perception and not the nature of the state of perception. However, a representationalist can capture this even so interpreted. Take your favourite representationalist account of the representation of properties in the environment,

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

199

for example, let a means of representation M₁ represent P₁ by being causally covariant with it in optimal circumstances (where ‘M₁’ is a particular type of means of representation in the brain and P₁ is a particular property in the environment; compare Tye 1995, 101). Such an account does not cover representation of a particular instance of P₁. A minimum condition on the representation of particular instances of P₁ would be that the instance in question cause the instantiation of the token M₁ that represents it. This is unlikely to be sufficient; however, it would be a mistake to require the representationalist to provide an analysis of the more complex kind of causal relationship that might hold without feeling the urge to press the naïve realist to say as much about the primitive notion of awareness to which they appeal. Indeed, the representationalist may claim that whatever physical conditions are required to be present for a subject to be aware of an instance of P₁ will be the story to which they appeal to characterize the more complex causal relationship. Call these representational properties concerning particular property instances and objects confrontational properties. Then the confrontational properties are representational properties that cannot be mistaken when appropriately combined. If a different property instance stood in the kind of causal relationship required, then the token M₁ would represent that instead. If no such property instance stood in the kind of causal relationship required, then no property instance would be represented. If perception involves the presentation of particulars, then the confrontational properties in virtue of which they are represented make perception essentially relational (the feature of naïve realism emphasized by Allen 2019, 2–3; Logue 2013, 107–10). Such properties suggest that what is an essential feature of perception need not give its fundamental nature. Confrontational properties explain why we are aware of objects and properties in the world and not vice versa (see Logue 2013, 128–31). Equally, the representation of particular instances of properties that objects possess demonstrate that it is a mistake to insist all representation involves generality (as Travis does, e.g. 2018, 339–40, 344). The second claim (conveyance) was that perception is a relation between nonrepresentational objects and properties in the world and representations. This is compatible with the relation in question holding in virtue of particular kinds of representational properties. Conflict will only occur if the naïve realist is committed to denying that the resulting representations are part of the perception, in virtue of the fact that they possess the representational properties responsible for the relationship, as opposed to a cognitive response to the perception. The third claim was that perception involves unmediated awareness. The appeal to states with confrontational properties may appear to make awareness mediated by the possession of these properties. But we may distinguish two ways in which awareness is mediated. According to the first, a subject’s awareness of X is mediated if they are only aware of X by being aware of Y (where Y ≠ X and Y is not a part of X; compare Jackson 1977, 20). According to the second, a subject’s

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

200

 

awareness of X is mediated if the subject would not be aware of X if it were not the case that some distinct property of the subject were instantiated. The difference is that it is not being claimed that the subject needs to be aware of this distinct property. Confrontational properties are not objects of awareness. So they do not mediate in the first sense but only in the second. Naïve realists are in no better position since they will acknowledge that there are properties of the brain (and, indeed, sub-personal psychological properties) which, if they had not been instantiated, the subject would not be aware of the objects and properties their perception concerns. The combination of the fact that appeal to confrontational properties make perceptions essentially related to the objects and properties they concern without the awareness of them being mediated in the first sense explains how the fourth claim (constitution) holds. Objects and properties in the environment constitute what the perceiver’s perception of the environment is like. The final claim was that sight is ontologically dependent upon its non-defective exercise and that is why the good case, seeings, have priority. Seeings count as non-defective because they are that for the sake of which the subject possesses the capacity (Kalderon 2018, 231). Although Kalderon takes this point to favour naïve realism, there is nothing in this characterization to rule out characterizing perception in terms of a state with confrontational properties instead. A state’s confrontational properties are not shared with defective cases such as hallucination and it is states with confrontational properties that explain why perceivers possess the capacity of perception. It is time to complete my first argument. Naïve realists will reject representationalist approaches that appeal to confrontational properties as incompatible with their own. In so doing, they have to take confrontational properties to be the kind of representational properties that makes a relation of awareness representational. And yet, at the same time, if their position is compatible with prediction error theory, they must claim that even if perception involves an updated representation of the world to minimize prediction error, this is part of the realization of a subject’s non-representational awareness of objects and properties in the world. Moreover, they must deny that the sub-capacities that constitute this process throw into question their claim that perception is the capacity to confront us with objects and properties in the environment. The charge of asymmetric discreditation is that they have no grounds for this asymmetric attitude to the involvement of representational properties, and the particular role they play in perception, regarding representationalist positions on the one hand and prediction error theory on the other. If they seek to avoid the argument by allowing that there is no problem with confrontational properties either, then, in effect, they acknowledge that they never offered a position that was in conflict with how representationalism was, in fact, formulated. However, the conflict with naïve realism does not just arise from the particular use to which

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

201

representations are put in prediction error theory. There is in addition, the treatment of hallucination and its implications for perceptual explanatory kinds to which we now turn.

3.3 Hallucination and Response-Dependent Properties The present argument focuses on the idea that the difference between hallucination and perception is not to be found in the nature of the state, which represents a hypothesis about the way the world is, but rather whether its occurrence was the result of a prior prediction error–checking process concerning the stimulations that would be received (Clark 2016, 14, 308, n. 3). Other psychological theories involving representations don’t imply the same level of representational similarity. Naïve realists have been concerned that if hallucination and perception involve the same kind of mental state, then the distinctive features of perception, by the lights of the naïve realist, will be explanatorily redundant (e.g. Martin 2004, 46). The cognitive and behavioural consequences of perception will be explained by what is common to the two cases. For this reason they make two moves. First, they provide a characterization of perceptual experience that covers both perception and hallucination without implying that they fall under a common mental kind. Thus in the case of vision they say Visual experience of O being F is that of a situation being indiscriminable through reflection from a veridical visual perception of an O being F as an O being F. (compare Martin 2006, 363)

Second, they make a negative claim about the case of hallucination. For causally matching hallucinations of O being F, there is no more to the phenomenal character of such experience than that of being indiscriminable from corresponding visual perceptions of an O being F as an O being F. (Martin 2006, 369)

I have appealed to Martin’s formulation with some adjustments for clarification. Other naïve realists say much the same thing (e.g. Fish 2009, 81; Travis 2013, 11). Although the temptation is to argue that there must be something that explains why hallucinating subjects make the judgements they do about their experience, for example, that there is a dagger (which, in fact, they are hallucinating) before them, this is incorrect. Their judgements are not properly constrained by any grounds in experience (Martin 2006, 389).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

202

 

Some naïve realists emphasize that what is common to cases of perception and hallucination can be understood in terms of the subjects’ responses (e.g. Travis 2013, 11; Fish 2009, 94, 101–2). These rely upon there being identifiable responses in common that don’t require sophisticated concept possession, on pain of not being able to characterize what is common to dogs’ perceptions and hallucinations or, for that matter, those of very young children. One difficulty with such an approach is the more primitive the response, the more plausible it is to identify it as a state that might play the same explanatory role as that which is attributed to the perceptual states to which they are a response. Naïve realists’ concern with explanatory redundancy is once more pressing. Martin avoids the difficulty of identifying a primitive kind of response by taking what is common to hallucination and perception to be characterized by an impersonal epistemological viewpoint. The point is not to focus on the actual cognitive states subjects might go into, or would go into if the situation were different, but rather provide a characterization of how things might seem to subjects, as a result of these states being indiscriminable, that abstracts away from the circumstances of any particular subject (Martin 2006, 383). A natural way to make this clearer is to take the characterization of subjective indiscriminability from an impersonal standpoint as the description of an idealized response-dependent property. We may understand them schematically as follows, setting aside whether or not it is an a priori truth. A subject has a visual experience of O being F if and only if, in circumstances in which response to the experience arises, a typical subject possessing the appropriate concepts would, on the basis of reflection alone, either respond with the same cognitive states, or respond with cognitive states that themselves, although different, are subjectively indiscriminable to the subject (call these the SI-cognitive responses). The latter qualification regarding cognitive states concerns those which would differentiate between perception and hallucination because of their objectdependent character, for example the belief that that object is F (which would fail to be met in some cases of hallucination). Without this qualification, we would have no way of capturing what is common to perceptions and the corresponding hallucinations because their cognitive effects would always, in principle, be different (Martin 2013, 42–3, floats such a proposal). Appeal to subjective indiscriminability at the cognitive level to capture subjective indiscriminability at the level of perceptual experience is not illicit because we can explain cognitive indiscriminability in terms of the same demonstrative-indexical capacities being at work in perception and hallucination regarding the states in question. Their exercise does not depend upon demonstrative or indexical reference being successful.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

203

A necessary condition upon a property being response-dependent is that no property in the world explains the specified responses but rather the common responses explain why entities in the world fall under a certain predicate that, thereby, expresses the response-dependent property (compare Wright 1992, 123–4). Prediction error theory potentially identifies such a property in the world, namely the representational (though not confrontational) properties that encode the perceptual hypotheses controlled to a greater or lesser extent depending upon whether the subject is perceiving or hallucinating. Moreover, this property explains how things could seem the same way to subjects when they lack the cognitive resources to respond in the identified ways, for example, because they are a dog. Martin himself allows that ‘further facts concerning the operation of the human visual system’ may explain why at least some hallucinations are indiscriminable from their corresponding perceptions (Martin 2013, 46). But this stops short of recognizing that impersonal subjective indiscriminability is not a responsedependent property for, at least, three reasons. First, it isn’t claimed that there is a common explanation of all such responses, so there is still something for commonality of response to bind together. Second, reference is made to the sensory areas of the human visual cortex rather than something distinctively psychological. Third, it is suggested that the effects in the hallucinatory case, for example believing that there is a pig before one, are parasitic on the veridical case because the hallucinatory case is indiscriminable from the case in which a pig is presented to one in experience (Martin 2013, 46). The problem for naïve realism is that prediction error theory identifies candidate representational states which, when and because they account for subjective indiscriminability, are plausibly thought of as at the personal psychological level. Psychological theories postulate representational states at many levels. The question of whether or not these states are states of a person with an impact upon their conscious mental life turns on the extent to which they are available to, and explain the cognitive responses of, the subjects in which, to put it neutrally, these representational states are instantiated. The SI-cognitive responses to these states are the basis for attributing to these states a personal character. Of course, it is possible to deny that there is this kind of common explanation of the subjective indiscriminability of hallucinations and their corresponding perceptions. As Martin notes, doing so will mean that the modality involved in the primitive subjective indiscriminability claim cannot be understood in terms of possible worlds (Martin 2013, 47). However, the plausibility of this claim should be evaluated in standard ways in which we evaluate the claim that a certain property is response-dependent. At the least, prediction error theory puts this under pressure, and if naïve realists attempt to discredit the kind of representational states that prediction error theorists postulate as appropriate, they once more are subject to the charge of asymmetric discreditation.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

204

 

3.4 Perception–Cognition Divide Prediction error theory only recognizes one type of processing: Bayesian processing involving updating to the hypothesis that minimizes prediction error. The difference between so-called perceptual processing and so-called cognitive processing is that the former concern predictions involving shorter timescales whereas the latter concern predictions with longer timescales. This corresponds to the relatively immediate prediction errors detected concerning sensory properties such as colour and shape compared to the errors revealed over a longer timescale concerning the more complex categorization of objects. In brief, there is a pretty immediate way that we can find out whether we are right and wrong about the boundaries of a cat; whether or not it is a cat, requires more lengthy observation (Hohwy 2013, 27–32). The implications of these features of prediction error theory are two-fold. First, it is not appropriate to consign its subject matter to sub-personal processes as naïve realists try to do. It applies as much to personal level cognition. Second, and perhaps more important for our current discussion, it takes the transition from perception to cognition to be a continuum with no sharp demarcation (e.g. Fletcher and Frith 2009, 48). If there are representations at a cognitive level, then there are representations generating predictions at a shorter timescale at perceptual level. There is no obvious room in this picture for recognizing a different, non-representational, kind of state.

4. Sensory Kinds as Natural Kinds Part of the clash between naïve realism and prediction error theory rested upon there being a common explanatory kind that prediction error theory recognized and naïve realism did not. It undermined the attempt to provide a non-ontological reading of subjective indiscriminability. The case of sensory kinds is different. Here the sense-related sciences are taken to support a proliferation in the number of senses (or sensory modes) recognized compared to the five that we tend to recognize in everyday life: vision, audition, olfaction, gustation, and touch. Some of the proliferation is due to the recognition of additional senses relating to posture and our internal organs. My focus will be on a line of thought that seeks to splinter or deny the division between the five senses just mentioned. There is evidence that some of traditional senses involve the workings of more than one capacity. For example, it is suggested that vision actually involves two capacities: the first, operating via the ventral stream, involves judgements of object identification; the second, operating via the dorsal stream, relates to action and, specifically, the location of objects (Goodale and Milner 1992, 21–4; Milner and Goodale 1995). Equally, there is evidence of inter-modal interaction. There are

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

205

cross-modal illusions like the McGurk effect. If you hear a Ba and see somebody pronouncing Ga, then you hear the Ba as a Da. This seems to be an example of two sensory modalities seeking to give a consistent experience where one might have expected audition to have priority (McGurk and MacDonald 1976, 746–8; Nudds 2011, 335). There can also be auditory effects on vision too. Two beeps occurring at the same time as a flash of light can make the flash seem like two flashes (Shams, Kamitani, and Shimojo 2000, 788; 2002, 148–152, noted by O’Callaghan 2012, 98–9). There is also the way in which taste and smell are integrated in order for us to identify flavours such as cherry. Retronasal stimulation of olfactory receptors produces referred taste experience on the tongue. Richardson argues that we need neither suppose that olfactory information is integrated into taste experience as a cross-modal effect or that some flavours are smells (2013, 222–3, 332–9). A response to this line of thought, seeking to preserve the traditional recognition of five senses, takes these sensory kinds to be conventional rather than natural kinds. My principal aim will be to show that this response is unnecessary and, indeed, that the way it has been developed supports the recognition of common explanatory kinds corresponding to the traditional senses. The argument I want to consider goes as follows. (1) If the traditional five senses are natural kinds, then they correspond to distinct perceptual mechanisms involving a modular structure or, at the minimum, particular specialized capacities that may, in turn, be broken down into sub-capacities (see section 2). (2) If a sense is modular, then it is relatively informationally encapsulated. (3) If a sense involves a distinctive specialized capacity, then there will be particular psychological processes that are responsible for that capacity. (4) Some of the traditional senses involve two or more modules or particular psychological processes subserving distinct capacities. (5) If senses involve complex modules or capacities, then there is a difference in kind between traditionally intra-modal interactions and inter-modal interactions of these modules or capacities. (6) There is no difference in kind between traditionally intra-modal interactions and inter-modal interactions of these modules or capacities. Therefore, (7) The traditional five senses aren’t natural kinds. I have put this argument together from a number of sources. Nudds places the emphasis upon the fact that each sense doesn’t correspond to a single capacity processing a distinctive subject matter (Nudds 2004, 35, n. 12; Nudds 2011, 322–38). But this ignores the possibility canvassed in (5) that, although there is more than one capacity involved in a particular sense, together they constitute a complex capacity. (5) and (6) is suggested by work of Casey O’Callaghan (2012,

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

206

 

98–9, 112). From this argument, it is urged that we have a choice of rejecting the standard way of counting the senses or taking the traditional five senses to be conventional kinds. Let me explain some of the key terms in the argument a little more. If a sense is informationally encapsulated, it takes information from proprietary transducers, processes this information, yielding an output independently of the other senses and generally available background information. A standard illustration of the purported informational encapsulation of vision is the Müller-Lyer illusion (Fodor 1983, 66–8). Even though we know the lines are the same length, we still see them as having different lengths. Taking senses to be capacities more generally allows that there is a proprietary neural mechanism, involving the processing of the input of distinctive sensory transducers and retrieving information about the world, even if there is some influence, from other senses and background knowledge, in the output. My response to the argument will have a number of distinct elements. First, and most directly, I will identify a difference between traditionally intra- and intermodal interactions. Second, I will argue that the difference is important for understanding the difference between attributing different types of sensory capacity and attributing different types of senses. Putting these two elements together, we can see why the existences of inter-modal interactions, together with intramodal interaction between the capacities involved in one sense, doesn’t imply that we should either splinter or deny the divisions between the senses. However, these elements are compatible with taking sensory kinds to be manifest kinds in the sense of having distinctive bundles of readily available properties on the basis of which we classify them—lakes, roads, bicycles, and jewels would be other examples—without them being natural kinds (Richardson canvasses this possibility—2013, 327–8). The third element is the case for them being natural kinds drawing upon their distinctive explanatory role. I start with the second element. It is easy to think if we identify a sensory capacity that responds to a particular kind of property, then that’s all we need to do to establish that there is one token sense which corresponds to it. But that’s not the case. Suppose that a creature has eyes at the front and the back of their head. Then there are two options. First, the creature has one token sense. Sensory input from these visual senses may be integrated into one 360o visual experience. Second, there are two token visual senses between which the creature may shift their attention. Note that the two streams need not have the same character and yet still be visual senses. The eyes at the back of the head may give information only in black and white and fail to identify objects as opposed to provide information for the creature to act upon. So, there is a difference between the creature in which this second option holds and ourselves with regard to vision. Even if there are two capacities at work in our case, the two capacities have an integration that is part of the reason why identifying vision as one token sense seems plausible. This is quite compatible with also allowing that if only one of the

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

207

capacities were in play in a creature, they would have a visual sense. It would just not be like ours. The traditional demarcation of our senses into five does not focus on whether some of our senses involve capacities that might have been instantiated in isolation and consider how many of these there are. Under what circumstances do two sensory capacities constitute one token sense? My suggestion is that two elements are in play. The first, which is implicit in the creature discussed above, is when the two capacities are integrated into one spatio-temporal sensory field so that cross-capacity identifications and adjustments are made. Two token senses are distinct, even if they are of the same type, if they are not so integrated. However, this is insufficient for two capacities to be integrated into one token sense. The cross-modal illusions above would otherwise suggest that our senses are collectively integrated into one token sense. So there is a second feature. To understand what it is, I need to introduce some terminology. First, there is the idea of sensory duplication. Two objects are sensory duplicates for a modality M iff, for any subject with the sensory modules or capacities distinctive of M, the subject cannot discriminate between these objects by M alone without appeal to their distinct spatiotemporal position. We can then define the sensory properties of M as those that must be shared by sensory duplicates for M. No property is a sensory property for M if two objects may differ in that property and yet the subject cannot discriminate between them. The second feature of my proposal may then be put like this. Two modules or capacities constitute one token sense if the sensory field in which they are integrated has the properties that are presented in the sensory field from each module or capacity as sensory properties for the token sense in question. The two capacities identified in vision involve the location of objects and their visual sensory properties within a spatio-temporal field. Causal interaction between modules or capacities that fail to have this result fail to constitute a single sense. Let’s consider how this works in the inter-modal cases. First, consider the McGurk effect. The cross-modal influence of seeing Ga pronounced on what is heard doesn’t make Ga presented as an auditory sensory property. Instead, what is heard is Da. So this is not an example of where the visual sense modality is integrated into one sensory field with the information it supplies, namely that Ga is pronounced. Instead, it is the straightforward causal interaction between one sense and another. Second, consider the Ventriloquist illusion (Bertelson and Radeau 1976; Bertelson and Aschersleben 1998; Bertelson 1998). We see the sound coming from the ventriloquist’s dummy. However, in seeing this, the auditory properties are not presented as the sensory properties of the visual sense. Instead, part of our visual experience involves non-sensory properties for that sense being represented

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

208

 

by it. The precise way to characterize this is controversial but we might take the sound to be part of the non-sensory but manifest content of our visual experience; others may talk of the sound being amodally represented (Noe 2004, 67–72; Noordhof 2018, 84–6). Vision and audition are not integrated in the required way. Note this is distinct from the issue of whether the representations involved in the manifest or amodal content are sensory specific or amodal (for the latter see Nudds 2015). Consider now the case of taste. The receptors on the tongue only detect the following: bitter, sour, sweet, salty, umami, and, possibly, metallic. Olfactory receptors approached retronasally rather than orthonasally pick up flavours like cherry and vanilla. This has led some researchers to conclude that olfaction is a dual modality detecting two kinds of sensory property: orthonasally—smells in the environment; retronasally—flavours of food (Rozin 1982). The problem for my present proposal is that retronasal olfactory experience seems well integrated into our gustatory sensory space and the integration involves sensory properties: flavours. There is a clear difference between experiencing a certain sweetness to be from a cherry (which might be a non-sensory manifest property of what we are tasting) and experiencing the cherry flavour. In this case, there are two possibilities compatible with my proposal. First, we could take retronasal olfactory reception as part of gustation because distinct neural processes are involved in its processing. There is no particular reason why two senses might not share sense receptors while using the information in different ways. Second, we could argue that the original characterization of the five senses should, given current use, have appealed to flavour rather than taste, given scientists identified taste receptors in a particular way. In brief, none of these cases (in contrast to the traditional senses) displays the key feature of causal interaction between the capacities as a result of which there is an integrated spatio-temporal sensory field in which sensory properties detected by either capacity are displayed. Apart from any independent plausibility my proposal has, it is natural to wonder whether there is anything further to be said in favour of these moves. Together they might form a defence of the claim that the traditional sensory kinds are manifest kinds. But is there any reason to suppose that the sensory kinds are natural? This brings me to the third element of my response to the argument: setting out the distinctive explanatory role for the traditional discrimination of sensory kinds. As part of his defence of sensory kinds as conventional, Nudds suggests that the reason why we discriminate between the senses is because it has explanatory significance. The traditional senses provide characteristic bundles of information about objects and properties in the world that have predictable consequences for explaining and understanding behaviour, and these characteristic bundles of information are produced in the same way (Nudds 2004, 44–7; 2011, 337). This

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

209

might explain why the traditional discriminations persist, but it does not explain why they won’t be superseded. After all, the same can be said of the sensory kinds some scientists and philosophers suggest that we should recognize involving subcapacities of the senses, e.g. object classification and location. The key difference, according to Nudds, is that the traditional classification is an established explanatory practice and it is the fact that we all draw the discriminations in the way that we do that gives the continuing utility to the practice (2004, 48, drawing on Lewis 1969 and Burge 1975). However, it is instructive to consider why it is an established explanatory practice. Nudds himself identifies at least two potential reasons (2004, 48–9). First, the traditional sensory classification supports relatively easily attributable states of information about the world on the basis of, for example, eye direction, the situation and uncovered character of ears, limbs making contact with objects, and so on. Second, they characterize in ways salient to the perceiving subject what information they can discover about the world and what they can fail to discover in a behaviourally salient way. The sensing creature knows what to do to discover further information about the world. As Mohan Matthen puts it, the senses, as we traditionally conceive them to be, are modes of exploration of the world (2015, 580). In the case of what we traditionally count as taste, there may be two to three kinds of receptors that provide information about flavour, but there is one distinctive mode of exploration: putting the food and drink in your mouth. It is less easy for a creature to think in terms of the enlisting of a particular type of colour receptor or heat receptor alone. These are a good stab at an explanation of why the explanatory practice of discriminating in terms of the traditional senses has built up. However, in fact, they are a little too good. The merits of the explanatory practice make it much harder to see how there could have been an incompatible regularity picked out by sensory kinds that would have served substantially the same functions just as well (Burge 1975, 254). Instead, they characterize an interesting explanatory level in themselves, with sufficient explanatory value to constitute a natural kind. They are the basis for a discrimination of senses that both enables creatures to attribute information states to others, predict their behaviour as a result of it, and also find out further information themselves. Whatever sub-capacities we recognize, there is no reason to abandon the traditional sensory classification demarcating an important explanatory role that, in itself, establishes the status of sensory kinds as natural rather than conventional kinds.

5. Concluding Remarks Although philosophical theories of perception operate at a high level of abstraction, conflicts with empirical theories may arise because of the structure of the

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

210

 

debate between different philosophical theories (naïve realism vs representationalism) and constraints that arise from the coherent development of naïve realism within a wider established framework (issues relating to response-dependence and fundamentality). As we saw, one consequence of this may be a reconfiguration of the philosophical debate so that, for example, naïve realism and representationalism properly developed are not in competition. The relative shallowness of the phenomena philosophers are seeking to characterize is no protection against such consequences. In the case of the sensory kinds, we found that philosophical discussion helped to identify an explanatory level that might be overlooked if we just took information over from the sciences in a straightforward and uncritical fashion. Identification of this additional explanatory level did not deny the relevance of scientific considerations but set it in a further context. There was no need to take sensory kinds to be conventional to allow for this possibility. The comparison with naïve realism is instructive. The sense-related sciences need not reject the explanatory value of distinguishing between perception and hallucination. No simple application of the proximate cause principle (same proximate sensory stimulation, same perceptual kind) has the consequence that perceptions and their corresponding hallucinations are of the same kind (e.g. Burge 2005, 25). However, conflict does arise when naïve realists deny the existence of a mental kind, inclusive of perception and the corresponding hallucination, that the sense-related sciences judge to be of psychological significance. The discussion of the characterization of perception offered by prediction error theorists and its implications for a response-dependent analysis of subjective indiscriminability was one way of bringing this point out. Perhaps the lesson is that philosophy is under less pressure from empirical results in the sense-related sciences when it is focused upon identifying issues for further exploration, and setting out how they might be resolved, than when it is seeking to resist the relevance of certain scientific explanations, supporting a certain ontology, while self-consciously putting nothing in its place due to philosophical pressures. An appropriately moderate naturalism about perception counsels the merits of recognizing these conflicts and sometimes just wading in.¹

¹ I’m very grateful for the feedback from audiences in a workshop organised as part of the Templeton funded Purpose and Procedure in the Philosophy of Perception project (I remember Zoe Drayson, Ivan Ivanov and Keith Wilson being particularly helpful and encouraging) back in March 2017, and research seminars in the Philosophy Departments of the Universities of Bristol and Glasgow, in 2019 and 2020, who commented on the second and first parts of the paper respectively, the other members of the SPIN team (Keith Allen and Clare Mac Cumhaill), and especially to the members of the SPIN team and editors of this volume, Heather Logue and Louise Richardson, both for their more detailed feedback and opportunity to publish a summary of the more extensive research and paper in the present volume.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

211

References Allen, Keith. 2019. ‘The Value of Perception’. Philosophy and Phenomenological Research 100 (3): 633–56. Bertelson, Paul and Radeau, Monique. 1976. ‘Ventriloquism, sensory interaction, and response bias: Remarks on the paper by Choe, Welch, Gilford, and Juola’. Perception and Psychophysics 19 (6): 531–5. Bertelson, Paul and Aschersleben, Gisa. 1998. ‘Automatic visual bias of perceived auditory location’. Psychonomic Bulletin and Review 5 (3): 482–9. Bertelson, Paul. 1998. ‘Starting from the ventriloquist: The perception of multimodal events’. Advances in Psychological Science Volume 2: 827–70. Burge, Tyler. 1975. ‘On Knowledge and Convention’. The Philosophical Review 85: 249–54. Burge, Tyler. 1991. ‘Vision and Intentional Content’. In Ernest LePore and Robert Van Gulick (eds.) John Searle and His Critics. Oxford: Basil Blackwell, 195–213. Burge, Tyler. 2005. ‘Disjunctivism and Perceptual Psychology’. Philosophical Topics 33 (1): 1–78. Clark, Andy. 2013. ‘Expecting the World: Perception, Prediction, and the Origins of Human Knowledge’. The Journal of Philosophy 110 (9): 469–96. Clark, Andy. 2016. Surfing Uncertainty. Oxford: Oxford University Press. Cummins, Robert. 1983. The Nature of Psychological Explanation. Cambridge, MA: The MIT Press. Dennett, Daniel C. 1969. Content and Consciousness. London: Routledge and Kegan Paul. Dennett, Daniel C. 1978. ‘Toward a Cognitive Theory of Consciousness’. In Brainstorms. Hassocks, Sussex: Harvester Press, 149–73. Dretske, Fred. 1995. Naturalizing the Mind. Cambridge, MA: The MIT Press. Dretske, Fred. 2003. ‘Experience as Representation’. Philosophical Issues 13: 57–82. Fish, William. 2009. Perception, Hallucination, and Illusion. Oxford: Oxford University Press. Fletcher, P. and Frith, C. 2009. ‘Perceiving is Believing: A Bayesian Approach to Explaining the Positive Symptoms of Schizophrenia’. Nature Reviews: Neuroscience 10: 48–58. Fodor, Jerry A. 1968. Psychological Explanation. New York: Random House. Fodor, Jerry A. 1983. The Modularity of Mind. Cambridge, MA: The MIT Press. Goodale, Melvyn A. and Milner, A. David. 1992. ‘Separate Visual Pathways for Perception and Action’. Trends in Neuroscience 15 (1): 20–5. Hohwy, Jakob. 2007. ‘Functional Integration and the Mind’. Synthese 159: 315–28. Hohwy, Jakob. 2013. The Predictive Mind. Oxford: Oxford University Press.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

212

 

Hornsby, Jennifer. 1997. Simple Mindedness. Cambridge, MA: Harvard University Press. Jackson, Frank. 1977. Perception. Cambridge: Cambridge University Press. Jackson, Frank. 1994. ‘Armchair Metaphysics’. Reprinted in his (1998), Mind, Method and Conditionals. London: Routledge, 154–76. Jackson, Frank. 1998. From Metaphysics to Ethics. Oxford: Oxford University Press. Kalderon, Mark. 2018. ‘Experiential Pluralism and the Power of Perception’. In John Collins and Tamara Dobler (eds.) The Philosophy of Charles Travis. Oxford: Oxford University Press, 222–36. LaPorte, Joseph. 2004. Natural Kinds and Conceptual Change. Cambridge: Cambridge University Press. Levine, Joseph. 1983. ‘Materialism and Qualia: The Explanatory Gap’. Pacific Philosophical Quarterly 64: 354–61. Lewis, David. 1969. Convention. Oxford: Basil Blackwell. Logue, Heather. 2013. ‘Good News for the Disjunctivist about (one of) the Bad Cases’. Philosophy and Phenomenological Research 86 (1): 105–33. Logue, Heather. 2015. ‘Disjunctivism’. In Mohan Matthen (ed.) The Oxford Handbook of Perception. Oxford: Oxford University Press, 198–206. Martin, M. G. F. 1997a. ‘The Shallows of the Mind’. Proceedings of the Aristotelian Society Supplementary Volume, 71: 75–98. Martin, M. G. F. 1997b. ‘The Reality of Appearances’. In Mark Sainsbury (ed.) Thought and Ontology. Milan: Franco Angeli, 81–106. Martin, M. G. F. 2004. ‘Limits of Self-Awareness’. Philosophical Studies 120: 37–89. Martin, M. G. F. 2006. ‘On Being Alienated’. In Tamar Szabo Gendler and John Hawthorne (eds.) Perceptual Experience. Oxford: Oxford University Press, 354–410. Martin, M. G. F. 2013. ‘Shibboleth: some comments on William Fish’s “Perception, Hallucination and Illusion” ’. Philosophical Studies 163 (1): 37–48. Matthen, Mohan. 2015. ‘The Individuation of the Senses’. In his Oxford Handbook of the Philosophy of Perception. Oxford: Oxford University Press, 567–86. McDowell, John. 1994. ‘The Content of Perceptual Experience’. The Philosophical Quarterly 44 (175): 190–205. McDowell, John. 2009. ‘Avoiding the Myth of the Given’. In his Having the World in View. Cambridge, MA: Harvard University Press, 256–72. McGurk, Harry and MacDonald, John. 1976. ‘Hearing Lips and Seeing Voices’. Nature 264: 746–8. Milner, A. David and Goodale, Melvyn. 1995. The Visual Brain in Action. Oxford: Oxford University Press. Noë, Alva. 2004. Action in Perception. Oxford: Oxford University Press.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

213

Noordhof, Paul. 1995. ‘Scientism’. In Ted Honderich (ed.) The Oxford Companion to Philosophy. Oxford, Oxford University Press, 814. Noordhof, Paul. 2003. ‘Not Old . . . but Not That New Either: Explicability, Emergence and the Characterization of Materialism’. In Sven Walter and Heinz-Dieter Heckman (eds.) Physicalism and Mental Causation: The Metaphysics of Mind and Action. Charlottesville: Imprint Academic, 85–108. Noordhof, Paul. 2010. ‘Emergent Causation and Property Causation’. In Cynthia MacDonald and Graham Macdonald (eds.) Emergence in Mind. Oxford: Oxford University Press, 69–99. Noordhof, Paul. 2018. ‘Evaluative Perception as Response Dependent Representation’. In Anna Bergqvist and Robert Cowan (eds.) Evaluative Perception. Oxford: Oxford University Press, 80–108. Nudds, Matthew. 2004. ‘The Significance of the Senses’. Proceedings of the Aristotelian Society 104: 31–51. Nudds, Matthew. 2011. ‘The Senses as Psychological Kinds’. In Fiona Macpherson (ed.) The Senses. Oxford: Oxford University Press, 311–240. Nudds, Matthew. 2015. ‘Is Audio-Visual Perception “Amodal” or “Crossmodal” ’. In Dustin Stokes, Mohan Matthen, and Stephen Briggs (eds.) Perception and its Modalities. Oxford: Oxford University Press, 166–88. O’Callaghan, Casey. 2012. ‘Perception and Multimodality’. In Eric Margolis, Richard Samuels and Stephen P. Stich (eds.) The Oxford Handbook of the Philosophy of Cognitive Science. Oxford: Oxford University Press, 92–117. Rescorla, Michael. 2015. ‘Bayesian Perceptual Psychology’. In Mohan Matthen (ed.) The Oxford Handbook in the Philosophy of Perception. Oxford: Oxford University Press, 694–716. Richardson, Louise. 2013. ‘Flavour, Taste and Smell’. Mind and Language 28 (3): 322–41. Rozin, Paul. 1982. ‘ “Taste-smell confusions” and the duality of the olfactory sense’. Perception and Psychophysics 31 (4): 397–401. Shams, Ladan, Kamitani, Yukiyasu, and Shimojo, Shinsuke. 2000. ‘What you see is what you hear’. Nature 408: 788. Shams, Ladan, Kamitani, Yukiyasu, and Shimojo, Shinsuke. 2002. ‘Visual illusion induced by sound’. Cognitive Brain Research 14: 147–52. Soteriou, Matthew. 2000. ‘The Particularity of Visual Perception’. European Journal of Philosophy 8 (2): 173–89. Sterelny, Kim. 1990. The Representational Theory of Mind. Oxford: Basil Blackwell. Travis, Charles. 2004. ‘The Silence of the Senses’. Mind 113 (449): 59–94. Travis, Charles. 2013. Perception. Oxford: Oxford University Press. Travis, Charles. 2018. ‘Reply to Keith A. Wilson’. In John Collins and Tamara Dobler (eds.) The Philosophy of Charles Travis. Oxford: Oxford University Press, 338–50.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

214

 

Tye, Michael. 1995. Ten Problems of Consciousness. Cambridge, MA: The MIT Press. Wittgenstein, Ludwig. 1967. Zettel. G. E. M. Anscombe and G. H. Von Wright (eds.), translated by G. E. M. Anscombe. Oxford: Basil Blackwell. Wright, Crispin. 1992. Truth and Objectivity. Cambridge, MA: Harvard University Press.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

11 Naturalism and the Metaphysics of Perception Zoe Drayson

1. Introduction What is the relationship between philosophical and scientific theories of perception? We should not expect there to be a straightforward answer to this question, not least because of the multitude of subdisciplines and methodologies involved. The philosophy of perception, for example, includes research on the epistemic role of perceptual states, the subjective qualities of perceptual experience, and the metaphysical relation between perceivers and the world. Research in the science of perception includes computational models of perceptual algorithms, neurobiological models of perceptual systems, and behavioural models of the relation between perception, attention, and action. In this paper, I will be narrowing my focus to the relationship between two particular debates about perception: the debate between naïve realists, intentionalists, and others, which I will refer to as the “Metaphysical debate”, and the debate between ecological theorists and constructivist theorists, which I will refer to as the “Psychological debate”.¹ My interest in the relationship between the philosophy and science of perception comes largely from thinking about recent work in naturalistic metaphysics, which explores the role of scientific theories and data in our metaphysical theorizing.² Although I hope this paper paves the way for further work in the naturalistic metaphysics of perception, I will not be making any particularly naturalistic assumptions or arguments here. I am proposing that, regardless of

¹ Throughout this paper, I will use the term “Metaphysical” to describe both the particular debate and the individual theories it involves: intentionalism, naïve realism, sense datum theory, and adverbialism. I will use the term “Psychological” to describe both the particular debate between ecological and constructivist theories, and the theories themselves. As will become apparent, I want to leave open the possibility that the Psychological debate is as metaphysical as the Metaphysical debate. ² The commitments of naturalistic metaphysicians vary widely: for a representative sample see Hawley (2006), Ladyman and Ross (2007), French and Mackenzie (2012), Paul (2012), Chakrabartty (2017). Zoe Drayson, Naturalism and the Metaphysics of Perception In: Purpose and Procedure in Philosophy of Perception. Edited by: Heather Logue and Louise Richardson, Oxford University Press (2021). © OUP. DOI: 10.1093/oso/9780198853534.003.0011

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

216

 

how you characterize the relationship between metaphysics and science, Metaphysical and Psychological theories of perception present a particularly interesting case study. Philosophers debating naïve realism and intentionalism take themselves to be engaged in a debate which is distinct from and orthogonal to the psychological debate between ecological and constructivist theories of perception. McDowell (2010), for example, clearly distances the philosophical study of perceivers from the scientific study of their perceptual systems. Even philosophers who acknowledge a relationship between Metaphysical and Psychological theories, such as Lowe (2000), assume that the debate between naïve realism and intentionalism does not stand or fall with the debate between ecological and constructivist theories. When we try to specify the grounds for distinguishing between the two debates, however, we face a challenge: the usual strategies for distinguishing between metaphysical and scientific theories (e.g. appealing to differences in their modal strength, their methodology or their explanatory features) don’t seem to apply. I will argue that any differences we find between the two approaches do not license the conclusion that the two debates are orthogonal. I will suggest that the two debates are engaged in the same general project concerning the nature of perception, and that the Psychological theories are no less metaphysical than the Metaphysical theories. I am not arguing that the Psychological theories are superior to the Metaphysical theories, but merely that the Psychological theories should be recognized as part of the debate about the metaphysical nature of perception. In the next part of the paper, I will introduce the two debates under consideration: the Metaphysical debate between theories such as naïve realism and intentionalism, and the Psychological debate between ecological and constructivist theories. I will then consider the standard strategies for distinguishing between different kinds of theories in virtue of their commitments or methodologies. One might attempt to distinguish between the modal strength of different theoretical claims (e.g. necessary versus contingent truths), different methodological approaches (e.g. conceptual analysis versus inference to the best explanation), or different explanatory strategies (e.g. constitutive versus causal explanation). I will argue that none of these strategies work: both the Metaphysical and Psychological debates rely on inference to the best explanation to draw contingent conclusions about the constitutive nature of perceptual experience. Finally, I will consider and reject the suggestion that the relevant distinction is between personal and subpersonal explanations. Even if this characterizes some of the differences between Metaphysical and Psychological debates, I will argue that it does not establish that the two debates are orthogonal, and it leaves open that Psychological theories are addressing the same metaphysical questions as Metaphysical theories.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

217

2. Two Approaches to Theorizing about Perception 2.1 Metaphysical Theories of Perception A key project in the metaphysics of perception is the attempt to specify the nature of perception. With further qualifications to come, the main positions in the familiar Metaphysical debate can be summarized as follows: Naïve realism. Perception is a non-representational relation to mind-independent objects. Intentionalism. Perception is a matter of representing mind-independent objects in certain ways. Sense-Datum Theory. Adverbialism.

Perception is a relation to a mental object.

Perception is an adverbially modified non-relational mental state.

Each of these positions has been put forward as a Metaphysical claim about the nature of perception, but they can also be used specifically to account for the phenomenal character of perceptual experience.³ I will focus in this paper on the metaphysical positions themselves rather than their accounts of phenomenal character. I will be using the terms “perception” and “perceptual experience” interchangeably. The Metaphysical debate about perception is usually taken to include the four positions outlined above. In what follows I will be restricting my attention to naïve realism and intentionalism, which I take to be the two most popular contemporary theories. I will assume that naïve realism and intentionalism are both “direct” theories in the sense that they both allow that we have veridical perceptual experience of the mind-independent world, and they both deny that we do so in virtue of first perceiving some mental object or sense-datum. The main difference between the two theories is in how they understand the perceiver’s connection to the mind-independent world: naïve realists claim that our relation to the worldly objects of our perceptual experience is not mediated by representation, whereas intentionalists claim that we perceive the world by representing it (accurately or otherwise) as being a particular way.⁴

³ These different roles are captured by Crane and French’s (2017) distinction between two levels at which a perceptual theory can operate, according to which Level 1 theories tell us about the nature or structure of experience, and Level 2 theories tell us how Level 1 theories account for phenomenal character. See also Fish’s (2009) distinction between the ontological and phenomenal formulations of naïve realism. ⁴ There are naïve realists who argue that intentionalism should be considered “indirect” in the sense that is not essentially world-involving, because we can be in the same type of representational perceptual state whether the relevant worldly object is present or not. Further discussion of the matter

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

218

 

2.2 Psychological Theories of Perception The science of perceptual psychology, and vision in particular, has been dominated over the past fifty years by a debate between two theoretical approaches: constructivist and ecological.⁵ It is widely known that the data on our retina are compatible with an infinite number of distinct percepts: one pattern of retinal data could be caused by two objects of different sizes at the same distance from us, for example, or by two objects of the same size at different distances from us. (In vision science, this is also known as the ‘inverse problem’ of optics.) In each case, however, we experience just one determinate scenario, rather than a set of underdetermined possibilities. Constructivist and ecological theories propose very different ways to account for this. Proponents of constructivist theories acknowledge that our retinal data underdetermine our perceptual experience. They propose that our minds supplement the sparse retinal information with additional stored information, allowing us to reconstruct which of the many compatible scenes is actually causing our experience. The earliest constructivist theory of visual perception was put forward by Helmholtz (1878), who argued that when we perceive, we draw ‘unconscious inferences’ from the retinal data to their source: our perceptual experiences are the conclusions of these inferences. Any notion of inference requires positing representations, in the sense of semantically evaluable states which function as the premises and conclusions of the inferences.⁶ Constructivist theories take many different forms: computational models which recreate the causes of sensory input bottom-up look very different from top-down computational models which use stored priors to predict the probability of sensory inputs, and Helmholtz’s own model predated computational approaches to the mind. What they all share is the basic constructivist commitment that we perceive the world by drawing inferences from our sensory input. Constructivist theories can be contrasted with ecological theories of perception. Proponents of ecological theories deny that perception faces an underdetermination problem, because they do not take our sensory input to be restricted to static data on individual sense organs (e.g. the retina). They broaden the notion of sensory input to include dynamic information, such as the way that the ambient light array changes as we move. Ecological theories of perceptual psychology were first proposed by Gibson (1967, 1979), who argued that dynamic information can be found in McDowell (1998), Fish (2004), and Drayson (2018). (Some of the claims I make in the current paper diverge from those in the 2018 paper.) ⁵ In what follows I will frame the difference between constructivist and ecological approaches in terms of visual perception, but the two positions apply to perception in general and can be applied to all its modalities. Constructivist and ecological approaches tell very different stories about haptic touch, for example. ⁶ Constructivist theories differ as to whether they understand the representations as propositional contents, formal symbols, or imagistic models.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

219

from the light array can account for our determinate perceptual experience in a way that static retinal data cannot, and without any need for inference. Gibson proposed that when a perceiver actively explores their environment, some aspects of the light array will change while others will remain invariant. These invariances supposedly provide information about the world which the perceiver can detect without any need for inference. Ecological theories take a non-representational approach to perception: without the need for inference, there is no need to posit the sorts of representations over which inference would take place. There is, of course, much more to the psychology of perception than what I am calling the Psychological debate. My interest here is in the debate between constructivist and ecological theories in general, rather than in different versions of them, or in the psychology of perception more generally.

2.3 Comparing the Metaphysical and Psychological Debates There is a prima facie resemblance between the Metaphysical and Psychological debates about perception. It has been noted that the differences between constructivist and ecological theories ‘echo, to some extent, disagreements amongst contemporary philosophers of perception’ (Lowe 2000, 131) concerning naïve realism and intentionalism. Both the Metaphysical and Psychological debates concern whether or not our perceptual experience of the world is mediated by some form of representation. But proponents of Metaphysical theories like naïve realism and intentionalism generally take themselves to be engaged in a different debate from proponents of Psychological theories. In the rest of this paper, I will explore the various ways we might try to distinguish between the two debates. First, one might look at the modal strength of the theories: are Metaphysical theories of perception proposing necessary truths, for example, and are Psychological theories of perception proposing contingent truths (3.1)? Second, one might consider the methodology which produces the theories in question: are Metaphysical theories the result of a priori conceptual analysis, for example, while Psychological theories use empirical evidence to draw an inference to the best explanation (3.2)? I will argue that proponents of both Metaphysical and Psychological theories use inference to the best explanation to draw contingent conclusions about the nature of perception, and that neither strategy relies on empirical evidence. Third, I explore whether we can distinguish between two different approaches to inference to the best explanation: perhaps Metaphysical theories concern constitutive explanation while Psychological theories concern causal explanation (3.3)? I will argue that both Metaphysical and Psychological theories of perception should be understood as constitutive explanations of psychological capacities rather than causal explanations of psychological events. I then consider whether we might understand the two debates as concerned with

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

220

 

different ‘levels’ of constitutive explanation, using the personal/subpersonal distinction (3.4). I think this is perhaps the most promising strategy for highlighting how some Metaphysical and Psychological theories differ, but I will argue that it fails to demonstrate that the two debates are in any sense orthogonal to each other.

3. Distinguishing the Metaphysical and Psychological Debates 3.1 Is There a Difference in the Modal Strength of the Theories? Philosophers have traditionally separated philosophical claims from scientific claims in terms of their modal status, arguing that science gives us facts about how the world actually is, while philosophy delivers truths about how the world could be or must be. Many philosophers understand metaphysical truths to be true of necessity: metaphysical truths are thus assumed to be truths that hold in all possible worlds.⁷ If we apply this framework to the case of perception, we might be tempted to assume that Metaphysical theories propose necessary truths about properties that perception has in all possible worlds, while Psychological theories merely offer us contingent truths about properties that perception has in the actual world. There are several reasons, however, to think that this modal framework will not help us to distinguish between Metaphysical and Psychological theories of perception. Most importantly, this is not an accurate depiction of metaphysical and scientific claims more generally: there are metaphysical truths which are not necessary, and there are scientific truths which are not contingent. Many metaphysicians acknowledge that a metaphysical model of the world ‘does not need to rely on claims involving necessity’ (Paul 2012, 15), and allow that at least some metaphysical claims about the nature of the world are merely contingently true.⁸ Conversely, scientifically discovered truths (such as the those concerning the molecular structure of water and the atomic number of gold) are often claimed to be necessary truths. Even if metaphysical and scientific claims more generally could be categorized respectively as necessary and contingent, this would not help us to understand the difference between Metaphysical and Psychological theories of perception. Notice that most Metaphysical theories of perception are not framed in terms of necessity. Very few naïve realists can be found explicitly committing to the claim that perception is necessarily an unmediated relation to worldly objects, and even ⁷ Even contingentists acknowledge that it is a widely held belief that ‘metaphysical truths are not just truths about our world, but are truths about every world: they are metaphysically necessary’ (Miller 2009, 23). ⁸ Paul emphasizes that understanding the nature of the world is the sort of project that ‘may take its claims about the world to be contingently true in the actual world and worlds relevantly similar to our world’ (Paul 2012, 8).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

221

fewer intentionalists explicitly claim that perception is necessarily a matter of representing the world a certain way. (There is a subset of naïve realists who do sometimes appear to be using transcendental arguments to make stronger modal claims. I will not address them in this paper, because their necessity claims do not concern the mind-independent world and thus cannot establish the version of naïve realism under consideration here.⁹) A standard way to argue for a metaphysically necessary truth is to who show that its negation is metaphysically impossible, but we do not generally find proponents of Metaphysical theories of perception characterizing the negation of their view as impossible, inconceivable, or leading to contradiction.¹⁰ In the absence of explicit modal commitments, we might nevertheless have grounds for interpreting Metaphysical theories of perception as necessity claims on the basis of the arguments involved: strategies relying on conceptual analysis, two-dimensional semantics, the logic of counterfactual conditionals or the cognitive ability to detect metaphysically necessary truths, for example, are associated with the epistemology of modality. But the Metaphysical debate about perception rarely appeals to such strategies, either a priori or a posteriori. One might think that philosophers of perception ought to be making stronger modal claims about the nature of perception than they do in fact make. In this paper I take no stance on this matter: my aim here is to give an accurate characterization of how proponents of Metaphysical theories describe their own views. To understand why Metaphysical theories of perception are rarely framed as necessary truths, we can take a brief detour into the history of the debate. Each of the Metaphysical theories of perception (intentionalism, naïve realism, sensedatum theories, and adverbialism) was introduced to show how perceptual experience is possible in the face of the worry that perception, as we ordinarily understand it, might be impossible. Also known as the ‘Problem of Perception’, this worry arises when we try to reconcile the apparent properties of perception with each other. On one hand, our ordinary perceptual experiences seem to give us access to mind-independent features of the world. But on the other hand, at least some of our perceptual experiences are illusory or hallucinatory and thus ⁹ Some naïve realists use transcendental arguments to claim that certain features of our perceptual phenomenology necessitate that we are directly (non-representationally) related to worldly objects. But naïve realism is a position about the mind-independent objects of perception, and it is widely acknowledged that transcendental arguments cannot provide us with necessary truths about the mind-independent world rather than how we take the world to be (Stroud 1968). In a recent defence of transcendental naïve realism, Allen (2020) acknowledges this point and focuses on how a transcendental approach can reveal how naïve realism is possible, or how it is practically (rather than rationally) compelling. See also Allen (this volume) for discussion of different ways to interpret the naïve realist position. ¹⁰ When Lowe argues against naïve realism, for example, he emphasizes that he is not trying to show that naïve realism is incoherent, absurd, inconsistent, or inherently impossible. He explicitly proposes his own alternative theory as a contingent matter of fact rather than a necessary truth (Lowe 1981).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

222

 

inaccurate with respect to the mind-independent features of the world. The Problem of Perception is the problem of providing an account of perceptual experience that can make sense of both of these features: perception’s ability to reveal the world to us, and its ability to mislead us.¹¹ The Problem of Perception is that if illusions and hallucinations are possible, then perception, as we ordinarily understand it, is impossible. [ . . . ] if these kinds of error are possible, how can perception be what we ordinarily understand it to be, an openness to and awareness of the world? (Crane and French 2017)

I propose, therefore, that Metaphysical theories of perception should initially be understood as theories of how perception might possibly be: they are attempts to offer an intelligible account of perception which solves or dissolves the Problem of Perception. There is nothing to prevent one from trying to solve the Problem of Perception with claims of metaphysical necessity, but since solving the problem requires only claims of possibility, we should not expect to find Metaphysical theories of perception standardly proposed as claims of metaphysical necessity. Contemporary philosophers of perception, however, generally attempt to do more than demonstrate that a particular theory of perception is possible. The debate over Metaphysical theories of perception seems to be concerned with figuring out which of these theories of perception is the right one: intentionalists, for example, take their theory to be correct and naïve realism to be wrong; while naïve realists take their own view to be correct and intentionalism to be wrong. In this respect, the Metaphysical debate appears to be strikingly similar to the Psychological debate, which focuses on which possible account of perception (ecological or constructivist) is preferable. Psychological theories of perception are also proposed as contingent rather than necessary: their proponents allow that in a different sort of environment, our perceptual systems could have evolved differently. There is no obvious justification, therefore, for interpreting the Metaphysical and Psychological theories of perception as differing in the modal strength of their claims.

3.2 Do the Theories Result from Different Methodologies? Scientific and philosophical theories have often been supposed to differ in methodology as well as modal strength—particularly prior to Quine’s (1951) rejection of the analytic/synthetic distinction. Philosophical methodology is sometimes

¹¹ For further discussion of the supposed conflict between different features of perceptual experience and how it challenges the possibility of perception, see Crane (2005).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

223

characterized as employing a priori reflection on our concepts, for example, while scientific methodology uses a posteriori means to test the empirical predictions of our scientific hypotheses. Might we distinguish between the Psychological and Metaphysical debates about perception along these methodological lines? An obvious problem with this approach is that, post-Quine, philosophers generally agree that there is nothing essentially a priori about conceptual analysis. Conversely, science does not always proceed by empirical inquiry: scientists often compare and evaluate theories which make the same empirical predictions. They do so by drawing an inference to the best explanation, taking into account theoretical virtues such as simplicity, parsimony, fruitfulness, and compatibility with other theories, in addition to virtues such as empirical adequacy and predictive strength. Not all scientific theories make empirical predictions: some use hypothetical toy models to capture aspects of the world that are ‘often unobservable, indirectly confirmable, and abstract’ (Paul 2012, 9). Even if one were to reject Quinean naturalism, however, it would be difficult to find a methodological distinction between the Metaphysical and Psychological debates about perception. Notice that Psychological theories of perception are not generally engaged in making empirical predictions: constructivist and ecological theories are examples of the toy models mentioned above, which don’t make observable or directly confirmable predictions. And as I have already observed, proponents of Metaphysical theories are not generally engaged in analysing the concept of perception into necessary and sufficient conditions. They are instead putting forward possible theories of perception and debating their relative merits, in a process which resembles the methodology of inference to the best explanation.¹² This methodology is explicitly endorsed in the literature. When the naïve realist, for example, argues that perception is a non-representational relation to the world, they are claiming that their theory offers a better explanation of certain perceptual features than the rival theories: What reason is there for thinking that naïve realism about visual experiences is true? The NR [naïve realist] claims that the best explanation of the fact that visual experiences introspectively seem to have the NR property [ . . . ] is that veridical experiences actually do have it: having the NR property explains the way visual experiences introspectively seem. (Nudds 2009, 335, my italics)

Those who reject naïve realism tend to present themselves as offering a ‘a better explanation’ than naïve realism of perceptual experience, often appealing to

¹² There is nothing singular in this respect about perceptual metaphysics: many metaphysical arguments rely on inference to the best explanation. For further discussion, see Hawley (2006), Paul (2012), Williamson (2016).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

224

 

illusions as ‘powerful abductive support’ for intentionalism (Philips 2016, 355, author’s italics). Cavedon-Taylor, for example, explicitly acknowledges in his own proposal for an intentionalist claim that his ‘argument for this claim is an inference to the best explanation’ (Cavedon-Taylor 2018, 391, my italics). Proponents of Metaphysical theories of perception can generally be understood as employing inference to the best explanation. Unsurprisingly, Psychological theories of perception are also the result of inference to the best explanation: constructivists propose that the best way to account for perception is to assume that our sensory input is supplemented by stored information and inferential processing; while ecological theorists propose that perception is best explained by our active role in our sensory environments, which removes any role for inference. We cannot easily distinguish between the Metaphysical and Psychological debates, therefore, by appealing to their broad methodological approaches.¹³ One might object that there is a more fine-grained methodological difference between the two debates. Despite my characterization of constructivist and ecological theories as hypothetical toy models, one might think that Psychological theories have a place for empirical evidence which Metaphysical theories lack. But any characterization of the Metaphysical debate as non-empirical seems to be at odds with the very idea of inference to the best explanation, which is usually understood as sensitive to all the available evidence (empirical or otherwise). When inference to the best explanation is used to construct and select between philosophical theories, presumably the evidence base should include all of our relevant knowledge, just as it does when it applies to scientific theorizing. There should be, in particular, ‘no restriction to knowledge gained in some special “conceptual” or “a priori” or “intuitional” or “armchair” way’ (Williamson 2016, 268). It looks like we cannot appeal to any peculiarly philosophical methodology to distinguish Metaphysical theories of perception from Psychological theories. Both use the methodology of inference to the best explanation, and neither makes easily testable claims about empirical matters of fact. A more fruitful way to address the distinction might be to allow that both debates are using inference to the best explanation, but with a different kind of explanation. Next I will consider whether Metaphysical theories are proposing an inference to the best constitutive explanation and Psychological theories are proposing an inference to the best causal explanation.

¹³ There may, however, be a way to distinguish the transcendental arguments mentioned previously from inferences to the best explanation. See Gomes (2017) and Allen (2020) for discussion of the relationship between transcendental argument and inference to the best explanation in the philosophy of perception.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

225

3.3 Is There a Difference in Whether the Explanations Are Causal or Constitutive? Some philosophers have been inclined to see the distinction between metaphysics and science as a difference between constitutive and causal explanations. Many metaphysical explanations are indeed constitutive: they account for something’s existence or occurrence by appealing to the kind of thing it is or consists in, that in virtue of which it has certain properties. Constitutive explanations like these are synchronic: if a change to an object or event is explained in terms of a change to its constitutive basis, these changes occur at the same instant rather than over a period of time. The explanandum and explanans of constitutive explanations are not independent existences. Many scientific explanations, on the other hand, are causal: they account for something’s existence or occurrence by appealing to the etiological sequence of events which brought it about. Causal explanations are diachronic (i.e. they take time), and the cause (explanans) and effect (explanandum) are distinct existences.¹⁴ Could we argue, therefore, that both Metaphysical and Psychological approaches to perception use the same broad strategy of inference to the best explanation, but applied to constitutive and causal explanations respectively? When we look at the Metaphysical debate about perception, it certainly seems to be primarily concerned with the constitution of perception: naïve realism is characterized as the claim that perceptual experience consists in being related to mind-independent objects; while intentionalism is the view that we perceive in virtue of representing the world in certain ways. These Metaphysical theories are proposed as synchronic explanations of what perception consists in, rather than diachronic explanations of the temporal antecedents to perceptual experience: ‘that in virtue of which it [perceptual experience] has all the other psychological properties it does’ (Logue 2011, 269, my italics); ‘the kind in virtue of which [it] has the nature it does’ (Martin 2004, 60, my italics).¹⁵ Further evidence of this commitment to constitutive explanation can be found throughout the philosophy of perception: for a representative sample see Brewer (2011), Stazicker (2015), Crane and French (2017), and Brogaard (2018). We should therefore understand naïve realism, intentionalism, and other Metaphysical theories of perception as offering constitutive explanations of perceptual experience.¹⁶ ¹⁴ See Ylikoski (2013) for a fuller discussion of the differences between constitutive and causal explanations. I am not taking any particular stance on the nature of causation or constitution or the dependence relations involved. I remain neutral as to the relation between constitution and identity. ¹⁵ I take it that claims about the constitution of perception need not be understood as claims of metaphysical necessity. As Leuenberger (2014) and others have argued, metaphysical determination relations can be contingent. ¹⁶ There is no tension between saying that an explanation is both constitutive and an inference to the best explanation. The methodology of inference to the best explanation is widely understood to apply to both causal and non-causal explanation: Williamson (2016) gives the example of Newton’s laws, which

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

226

 

If Psychological theories of perception offer causal (non-constitutive) explanations, then we would have a way to contrast the Metaphysical and Psychological debates. Some philosophers of perception do indeed characterize the debate in this way, proposing that scientific psychology provides information only about the causes of perceptual experience. McDowell, for example, claims that science addresses the ‘causal or enabling question’ of perception, and cautions us to resist the ‘temptation to suppose that the question is rather a constitutive one’ (McDowell 1994, 199). This is a controversial way to characterize the debate, however, as I will now argue: the mind-sciences are concerned primarily with explaining our cognitive capacities (rather than particular events or behaviours) and explanations of capacities are usually understood as constitutive (rather than causal). The mind-sciences are not primarily engaged in explaining particular events, such as single instances of cognition or behaviour. Instead, their aim is ‘to explain the human cognitive capacities—what they are, how they are exercised, in virtue of what we have them, and how they interact’ (Von Eckardt 1993, 258). When cognitive scientists, for example, offer explanations of how we are able to comprehend language, to attribute mental states to others, or to perceive the world around us, they are seeking to account in each case for a particular cognitive capacity. There is a general consensus in the literature that the explananda of the mind-sciences are cognitive capacities: see Wallis (1994), Rupert (2009), Samuels, Margolis, and Stich (2012), Miracchi (2017) for further evidence. Explanations of capacities are constitutive explanations: they are not diachronic explanations of one event or phenomenon in terms of the causal processes leading to it, but rather synchronic explanations concerning what gives rise to the capacity: what it consists in. In this respect, explanations of capacities are similar to explanations of abilities, dispositions, and propensities (Ylikoski 2013). When we try to account for the fragility of a vase, for example, we are not usually trying to give an etiological explanation of how the vase came to acquire the property of fragility. Instead, we are trying to say what the vase’s fragility consists in: which are the properties of the vase in virtue of which it has the simultaneous property of being fragile? Something similar is going on in our explanations of cognitive capacities, when we ask what our ability to attribute mental states to others consists in and in virtue of what we can comprehend language.¹⁷ provided the best explanation of Kepler’s laws, but certainly not because Newton’s laws caused Kepler’s laws. ¹⁷ It is not particularly surprising to learn that scientific explanations can be constitutive explanations, because this is not confined to the mind-sciences. Facts about water’s behaviour at different temperatures, for example, can be scientifically explained in virtue of its molecular properties, which in turn can be scientifically explained in virtue of atomic properties and physical forces. Science tells us what atoms, molecules, ecosystems, and weather fronts are, not merely what causes them. There has been much interest of late in the role played by non-causal (including constitutive) explanations in science: for the current state of play, see Reutlinger and Saatsi (2018).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

227

Even when scientific psychologists do want to give a causal explanation of the source of a particular token mental state (e.g. a belief or a perceptual experience), the methodology is usually one of explaining how that token mental state can be a manifestation of a mental capacity.¹⁸ The primary focus of the mind-sciences is not the causal explanation of individual mental states, but rather the constitutive explanation of mental capacities: what they are, what they consist in, that in virtue of which we have them. This general lesson transfers to Psychological theories of perception, and is presumably what motivates Burge to claim that constructivist Psychological theories of perception are offering constitutive explanations: he takes it that the aim of perceptual psychology is to explain the structure of perception, and not merely to provide enabling conditions of perception (Burge 2005, 21). (See Fish, this volume, for discussion of the debate between Burge and McDowell on constitutive and causal-enabling explanations.)¹⁹ It seems plausible, therefore, that Psychological explanations of perception are attempts to give a constitutive account of our capacity to perceive. If this is correct, then both the Metaphysical and Psychological debates about perception involve inferences to the best constitutive explanation.

3.4 Are There Different Kinds of Constitutive Explanation? Even if both Psychological and Metaphysical theories offer similarly constitutive explanations of perceptual experience, it might be possible to distinguish them by appealing to two different kinds of constitutive explanation. We might explore, for example, whether Metaphysical and Psychological theories are relying on different concepts of constitution, and whether these concepts pick out different kinds of dependence relations.²⁰ But notice that even if there are different notions of constitution, this does not entail that only one of them has a place in metaphysical explanations. Some philosophers of perception seem to suggest that the difference between the Metaphysical and Psychological debates is in the “level” of explanation concerned. There is a distinction in philosophy of mind between personal and subpersonal levels, which corresponds to the following two ways of ascribing mental states. When we speak of someone calculating a sum, representing a stick in water as bent, or predicting the outcome of an election, we are attributing the calculations, representations, and predictions to the person. To give an explanation of mental phenomena by attributing psychological states to persons ¹⁸ See Wallis (1994), Rupert (2009). ¹⁹ One might argue we nevertheless ought to draw the distinction between metaphysical and scientific explanations in constitutive/causal terms, but this would be a particularly revisionary claim. ²⁰ See Harbecke (2016) for a discussion of the relationship between material constitution and mechanistic constitution, for example.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

228

 

in this way is to give a personal-level explanation. In the mind-sciences, however, it is common to ascribe the same psychological terms to cognitive subsystems of the person: we might speak of the visual system calculating the Laplacian of the Gaussian, a neural structure representing left-to-right motion, or a Bayesian network predicting the next input. To give an explanation of phenomena by attributing psychological states to subsystems of the person like this is to give a subpersonal-level explanation. Metaphysical theories of perception appear to offer personal-level explanations: they account for perceptual experience by appealing to how the person relates to or represents the world. Some philosophers take this feature to be definitive of metaphysical theories of perception: ‘What my experience fundamentally consists in (i.e. its metaphysical structure) is that which provides the ultimate personal-level psychological explanation of the phenomenal, epistemological and behavioural facts’ (Logue 2012, 212). Logue contrasts personal-level theories such as naïve realism with scientific theories which appeal to ‘subpersonal psychological facts (e.g. the perceptual processing in the brain that takes place between stimulation of the sensory organs and experience)’ (Logue 2012, 212). This description seems to accurately capture some Psychological theories of perception: many computational constructivist theories, for example, ascribe representations of low-level visual features to the person’s early visual system rather than to the person. But not all constructivist theories offer subpersonallevel explanations: in Helmholtz’s (1878) original constructivist theory, it is the person who draws the inferences (albeit unconsciously) rather than some visual subsystem. Furthermore, ecological theories are particularly difficult to characterize as subpersonal-level explanations of perception, because they do not attribute representations to any part of the cognitive system.²¹ So we cannot easily map the personal/subpersonal distinction onto the distinction between Metaphysical and Psychological theories of perception. For the sake of argument, however, I wish to grant that there may be some way to make sense of Metaphysical theories as personal and Psychological theories as subpersonal. Acknowledging this, however, would not yet explain why proponents of Metaphysical theories of perception take themselves to be engaged in a project which is separate from and orthogonal to the Psychological debate. When

²¹ Notice that not all ways of explaining the brain are subpersonal-level explanations. Subpersonal explanation is a kind of psychological explanation which ascribes psychological states (e.g. calculations, representations, prediction) to subsystems of the person. A neural explanation in terms of firing rate or neurotransmitter uptake is not a subpersonal psychological explanation, but rather a neurological (non-psychological) explanation. The particular usefulness of subpersonal explanations comes from their being psychological explanations: they allow us to ascribe contentful mental states where the person would not necessarily endorse the content or even have the concepts to express it. The use of subpersonal explanation is consistent with different ontological interpretations: subpersonal explanations might refer to genuinely contentful mental states, some deflationary kind of computational processing, or nothing at all.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

229

two explanations of one phenomenon focus on different properties or levels of generality, we cannot generally conclude that the resulting explanations are independent from each other in any interesting sense. Consider the case of a chemist and a physicist giving constitutive explanations of properties of water by appealing respectively to its molecular and atomic structure. We do not assume that the atomic explanation is unconnected to the molecular explanation; in fact, we might even think that the molecular explanation holds in virtue of the atomic explanation. There are many philosophers who think that personal and subpersonal psychological explanations stand in a comparable relationship to molecular and atomic explanations, and who propose that personal-level explanations are reducible to subpersonal explanations.²² Those philosophers who think that the Psychological debate is orthogonal to the Metaphysical debate seem to be making a certain assumption about the relationship between personal and subpersonal explanation. They believe that personal-level explanations of perceptual experience play a privileged role in providing a metaphysical explanation of the nature of perception, and that these personal-level explanations are independent from subpersonal-level explanations. This is a familiar position in philosophy of mind, most commonly associated with the Sellarsian idea that explanations of mentality belong to a normative ‘space of reasons’, and not with our descriptive explanations of the scientific world.²³ On this view, personal-level psychological explanations are constrained by norms of rationality, which makes them autonomous from and irreducible to the causal explanations offered by the mind-sciences.²⁴ The Sellarsian position, however, is just one way to interpret the relation between personal and subpersonal explanations. As I have already observed, there are philosophers who think that personal-level explanations can be neatly reduced to subpersonal-level explanations, and there are alternatives beyond these two approaches. Those taking a Sellarsian approach to perception, however, seem to think that only personal-level explanations provide metaphysical explanations; subpersonal explanations are dismissed as belonging to some distinct and nonmetaphysical project. I have no interest here in settling whether personal-level explanations are autonomous from or reducible to subpersonal explanations: I’m simply suggesting that the question itself is a matter for metaphysical discussion, and recommending that metaphysicians of perception should at least consider the possibility that reductive theories can still be metaphysical theories. ²² This is probably the default view among philosophers of cognitive science. It has been articulated as ‘homuncular functionalism’ by Lycan (1987); see also Fodor (1975). ²³ The space of reasons is introduced in Sellars (1956). For further discussion of the Sellarsian interpretation of the personal/subpersonal distinction, see Drayson (2014). ²⁴ The Sellarsian position sits uneasily with work in scientific psychology which gives subpersonal content an explanatory role. By characterizing subpersonal explanation as merely causal, these philosophers must interpret their content ascriptions as purely instrumental and therefore non-causal. For more on different approaches to the personal/subpersonal distinction, see Drayson (2012, 2014, 2017).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

230

 

4. Conclusion I have argued that Psychological theories of perception, such as ecological and constructivist theories, are not clearly distinguishable in terms of their modal strength or methodology from Metaphysical theories of perception, such as naïve realism and intentionalism. I have not tried to provide an exhaustive refutation of all possible ways to distinguish the two projects. I happily acknowledge that there are differences between the two projects: differences in how the explanandum is framed, for example, and differences in the motivations and background assumptions of the theorists.²⁵ But I think that we should be wary of assuming that these differences make the two projects orthogonal to or independent from each other.²⁶ I propose that both Metaphysical and Psychological theories of perception can be understood as trying to answer the same metaphysical question about the nature of perception: what is it that constitutes perception? Notably, my argument for this does not rely on any taking any particularly naturalistic stance. I have not denied a role for a priori theorizing, for example, or claimed that our scientific judgements are evidentially stronger than our intuitions. I have pointed out that that there is an ongoing debate about the role that scientific theories play in metaphysics, but I have not taken a side in this debate. In this paper, I am merely suggesting that we should acknowledge the possibility of more naturalistic metaphysical positions within the philosophy of perception, instead of ruling that those positions are engaged in a completely different project. Why do some philosophers of perception deny Psychological theories a place in the debate over the metaphysical nature of perception? I do not that think simple anti-naturalism is to blame: I suspect that many people who deny a role for scientific theories in the metaphysics of perception would happily concede that being H₂O is a metaphysical fact about water, and that having atomic number 79 is a metaphysical fact about gold. Philosophers might allow that water and gold can have scientifically discovered metaphysical properties, but deny that the same is true of mental phenomena. I think that this is the role played by the Sellarsian interpretation of the personal/subpersonal distinction in the metaphysics of perception, with the resulting claim that perceptual experience, as a mental phenomenon, is governed by different explanatory norms than water or gold. I won’t attempt to argue against such approaches here. My aim is simply to highlight that other metaphysical positions are available, and that the philosophy ²⁵ See Fish (this volume) for discussion of the differences between the projects of Burge and McDowell, for example, with respect to whether explanations of perception must account for conscious experience. (I am assuming that both Metaphysical and Psychological theories are concerned with explaining conscious perceptual experience, and that both should allow for the metaphysical possibility of unconscious perception.) ²⁶ A theory which explains the boiling point of water has a different explanandum from a theory which explains the pH value of water, for example, but we would not conclude from that the theories are orthogonal.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

231

of perception seems to be an outlier in its reluctance to accommodate them. In the metaphysics of mind more generally, there is an active debate about whether personal-level explanations are autonomous from, reducible to, or eliminated by subpersonal-level explanations.²⁷ In the metaphysics of perception, however, those who take the subpersonal level to play a role in determining the nature of perception are sometimes considered, as a result, not to be engaging with the metaphysical debate. I think that we do a disservice to both philosophy and science if we assume without argument that the metaphysics of mind must be conducted solely at one particular level of explanation.²⁸

References Allen, Keith. 2020. ‘The Value of Perception’. Philosophy and Phenomenological Research. 100 (3): 633–56. Brewer, Bill. 2011. Perception and Its Objects. Oxford: Oxford University Press. Brogaard, Berit. 2018. Seeing and Saying: The Language of Perception and the Representational View of Experience. New York: Oxford University Press. Burge, Tyler. 2005. ‘Disjunctivism and perceptual psychology’. Philosophical Topics 33 (1): 1–78. Cavedon-Taylor, Dan. 2018. ‘Naïve Realism and the Cognitive Penetrability of Perception’. Analytic Philosophy 59 (3): 391–412. Chakravartty, Anjan. 2017. Scientific Ontology: Integrating Naturalized Metaphysics and Voluntarist Epistemology. New York: Oxford University Press. Crane, Tim. 2005. ‘What Is the Problem of Perception?’ Synthesis Philosophica 20 (2): 237–64. Crane, Tim and French, Craig. 2017. ‘The Problem of Perception’. In The Stanford Encyclopedia of Philosophy (Spring 2017 Edition), Edward N. Zalta (ed.). https:// plato.stanford.edu/archives/spr2017/entries/perception-problem/ Drayson, Zoe. 2012. ‘The uses and abuses of the personal/subpersonal distinction’. Philosophical Perspectives 26 (1): 1–18. Drayson, Zoe. 2014. ‘The Personal/Subpersonal Distinction’. Philosophy Compass 9 (5): 338–46. ²⁷ There are further discussions to be had about the relationship between subpersonal-level psychological explanations and lower-level neurological explanations. ²⁸ The original version of this paper was presented at a workshop on naturalism in the philosophy of perception at the University of York in March 2017. Subsequent versions have been presented at UC San Diego, Stanford University, Macquarie University, the University of Wollongong, and the University of Chicago. I also presented this paper at the Southern Society for Philosophy and Psychology (Cincinnati, 2018) with commentary from Andrew Lee. I have benefited from the feedback I received on each of these occasions, and from discussion with Arieh Schwartz on the final draft. I am particularly grateful to Heather Logue, Clare Mac Cumhaill, and Louise Richardson for their detailed and helpful comments on earlier drafts.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

232

 

Drayson, Zoe. 2017. ‘Psychology, personal and subpersonal’. In The Routledge Encyclopedia of Philosophy. Taylor and Francis. Retrieved 24 Jul. 2020, from https:// www.rep.routledge.com/articles/thematic/psychology-personal-and-subpersonal/v-1. doi:10.4324/9780415249126-V044-1 Drayson, Zoe. 2018. ‘Direct perception and the predictive mind’. Philosophical Studies 175 (12): 3145–64. Fish, William. 2004. ‘The Direct/Indirect Distinction in Contemporary Philosophy of Perception’. Essays in Philosophy 5 (1): Article 5. Fish, William. 2009. Perception, Hallucination, and Illusion. New York: Oxford University Press. Fodor, J. A. 1975. The Language of Thought. Cambridge, MA: Harvard University Press. French, Steven and McKenzie, Kerry. 2012. ‘Thinking Outside the Toolbox: Towards a More Productive Engagement Between Metaphysics and Philosophy of Physics’. European Journal of Analytic Philosophy 8 (1): 42–59. Gibson, James J. 1967. ‘New reasons for realism’. Synthese 17 (1): 162–72. Gibson, James J. 1979. The Ecological Approach to Visual Perception. Boston, MA: Houghton Mifflin. Gomes, Anil. 2017. ‘Perception and reflection’. Philosophical Perspectives 31 (1): 131–52. Harbecke, J. 2016. ‘Is Mechanistic Constitution a Version of Material Constitution?’ In K. Aizawa and C. Gillett (eds.) Scientific Composition and Metaphysical Grounding. London: Palgrave-Macmillan, 91–121. Hawley, Katherine. 2006. ‘Science as a Guide to Metaphysics?’ Synthese 149 (3): 451–70. Helmholtz, Hermann. 1878. ‘The Facts in Perception’. In R. Kahl (ed.) Selected Writings of Hermann Helmholtz. Middletown, CT: Wesleyan University Press. Ladyman, James and Ross, Don. 2007. Every Thing Must Go: Metaphysics Naturalized. Oxford: Oxford University Press. Leuenberger, Stephan. 2014. ‘Grounding and Necessity’. Inquiry: An Interdisciplinary Journal of Philosophy 57 (2): 151–74. Logue, Heather. 2011. ‘The skeptic and the naïve realist’. Philosophical Issues 21 (1): 268–88. Logue, Heather. 2012. ‘Why Naive Realism?’ Proceedings of the Aristotelian Society 112 (2pt2): 211–37. Lowe, E. J. 1981. ‘Indirect perception and sense data’. Philosophical Quarterly 31 (October): 330–42. Lowe, E. J. 2000. An Introduction to the Philosophy of Mind. Cambridge: Cambridge University Press. Lycan, William G. 1987. Consciousness. Cambridge, MA: MIT Press. Martin, Michael G. F. 2004. ‘The limits of self-awareness’. Philosophical Studies 120 (1–3): 37–89.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

233

McDowell, John. 1994. ‘The content of perceptual experience’. Philosophical Quarterly 44 (175): 190–205. McDowell, John. 1998. Meaning, Knowledge, and Reality. Cambridge, MA: Harvard University Press. McDowell, John. 2010. ‘Tyler Burge on Disjunctivism’. Philosophical Explorations 13 (3): 243–55. Miller, Kristie. 2009. ‘Defending Contingentism in Metaphysics’. dialectica 63 (1): 23–49. Miracchi, Lisa. 2017. ‘Generative explanation in cognitive science and the hard problem of consciousness’. Philosophical Perspectives 31 (1): 267–91. Nudds, Matthew. 2009. ‘Recent work in perception: Naïve realism and its opponents’. Analysis 69 (2): 334–46. Paul, L. A. 2012. ‘Metaphysics as modeling: the handmaiden’s tale’. Philosophical Studies 160 (1): 1–29. Phillips, Ian. 2016. ‘Naive Realism and the Science of (Some) Illusions’. Philosophical Topics 44 (2): 353–80. Quine, Willard V. O. 1951. ‘Two Dogmas of Empiricism’. Philosophical Review 60 (1): 20–43. Reutlinger, Alexander and Saatsi, Juha (eds.). 2018. Explanation Beyond Causation: Philosophical Perspectives on Non-Causal Explanations. Oxford: Oxford University Press. Rupert, Robert D. 2009. Cognitive Systems and the Extended Mind. New York: Oxford University Press. Samuels, R., Margolis, E., and Stich, Stephen. 2012. ‘Introduction: Philosophy and Cognitive Science’. In Eric Margolis, Richard Samuels, and Stephen Stich (eds.) The Oxford Handbook of Philosophy of Cognitive Science. New York: Oxford University Press, 3–18. Stazicker, James. 2015. ‘Introduction’, in James Stazicker (ed.) The Structure of Perceptual Experience. Malden, MA: Wiley-Blackwell. Stroud, Barry. 1968. ‘Transcendental arguments’. Journal of Philosophy 65 (9): 241–56. Von Eckardt, Barbara. 1993. What is Cognitive Science? Cambridge, MA: MIT Press. Wallis, Charles. 1994. ‘Using representation to explain’. In Eric Dietrich (ed.) Thinking Computers and Virtual Persons. San Diego, CA: Academic Press. Williamson, Timothy. 2016. ‘Abductive Philosophy’. Philosophical Forum 47 (3–4): 263–80. Ylikoski, Petri. 2013. ‘Causal and Constitutive Explanation Compared’. Erkenntnis 78 (2): 277–97.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

12 Phenomenology as Radical Reflection Dave Ward

1. Introduction ‘What is phenomenology?’, asks Merleau-Ponty in the opening sentence of his Phenomenology of Perception (Merleau-Ponty 1945/2012 (henceforth PhP), lxx). The next sentences list a series of apparently antinomial characterizations of phenomenology: as the study of ‘essences’ versus the study of facts; as a transcendental philosophy versus a philosophy based on our empirical contact with the world; as an ‘exact science’ versus an account of the world as we live and experience it; as an attempt to provide a theoretically unmediated characterization of experience versus an attempt to investigate the ‘genetic’ processes via which experience is ‘constructed’ (PhP lxx). Why begin a characterization of phenomenology in this way? Merleau-Ponty’s intention, I will suggest, is neither to show that phenomenology is a hopeless, self-contradictory endeavor, nor to pick between these apparently opposing aspects of phenomenology. Instead, he aims to show that phenomenology, as he pursues it, is defined by the task of understanding our perceptual and cognitive contact with the world in a way that faces up to and attempts to transcend these apparent contradictions. It is defined by this task because, as we shall see, phenomenological philosophy as Merleau-Ponty understands it emerges dialectically from tensions within attempts to understand our perceptual and cognitive contact with the world in ways that emphasize only one side of the antinomies above. This chapter traces that dialectic, with a view to illuminating the distinctive Merleau-Pontian methodology for philosophy of perception that results. As with any good dialectic, the story unfolds by examining a particular way of understanding some domain (here, perceptual experience and its relation to the world), showing that it falls prey to internal tensions or contradictions, then using these tensions to motivate an alternative understanding. For example (in an instance of this dialectic that will be familiar to readers of PhP), Merleau-Ponty argues that ‘empiricist’ accounts of perception as a straightforward record of the impacts of the world on our sensory periphery fail, in a way that motivates ‘intellectualist’ accounts that emphasize the perceiver’s active role in structuring their experience of the world. The dialectical twist in the tale is that the contradictions of the first accounts are only apparently resolved by the second—intellectualist Dave Ward, Phenomenology as Radical Reflection In: Purpose and Procedure in Philosophy of Perception. Edited by: Heather Logue and Louise Richardson, Oxford University Press (2021). © OUP. DOI: 10.1093/oso/9780198853534.003.0012

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

235

accounts succumb to their own internal contradictions, motivating a new kind of account that aims to develop the best bits of empiricist and intellectualist approaches while avoiding their problems. As we will see, an instance of this empiricist/intellectualist dialectic plays out in Merleau-Ponty’s methodological reflections in PhP and throughout his career. The upshot, I argue here, is that Merleau-Ponty understands phenomenology of perception as committed to engaging in radical reflection—a kind of reflection on the structure of perceptual experience that simultaneously ‘attempts to understand itself ’ (PhP 251/288) by understanding its own origins and limitations. Seeing what this means in practice involves understanding how Merleau-Ponty is led to this view via the failures of alternative ways of pursuing phenomenological philosophy. The plan for the next sections is thus to outline various ways of conducting phenomenological philosophy, which I shall label ‘Humean’, ‘Kantian’, and ‘Husserlian’ (§2), and show how the explanatory credentials of each are rendered problematic (§3, §4) in ways that ‘Merleau-Pontian’ phenomenology aims to resolve (§5). As well as affording a clearer vision of the shape of Merleau-Ponty’s philosophy and its motivations, the pay-off of the account I present here is the promise of a richly interdisciplinary way of doing philosophy of perception, according to which it is necessarily integrated with many other branches of philosophy and the sciences.

2. Humean, Kantian, and Husserlian Phenomenology of Perception This section works through three competing ways of pursuing phenomenological philosophy of perception. But what does it mean to adopt a ‘phenomenological’ methodology when attempting to understand perception? To give us something to work with, let’s start with the following broad construal: Phenomenology of perception is the attempt to draw philosophical conclusions about the structure of perception from careful reflection on experience. Talk of the structure of perception should be interpreted broadly, as encompassing any invariant feature of perceptual experiences—for example, the kinds of content, phenomenal character, temporal structure, or relationship to the world, that are found in perceptual experiences, or some well-defined subset of perceptual experiences.¹

¹ In what follows, we will see that this definition of phenomenology needs to be deepened and complicated if phenomenology is to be a viable strategy for doing philosophy of perception. But I nonetheless adopt it for now—seeing the problems it engenders will prove useful for understanding the motivations of Husserlian and Merleau-Pontian phenomenology.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

236

 

On this broad construal, the history of philosophy is filled with phenomenologists. Consider, for example, some views on perception and causation that might be extracted from Hume and Kant. How, if at all, do causal relations between events figure in our perceptual experience? Hume’s sceptical views about causation (as well as induction, the self, and more) are grounded in his reflections on the relations between different elements of our experience. Hume distinguishes between impressions—sensory experiences, including internal sensations of desiring, willing, and feeling—and ideas, experiences resulting from abstracting from, or reflecting on, impressions. This distinction, Hume thinks, is itself phenomenologically obvious—‘as evident as that betwixt feeling and thinking’ (Hume 1740/ 1978, p.647). He goes on to argue that our perceptual awareness of causal relations between events is an idea, rather than an impression of causation. In any given instance where we experience a causal connection between events, we can discern only an impression associated with the perceived cause, an impression associated with the perceived effect, and an idea corresponding to the perceived causal relation between the two. For our purposes here, the most important feature of this argument is that Hume can be understood as simply reading off these conclusions from disciplined reflection on the structure of experience. This reflection first reveals the distinction between Hume’s two basic ingredients of experience, then shows how these ingredients are structured in experiences of causal relations. Read in this way, Hume is an example of what we might call an empirical phenomenologist. Empirical phenomenologists hold that there are empirical facts of the matter about the structure of experience which serve as key premises in philosophical arguments, and that all an aspiring phenomenologist need do to uncover these facts is attend carefully. In what follows, I’ll refer to phenomenologists of this kind as ‘Humean’.² According to the orthodox reading of Hume, the above view of the structure of causal experience reveals that our irresistible compulsion to experience certain contiguous events in terms of causal relations reflects a mere quirk of human psychology, rather than any insight into the metaphysical structure of events. Famously, Kant claimed that these sceptical conclusions awoke him from the ‘dogmatic slumbers’ of rationalist metaphysics. Kant agrees with Hume that the structure of our empirical experience cannot ground the claims to legitimacy of our compulsive application of causal concepts to the world. But he nonetheless argues that this application is justified. His strategy is transcendental—here and elsewhere, he proceeds by attempting to specify structures that perceptual experience and judgement must have in order for it to be intelligible that our experience

² The arguments and methodologies I label as ‘Humean’ and ‘Kantian’ are not intended as highfidelity reconstructions of the positions of Hume and Kant themselves. This way of labelling positions is partly inspired by Husserl’s (1954/1975) own historical origin story for phenomenology, as well as the rhetorical use Merleau-Ponty makes of Hume and Kant.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

237

or thought instantiates a property that it manifestly possesses. We can illustrate this via considering his response to Hume, which aims to show that applying causal concepts to suitably contiguous events is no mere quirk of human psychology, but required as a matter of rational necessity. Perhaps the simplest version of an anti-Humean argument can be extracted from the second ‘Analogy of Experience’ in the Critique of Pure Reason. There Kant argues that the very experience of something happening in the perceiver’s world presupposes the application of causal concepts: I render my subjective synthesis of apprehension objective only by reference to a rule in accordance with which the appearances in their succession, that is, as they happen, are determined by the preceding state. The experience of an event [i.e. of anything as happening] is itself possible only on this assumption. (Kant 1787/1997, A195/B240)

Why? Kant’s reasoning appears to be as follows. Consider two ways in which you might experientially undergo a successive stream of appearances—first, as a Jamesian ‘blooming, buzzing confusion’ in which it seems that the appearances composing your ongoing experience are arbitrary, chaotic, and unstructured; second, as an experience of a perceptible event unfolding in the perceiverindependent world. Kant’s argument relies on the insight that the first type of experience need not present itself to you as disclosing any events that are external to your subjectivity—such a sequence of appearances could be experienced merely as a series of arbitrary and unstructured modifications of that subjectivity. To have an experience of an unfolding external event, you must understand the appearances composing your experience as connected—you must take there to be some rhyme or reason as to why this appearance is succeeded by that one. Kant argues that taking one’s subjective appearances to be connected in this way (i.e. to experience things in terms of ‘a rule in accordance with which the appearances in their succession . . . are determined by the preceding state’ (Kant 1787/1997)) simply is to experience them in terms of their causal connections—that is, in terms of the way in which one appearance necessarily follows from its predecessors. Thus, Kant has argued, having experiences as of external unfolding events presupposes that our experiences are structured by an understanding of how current appearances are determined by past ones; and this simply amounts to understanding the relations between succeeding appearances causally. The argument just sketched counts as a piece of transcendental phenomenology. It is a phenomenological argument (at least on the broad construal we are currently working with) because it reaches its conclusion by starting from a manifest fact about experience—that we have experiences as of things happening. Empirical (or Humean) phenomenology purports to read off structural properties of experience, such as the way in which causation figures in our perception of the world, from the

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

238

 

experiences themselves. Transcendental phenomenology, by contrast, reflects on the conditions of possibility that must be fulfilled in order for experience to manifest itself to us in the way it does, attempting to identify the structural properties of experience that are required for it to possess its manifest properties. Hume argues that empirical reflection on perceptual experience reveals that causation is nothing but an idea that we compulsively associate with appropriately-structured impressions; Kant responds by arguing that transcendental reflection reveals that our experience couldn’t intelligibly possess a property that it manifestly does possess (the property of seeming to disclose the occurrence of perceiver-independent events) without its constituent appearances being structured for the subject in terms of their causal interrelations. Thus, contra Hume, our compulsive application of the concept of causality in experience is not a rationally arbitrary psychological habit or custom, but a necessary precondition on the possibility of having a mind that could be in touch with unfolding external events. If the Kantian argument is sound, then, the causal structure of perceptual experience cannot be rationally arbitrary, since it is a prerequisite for the capacity to experience or cognize events at all. So far I have been using a permissive definition of ‘phenomenological’ philosophy that allows me to encompass these Humean and Kantian arguments (for some authoritative support for this appropriation, see Husserl: CES 87–100; Merleau-Ponty: PhP 226–9). Suppose we instead adopted the more restrictive (and more common) construal of phenomenology as picking out a philosophical movement inaugurated by Husserl, running through Heidegger, Sartre, de Beauvoir, Merleau-Ponty and others, and continuing to the present day. Are there any Humeans among the members of that movement? Whilst phenomenologists are sometimes criticized in terms that depict them as simply ‘reading off ’ philosophical conclusions from the inspection of experience (see e.g. Dennett 1991; Metzinger 2003; Bayne 2004), I know of no major phenomenologists who explicitly avow the use of this kind of methodology. The best someone searching for such an avowal can do is, I think, to wrest phrases or passages out of important context—for example, by citing Merleau-Ponty’s claim that ‘[Phenomenology] is the attempt to provide a direct description of our experience such as it is . . . ’ (PhP lxx). We’ll return to the context in which Merleau-Ponty makes this statement below, where we’ll see how and why it allows us to understand his distinctive approach to phenomenological methodology. This lack of bona fide Humean phenomenologists can be traced to the fact that Husserl, in the phenomenological movement’s inaugural works, defines phenomenology in transcendental terms, as an attempt to uncover the essential a priori structures governing thought and experience. In his Logical Investigations, Husserl defines the subject matter of phenomenology as ‘experiences intuitively seizable and generalizable in the pure generality of their essence, not experiences empirically perceived and treated as real facts, as experiences of human or animal

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

239

experients’ (LI 86), and goes on to say that each statement of such an essential property of experience is ‘an a priori statement in the highest sense of the word’ (LI 86). In the subsequent Ideas, he complains about those who have misinterpreted his phenomenology as a ‘sub-domain of empirical psychology’ (I 2), and contrasts ‘phenomenology’ that merely elucidates empirical facts about experience with ‘pure or transcendental phenomenology . . . which aims exclusively at establishing “knowledge of essences” (Wesenserkenntnisse) and absolutely no “facts” ’ (I 3, original emphasis). Husserl makes this contrast between ‘essence’ and ‘fact’ here because he sees himself as engaged in a Kantian transcendental project (LI 14) which aims to show how the very intelligibility of there being facts of the kind that Humeans aim to catalogue rests on the essential structural properties that the transcendental Phenomenologist aims to uncover. It is because he thinks that the essences revealed by the Phenomenologist are the transcendental basis of the possibility of empirical facts that he thinks of statements of essence as a priori ‘in the highest sense of the word’ (LI 14). One way in which we can contrast the Humean with the Kantian and Husserlian approaches to phenomenology is by comparing their respective strategies for identifying the essences, or essential properties, of kinds of perceptual experience. Broadly construed, something’s essence is that which makes it what it is; its essential properties are those necessary for its being the kind of thing it is. Insofar as we can understand the Humean as attempting to uncover essential properties at all, these are mere empirical regularities—such as the exceptionless regularity that experiences of causation are always occasioned by a concatenation of impressions with a particular structure. For the Kantian, an experience’s essential properties are those that are transcendentally (rather than merely empirically) necessary—causal structure, for example, is an essential property of experiences that seem to disclose events. However, Husserl (at least by the time of Ideas) recommends a route from an experience’s manifest properties to its essential properties that looks slightly different from Kant’s. Whilst Kant aimed to identify experience’s a priori structures via constructing arguments about the conditions under which experience could possess its manifest properties, Husserl’s phenomenology aims at intuiting essences—that is, gaining direct, self-evident experience of fundamental structural properties of experience—via the activity of eidetic variation. In eidetic variation we home in on the eidos (essence) of a type or property of experience through imaginatively varying aspects of the experience, and seeing what must remain invariant if the type or property of experience is to be preserved (I 14, 134–7). Consider a perceptual experience of a dog. What could you imaginatively remove from the way in which the experience presents its object to you whilst preserving its dog-presenting character? Tail-wagging? Fur? Three-dimensionality? Whatever you couldn’t consistently imagine subtracting from the experience whilst leaving its dog-presenting character intact is part of the eidos of your perceptual experiences of dogs. We might see Kant’s transcendental arguments as a sophisticated instance

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

240

 

of this kind of eidetic variation. The Kantian argument sketched above invited us to consider whether an experience as of external unfolding events could survive the subtraction of the experiencer’s grasp of the way in which successive appearances were determined by their predecessors. The suggested conclusion was that it couldn’t, thereby showing that causal structure is an essential property of experiences of external happenings. Despite this commonality, Husserl and Kant diverge importantly. Whereas the ultimate aim of Husserl’s analyses are intuitions of experience’s essential properties, Kant aimed instead at a discursive understanding of those properties. The contrast between these options will become clearer in the next section, when we examine how Husserlian phenomenology is motivated by a dissatisfaction with the Kantian approach. In addition to Humean phenomenology, which aims to read perception’s structural properties off the surface of experience, we now have two kinds of transcendental phenomenology: Kantian phenomenology which aims at discursive understanding of the necessary preconditions for manifest features of our experience, and Husserlian phenomenology which aims at an intuitive grasp of the essential properties of experience, achieved via eidetic variation. This latter conception of phenomenology, however, is transformed by the end of Husserl’s career (though I’ll continue to refer to it simply as ‘Husserlian’ in what follows), and the transformation is inherited and continued by Merleau-Ponty. We can start to understand this transformation and its motivations by considering a pressing question that phenomenologists face—from where do the standards that guide phenomenological reflection derive their authority?

3. Problems with Authority So far, I’ve said nothing to support my initial claims that there is a dialectical relationship between the Humean, Kantian, and Husserlian approaches sketched above, and that this relationship is important for Merleau-Ponty’s methodology. In this section I suggest that we can see how Humean, Kantian, and Husserlian phenomenology stand in a dialectical relationship by considering how each of them attempts to ground the authority of their own phenomenological claims. As we will see, the Husserlian approach is motivated by problems with the sources of authority appealed to by the Humean and Kantian strategies. This will put us in a position to see, in the subsequent sections, how Merleau-Ponty’s phenomenology is motivated by similar problems for the Husserlian approach. The Humean, Kantian, and Husserlian as presented above each have a different strategy for characterizing the structural properties of perceptual experience. Which strategy should we prefer? One way of attempting to answer this question is to consider the authority on which the characterizations of each strategy are supposed to rest. Consider first how this issue of authority arises for the Humean.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

241

As we saw above, Hume took his view of the way in which causality figures in our experience to motivate the sceptical conclusion that our experience gives us no insight into the metaphysical structure of reality with respect to causal relations. Once we see that our compulsive association of causal ideas with suitably structured impressions is a rationally arbitrary psychological custom, we have no reason to think that this custom reflects anything but a quirk of our own psychological structure. There were several parts to this Humean argument: an initial claim that the elements of experience can be partitioned into impressions and ideas; the claim that our experiences never involve an impression of causation, only conjunctions of impressions that reliably give rise to the idea of causation; and the concluding claim that this shows that we have no reason to think that our experiences of causation track some experience-independent structural feature of reality. Why should we accept these claims? As I’ve depicted Humean phenomenology above, it invites us to simply ‘read off ’ the structure of perception from our experiences—and Hume invites his readers to consider their own experiences in support of his split between impressions and ideas, and his claim that experiences of causation consist in ideas occasioned by impressions. So part of the authority of the Humean phenomenology of causation is supposed to derive directly from its match with experience itself. But note that the Humean claims above concern more than just the manifest properties of particular experiences—they include claims about a general distinction between impressions and ideas that is supposed to apply to all experiences, about the structure of all experiences of causation, and about what this shows about the relationship between experiences of causation and the world. These are conclusions that can’t simply be ‘read off ’ from individual experiences, since they aspire to a generality that goes beyond any individual experience we might examine. Instead, Hume arrives at these more general conclusions by reasoning and extrapolating from the specific experiences he appeals to. The problems for the Humean arise when we ask after the authority of these extrapolations and episodes of reasoning. As Husserl points out, Hume did not ask the question, or at least did not say a word, about the status of the reason—Hume’s—which established this theory as truth, which carried out these analyses of the soul and demonstrated these laws of association. . . . Even if we knew about them, would not that knowledge itself be another datum on the tablet? (CES 88)

That is, the Humean’s demonstration of the lack of a rational relation between our ideas and the concatenations of impressions with which they are associated appears to undermine its own authority. If the moves in Hume’s argument strike us as compelling or unavoidable, the worry arises that they do so only in virtue of other rationally arbitrary psychological habits. We have no reason to think that

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

242

 

the episodes of reasoning involved in Hume’s argument are any more than a series of ideas occasioned by the impressions on which the Humean was reflecting—that is, we have no more reason to think that they disclose a fundamental structural feature of perception than we have to think our experiences of causation disclose a structural feature of reality. By limiting their methodology to reading psychological structure off the surface of experience, Humeans thus appear to deprive their read-outs of any plausible source of authority. When faced with the question of why their descriptions should compel assent, or why they represent an important epistemic gain, the only resources available for an answer will be more rationally arbitrary psychological facts—more ‘data on the tablet’. Humeans might invite us to reflect further on our experiences of billiard-ball collisions, or to gather more such experiences; or they might appeal to how rationally compelling the moves in their argument seem. But no particular experience of causation, or of rational compulsion, can mandate the general claims about the structure of perception that Humeans seek to establish. This is not yet to say that the Humean claims should be rejected—only that, in considering how those claims can be supported, we must look beyond the surface properties of experience to which Humeans appeal. Thinking through the credentials of Humean phenomenological claims thus motivates us to consider what sources of authority there might be, other than appeal to the surface properties of experience, for general claims about the structure of perception. Does the Kantian approach fare better? From where do its phenomenological claims draw their authority? We saw above that the Kantian’s transcendental claims are supposed to have a straightforwardly rational grounding—Kantians start with uncontroversially manifest features of experience (such as the facts that perceptual experience seems to present us with a mind-independent world, or takes place from a perspective, or can involve experiences as of things happening) and attempt to provide a priori arguments about the necessary conditions for experience’s possession of those features. They traffic in a priori claims about which structural properties of perception are necessarily required for it to be intelligible that experience possesses a given manifest property. The focus on rational necessity and a priori entailment means that the Kantian’s claims should automatically have the generality and universality that eluded the Humean. However, the Kantian’s appeal to rational authority faces a different challenge: why should the rational authority that is grounded in the cogency of the Kantian’s theoretical arguments and reconstructions have any jurisdiction within the domain of untheorized, lived experience that is the subject matter of phenomenology? Before we start to philosophize, after all, experience isn’t presented to us in Kantian terms of syntheses, apperceptions, and mutually interdependent faculties of cognition. If it were, the Critique of Pure Reason would read like a long-winded reminder of facts that were already perfectly familiar to us. Instead, these terms are theoretical constructions via which Kantians attempt to make the structure of

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

243

experience intelligible to reflection. There is, of course, nothing intrinsically wrong with this—theorizing a domain inevitably involves imposing some general framework upon it that confers intelligibility in part by abstracting away from finegrained details and particularities. But we can always ask of a given theory whether it abstracts away from the right particularities—whether it ignores or glosses over important facts about the theorized domain that it cannot readily accommodate. In this case, Kantians attempt to understand the structure of perceptual experience by providing transcendental arguments about the structures such experience must have—as a matter of rational necessity—if it is to possess its manifest properties. In response, we may wonder whether this methodology abstracts away from important structural features of experience. Perhaps there is a gap between aspects of the immediate, untheorized experience from which Kantian phenomenology begins, and the theoretical reconstruction of the abstract structures of experience at which it aims. There may be important structural aspects of our experience which are simply unamenable to the Kantian’s reflective analysis, in that their contribution to our experience cannot be revealed via a rationally compelling transcendental argument. The problem I am raising here is not that there demonstrably are such aspects of experience which resist the Kantian’s rational reconstructions—merely countenancing the possibility that there might be poses a serious problem for their methodology. To see why, consider how a Kantian might respond to this possibility. The simple strategy of taking an introspective inventory of one’s experience to check for the presence of such rationally unanalysable structural features is unavailable to them, since their methodology is defined by the reflective attempt to abstract away from the surface features of experience to the a priori structures that are their transcendental requirements. Because Kantian phenomenology is motivated by the attempt to provide a more compelling source of authority for phenomenological claims than the Humean appeal to the surface properties of experience, it cannot now fall back on an appeal to its fit with those properties to ground the authority of its transcendental claims. An alternative response on behalf of the Kantian would be to claim that their a priori reflections must be accurately revealing the structural properties of experience—for how could it intelligibly be otherwise? Their transcendental arguments, after all, show that perceptual experience could not intelligibly possess some universally agreedupon manifest property without having these structural features. But in this context, such a response is clearly question-begging—the ‘must’ above is the ‘must’ of rational necessity. But the possibility we are countenancing is that the Kantian’s project of telling us, via armchair rational reconstructions, about the structure that experience (rationally) must have might obscure important facts about the structure that it does have. Husserl (CES 97, 115) and Merleau-Ponty (PhP 228/265, PrP 18–19) both suggest that Kant fails to break completely from his dogmatic rationalist roots by neglecting to properly consider this question of

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

244

 

the authority of his rational reconstructions over the domain of untheorized experience. We have been asking about the source of the authority of phenomenological analyses—why should we think that these analyses accurately reflect the reality they aim to capture, or that they represent an epistemic gain? In answering this question, we have seen that Humeans can appeal only to empirical experience, Kantians only to rational intelligibility. But the scepticism that flows from Humean phenomenology makes the first answer unconvincing, and in the context of Kantian phenomenology the second answer is question-begging—Kantians lack the materials to build a bridge between the character of lived experience and their theoretical descriptions. We can see a dialectical relationship between the Humean and Kantian positions and their respective shortcomings. The problem for the Humean was that appeal to the surface properties of particular experiences cannot ground their general claims about the structural properties of experience. We can see the Kantian position as a natural reaction to this problem. Because it relies on a priori theorizing about the necessary preconditions of experience, the a priori structures it identifies should be generally valid—they should apply to all experiences that we can make intelligible sense of. But we asked above why the fact that these structures should apply entitles us to conclude that they do apply to the particular experiences we in fact have. Surely the right test for whether they do apply is simply to look and see? But to endorse this test would, it seems, be to fall back on the Humean position from which the Kantian was attempting to escape. The Husserlian approach described above can be seen as attempting to resolve this dialectic by synthesizing the Humean and Kantian approaches—building on each of them in a way that keeps the good bits and throws away the bad.³ Husserlians assure us that we can trust their claims about the essential structures of our experience because those essences are intuited—grasped directly and selfevidently. Like the Humean, the Husserlian grounds the authority of their analyses in the experience on which they are reflecting—when discovering a Husserlian essence we intuit, rather than discursively understand, a property that our experience really has. But Husserlians can say more than Humeans about why we should take the psychological properties we intuit as revealing important structural truths, since we intuit them as features of experience that are reflectively necessary ³ I’m not arguing here that the development of Husserl’s philosophy is explicitly motivated by the failings of Humean and Kantian approaches—only that this way of labelling and understanding the tensions at work in Husserl’s early phenomenology (that is, up to and including Ideas) helps us understand the Merleau-Pontian methodology outlined in the next sections. But a plausible case could be made that something like the dialectic sketched in this section plays out in the development of Husserl’s early thought. It might be argued that the criticisms of the ‘psychologistic’ project of his Foundations of Arithmetic (see Frege and Kluge 1972) parallel the problems for Humean phenomenology sketched above, and motivate the transcendental methodology of his Logical Investigations which develops into the Husserlian phenomenology of the Ideas.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

245

and self-evident. By appealing to an intuitive contact with the necessity and self-evidence of the essential structures of perception, the Husserlian thus attempts to simultaneously ground the authority of their claims in experience and in reflection. However, it appears that this attempted synthesis of Humean and Kantian phenomenology inherits rather than transcends the problems with each position. For any intuited essence we can ask the Husserlian why we should be impressed either by the authority of its experiential character, or by the way it appears to reflection. Why isn’t the fact that I experience a purported essential property as apodictic or self-evident just another ‘datum on the tablet’ to be set alongside all the other arbitrary empirical properties of my experience? If the Husserlian answers via appeal to the rational credentials of the intuition—the way in which the experienced self-evidence makes good reflective sense—then they can be asked the same questions we used to embarrass the Kantian, above. Instead of being stuck in the unsatisfactory positions of either the Humean or the Kantian, the Husserlian is now see-sawing between them—a different predicament, but not obviously a better one. We are now in a position to see why the later works of Husserl, and the phenomenology of Merleau-Ponty, come to see phenomenology as defined by the need for ‘radical reflection’ that involves a critical awareness of its own presuppositions. We can see each of the positions sketched above as attempting to secure general philosophical conclusions about perception via a particular kind of reflection. The Humean attempts to attend carefully to the surface features of particular experiences. The Kantian attempts to provide theoretical reconstructions of the rationally necessary preconditions for uncontroversially manifest features of general classes of experience. And the Husserlian occupies a halfway house between the Humean and Kantian positions, attempting to intuit essential properties of types of experience by grasping them directly and self-evidently. We have just questioned the authority of each kind of reflection, asking why it should count as a reliable source of insight into perception’s structural properties. In the next section we will see how these questions motivate the later Husserl and Merleau-Ponty to build a commitment to an ongoing assessment of the genesis and authority of our reflective capacities into phenomenological methodology.

4. From Husserlian Phenomenology to Radical Reflection Let’s revisit some of the antinomial characterizations of phenomenology with which Merleau-Ponty begins his Phenomenology of Perception, alluded to at the start of this chapter. Phenomenology is a ‘study of essences’, but one which ‘places essences back within existence and thinks that the only way to understand man and the world is by beginning from their “facticity” ’ (PhP lxx); and it

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

246

 

is a ‘transcendental philosophy’ which nonetheless seeks to rediscover a ‘naïve contact’ with a world which is ‘always “already there” prior to reflection’ (PhP lxx). These are the tensions we have just articulated for Husserlian phenomenology, as it see-saws between the unsatisfactory Humean attempt to ground its claims in untheorized (intuitive) experience, and the unsatisfactory Kantian attempt to support those claims via reflective reconstruction of transcendental conditions. As we will see below (§5), there is a lot packed in to Merleau-Ponty’s claim that phenomenology ‘places essences back within existence’ and emphasizes the ‘facticity’ of perceivers and world. But it in part alludes to the tension we have seen between Humean and Kantian approaches—between the Humean impulse to ground their claims in the empirical regularities that characterize particular experiences and the Kantian impulse to ground their claims in the transcendental necessity of the perceptual structures they describe. Likewise, we have just seen that a potentially problematic gap opens up between the Kantian’s ‘transcendental philosophy’ and the ‘naïve’ untheorized experience that exists ‘prior to reflection’—but also that the Humean’s sole emphasis on the surface properties of experience appeared to thwart their aspiration to articulate general structural features of perception. The Husserlian attempts to respond to these problems by amalgamating the Humean and Kantian positions, appealing to a capacity to ‘intuit’ essential properties of perception that is supposed to be simultaneously experiential (grounded in an experience of insight arrived at via eidetic variation) and rational (the insight is experienced as reflectively apodictic). In this section I further develop the problems that Merleau-Ponty, drawing on Husserl’s later work, sees for the Husserlian position, and the way in which these problems motivate a conception of phenomenology as involving ‘radical reflection’. A good starting point is the final antinomy Merleau-Ponty lists. Phenomenology is, he says: the attempt to provide a direct description of our experience such as it is, and without any consideration of its psychological genesis or of the causal explanations that the scientist, historian, or sociologist might offer of that experience; and yet in his final works Husserl mentions a ‘genetic phenomenology’ (PhP lxx)

Whereas the ‘static’ Husserlian phenomenology we have been considering aims only to grasp the essential structures that our experience has, the ‘genetic Phenomenology’ mooted in Husserl’s Cartesian Meditations and developed in his Crisis of European Sciences and other late writings aims to uncover the genesis of those structures in our bodily, historical, and cultural situation. Because, in my view, this kind of genetic phenomenology receives its fullest articulation in Merleau-Ponty’s work, I’ll refer to it simply as Merleau-Pontian phenomenology in what follows. This move from static to genetic phenomenology should be understood, I suggest, as a response to the worries about the credentials of

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

247

Humean, Kantian, and Husserlian modes of reflection on experience summarized at the end of the last section. Genetic phenomenology aims to investigate the origins of what seems reflectively evident or apodictic to us, and does so in a way that problematizes the kinds of phenomenological reflection we’ve considered so far. Considering the ways in which bodily, historical, and cultural factors condition our experience suggests that we can’t properly identify experience’s essential structures via the armchair reflections of the Kantian or Husserlian. And considering the way those same factors condition our reflective capacities problematizes the idea that those capacities are apt to put us in touch with the kinds of essences that Kantians and Husserlians seek. This appears to send us back into the arms of the Humean. But we have just seen that Humeans cannot motivate conclusions about essential structures of experience that have the generality which phenomenologists seek. This all sounds like bad news for phenomenologists. Indeed, the rest of this section draws on Merleau-Ponty to note the negative consequences that the contingent bodily, linguistic, and cultural origins of our experiences and reflections appear to have for Husserlian phenomenology. But don’t despair— the negative consequences surveyed in this section allow us to see, in the next section, the positive shape that a Merleau-Pontian phenomenology of perception must take. Merleau-Ponty has the most to say about the ways in which bodily and psychological contingencies shape our reflective capacities and intuitions about essential properties of experience. I’ll say more about this in the next section, since Merleau-Ponty’s views here are particularly helpful for understanding his positive suggestions about how Husserlian phenomenology should be superseded. But the simple idea is familiar from some Humean worries considered earlier. Recall that Hume argued that we lacked reason to think that the apparent ineliminability of some ideas, such as causal connectedness, from certain perceptual experiences reflected any more than a quirk of our psychology. If a Husserlian phenomenologist, engaged in eidetic variation, finds themselves unable to imaginatively subtract causal structure from their experiences of unfolding events then this might reflect only contingent features and limitations of their psychology, rather than any deep insight into transcendentally necessary structures of experience. The contingencies of our bodily (including neural and thus, I am supposing, psychological) organization constrain what is reflectively and imaginatively possible for us. Can you imagine, for example, having a perceptual experience without simultaneously experiencing yourself as existing? Probably not—but this is just how some sufferers of Cotard’s syndrome sincerely describe their experience (Debruyne et al. 2009). When you engage in phenomenological reflection on whether a sense of one’s own existence as a perceiver is a necessary structure of perceptual experience, your best introspective, reflective, and imaginative attempts to answer this question might be mere reflections of the happy neural contingency that you don’t have Cotard’s syndrome.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

248

 

As well as worrying the Husserlian with the possibility that the results of their eidetic variation reflect mere contingent psychological quirks, we can worry them with the possibility that their intuitions of reflective apodicticity or self-evidence are results of their contingent linguistic, cultural and historical situation. MerleauPonty writes: It is possible for me to believe that I am seeing an essence when, in fact, it is not an essence at all but merely a concept rooted in language, a prejudice whose apparent coherence reduces merely to the fact that I have become used to it through habit. (Phenomenology and the Sciences of Man (henceforth PSM), PrP 75)

Abstracting away from the nuances of Merleau-Ponty’s account of language, we can read his argument here as drawing on two uncontroversial claims. First, the phenomenologist’s capacity for sophisticated reflective thought depends on facility with language. Making reflective judgements based on the results of eidetic variation is something that we can only do in virtue of the fact that we are language users. Second, this facility with language is a contingent, culturally mediated acquisition—the linguistic capacities on which the phenomenologist’s reflective thought depends are available to them only because they are born and enculturated into existing practices for using language. It follows from these two claims that the Husserlian’s reflections on the essential structures of experience can no longer be understood as uncovering a priori truths that are straightforwardly independent of the empirical contingencies of their situation. We saw above that if phenomenology consists in attempting to grasp the essential structures of experience, then it is susceptible to the question of how we can be certain we’ve attained such a grasp. We also saw that attempting to answer this question via simple appeal to the way in which such a grasp is experienced by the phenomenologist courts Humean sceptical worries—why isn’t such an experience just one more datum on our tablet of contingent psychological phenomena? Now we have seen that the more sophisticated attempt to answer this question in terms of the way in which an intuited essence presents itself to reflection as apodictically certain also faces a challenge: if we agree that the Husserlian’s reflective capacities depend on the contingent fact that they have been enculturated into a specific system of linguistic practices, then their reflective judgements of apodicticity appear to be tainted with this contingency. It thus becomes hard to see how such judgements can be taken to reveal essential as opposed to contingent structural properties of experience. More bluntly, the worry is that the Husserlian’s putative grasp of an essence fails to show anything interesting about the fundamental structure of perception; instead, it merely shows us something about the way they happen to think, which is in turn a mere function of a way they happen to talk.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

249

The linguistic mediation of thought is one way in which the Husserlian’s cultural and historical situation threatens to infect their claims about experience’s essential structures with contingency. Beyond this, we might worry that empirical evidence of sociological and anthropological diversity suggests that what strikes us as an apodictic or essential feature of experience is a mere cultural quirk—as when, for example, intuitions about the way in which numerical quantity figures in our experience are called into question by learning that some Australian or Amazonian cultures have no fine-grained system for describing numerical quantity (e.g. Frank et al. 2008). In this connection, Merleau-Ponty reports Husserl changing his mind about the relevance of empirical anthropological data to phenomenology upon reading Lucien Lévy-Bruhl’s ethnological study Primitive Mythology. ‘Before this,’ he writes, ‘Husserl had maintained that a mere imaginative variation of the facts would enable us to conceive of every possible experience we might have’ (PSM, PrP 90). But in a subsequent letter to Lévy-Bruhl Husserl admits that: the philosopher could not possibly have immediate access to the universal by reflection alone—that he is in no position to do without anthropological experience or to construct what constitutes the meaning of other experiences and civilizations by a purely imaginary variation of his own experiences. (S, 107)

The worry thus arises that the Husserlian’s contingent cultural inheritance places arbitrary restrictions on their reflections in two ways—via constraining the vocabulary within which those reflections can take place, and via local cultural idiosyncrasies influencing reflection on what is or is not an essential structural property of experience. Finally, the particular point in intellectual history at which we are situated can make a difference to the truths we take to be apodictically certain, as when the formalization and empirical application of non-Euclidean geometries reveals that ‘The supposed evidentness of Euclidean geometry’ was ‘evident merely for a certain historical period of the human spirit’ (PhP 414/454). RomdenhRomluc’s example (2018, this volume) of Carmita Wood illustrates how phenomenological reflection can be tied to cultural and historical contingencies. Wood, a lab worker in Cornell’s department of nuclear physics during the 1960s, quit her job after being subjected to a sustained campaign of sexual harassment by a senior scientist. But it was only years after her resignation that Wood was able to make full sense of the pattern of behaviour and experiences that drove her from her job. The reflective understanding of that pattern as a manifestation of sexual harassment was not readily available in Wood’s 1960s environment; her intuitive grasp of that pattern only became possible retrospectively, once the concept of sexual harassment had become common currency. Like Carmita Wood, a phenomenologist’s best attempts at grasping the essential structures of their experience will go

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

250

 

only as far as the reflective apparatus afforded by their historical and cultural situation can take them. So far, then, we have seen that the problems for Humean, Kantian, and Husserlian phenomenology point towards the need for a ‘radical reflection’—a way of doing phenomenology that critically examines the origins of its own reflective capacities and convictions. But the suggestions about these origins canvassed here look like bad news: the ‘human sciences’—including psychology, linguistics, sociology, and history—reveal multiple ways in which the phenomenologist’s grasp of putatively essential structures of experience is inescapably influenced by contingent features of their situation.

5. Merleau-Pontian Phenomenology of Perception The sources of contingency surveyed in the last section simultaneously motivate scepticism about the essences that the phenomenologist tries to grasp, and the reflective capacities used to grasp them. Both now appear to be reflections of various contingencies of our factual situation. The Merleau-Pontian phenomenologist, however, views these contingencies not as problems but as opportunities. Merleau-Ponty argues that we should reconceive our understanding of both essences and reflection in ways that acknowledge and embrace these contingencies, putting them to work in accounts of how particular experiential structures come to count for us as essential to our subjectivity—that is, in accounts of the genesis of essences and their reflective credentials. For reasons of space, I’ll limit myself to one example of this strategy—the way in which Merleau-Ponty’s use of a specific empirical case study figures in an argument that an implicit sense of the possibilities for active engagement with our environment is an essential property of world-disclosing perception. This illustration, however, does not fully capture the way in which Merleau-Pontian phenomenology involves a reconception of reflection itself as well as the essences at which reflection aims. The remainder of this section briefly brings the general shape of this ‘radical reflection’ into clearer view. Merleau-Pontian phenomenology remains committed to the search for essences, but—as we will now see—its vision of what those essences are, and how to find them, changes. One source of this change is the recognition of a close relationship between the essential structures and the factual contingences of the ways in which they are realized by our body and its relation to the world. MerleauPonty argues that the Husserlian’s transcendental methodology already entails this relationship: If in Husserl’s view the knowledge of facts is impossible without some insight into essence and is always helped by this, it follows that all sound knowledge of

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

251

facts must include, at least implicitly, some insight into essences, and that Husserl must admit, as he does in effect, that those psychologists who have been preoccupied with facts have nevertheless been able to find out something concerning essences. (PSM, PrP 66)

We saw above (§2) that Husserl makes the transcendental claim that our access to empirical facts is necessarily mediated by the essential structures that phenomenologists aim to grasp and catalogue. Merleau-Ponty’s claim above is that this commits us to thinking that, insofar as we agree that psychologists are successfully uncovering empirical facts about our minds, these facts can function as a source of insight into the essences to which they are necessarily related. This is why Merleau-Ponty took such a keen interest in a wide range of scientific disciplines, including Gestalt psychology, Saussurian linguistics, and Freudian psychiatry. Whilst he disagreed in important ways with the commitments of these approaches, he nonetheless argued that each domain yields new insights into the structures of experience that must be integrated into an adequate phenomenology. Empirical psychology thus becomes a key catalyst for eidetic variation— each of the above domains prompts us to consider, and find truth in, new ways of understanding ourselves. Consider, for example, Merleau-Ponty’s discussion of the brain-damaged World War I veteran, ‘Schneider’ (PhP Pt. I ch.3). As a result of his injuries, Schneider exhibited an unusual patchwork of sensorimotor and cognitive impairments, including an inability to visually recognize objects coupled with intact visuomotor abilities to perform habitual and automatic actions, such as negotiating his environment, opening doors, and even making leather wallets. The most important feature of Schneider’s syndrome of impaired and intact capacities, for Merleau-Ponty, is the contrast between Schneider’s intact sensorimotor habits and his impaired capacity to put those habits to use in flexible, open-ended interaction with his environment—what Merleau-Ponty calls a power to ‘reckon with the possible’ (PhP 112/139; Romdenh-Romluc 2007). Schneider can interact with his perceptible environment in rigid, task-specific ways, but lacks, on MerleauPonty’s diagnosis, a sense of the way in which his relation to that environment affords the satisfaction or frustration of an open-ended range of actual and potential projects. The sensorimotor actions Schneider can successfully perform are habituated and task-specific ones that are automatically drawn from him by his environment. The actions which he cannot, or struggles to, perform—such as pointing, pantomiming, or playful and creative ways of engaging with his surroundings—are those not directly linked to a current, habituated task. The case of Schneider is one of many in PhP that is intended to reveal the way in which our awareness of our surroundings is pervasively structured by an implicit sense of the way in which they afford the satisfaction or frustration of both our current and our potential projects. We might judge, from our armchairs, that our perceptual

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

252

 

experience of objects could survive the subtraction of a background sense of the open-ended range of possible ways we might engage with them. But the empirical facts about the way in which this sense is impaired for Schneider help us intuitively grasp the role that it plays in structuring our own experience—a role that we might otherwise have missed. At the same time, these facts reveal a sense in which this essential structure of our experience has its basis in empirical contingencies—the happy contingency that we lack the brain lesions that are responsible for Schneider’s symptoms. Importantly, Merleau-Ponty’s use of Schneider’s case is just one strand of a central argument of PhP which aims to demonstrate that in perception, ‘a system of possible movements, or “motor projects” radiates from us to our environment’, and that all perceptual experiences are conditioned by this ‘global, practical, and implicit notion of the relation between our body and things’ (PrP 5). Abstracting away from its details, we can call this the ‘bodily-structure claim’ (BSC): BSC: Perceptual experience necessarily involves an implicit awareness of the perceiver’s embodied practical relationship to the perceived. Why should we think that BSC constitutes the kind of claim about the kind of essential structure of experience that phenomenologists seek, rather than one of the contingent psychological quirks catalogued by Humeans? Considering the empirical facts of Schneider’s case, or of our own experiences, cannot answer this question. Instead, a sustained transcendental argument is woven through the rest of PhP that BSC is a condition of the very possibility of perceiver/world relations, since our experience of the reality of the objects of perception must be understood in terms of their positive and negative relations to our actual and potential exploratory activities and projects. In brief outline, the argument aims to show that perceiving something as ‘real’—as enjoying a perceiver-independent existence—requires being aware of it ‘as the infinite sum of an indefinite series of perspectival views in each of which the object is given, but in none of which is it given exhaustively’ (PrP 15). As finite, embodied perceivers, any vantage point on an object reveals particular aspects while obscuring others—as when we see the facing side of a cube better than its partially visible or hidden sides. Merleau-Ponty argues, however, that this partial access does not impede our perception of the cube’s objective properties, but makes it possible: ‘It is not accidental for the object to be given to me in a “deformed” way, from the point of view which I occupy. That is the price of its being “real” ’ (PrP 15–16). Unlike your inner speech or sensory imaginings, your experience of external objects is given to you as provisional and incomplete, in a way that permits and invites further bodily exploration: ‘When I see an object, I always feel that there is still some being beyond what I currently see . . . a depth of the object that no sensory withdrawal will ever exhaust’ (PhP 224/261). This

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

253

experienced provisionality is part of what it is to experience those objects as having an objective, rather than purely subjective, existence. And we experience this provisionality in the form of an implicit awareness of the ways in which changes in our bodily relationship to the object will bring other aspects into view. In the interests of space, I am suppressing many of the details of this argument as it figures in PhP.⁴ Most important for our purposes is its general structure, which closely parallels the Kantian strategy we encountered above. Faced with the challenge of demonstrating that BSC reflects an essential structure of perceptual experience, Merleau-Ponty argues that it is a prerequisite for an uncontroversially manifest property of perceptual experience—its seeming to present us with real, perceiver-independent things. The aim is to show that BSC inherits its essentiality from the world-disclosing character of perception, an essential structural feature of perceptual experience that is not in dispute. However, this is not the armchair transcendentalism of the Kantian. Like the Husserlian, Merleau-Ponty attempts to bring this structure into view for us via eidetic variation, inviting us to test his claims against specific experiences of our own. Unlike the Husserlian, MerleauPonty’s eidetic variation also makes use of empirical cases involving experiences his readers are unlikely to have had (such as Schneider’s) as catalysts. The reported experiences of those for whom implicit understanding of the systematic and openended range of possibilities for bodily exploration of the perceived has gone missing (as in Merleau-Ponty’s discussion of Schneider) or has yet to be learned (as in his discussion of the experiences of post-operative congenital cataract patients) (PhP 231–2) provide contrast cases that illuminate the way in which this bodily understanding structures our own perception. Once we have homed in on BSC in this way, Merleau-Ponty (like the Husserlian and Kantian) argues that it qualifies as an essential structure of experience in virtue of its status as a necessary prerequisite for an uncontroversial manifest feature of experience—its world-disclosing status. The above argument thus involves a reciprocal interplay between empirical and transcendental considerations in the service of identifying an essential structure of experience. In doing so, it sets Merleau-Pontian phenomenology apart from the other positions canvassed above. For our purposes here, the most important consequence of this form of argument is a blurring of the boundaries between empirical facts and transcendental structures. Kantian and Husserlians are committed to a sharp distinction between the essential structure we have identified (as expressed in BSC) and the contingent facts in virtue of which experience possesses that structure. The contingent facts in this case would include the particular details of our embodiment, and the ways in which Schneider’s sensorimotor capacities were impaired. Merleau-Ponty’s claim, here and elsewhere, is that fact

⁴ See in particular PhP 242/280; 331–41/373–84.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

254

 

and essence cannot be pulled apart—the reflective purchase we have on BSC exists only in virtue of the contingent facts of our embodiment that give our experience the structure BSC describes. An articulation of an essence like BSC ties together and reveals the significance of particular contingencies of the facts of our embodied perceptual lives—but that significance is only available to the phenomenologist’s reflection in virtue of their living such a life. As Merleau-Ponty puts it in PhP’s preface, ‘essences must bring with them all of the living relations of experience, like the net that draws up both quivering fish and seaweed from the seabed’ (lxxix, 15–16). I started this section by saying that Merleau-Pontian views the contingencies we noted for the Husserlian (§4) as opportunities rather than problems—we have now seen one way in which this is so. The contingent facts of our embodied relation to the world are put to work in an account of how the phenomenologist can grasp a particular structural property as essential to their subjectivity, since those facts constitute the phenomenologist’s mode of reflective access to that property. This is the sense in which Merleau-Pontian phenomenology involves ‘radical reflection’—it focuses not just on identifying essences that are experienced as reflectively apodictic, but on accounting for the origins of that reflective status via connecting it to the contingent facts of our lives and situation. The example I have just sketched drew only on the facts of our bodily situation—but we saw above that the phenomenologist’s reflections are also shaped by linguistic, sociological, and historical facts. Whilst bodily contingencies of the kinds we’ve just alluded to shape the perceptual experience on which the phenomenologist reflects, linguistic, social, and historical contingencies chiefly shape the reflective capacities that the phenomenologist trains on their experiences. Thus, a Merleau-Pontian ‘radical reflecter’ is committed to understanding their own reflective capacities and activities by, in part, studying and understanding these contingencies and their role in our reflective lives. Merleau-Ponty’s sustained engagement with linguistic theory, history, politics, sociology, and anthropology speaks to this ambition. The task of extracting a stable theory of reflection from this huge body of work is a daunting one. Instead of attempting it here, I will end by considering some remarks that suggest the shape which Merleau-Ponty thought such a theory should take. A good place to start is with a line of thought in PhP on the dependence of the ideals of truth and objectivity at which reflection aims on an intersubjectively shared set of linguistic capacities. Speech, argues Merleau-Ponty, involves ‘a power of thinking according to others, which enriches our own thoughts’ (PhP 184/218–19, original emphasis; compare CES 360–1), and does so by leading ‘to a thought which is no longer ours alone, to a thought which is presumptively universal’ (Merleau-Ponty 1962/1964, 8). Our acquisition of unthinking facility with speech and the intersubjectively accessible and situation-independent meanings at which it aims gives rise to the

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

255

ideal of a locus of meaning and truth that is independent of particular contingent acts of expression: Language has, in fact, installed in us this certainty that we have of reaching, beyond its expression, a truth separable from that expression, and of which this expression is only the clothing and the contingent manifestation. (PhP 422/460, see also 196/231)

Just as Merleau-Ponty argues that the world-disclosing nature of our experience is founded on, rather than undermined by, our inevitably incomplete bodily relation to the world, he wishes to argue that our ideal of truth is founded on, rather than undermined by, the complex web of intersubjective conventions and relations that composes the cultural institution of language. This line of thought persists throughout Merleau-Ponty’s career. In a late essay, he writes: Since we are all hemmed in by history, it is up to us to understand that whatever truth we may have is to be gotten not in spite of but through our historical inheritance. Superficially considered, our inherence destroys all truth; considered radically, it founds a new idea of truth . . . [I]f I have once realized that through it [my inherence in a historical situation] I am grafted onto every action and all knowledge which can have a meaning for me, and that step by step it contains everything which can exist for me, then my contact with the social in the finitude of my situation is revealed to me as the point of origin of all truth. (S, 109)

An aim of phenomenology as radical reflection, then, is to understand the ways in which the very ideal of truth—the ideal which guides our assessment of the veracity of the phenomenologist’s descriptions and reflective reconstructions—is not undermined by, but is dependent on, the facts of our social, cultural, and historical situation. Much of Merleau-Ponty’s work after the completion of PhP, variously focusing on linguistics, sociology, anthropology, and history, affords suggestive hints about how he might have wished to fill in the details of this picture. But they remain only hints, as Merleau-Ponty’s work was cut short by his early death. If we don’t know the exact shape that a Merleau-Pontian phenomenology of radical reflection would take, however, we do know its general structure. As well as identifying the essential structures of perceptual experience, it must simultaneously identify the essential structures of the reflective capacities involved in its own execution. And it must do both of these things in a way that shows how these essential structures emerge from the contingencies of our worldly situation, in a way that must be informed by a richly interdisciplinary engagement with multiple empirical and human sciences. If the Merleau-Pontian argument above is right, then phenomenological philosophy of perception must take this shape—the

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

256

 

failures of Humean and Kantian approaches that appeal only to the concrete matter or the abstract structure of experience, and of Husserlian approaches that appeal to an undertheorized capacity to intuit apodictically necessary essential structures, lead us inexorably to a conception of phenomenology as radical reflection. Even if Merleau-Ponty has, understandably, left us short of the destination of a truly radically reflective phenomenology, he has provided us with a good head start, a detailed map, and plenty to do on the journey.⁵

List of Abbreviated Works LI: Husserl, E. (1900/2001) Logical Investigations. Trans. J.N. Findlay, ed. D. Moran. UK: Routledge. I: Husserl, E. (1913/2012) Ideas: General Introduction to Pure Phenomenology. Trans. W.R. Boyce Gibson. UK: Routledge. CES: Husserl, E. (1954/1970) The Crisis of the European Sciences and Transcendental Phenomenology: An Introduction to Phenomenological Philosophy. Trans. D. Carr. Evanston: Northwestern University Press. PhP: Merleau-Ponty, M. (1945/2012) Phenomenology of Perception. Trans. D. Landes. UK: Routledge. PrP: Merleau-Ponty, M. (1964) The Primacy of Perception and Other Essays on Phenomenological Psychology, the Philosophy of Art, History and Politics. J.M. Edie (ed.). Evanston: Northwestern University Press. S: Merleau-Ponty, M. (1960/1964) Signs. Trans. R.C. McLeary. Evanston: Northwestern University Press.

References Bayne, T. 2004. ‘Closing the gap? Some questions for neurophenomenology’. Phenomenology and the Cognitive Sciences, 3 (4): 349–64. Debruyne, H., Portzky, M., Van den Eynde, F., and Audenaert, K. 2009. ‘Cotard’s syndrome: a review’. Current Psychiatry Reports 11 (3): 197–202. Dennett, D. C. 1991. Consciousness Explained. Boston: Little Brown. Frank, M. C., Everett, D. L., Fedorenko, E., and Gibson, E. 2008. ‘Number as a cognitive technology: Evidence from Pirahã language and cognition’. Cognition 108 (3): 819–24.

⁵ Thanks to the editors of this volume and Clare Mac Cumhaill for helpful comments on previous drafts of this essay. Thanks also to several generations of students in my class on Phenomenology at the University of Edinburgh, conversations with whom have shaped my thinking about Merleau-Ponty’s methodology.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   

257

Frege, G. and Kluge, E. W. 1972. ‘Review of Dr. E. Husserl’s Philosophy of Arithmetic’. Mind 81 (323): 321–37. Hume, D. and Selby-Bigge, L. A. 1740/1978. A Treatise of Human Nature. Oxford: Clarendon Press. Kant, I. 1787/1997. Critique of Pure Reason. Translated and edited by P. Guyer and A. W. Wood. Cambridge: Cambridge University Press. Merleau-Ponty, M. 1962/1964. ‘An Unpublished Text’. In Merleau-Ponty, M. (1964) The Primacy of Perception and Other Essays on Phenomenological Psychology, the Philosophy of Art, History and Politics. J.M. Edie (ed.). Evanston: Northwestern University Press. Metzinger, T. 2003. Being No One: The Self-Model Theory of Subjectivity. Cambridge, MA: MIT Press. Romdenh-Romluc, K. 2007. ‘Merleau-Ponty and the Power to Reckon with the Possible’, in T. Baldwin (ed.) Reading Merleau-Ponty. London: Routledge, 44–58. Romdenh-Romluc, K. 2018. ‘Science in Merleau-Ponty’s phenomenology: from the early work to the later philosophy’. In D. Zahavi (ed.) Oxford Handbook of the History of Phenomenology. Oxford: Oxford University Press, 340–59.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

13 Merleau-Ponty Perception and Methodology Komarine Romdenh-Romluc

1. Introduction It is well known that Merleau-Ponty was particularly interested in perception—as the name of one of his most important texts, the Phenomenology of Perception, implies. However, it is also perhaps fair to say that he is not particularly known as a philosopher of perception—at least in the contemporary Anglo-American tradition. Maybe because he studied perceptual experience in the service of grander aims: to provide a new conceptual framework for understanding consciousness, the world, and their relation. Nevertheless, his work contains an account of perception, and it is this that is my concern here. My focus is Merleau-Ponty’s methodology. A methodology is a theory of method. In asking for a methodology, one wants to know which method(s) we should use for investigating some topic, and why. I will begin by considering Merleau-Ponty’s philosophical methodology in general terms, before turning specifically to his studies on perception.

2. Merleau-Ponty’s Phenomenology Merleau-Ponty is a phenomenologist. Phenomenology is not a unified movement—it’s more like a tradition, where figures are linked by historical lineages: the inheritance and taking up of particular problems, ways of framing issues, certain themes and ideas. Nevertheless, it self-consciously bills itself as having a distinct method. It will be useful to think of it as having four components: Experience: the phenomenologist’s data are experiences as they are undergone by their subject—i.e., considered from a first-person perspective. The Transcendental-Phenomenological Reduction: the phenomenologist must perform this to consider experience from the correct perspective. It involves suspending some attitude or way of thinking about things. Komarine Romdenh-Romluc, Merleau-Ponty: Perception and Methodology In: Purpose and Procedure in Philosophy of Perception. Edited by: Heather Logue and Louise Richardson, Oxford University Press (2021). © OUP. DOI: 10.1093/oso/9780198853534.003.0013

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-:   

259

Description: phenomenology seeks to describe the experiences that are its starting point. Analysis: this is a controversial label, since phenomenologists often contrast what they are doing with analysis. But I use the term to indicate that phenomenologists do not stop with descriptions of experiences. They also seek a further understanding of those experiences. In particular, they seek to uncover their essences—roughly, their underlying structure, but more of this shortly. As it stands, this general description of phenomenology’s method is not very informative. It can—and has been—understood in different ways. A common misconception takes phenomenology’s focus on experience to consist in introspecting the contents of one’s own mind. Indeed, the Transcendental-Phenomenological Reduction in Husserl’s early thought consisted in suspending one’s commitment to the worldly existence of the objects of one’s experience. Yet, this is by no means the only way to understand the phenomenological method, and it certainly is not Merleau-Ponty’s understanding of it. I will go through each component in turn and explain how it is formulated in his framework, before turning my attention to his account of perception.

2.1 Experience All phenomenologists start with experience as it is undergone from the perspective of its subject. However, this alone does not tell us an awful lot because there are different conceptions of experience available. We can identify three broad stances one might take on the status of perceptual experience: one might think of it as an inner content of consciousness that can remain just as it is, regardless of the nature or existence of the world it seemingly presents to its subject; one can think of experience as being directly of an objective reality; or one can adopt a position ‘midway’ between these. Merleau-Ponty takes the latter view, and equates perceptual experience with what he calls ‘the phenomenal field’. The phenomenal field is analogous to the visual field. One way to understand the visual field is as the region of the world that is currently seen. On this understanding, to describe the visual field is to describe the worldly objects of visual perception as they are visually experienced. The phenomenal field encompasses all the senses; so to describe it is to describe the worldly objects of perception as they are experienced in all perceptual modalities. Merleau-Ponty holds that if one is conducting an investigation that involves describing perceptual experience, one needs to take some initial stance on what it is that one is describing. Furthermore, one will need to pick one of the positions described above because they exhaust the available options (although there will be many different ways of cashing out each of these broad stances).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

260

 -

Philosophers have often tended to assume, at least as an initial starting point, the first position—experience is an inner content of consciousness that can remain just as it is, regardless of the nature or existence of the world it seemingly presents to its subject. This is the view taken—at least on standard interpretations—by, e.g., Descartes in his Meditations (Descartes 1996). But this position is not inevitable. Of course, in that text, Descartes relies on the claim that our senses may deceive us. Such cases seem, at first glance, to be exactly those where the content of experience and the world diverge. So the possibility of our senses being mistaken seems to support the first conception of perceptual experience. However, matters are not so straightforward, because there are different ways to account for perceptual mistakes—e.g., by holding that illusory and veridical perceptions are actually different sorts of states. Merleau-Ponty’s investigation is phenomenological, and as such, the highest ‘court of appeal’ for him is perceptual experience. What experience tells us is the case has to be accepted. This is not to say that he denies the occurrence of perceptual mistakes. Experience can sometimes lead us astray. However, Merleau-Ponty contends that any perceptual mistake can be rectified by further perceptual experience; and it is only because such errors can be corrected that we can classify them as mistakes in the first place. (He holds, in other words, that wholesale scepticism about the world presented to us in perception is not a real possibility for us.) If we describe our perceptual experience faithfully, then one thing that presses upon us is the impossibility of doubting that it presents us with the real world. For if I am able to speak about ‘dreams’ and ‘reality’, to wonder about the distinction between the imaginary and the real, and to throw the ‘real’ into doubt, this is because I have in fact drawn this distinction prior to the analysis, because I have an experience of the real as one of the imaginary . . . we must not wonder if we truly perceive a world; rather, we must say: the world is what we perceive . . . I am open to the world, I unquestionably communicate with it. (2012, lxxx–lxxxi)

Whilst it might be possible to both conceive of perceptual experience as an inner content of consciousness and give an account of our certainty of the world, Merleau-Ponty holds that this is less faithful to the way that experience strikes us. Instead, to capture the experience of immediate openness to the world, we should describe experience as being directly of it. Yet at the same time, MerleauPonty holds that it would be a mere assumption to hold that experience presents that world as it is, in itself. Thus, he also rejects the second stance on experience. This leaves the last option—the claim that when we describe perceptual experience, what we are describing are the worldly objects of perception as they are experienced. Merleau-Ponty’s investigations into perception adopt this general starting point. With respect to any particular experience, however, he leaves it

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-:   

261

open that we might discover that it is not of something real after all, in which case we should obviously revise our description of it to reflect what we have discovered. But this possibility does not alter his methodological starting point. For Merleau-Ponty, the experiences that the phenomenologist studies are the experience of any human subject. They include one’s own perceptions; but also reports of experience collected under experimental conditions; the experiences of children; those of peoples studied by anthropologists; the experiences of those classified as mentally ill; the perceptions of people on psychoactive drugs; and so on. Merleau-Ponty always starts from actual experiences, and is not interested in those that are merely conceivable. This is because his aim is to understand the nature of actual phenomena, which are not revealed just by what people can imagine from the armchair. Just as a scientist cannot gain knowledge of the world by merely imagining what it is like, she needs to go out and study it; so too, Merleau-Ponty holds that if we want to know about the nature of actual phenomena, we need to examine the experiences people actually have. (This is not to say that imagination plays no role in Merleau-Ponty’s method. I will say more about this below.) One place where we see this commitment at work is his treatment of illusion and hallucination. These are perception-like states that do not present the world in a reliable manner. Typically, this is because they present it incorrectly. (It is also possible for there to be illusions and hallucinations that present the world correctly. The difference between these states and veridical perceptions is that the former are unreliable. It is only a coincidence that their content is momentarily accurate. They do not reliably track how things are in the world.) It is often claimed that illusions differ from hallucinations merely in their degree of error, so that illusions are usually inaccurate experiences of an existing object, whilst hallucinations typically present their subject with an entity that does not exist. Moreover, it is claimed that, even setting global scepticism aside, such states are phenomenologically indistinguishable from veridical perception. Theorists then seek to provide an account of perception that accommodates such states.¹ Merleau-Ponty repudiates such an approach. If we want to understand illusion and hallucination, then we need to examine actual examples of these states. We find if we examine experience that illusions are errors that sometimes occur in the ordinary course of perception in cases where the subject is not in the best context for perceiving the thing in question—when she has not got what Merleau-Ponty calls ‘maximum grip’ on it. Illusions can always be corrected by getting in a better viewing context—e.g., moving closer to the object. If the subject is not in the best context for perceiving an object, she may momentarily be unable to distinguish whether she is having a veridical experience or an illusion. But since perception is ¹ Soteriou’s entry on ‘The Disjunctive Theory of Perception’ for the Stanford Encyclopedia of Philosophy (2014) reflects these ideas.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

262

 -

a dynamic, ongoing exploration of the world, rather than a static state, this inability is only fleeting. Hallucinations, in contrast, are extraordinary experiences. Someone might hallucinate as a result of suffering from some pathological condition (Merleau-Ponty focuses particularly on schizophrenia),² or as a result of taking psychoactive drugs (Merleau-Ponty often mentions mescaline).³ What we find when we examine such cases is quite different from what we might predict from the armchair. For example, Merleau-Ponty tells us that in many cases, subjects undergoing schizophrenic hallucinations distinguish between their perceptions and their hallucinations, and in some cases, indicate that they can tell which experiences are real, yet are nevertheless misled by their hallucinations. There is no space here to assess whether or not Merleau-Ponty’s claims about illusion and hallucination are correct. But they illustrate his general approach.⁴

2.2 The Transcendental-Phenomenological Reduction To reach the experiences she studies, the phenomenologist must perform the Transcendental-Phenomenological Reduction, which consists in suspending some attitude or way of thinking. For Merleau-Ponty, something he calls ‘Objective Thought’ is to be suspended. This is a conception of the world as causally determined, and made up of its basic components in such a manner that a reductive analysis of it can be given. One variant of Objective Thought (Empiricism) takes consciousness to be included amongst the physical items of the world—just so much causally determined matter. Whilst another variant (Intellectualism) takes consciousness to be radically different from the rest of the universe and not bound by its causal laws. Merleau-Ponty takes Objective Thought’s characterization of the world to be our common-sense way of thinking, and to be presupposed, in most cases, by the science of his day. However, rather than being something of which we can be absolutely certain, this conception of the world is, for Merleau-Ponty, a hypothesis about its nature that should be subjected to the court of experience. Rather than assuming this is what the world is like at the start of inquiry, we should ask whether this conception is supported by the available data. It is a well-known truism that inquiry is theory-laden. For instance, what one looks for, the experiments one conducts, one’s descriptions of the data one collects, can all be affected to a greater or lesser extent by the theories one adopts. It follows that in order to assess whether Objective Thought is supported by

² See, e.g., (Merleau-Ponty 2012, 349–60). ³ See, e.g., (Merleau-Ponty 2012, 237). ⁴ I provide a detailed account of Merleau-Ponty’s theory of illusions and hallucinations in (Romdenh-Romluc 2009).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-:   

263

evidence, we need data that is not distorted by the assumptions of this hypothesis. Thus to collect the data, we need to put the hypothesis out of play or suspend it. Merleau-Ponty does not think that the Transcendental-Phenomenological Reduction is easy to perform. The way of thinking we are supposed to suspend is—if Merleau-Ponty is right—our common-sense view of the world. As such, it’s difficult to see how far this conception of the world ‘infects’ inquiry so ‘polluting’ our data. This difficulty is compounded by a further worry. Merleau-Ponty’s data include our ordinary day-to-day perceptions. These will, of course, be so familiar, that they are often difficult to notice. As Merleau-Ponty remarks, ‘Nothing is more difficult than knowing precisely what we see’ (2012, 59). He develops a method to contend with both of these difficulties. One should proceed by examining what are, from the phenomenologist’s perspective, extraordinary cases. For many of us, this will include pathological experiences, the experiences of people from significantly different cultures, those had under the influence of psychoactive drugs, and observations made in the course of experiments by scientists. Thus we can now see that the phenomenologist is not just permitted, but required to include a range of different subjects’ experiences in her data. There are a few reasons why Merleau-Ponty adopts this method. First, there is the simple reason that if we want to know about some entity, it’s useful to observe it often, under different conditions—which is to say that it’s useful to collect lots of data about it. Second, by including the experiences of people from significantly different cultures to one’s own, one collects data about the phenomena under examination from people with very different perspectives and preconceptions than one’s own. Their experiences can be compared with one’s own, and with those of others from one’s culture. This helps reveal ways in which cultural assumptions may be informing one’s inquiries. Third, turning to pathological experiences and those induced by drugs, one may initially be puzzled by Merleau-Ponty’s inclusion of them in a general framework for inquiry, since we assume that these distort reality and are not a good guide to its real nature. Exactly what role these play in a particular inquiry varies somewhat depending on what one is investigating—I will have more to say about this below in connection with Merleau-Ponty’s inquiries into perception. But in general, one reason to include these is so that they can be compared with ‘normal’ experience and so bring features of it to our attention that might otherwise have gone undetected due to their familiarity. Fourth, the inclusion of the experimental observations made by scientists is completely unsurprising, given that science is clearly a good way to find out about the world. However, whilst Merleau-Ponty relies on the observations of scientists at various points in his inquiries, his reliance is not uncritical. One the one hand, scientific observations can tell us more about whatever it is that we’re interested in investigating, since scientists can collect observations of it that we do not, or cannot, have in the usual course of day-to-day life. But on the other hand, scientific

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

264

 -

observations may presuppose Objective Thought and be infected with its assumptions. Even where this is not a problem, the way in which the experiment is set up may mean that the experimental conditions distort the phenomenon we are investigating.

2.3 Description Phenomenology’s commitment to describing experience is nothing mysterious. For the theories the phenomenologist puts forward to be defensible, the observations they are based on must be captured as accurately as possible. Thus as far as possible, the data must initially be described, rather than interpreted or explained. There is, of course, a question concerning how far this is possible, given that the words we choose to describe something will inevitably be guided to some extent at least by our interpretation of events, and theories about the things we talk about are often embedded in the meanings of our words. But, for Merleau-Ponty, we just have to muddle along the best we can in the face of this worry.

2.4 Analysis Once the phenomenologist has collected descriptions of a phenomenon, the next stage is to discern its essence(s). Merleau-Ponty takes his notion of essence from Husserl. Whilst the term might imply an abstract ideal object that stands apart from messy concrete reality, this is not Merleau-Ponty’s conception. Essences, in his phenomenology, are instead the meaningful forms or structures that characterize experienced entities, and—importantly—cannot be separated from those experienced entities. At the same time, however, they are not simply reducible to a collection of our experiences of things. Merleau-Ponty uses the idea of Gestalt form (taken from the Gestalt psychologists) to explain this conception of an essence further. The Gestalt psychologists hold that our perceptual experience is characterized by meaningful structures— Gestalten. Even the simplest perception will be of something that is presented as having what Merleau-Ponty calls a ‘physiognomy’. To see how Merleau-Ponty thinks the idea of Gestalt form can illuminate the notion of an essence, we need to consider a different case—that of a Gestalt switch. As the name suggests, this is where the meaning or form of what one sees abruptly alters. Merleau-Ponty describes such an experience in the passage below: If I am walking on a beach towards a boat that has run aground, and if the funnel or the mast merges with the forest that borders the dune, then there will be a moment in which these details suddenly reunite with the boat and become

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-:   

265

welded to it . . . I merely felt that the appearance of the object was about to change . . . The spectacle was suddenly reorganized satisfying my vague expectation. (Merleau-Ponty 2012, 17–18)

Here, Merleau-Ponty’s visual experience is first characterized by a trees-Gestalt, but then flips to a boat-Gestalt. What is significant in the current context is that there is a sense in which nothing of what is seen changes. The vertical structures, in a sense, remain the same throughout. What changes is the meaning they have for him—in the first instance, they are seen as trees, then they are seen as masts. The Gestalt form of the experience is inseparable from it. The form just is the physiognomy of the things that are presented to the perceiver. We might say, it is the way they look to him. Yet the fact that the meaning can alter (the way things look can switch) without there being any corresponding change in the elements of the experience (in our example, the same vertical structures are seen throughout) means that the Gestalt form of the experience cannot be simply reduced to its elements. Instead, the Gestalt form is something that unifies disparate elements. The visual experience characterized by a boat-Gestalt contains elements such as the vertical structures seen as masts, the horizontal elements seen as the deck, the round part seen as the ship’s wheel, etc. The boat-Gestalt—the meaning or physiognomy—is not simply a collection of these elements, but is instead a meaning in which they all partake. In an analogous way, the essence of some phenomenon is a meaning that unifies disparate experiences of that phenomenon. It is something over and above those experiences insofar as it cannot simply be reduced to a collection of such experiences. Yet it does not exist in the absence of those experiences. To see what that means, consider the example of sexual harassment. This cultural phenomenon has not always been recognized, as the following passage, which describes events that took place in the US during the 1970s, makes clear. [Carmita Wood] had worked for eight years in Cornell’s department of nuclear physics, advancing from lab assistant to a desk job handling administrative chores . . . a distinguished professor seemed unable to keep his hands off her . . . [T]he eminent man would jiggle his crotch when he stood near her desk . . . he’d deliberately brush against her breasts while reaching for some papers. One night as the lab workers were leaving their annual Christmas party, he cornered her in the elevator and planted some unwanted kisses on her mouth . . . Carmita Wood went out of her way to use the stairs in the lab building in order to avoid a repeat encounter, but the stress of the furtive molestations and her efforts to keep the scientist at a distance while maintaining cordial relations with his wife, whom she liked, brought on a host of physical symptoms . . . She requested a transfer to another department, and when it didn’t come through, she quit . . . [S]he applied for unemployment insurance. When the claims investigator asked why she had

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

266

 -

left her job after eight years, Wood was at a loss to describe the hateful episodes. She was ashamed and embarrassed . . . Her claim for unemployment benefits was denied. (Brownmiller 1990, 280–1)

When these events took place, people tended to categorize behaviour of this sort as harmless flirting. However, through conversations with each other about their lives, women such as Wood started to discern a pattern in experiences of this sort. They called it ‘sexual harassment’. In Merleau-Ponty’s framework, to discern the pattern referred to as ‘sexual harassment’ is to grasp the essence of these episodes. Discrete instances of behaviour are unified as manifestations of the same cultural phenomenon by the meaning ‘sexual harassment’. This is analogous to the way in which different elements of the perceptual scene in the example above—the sails, the wheel, the wooden planks that make up the deck, the vertical structures of the masts— are unified by a meaning, the boat-Gestalt, such that they appear as parts of a single entity. Gestalten are constituted by their elements in such a manner that they are nothing over and above them, but cannot be reduced to them. Above, I explained how Merleau-Ponty takes this to be so in the case of something such as the boat-Gestalt. The same applies to a cultural phenomenon such as sexual harassment. It is nothing over and above episodes of behaviour such as that of the professor described in the above passage. Remove these episodes and there is no sexual harassment. Yet the cultural essence is not reducible to a mere collection of such episodes. This is evident in the way that Carmita Wood was first unable to express what was wrong with the way the professor treated her. Although she experienced the episodes as ‘hateful’, she could not articulate to others or even to herself why they were hateful. Not only did she lack adequate language to describe them; people had not yet identified the pattern manifest by these incidents. In Merleau-Ponty’s parlance, they had not grasped the essence of the cultural phenomenon. It follows that for Merleau-Ponty, a phenomenon such as sexual harassment is not just a collection of particular instances of behaviour; but a pattern discerned in them by consciousness. The second claim that Merleau-Ponty makes about essences is that they are not deducible from experience. Instead, they must be grasped by a sort of insight. This is also illustrated by the idea of Gestalt form. In the usual case, the form of a perceived entity is given ‘all at once’ as it were—the experience of the Gestaltswitch is exactly such a case. There is not a gradual move from trees to masts; the new boat-Gestalt is given all at once as the perceiver moves closer to the ship lying on the beach. Moreover, whilst one could calculate what it is that one must be looking at, this reasoning to a conclusion is not the same thing as having an experience characterized by the relevant Gestalt form. Concluding from the rustling sounds that there is a lizard camouflaged in the bushes right in front of one is not the same as actually seeing the form of the lizard. Reasoning can

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-:   

267

sometimes help one to see, but it is no substitute for actually seeing. Similarly, the essence of a phenomenon is grasped by insight. Reasoning about it may help, but real understanding always requires a final leap of realization. The early consciousness-raising sessions of women such as Carmita Wood may have involved them discussing—i.e., reasoning—about how to interpret the inappropriate behaviour to which they were submitted at work. But discerning the essence of these episodes nevertheless required a flash of insight, a sudden moment of recognition. The reasoning is no more than an aid to insight. Merleau-Ponty writes, ‘I grasp something through this experience . . . an intelligible structure that imposes itself on me whenever I think of the intentional object in question’ (1964a, 54).⁵ Imagination can sometimes help us to discern essences. By imagining alternatives to the actual situation, we can sometimes gain insight into its essential structure. For example, imagining the offending professor treating her respectfully could have helped Carmita Wood come to the realization that his treatment of her constituted what we now call sexual harassment. However, whilst imagining can play a role in discerning the essence of some phenomenon, one’s insight in that essence—one’s conclusion about its essential structure—must always be confirmed by experience. To summarize, Merleau-Ponty’s method involves the following. One should begin by collecting a number of different experiences of the phenomenon one is investigating. These should be experiences of it from different perspectives—as we have seen, Merleau-Ponty relies on both pathological experiences and scientific observations of the phenomena he examines, together with more mundane experiences. The aim here is to both learn more about the phenomenon in question by observing it (or collecting observations of it made by others) in different contexts, and also to unearth any unjustified assumptions one may have made about it so that they can be suspended, as required by MerleauPonty’s understanding of the Transcendental-Phenomenological Reduction. One should aim to describe the experiences one has collected as accurately as possible. In so doing, one should initially adopt the hypothesis that what one is describing is the phenomenon-as-experienced; in other words, one’s descriptions neither concern a thing-in-itself, but nor are they merely descriptions of some internal content of consciousness. They are instead descriptions of something-asexperienced. Subsequent investigations will then reveal whether the thing one is describing is part of the shared world or whether it is subjective and private. Finally, one aims to provide an analysis of the essence or essential structure of the phenomenon one is investigating.

⁵ I discuss Merleau-Ponty’s ideas about essences, including the Carmita Wood example, at greater length in Romdenh-Romluc (2018).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

268

 -

3. Application of Merleau-Ponty’s Method to Perception In the final section of the paper, I will focus on Merleau-Ponty’s investigations into perception. A full examination of this topic is beyond the scope of the present essay. I thus want to draw attention to some specific issues that arise for MerleauPonty’s inquiry into perception and indicate how he addresses them. The foregoing discussion makes clear how the phenomenologist should proceed: she should begin by collecting a number of experiences of the phenomenon she is investigating; describe them as accurately as possible; attempt to uncover and suspend any unjustified assumptions; and adopt the methodological starting point of taking those descriptions to be of the thing-as-experienced. However, an immediate complication when it comes to investigating perception itself is in determining what should be observed. Studying perception involves observing the means of observation itself. There are two different kinds of thing we might observe in order to learn about perception. One is perception as it is undergone from the first-person perspective, i.e., perceptual experience. Generally speaking, observing something involves going to the relevant bit of the world to look for the thing one wants to examine. Observing zebras, for example, involves going out into the veldt and finding some. Perceptions, however, are not objects that we can come across in the world like this. Thus if we want to examine perceptual experience itself, the challenge is to determine what we should observe. The second thing we might observe in order to learn about perception is the neurophysiological goings-on in the body and its interactions with the environment that, from a scientific point of view, underpin perceptual experience. The challenge in this case is to understand the status of scientific accounts of perception, given the phenomenologist’s transcendental understanding of perception: the world studied by science is (partially) constituted by the perceiver in her experience of it, which means that she cannot be a mere part of it, and a complete scientific account of perceptual experience is correspondingly impossible.⁶ The latter is a substantial and difficult issue that it would be impossible to address in this limited space. I will thus focus on Merleau-Ponty’s investigations of perception that are based on experience as it is undergone from a first-person perspective. Merleau-Ponty’s answer to the puzzle of how we can observe perceptual experience itself, given that we cannot come across experiences as objects in the world, is to hold that we can study perception by examining the ways in which worldly entities are presented in experience. He holds that we can distinguish three different aspects that can be studied.

⁶ For a discussion of this issue, see Ward, this volume.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-:   

269

First, we might examine the properties of perceived objects. The idea here is that when we perceive worldly entities, we are never presented with just their objective properties. Perception is an interaction between the world and the subject. The way we perceive objects is a result of this interaction, and this means that the way they are presented to us has both to do with their nature and also ours. MerleauPonty writes, ‘since the seer is caught up in what he sees, it is still himself he sees: there is a fundamental narcissism of all vision’ (Merleau-Ponty 1968, 139). The properties we perceive them as having thus vary from those that we are more inclined to think of as objective (such as size and shape) to those that we recognize as having more to do with us than with them. The latter includes what are usually classed as aesthetic properties such as the wistfulness of a crumbling wall viewed in the sunshine of a late summer afternoon; affect such as the scariness of an approaching tiger; through to such things as the fuzziness of an out-of-focus text. Merleau-Ponty also holds that we are presented with affordances in perception— we perceive the surrounding world as offering us opportunities to interact with it. Again, the affordances someone perceives are partly to do with what the world is like, but also determined by the kind of creature she is, and the sorts of capacities she has.⁷ Second, we might attend to the structural properties of the perceptual field, such as its orientation—the fact that it has direction, there is an up and a down to it, a right and a left, etc. The structural features of the perceptual field are not perceived as properties of individual objects. But awareness of them is nevertheless tied to awareness of objects perceived, because there would be no perceptual field without any perceived objects. We might say that to perceive is to have a point of view on the world. Clearly, a point of view is not something that can exist in and of itself. One can only have a point of view on something else. But once one has a perspective on something, one can examine the qualities of the perspective itself. Merleau-Ponty’s claim with respect to the structural properties of the perceptual field is very similar. Its structural features are not perceived as the qualities of particular objects, but they only appear in relation to perceived objects. They are general features of the way in which objects are presented in perception. The third thing we might examine is the way in which objects take shape for us in perception. Merleau-Ponty holds, very plausibly, that the function of perception is to give the perceiver a shared world of things. Our sensory experience immediately strikes us as opening us to a world inhabited by other people. The way in which this strikes us means that, for Merleau-Ponty, this is a fundamental feature of our existence, which cannot be seriously questioned or doubted. He thus holds that the role of perception is to present the perceiver with a world in which she can act, and which includes other people with whom she can communicate and join in ⁷ There is debate over the admissible contents of perception, i.e., the range of things immediately presented in perceptual experience. See Cavedon-Taylor, this volume, for an approach to this issue.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

270

 -

shared activities. However, the shared world that is the object of our perceptual experiences is not simply given in a way that admits of no further analysis. One thing we might study in examining perception is how the intersubjective world forms in our experience. There are two important points to note here. First, Merleau-Ponty holds that the perceptual field contains more than just perceived entities that are parts of a shared world. It also includes phenomena—experienced presences—that have not yet developed into objects of particular sorts that are part of a shared world. Examples of these might be the vague something-or-others that are seen out of the corner of one’s eye. I am aware that there is something there, but I experience it as an indeterminate presence. It is only when I turn my head and get a better look at it that it is presented as a being of a certain sort with a particular set of properties. The perceptual field also contains items that will never become part of a shared world such as hallucinatory items. As stated above, Merleau-Ponty thinks of hallucinations as extraordinary experiences that occur when the subject is in an unusual state. They often present their subject with entities that do not really exist. Yet those entities are not, in many cases, presented as somehow internal to her. She experiences them as being out there in the world. Even though the manner in which they are presented is such that the perceiver is, in some cases at least, aware that they are not real parts of an intersubjective world.⁸ The claim that hallucinatory entities are not presented as being internal to the subject, is just the claim that they are presented within the perceptual field. A further kind of item contained within the perceptual field is a ‘pseudo-presence’, where this is essentially an imagined entity. Merleau-Ponty’s work provides for a distinction between intellectual forms of imagination and a sort of imagination that is bodily. When the body imagines an object, it conjures up its pseudo-presence: When I imagine absent Pierre, I am not conscious of contemplating Pierre in an image numerically distinct from Pierre himself. As far away as Pierre might be, I aim at him in the world, and my power of imagining is nothing other than the persistence of my world around me. To say that I imagine Pierre is to say that I obtain a pseudo-presence of Pierre by triggering the ‘Pierre-behavior’. (Merleau-Ponty 2012, 186)

Merleau-Ponty suggests here that he has a set of ‘skills’ for interacting with his friend Pierre constituted by the habitual ways he engages with him. Just as with other motor skills, his ‘Pierre-skills’ have a behavioural and a perceptual component. He engages with him in a friendly, trusting manner. He perceives him as likeable, familiar, and as calling for certain kinds of trusting, friendly behaviour.

⁸ I discuss this at much greater length in Romdenh-Romluc (2009).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-:   

271

When Pierre is present, Merleau-Ponty uses these skills to perceive him as the particular, familiar, likeable, trustworthy being who is his friend. But MerleauPonty argues that he can also exercise these skills when Pierre is not present. Doing so is a kind of bodily imagining of Pierre. He uses his Pierre-skills to conjure up the demands for action that Pierre would make on him if he were really there. This is what Merleau-Ponty thinks of as a pseudo-presence. The second significant point is that perceptual experience has elements that combine in order to present us with objects, and we might inquire into what these elements are and the ways in which they must combine in order to give us a shared world of things. Some care is needed here. Merleau-Ponty repudiates the idea that perceptual experience can be reduced to a collection of basic ‘atoms’. He begins the Phenomenology of Perception (Merleau-Ponty 2012) with a series of arguments designed to show us that this cannot be the case. Perceptual experiences are instead made up of their parts in such a way that the whole is more than their simple sum. But they nevertheless have components. Since perception’s function is to provide us with a world of shared objects, some of these aspects of the phenomenal field are difficult to bring to our attention. Our natural focus is the world of shared things that perception delivers. We tend to pass over the vague presences that have not yet taken shape as things of certain sorts, and we are usually unaware of the components that make up perceptual experience. Merleau-Ponty says ‘perception hides itself from itself ’ (Merleau-Ponty 2012, 59), and references Scheler’s notion of a crypto-mechanism in the natural attitude that has to be disrupted to reach the phenomenal world (Scheler 1973, 73). Given the existence of this ‘crypto-mechanism’, the study of these aspects of perceptual experience requires some specific method for bringing them into view, in addition to the general method of investigation sketched in the previous section. A method used by Merleau-Ponty focuses on painting. The painter who produces figurative works must solve a problem: how to render the perceived world in canvas using shape and colour. In fact, this comprises two problems: the painter must identify the elements of perceptual experience, and then ‘translate’ them into paint on canvas. Her engagement with these problems means that she is particularly well-placed to investigate visual perception. The painter is restricted to using colour and shape to represent the perceived world; colour and shape are the constituents of vision. So in seeking to identify the elements of perceptual experience, the painter is asking specifically about vision. One may wonder, at this point, how this sits with Merleau-Ponty’s insistence that we should think of our senses as integrated, rather than functioning in isolation from one another. The manner in which they are integrated means that perceptual experience is always, to some extent, synaesthetic, such that each sense delivers properties that one might think prima facie correspond to the other senses. Given this view, it is perhaps surprising to find

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

272

 -

him claiming that the elements of visual experience are colour and shape. However, this air of inconsistency vanishes once we realize that the components of visual experience should not be identified with the properties of worldly objects presented to us by vision. Whilst visual experience is constituted by colour and shape, these elements combine to present us with a world of tactile, auditory, gustatory, etc. things. The reason that this is possible is because the seer is an embodied creature with senses other than vision. Whilst the painter asks primarily about the elements of visual experience, which she then tries to capture on the canvas using colour and shape, the result is (if the painting is any good) something that speaks to all of our senses, so that the pictured objects are ones that appeal to touch, taste, and so on. Indeed, it is only as such that they can be recognizable as worldly things. The painter’s task is one of abstraction. She must abstract from the fullness of perceptual experience to identify the elements of visual perception. There are different ways to abstract from any whole, and whilst I have spoken about the elements of vision, Merleau-Ponty is not committed to there being a finite collection of them that every painter must identify. Instead, different artists will pick out different components of visual experience. Similarly, there are different ways to render what is abstracted in paint on canvas. This accounts for the huge variety of artistic styles. An objection one might raise here is that there must be a definitive collection of elements that make up vision, because certain types of images are more like visual experience than others. One might suppose that photographs provide the best likeness of the perceived world, and so the elements of photographs (or paintings that closely approach them) are the closest analogue to the components of vision. However, Merleau-Ponty rejects this view. One worry he raises is that a photograph is a poor likeness of motion. Consider someone dancing. The configuration of his limbs undergoes a series of fluid changes over time. But a photograph can only capture his posture at a single instant, thereby losing the sense of movement. Merleau-Ponty writes, ‘the instantaneous glimpses . . . petrify the movement, as is shown by so many photographs in which an athlete-in-motion is forever frozen’ (1964b, 185). The artist who would create an image of motion is therefore faced with the problem of representing motion using that which does not move. Merleau-Ponty presents Rodin’s solution: Movement is given, says Rodin, by an image in which the arms, the legs, the trunk, and the head are each taken at a different instant, the image which therefore portrays the body in an attitude which it never at any instant held . . . The picture makes movement visible by its internal discordance. (1964b, 185)

The posture of the body is a composite—the parts of the body are presented in positions they took up at different times. The overall configuration is thus not one

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-:   

273

that it is possible for the body to really be in at a time. The impossible posture of the body thus creates a sense of tension in the viewer, even if they cannot say what is off-kilter about the image presented. Motion is suggested by this tension, which, of course, cannot be captured by an undoctored photograph. Merleau-Ponty is similarly sceptical of the claim that geometric perspective is a particularly accurate rendering of the visible world. ‘[T]he painters . . . knew from experience that no technique of perspective is an exact solution and that there is no projection of the existing world which respects it in all aspects and deserves to become the fundamental law of painting’ (1964b, 174). It is widely accepted that even where a perspective drawing has been constructed exactly in accordance with the relevant geometric principles, aspects of it nevertheless appear distorted. Also, objects that are nearby look smaller in real life than they do in either a photograph or a perspective drawing, whilst objects that are far away seem larger in real life than in the latter. As Merleau-Ponty remarks, ‘This can be seen in a movie, where a train approaches and gets bigger much faster than a real train would under the same circumstances’ (Merleau-Ponty 1964c, 14). Thus, one would be led astray if one took the elements of a perspective drawing to be the closest analogue to the components of vision. But none of this is to say, of course, that geometric perspective, photography, and other techniques for constructing an image are useless—far from it. These are all still ways of rendering the perceived world onto canvas, and so, ways of revealing the components of visual experience. Merleau-Ponty’s view is that the painter’s inquiry into vision must not take these techniques for granted. Instead, she must begin by performing a sort of Transcendental Reduction, setting aside theories of painting, to examine the perceived world afresh. From here, she must try to deconstruct visual experience, which involves training herself to notice what usually goes unseen. Merleau-Ponty writes, it is the mountain that [the painter] interrogates with his gaze. What exactly does he ask of it? To unveil the means, visible and not otherwise, by which it makes itself a mountain before our eyes. Light, lighting, shadows, reflections, colour, all the objects of his quest are not altogether real objects; like ghosts, they have only visual existence. In fact they exist only at the threshold of profane vision; they are not seen by everyone. The painter’s gaze asks them what they do to suddenly cause something to be and to be this thing, what they do to compose this worldly talisman and to make us see the visible. (1964b, 166)

It is exactly the painter’s training in deconstructing visual experience that means she is uniquely placed to identify the elements of vision. It remains an open question whether there are similar methods available for studying the other perceptual modalities.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

274

 -

4. Concluding Remarks My topic here has been Merleau-Ponty’s methodology. One of the interesting things about his investigation is the parallel with the scientific method. MerleauPonty’s phenomenology begins with experience. For him, this means the worldas-experienced, rather than some inner content of consciousness. This is also where scientific investigation begins. Scientists make observations, before explaining their experiences by forming hypotheses, which generate predictions for what will be observed in the future. In cases where these predictions conflict with experience, scientists either modify or reject the hypothesis in favour of one that captures more of the observed data. We are reminded of this by Merleau-Ponty when he writes, ‘Everything that I know about the world, even through science, I know from a perspective that is my own or from an experience of the world’ (Merleau-Ponty 2012, lxxii). As we have seen, Merleau-Ponty is committed to dismantling what he calls Objective Thought—the conception of the world as reducible to its most basic components, in its twin guises of Empiricism and Intellectualism. Empiricism is the view that the world is mind-independent, and that consciousness is just one of many things within it, all of which can be explained in terms of scientific laws. Many scientists take this for granted. However, Merleau-Ponty holds that this is really a hypothesis we posit to explain our experience, and as such, it is up for grabs. We can consider whether there is adequate evidence for these claims. We do this by suspending the hypothesis and collecting more data, i.e. by examining experience. Merleau-Ponty takes his phenomenological investigation to be continuous with that of science. For example, he says of phenomenology and psychology ‘[t]hey are not kinds of knowledge, but two different degrees of clarification of the same knowledge’ (1964a, 24). Nevertheless, his methods go beyond that of the scientist. He sees how the artist—often carelessly depicted as the antithesis of science—may have insights into the mysteries of vision. His work contains many discussions of what art can tell us about perceptual experience. There is much more that could be said about his methodology. But I hope to have at least shown here that Merleau-Ponty’s approach to perception and the insights it yields are worth exploring further, and offer a rich resource for philosophers working on perception.

References Brownmiller, S. 1990. In Our Time: Memoir of a Revolution. New York: Dial Press. Descartes, R. 1996. Meditations on First Philosophy. Translated by J. Cottingham. Cambridge: Cambridge University Press.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

-:   

275

Merleau-Ponty, M. 1964a. Phenomenology and the sciences of man. Translated by John Wild. In his The Primacy of Perception. Evanston: Northwestern University Press, 43–95. Merleau-Ponty, M. 1964b. Eye and mind. Translated by Carleton Dallery. In his The Primacy of Perception. Evanston: Northwestern University Press, 159–92. Merleau-Ponty, M. 1964c. ‘Cézanne’s doubt’. Translated by H. Dreyfus and P. A. Dreyfus. In his Sense and Non-Sense. Evanston: Northwestern University Press, 9–25. Merleau-Ponty, M. 1968. The Visible and the Invisible. Translated by A. Lingis. Evanston: Northwestern University Press. Merleau-Ponty, M. 2012. Phenomenology of Perception. Translated by Donald Landes. London: Routledge. Romdenh-Romluc, K. 2009. ‘Merleau-Ponty’s account of hallucination’. European Journal of Philosophy 17 (1): 76–90. Romdenh-Romluc, K. 2018. ‘Science in Merleau-Ponty’s phenomenology: from the early work to the later philosophy’. In D. Zahavi (ed.) Oxford Handbook of the History of Phenomenology. Oxford: Oxford University Press, 340–59. Scheler, M. 1973. ‘The idols of self-knowledge’. Translated by D. R. Lachterman. In his Selected Philosophical Essays. Evanston: Northwestern University Press, 3–97. Soteriou, M. 2014. ‘The Disjunctive Theory of Perception’. The Stanford Encyclopedia of Philosophy, Edward N. Zalta (ed.). https://plato.stanford.edu/entries/perceptiondisjunctive/

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

14 Sensation and the Grammar of Life Anscombe’s Procedure and Her Purpose Clare Mac Cumhaill and Rachael Wiseman

1. Introduction Analytic philosophy of perception of the last twenty or so years might reasonably be framed as a tug of war between two opposing theories—Representationalism (of various stripes) and Naïve Realism. A few dissenting voices come from the sidelines: that of the of the Sense Datum theorist (now usually empirically informed), the adverbialist, and a number of pluralists calling ‘halt’. The grammatical theorist, who had her heyday in Oxford in the 1950s, seems to have deserted the scene completely. This chapter is about one ‘grammatical theorist’ whose absence from the sidelines is, we think, a great loss to contemporary debates. Despite writing two groundbreaking papers on sensation,¹ G. E. M. Anscombe is hardly ever reckoned to be a philosopher of perception at all. A cursory sample of any clutch of the most influential monographs in the philosophy of perception over the last forty years will show her barely referenced (of ten of those on our combined bookshelves— the reader is encouraged to check their own!—Anscombe is mentioned only twenty-one times; five references are in footnotes. Nine references appear in a single text: Howard Robinson’s Perception (1994)). In the recent Oxford Handbook of Philosophy of Perception (Matthen 2015)—all 900 or so pages of it—Anscombe is cited only six times, again, in footnotes and mostly as Wittgenstein’s translator. The (lack of) influence of Anscombe’s philosophy of perception stands in sharp contrast to her dominance in philosophy of action. Her masterpiece Intention (1957/2000) has, at the time of writing, been cited over 5,000 times, and it is generally thought to provide the blueprint for the whole of that subdiscipline. But even when influential attempts have been made to understand perception as a kind of action, Anscombe’s work has not been thought relevant. The broad consensus that perceiving is an activity has led few current philosophers of ¹ Hereafter ‘The Intentionality of Sensation: A Grammatical Feature’ will be referred to as IS. See also ‘The Subjectivity of Sensation’ (1976/1981a). Clare Mac Cumhaill and Rachael Wiseman, Sensation and the Grammar of Life: Anscombe’s Procedure and Her Purpose In: Purpose and Procedure in Philosophy of Perception. Edited by: Heather Logue and Louise Richardson, Oxford University Press (2021). © OUP. DOI: 10.1093/oso/9780198853534.003.0014

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

277

perception to see the specifics of action theory as relevant to a philosophy of perception, at least in published work. Relatedly, until recently, questions proper to an ethics of perception have been rarely articulated—and if they have, they have been framed by moral theorists or philosophers of emotion anxious to pull the domain of value into the nexus of the perceivable, and not much by self-identified philosophers of perception. Here we focus only on making available some of the basic conceptual tools of Anscombe’s grammatical approach and on showing in what respects her procedure offers an alternative to the ontological approaches that still dominate the contemporary debate. Participants in the ‘tug of war’ are unified in their supposition that a theory of perception should tell us the fundamental nature of perceptual experience and, as part of that, what it is that we perceive. It is this supposition that Anscombe rejects and, as we explain, in doing so she reveals an ethical purpose for philosophy of perception. We begin by setting out a fragment of the theoretical background needed to understand Anscombe’s method in philosophy of perception before contrasting her grammatical procedure with the ontological procedure of her interlocutors. In later sections, we identify three mistakes that exegetes have made in approaching Anscombe’s philosophy of perception. These stem from failure to recognize the meaning and implications of Anscombe’s claim that the intentionality of sensation is, as she dubs it, ‘a grammatical feature’. Needless to say, the successful introduction of a new paradigm into philosophy of perception will require more space than a single chapter (see Fish, this volume). Readers should take this intervention as an encouragement to, and promise of, further work.

2. The Raging Nerve Extracted Anscombe was animated by questions concerning a philosophy of perception from her earliest days as an undergraduate at Oxford in the late thirties, long before she developed her interest in the philosophy of action. She attended H. H. Price’s lectures on perception, finding them ‘absolutely about the stuff ’ though she abhorred Price’s conclusions (‘I used to sit tearing my gown into little strips because I wanted to argue against so much that he said’ (Anscombe 1981, viii)). It is possible that such preoccupations made her peculiarly responsive to the force of Wittgenstein’s philosophy when she encountered it later as doctoral student in Cambridge. She recalls: I always hated phenomenalism and felt trapped by it. I couldn’t see my way out of it but I didn’t believe it . . . .the central nerve of it remained alive and raged

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

278

     

achingly. It was only in Wittgenstein’s classes in 1944 that I saw the nerve extracted. (ibid.)

Those classes contained Wittgenstein’s discussions of material that would come to be known as the ‘private language argument’ and which surfaces in those passages in the Philosophical Investigations (1958) that get us to see how our psychological verbs function.² Wittgenstein denies that to grasp the meaning of a psychological verb is to identify its referent—mental phenomena which, on an empiricist model of concept acquisition, such terms might be thought to pick out. Rather, it is to come to participate in forms of linguistic and non-linguistic behaviour. He called these forms or practices—which involve the interweaving of words and action into recognized patterns of human interaction—‘language-games’, and he seeks to describe them. In appealing to language-games that involve psychological verbs and sensation words, Wittgenstein wants to get us to see that such words express concepts that have a different, more complex, character than that of a straightforward classificatory concept—a concept that picks out some class of things.³ This complexity, Wittgenstein thinks, is revealed in, and reflected by, the structure of the practices in which those verbs have their home. Thus, in contrast to the mistaken empiricist impulse to classification (which Anscombe characterizes as: ‘I have got this, and I define “yellow” (say) as this’), participating in a linguistic practice involves far more than identifying an instance or token of a class or type. This is because it involves being acculturated—or trained, as Wittgenstein somewhat provocatively puts it—to continue in a way that is characteristic of the practice. It hence involves not only the use of words but forms of behaviour and interaction with others, as well as with the natural and material environment. Since the intelligible and intelligent use of a word is governed by the norms of such materially realized, temporally extended practices, sense or meaning cannot be privately established, even where the referents of terms used on occasion may be private, as when one is in pain. Anscombe’s contention is that a central task for philosophers of perception is to undertake a ‘grammatical investigation’. That is, to describe the structure of the concept of sensation, a structure that is revealed in the complex patterns of speech and action that manifest an individual’s grasp of the meaning of words like ‘see’, ‘colour’, and ‘appearance’. If they were to turn their attention to that task, two things would follow. First, they would discover that the ontological debate that characterizes the tug of war is ultimately misguided. Second, as we later explain, if briefly, the ethical dimension of our perceptual lives would come into view.

² For pertinent discussion of what ‘getting us to see’ amounts to, see McGinn 1997, especially p. 116. ³ For helpful discussion, see Frey and Frey (2017, 217–19).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

279

3. What Is an Object of Sight? To see what such a grammatical investigation might look like, we now turn to Anscombe’s philosophy of perception proper. Anscombe’s central claim is one that will have a familiar ring to philosophers of perception: verbs of sensation take intentional objects. Like many theorists, she sees this claim as key to unlocking the relationship between the subjectivity of sensation and perception’s epistemic role. The category of ‘intentional object’ reflects Anscombe’s intellectual debt to the Thomist tradition,⁴ and it is this phrase, and this heritage, that has seemed to some to place her philosophy of perception in the tradition associated with her fellow Thomist, Franz Brentano. But while Aquinas (and Aristotle) is indeed an unmistakable and important influence, her deployment of Wittgenstein’s method transforms Brentano’s notion of an intentional object into a grammatical category—a transformation this section of the chapter explains.⁵ The subtitle of her paper, ‘The Intentionality of Sensation’ is ‘A Grammatical Feature’: as we shall see, the switch from an ontological to a grammatical understanding of intentionality takes her away from any position which could be regarded as a forerunner of contemporary intentionalism. Anscombe’s grammatical procedure is evident from the outset: she sets about illuminating the status of an ‘object of sight’ not by reflecting on perceptual experience, but by examining the way a teacher might elicit understanding of the grammatical concept of a direct object in her pupils. She imagines the teacher proceeding as follows. The teacher gives her pupils a paradigm sentence: ‘John sent Mary a book.’ Then she asks them: ‘What does the sentence say John sent Mary?’ On getting the pupils’ response ‘a book’, the teacher replies: ‘That is the direct object.’⁶ We might imagine the teacher giving a wide range of sample sentences and asking the same question in respect of each. The pupils ‘catch on’ to the grammatical concept of a direct object when they know how to go on—they return the correct answer in novel cases and are able to construct cases of their own. The understanding that the teacher elicits is, in an important sense, already manifest in the pupils’ ordinary linguistic practice. That the pupils are competent language users, able to construct and employ sentences containing transitive verbs, is a precondition for her teaching. In that sense, the pupils already operate with the grammatical category, ‘direct object’. The teacher’s lesson makes explicit this understanding by introducing the term, ‘direct object’, and teaching them its

⁴ For discussion of Anscombe’s Thomism in the context of her philosophy of action, see Schwenkler (2019, 157–61). ⁵ See Poulvet 2008 and Geach 1957 for comparative discussions of Wittgenstein’s and Aquinas’s treatment of intentionality. ⁶ See IS, 6–9.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

280

     

use. Once they have mastered the use of this term, they can employ it to talk about features of their linguistic competence. We can see how different the character of the concept ‘direct object’ is from that of a classificatory concept (like ‘book’) by reflecting on the teachers’ procedure. In her lesson, the teacher draws attention to the relevant grammatical category by posing questions that require the pupils to reflect on the parts and structure of a familiar sentence. ‘What does the sentence say John sent Mary?’ The concept of a direct object is not, and could not be, ostensively taught: the teacher does not, that is, pick out things that form a class. One way to put this is to say that the word ‘That’ functions differently in the teacher’s response ‘That is the direct object’ than it would in an ostensive demonstration ‘That is a book’: only in the latter case is ‘that’ a referring expression. A pupil who raised the question, ‘Which book?’, would show that she had not understood this. This question lacks application because the response that gives the direct object—‘a book’—does not refer to a particular worldly book, but rather identifies a part of the sentence. But note that although the term ‘direct object’ identifies a part of the sentence (viz. ‘a book’), it does not refer to the words ‘a book’ either (another kind of thing that could be picked out with the ostensive use of ‘That’). As Anscombe notes, it is not the case that the sentence says that John gave Mary a bit of language! In sum, direct objects are not things that we find in the world; rather, when we are talking about our talk about the world, we can pick out a certain structure or pattern in that talk by pointing out a particular way in which verbs are combined with other terms. It is here that the concept of a ‘direct object’ is to be found, and this is what Anscombe means by calling it a ‘grammatical category’.⁷ Anscombe’s profound challenge to both sets of participants in philosophy of perception’s tug of war is the following: each treats the concept of an object of sight as if it were a classificatory category. But for Anscombe, an object of sight is the direct object of the verb ‘to see’. As such, it too is a grammatical category. To get a sense of what she has in mind here we can imagine our teacher giving a lesson to teach her class the meaning of the phrase ‘object of sight’. Teacher: Pupil: Teacher:

‘Mary sees a book.’ What does the sentence say Mary saw? A book. That is the object of sight.

The understanding that the teacher elicits is, once again, already manifest in her pupils’ ordinary linguistic practice; for example, in her ability to say what she sees and to ask another what they see. That the pupils are competent language users, able to construct and employ sentences with the grammatical structure of ‘Mary

⁷ Compare Frege on the concept ‘horse’; for discussion see Frey and Frey (2017, 221–3).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

281

sees a book’, is a precondition for her teaching. So, in this sense, the pupils already operate with the grammatical category, ‘object of sight’, before they are taught the use of that term. The teacher’s lesson, if it is successful, gives them a grammatical term, ‘object of sight’, which enables them to talk about the linguistic mastery they already possess. They can now not only talk about what they see but talk about their talk about what they see. Anscombe describes the use of ‘object’ as she means it here as the ‘old’ use, a use that nevertheless remains familiar in phrases such as ‘object of desire’. The ‘modern’ use, conversely, picks out objects of the sort that we might find in a person’s pocket—worldly stuffs like pencils and keys. Anscombe’s old use of ‘object’ is, however, grammatical—an object of sight is the direct object of the sensation verb ‘to see’. Accordingly, the pupil who responds to the teacher’s questioning by casting around looking for the worldly book that Mary saw—an object in the ‘modern’ sense—has not grasped the fact that ‘object of sight’ is a grammatical category. For Anscombe, both the Direct Realist and her historical opponent, the Sense Datum theorist make a mistake akin to that of such a child. In learning that the sentence says that Mary saw something and forgetting that ‘object of sight’ is a grammatical category, they ask instead after what this something is that Mary saw, or what it is that she really saw. That is, they ask, ‘What did Mary see?’, and treat the answer as returning an object in the modern sense: a worldly entity or an entity of some other kind. Anscombe notes that this is bound to lead to philosophical confusion in cases of error, illusion or hallucination, where Mary’s answer to the question ‘What do you see?’ does not describe anything in her vicinity.

4. Anscombe’s Procedure Suppose Anscombe were right in her claim that ‘object of sight’ is a grammatical, rather than a classificatory category. How should we then proceed in philosophy of perception? Above we emphasized that the teacher’s procedure works only for a pupil whose linguistic competence already involves mastery of the verb ‘to see’. If Anscombe is right, it is then this very mastery that is the proper subject matter for philosophers of perception, and the term ‘object of sight’ is one of the lexical tools that the philosopher will use in her description of that mastery. Armed with the term ‘object of sight’, the philosopher of perception can now proceed to describe this linguistic competence by asking what patterns the verb ‘to see’ imposes on its (grammatical) objects. One line of investigation will ask which putative ‘x’ in sentences of the form ‘Mary sees x’ or ‘I see x’ are intelligible (what are the possible objects of sight). Another will investigate whether, where some ‘a’ is an object of sight, linguistic competence commits us to accepting other objects, ‘b’, ‘c’, ‘d’. For example, whether ‘I see a rabbit’ commits a speaker to, e.g. ‘I see a mammal’, ‘I see

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

282

     

a grey fluffy shape’, or ‘I see a living thing’ (what are the relations between objects of sight).⁸ The first line of investigation will draw the limits on what we see from within the linguistic practice; the second will reveal patterns of inference within the conception of sensation (its geometry).⁹ In short: the task for the philosopher of perception is not to describe the character of what we see, but to describe the character of our descriptions of what we see. That is, it is to describe the character of ‘objects of sight’. Because the notion of linguistic competence includes more than mere use of words—as noted above, we are speaking here of language-games and not mere sentential structure—this investigation into patterns of intelligibility and inference will need to study words in use, and will not be limited to the question of whether a sentence is grammatically well formed or felicitous. Analogously, a study of the concept length would encompass not just a study of well-formed sentence-types containing the word ‘length’ and its cognates, but also of the practices of measuring, estimating, building, etc., into which the use of those sentences is interwoven. It might be of interest to such a study to discover, for example, that the question ‘How long?’ can be intelligibly asked of a corpse but not of a living human adult.¹⁰ To begin to see how such an investigation might proceed in relation to the verbs of perception, consider a non-perceptual transitive verb, ‘to donate’. There are conditions on the intelligible use of that verb that go beyond the simple sentential structure: subject, verb, object, object (e.g. ‘John donates food to the homeless’). For example: the subject must be sentient and capable of intentional action (a machine cannot donate its time to a cause, nor a germ its disease to a cat). To be a donor, one must also have an understanding of what is needed or wanted by those to whom one is donating. A monkey cannot donate its blood to medical science, nor a baby its rattle to a dog. This latter point brings into view constraints on intelligible objects of the verb ‘to donate’: a saucer of mud, a punch in the face, or a toenail clipping cannot typically be donated—the verb’s possible objects are limited to things wanted or needed. Here, further patterns and connections will come into view. These patterns and connections are part of what a person with linguistic competence knows—or part of what is encoded in the linguistic competence that is manifested in her participation in practices of giving and receiving. If I encounter you leaving vegetable peelings outside a fire station, your utterance ‘I am donating food to the homeless’, though well formed, will suggest something awry.

⁸ Compare Frey and Frey (2017). ⁹ Compare Intention (Anscombe 1957/2000). See Wiseman (2016, ch. 3) for a discussion of the form of philosophical enquiry in Intention. ¹⁰ See Anscombe (1976/1981b, 117).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

283

In ‘The Intentionality of Sensation: A Grammatical Feature’, Anscombe makes a start on just this investigation. She observes that objects of the verb ‘to see’ display a pattern that is found elsewhere and which, in fact, is characteristic of a large class of psychological verbs. She draws our attention to this pattern by saying: the verb ‘to see’ is an intentional verb and takes an intentional object. Note again that ‘object’ here has its grammatical and not its ontological meaning, so, to emphasize, what is being said to be ‘intentional’ is not a class of things but a class of direct objects—here is a first move in a philosophical description of linguistic mastery, made by employing ‘direct object’ as a grammatical category. This is why Anscombe calls the intentionality of sensation ‘a grammatical feature’. The use of the word ‘intentional’ to describe this pattern is Anscombe’s stipulation, but she chooses it because the pattern displayed by (grammatical) objects of sight is also found in descriptions of intentional action. This might at first seem astonishing, but a little reflection reveals the source of the shared form. Anscombe notes things about descriptions of intentional actions: 1.A Not any true description of what you do describes it as the action you intend: only under certain of its descriptions will it be intentional (‘Do you mean to be using that pen?’—‘Why, what about this pen?’—‘It’s Smith’s pen.’—‘Oh Lord, no!’) 2.A The descriptions under which you intend what you do can be vague and indeterminate. (You mean to put the book down on the table all right, and you do so, but you do not mean to put it down anywhere in particular on the table.) 3.A Descriptions under which you intend to do what you do may not come true, as when you make a slip of the tongue or pen. ‘You act, but your intended act does not happen.’ These features are likewise broadly found in descriptions of what is seen (that is, are features of objects of sight): 1.P Not any true description of what you see, is a description under which you see what is seen. (‘See that man in the red coat? He’s the mayor’—‘The mayor? I’d never have guessed!’ 2.P The descriptions under which you see something may be vague and indeterminate. (‘I see a vase of flowers’—‘How many flowers?’—‘Loads! More than 20’) 3.P Just as the action you intend to bring about may not occur, the object you see may not exist (the possible non-existence of the object is the analogue of the possible non-occurrence of the intended action). (‘I see pink elephants!’)

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

284

     

The following case illustrates the presence of these features in descriptions of action and perception (adapted from IS, 9–10). A hunter is hunting in a forest. Peering into the wood he says ‘I see a shadow moving in the trees!’. ‘What is it?’ asks his fellow hunter, ‘What can you see?’. ‘A stag!’, the hunter replies. He takes aim, and shoots. But what he saw was, in fact, his father out for a stroll in the forest. He shoots his father. Later, in a law court he says, weeping, ‘I saw my father’. His friend, called as witness, says, ‘He aimed and shot intentionally, and he did shoot his father. But he didn’t shoot his father intentionally—he never saw his father.’

It is true that the hunter saw his father, something he came to realize after the event and that he weepingly acknowledges in the law court. But while the description ‘his father’ was true of what he saw—what he saw was, in fact, his father—his first-person report ‘I see a stag’ is not untruthful. Indeed, it is the fact that he saw his father under the false description ‘stag’ that is explanatorily relevant: it explains why he took the shot. The vignette contains multiple intelligible uses of the sensation verb ‘to see’. Using the term ‘object of sight’, we can describe connections between these uses by describing relations between the objects of sight. Note that these relations will not be relations between two distinct relata (like the relation between a table and a book), but rather logical or inferential relations between direct objects (like the relation between ‘a gift’ and ‘a book’ in ‘John gave Mary a book’). Let us see how. The object of sight in the initial statement ‘He sees a shadow moving in the trees’ is ‘a shadow moving in the trees’. This object is both the description under which the hunter sees (‘a shadow moving in the trees’ is a description that he would give of what he sees) and a true description of something that is there in the situation and which might be seen by others (there is, indeed, a shadow moving in the trees). Anscombe calls the first kind of object (one that gives the description under which what is seen is seen) an intentional object, and it is with respect to these objects that the pattern found in descriptions under which what is done is done is replicated. In ‘intentional object’, ‘object’ is used in its old, grammatical, use. She calls the second kind of object a material object. The ‘material object’ is a true description of something there to be seen. So here, the intentional object, ‘a shadow moving in the trees’ is also a material object. The hunter’s tragic acknowledgement in the courthouse ‘I saw my father’ records what Anscombe calls a merely material use of the verb ‘to see’. Here, the description ‘my father’ is not an intentional object—the hunter did not see the shadow under that description—but is nevertheless a true description of what the hunter saw. Anscombe points out that whenever the verb ‘to see’ is employed with a merely material use, it is always legitimate to ask: ‘What did you see in that you saw (e.g.) your father?’ For a merely material use of the verb ‘to see’

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

285

to be intelligible, there must be some answer to that question which gives an intentional object. (Note: the previous sentence is an example of the sort of description that will constitute a philosophy of perception, on Anscombe’s view.) In contrast, the hunter’s first-person report, ‘I see a stag!’, gives the description under which he saw his father, and so ‘a stag’ is an intentional object. The tragedy of the case is that this intentional object was merely intentional, though the hunter took it to be also material. Anscombe’s vignette illustrates the three features which should be recognizable to any seer. Not any true description of what you see is a description under which you see it (P1). Like descriptions of intentional action, descriptions of what is seen essentially evoke the point of view of the agent or subject. While I may be able to give many true descriptions of what you, in fact, see, only some of those are descriptions under which you see what is seen. ‘A man born in Liverpool’, ‘Your father’, ‘A mammal weighing 80kg’ may be true descriptions of what the hunter in fact sees (they are all descriptions that are true of his father), but none gives the description under which he sees that which is, in fact, his father. Whether or not there were stags to be shot in the vicinity, there was no stag at the place at which the hunter took a stag to be: just as the action you intend to bring about may not occur, the (intentional) object of sight you see may not exist (P3). A cautious or more experienced hunter may have responded differently to his friend’s query. ‘I’m not sure, there is a moving shadow in the trees—but I can’t see more than that, perhaps there is something there.’ The descriptions under which you see something may be vague and indeterminate (P2). What Anscombe is concerned to reveal is that the question ‘What do you see?’ has its life in the language-game that is characteristic of the verb ‘to see’, and is answered by giving a description of what one sees (the object of sight), something the story also illustrates. In the normal case, however, the description will be true of something there to be seen, and so will be a material, as well as an intentional, object. This second claim, about what is normal, is also a contribution to a philosophical understanding of perception, being as it is one part of a description of the mastery that is its topic. We can now state explicitly the mistake Anscombe diagnoses in the positions of her mid-century dialectical opponents, although, as we shall see, she characteristically concedes that both ‘have a great deal of point’ (IS, 13). Both the Ordinary Language philosopher and the Sense-Datum Theorist extract the question ‘What do you see?’ from the language-game in which it has its home (and which, for Anscombe, they should be seeking to describe), and treat it instead as a classificatory question that is to be answered by returning a type of entity. One way of putting the observation that the Ordinary Language Philosopher and Sense-Datum Theorist take the question to be ontological (albeit while glossing over important differences between them even with respect to this) is as follows: both fail to recognize that in its primary use an answer to the question

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

286

     

‘What do you see?’ gives an intentional object, a (grammatical) object of the verb ‘to see’ that has the features described in (1.P)—(3.P). Failure to recognize this—a failure which extends to the Ordinary Language Philosopher despite his superficial methodological closeness to the grammatical theorist—leads to a search for a kind of thing. Now a theorist’s sensitivity to, and concern for, the three features Anscombe notes will determine which species of ontological theory he is drawn to. The Sense-Datum Theorist takes very seriously (1.P)—(3.P), and so refuses to identify ‘what is seen’ with (what Anscombe calls) the material object. But without the grammatical category of an intentional object, his only option is to reify intentional objects into descriptions of some other kind of thing—some thing that cannot but be there in the whole situation, whether or not that description matches anything that might be publicly seen. This is the role for sense data. The Ordinary Language Philosopher by contrast ignores or seeks to explain away (1.P)—(3.P), preferring instead to advert directly to the worldly objects that perceivers, when genuinely perceiving, confront. And this is why, according to Anscombe, though such a philosopher will recognize material uses of the verb to see—indeed, the method of such a philosopher is to chart such uses—he will not allow intelligible uses of ‘see’ that are merely intentional, or even those uses that are intended to be material but miss their mark. On this view, the hunter’s use of ‘see’ in ‘I see a stag’ is faulty. Concomitantly, such a theorist may well insist that the hunter saw his father, tout court; the testimony of the hunter’s friend must be dismissed, as well as that of the hunter. We will see the difficulty for this position shortly. To close this section, we circle back to the thought that, if the exposition we have offered is broadly right, Anscombe should in no way be counted as a forerunner to contemporary representationalism. Representationalism is an ontological theory of perception—it tells us that what is seen are represented properties, features, relations, perhaps under certain manners or modes of presentation, which may or may not be instantiated. That Anscombe supposes there is an intelligible use of the verb ‘see’ where the application is merely intentional—the use is not intended to give a material object of sight, an object that could in principle be available for others to see at a sensory context—implies nothing about what she takes the nature of hallucinatory (and perceptual states) to be. She offers no positive ontology of perception; this is not what she is up to.

5. Two Ontological Errors in Brief Anscombe’s position is, undoubtedly, subtle and complex. The unfamiliarity of the grammatical register of Anscombe’s philosophy of perception—contemporary philosophers without a background knowledge of Wittgenstein’s later work will find her moves neither natural nor logical—has led to a number of interconnected

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

287

exegetical errors. These errors have combined to prevent subsequent philosophers of perception from recognizing the significance of her intervention. We gloss two errors here, picking up the third in section 6. The first error can be quickly stated. A failure to recognize that ‘object of sight’ is a grammatical, rather than ontological, category for Anscombe may prompt questions about the mode of existence and relations between intentional and material objects of sight, as well as about the identity conditions of the former. What are the identity conditions for intentional objects? What are the modes of existence of intentional and material objects? In what relations do material and intentional objects stand? For Anscombe, the questions like ‘When is object a the same as object b?’ will be answered by a very different procedure depending on whether ‘object’ is used in its old or modern sense, and in some cases, attention to the linguistic practice that the philosopher of perception seeks to describe will show why the demand for an answer is misplaced.¹¹ Take the question ‘What are the identity conditions for intentional objects?’ There are, Anscombe says, cases in which questions concerning the identity of an intentional object have ‘some interest’—and note these questions are questions of practical and not philosophical interest to people engaged in language-games characteristic of our everyday commerce with the verbs of perception. Where the intentional object is also a material object, questions of the identity of the former are usually taken as questions of the identity of the latter: if you see your father (intentional object) and I see my boss (intentional object), the question of the identity of those objects reduces to the question of whether your father (material object) is my boss (material object). The question of the identity of intentional objects takes on a less prosaic form in areas of linguistic practice where, even in the presence of a shared material object, the reduction is ruled inappropriate. For example, if you and I are looking at together at a painting that strikes me as naïve and gauche and you remark ‘I see echoes of German Expressionism in this work’, I might ask, shocked: are we seeing the same thing? Is your visual impression very different to mine? Are you colour-

¹¹ Johnston (2004) makes such a demand. He finds Anscombe’s notion of intentional object ‘perfectly harmless’—he sees that it does not turn direct objects into things. Yet he also says it is ‘jejune’ because it fails for him to meet what he takes to be a primary explanatory task for a philosophy of perception. He writes: ‘Anscombe offers us no way of making sense of non-trivial claims of identity of intentional objects across [perceptual] reports.’ It is not clear what non-trivial might mean in this context, though it is plain that Johnston is looking for something more substantive than ‘that is just how our language goes.’ This suspicion is raised by the peculiar charge he raises against Anscombe. Johnston is taken by the familiar theoretical thought that hypothetical experiential transitions between hallucinatory and veridical states could be ‘seamless’. This raises epistemic questions—how can I know if I am now hallucinating or now genuinely perceiving? But Johnston’s problem is different. He wants to tell us in virtue of what it is that such transitions are seamless—what makes them so. Granted we may be apt to offer the same perceptual report with the same intended use in both cases, but, asks Johnston, what makes that so much as possible? This, he thinks, is what Anscombe’s grammatical account cannot give us. And he is right. Anscombe’s project is not ontological.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

288

     

blind, or wearing the wrong glasses, or not really looking? Or, to give another example, I might wonder on looking again at my lover after discovering his infidelity whether I any longer see the same face: his visage always struck me as innocently beautiful, but now I see his shifty expression and his arrogant gaze. In cases where we both see a pink elephant (intentional object) when no pink elephant (material object) is there to be seen, our ordinary linguistic practice offers a number of procedures for deciding whether we see ‘the same’ (causal, phenomenological, counterfactual) but the question is ultimately left open by the practice. A second way in which the failure to take on board the grammatical character of Anscombe’s investigation reveals itself, is in a lack of sensitivity to the importance of the first-personal character of perceptual reports in Anscombe’s methodology. This reveals itself, for example, where Anscombe’s position is reconstructed using only the third-person pronoun, as is relatively common in brief expositions of Anscombe’s position, or, relatedly, without due notice of the role of the interrogative mode (‘What do you see?’) in identifying objects of sight. For Anscombe, the question ‘What do you see?’ and its answer ‘I see x’ are essential parts of the exposition of the grammatical category ‘object of sight’; and a description of the formal character of this exchange is a central goal for a philosophy of perception.¹²

6. Anscombe’s Purpose Earlier in the paper, we outlined the background to Anscombe’s procedure. In this section we discuss Anscombe’s purpose by bringing her into fleeting dialogue with the disjunctivist, and concomitantly a third exegetical mistake. Our discussion will necessarily be brief, but it aims to highlight the richness of what Anscombe’s philosophy of perception offers. Recall, 3.P says that, just as the action you intend to bring about may not occur, the object you see may not exist. This, however, is something that the disjunctivist denies—in cases where the object does not exist, he says, you only ‘seem’ to see an object (or some other such locution). For him, cases of the kind in 3.P. are the ‘bad cases’ and involve error and illusion. Disjunctivism, as such and broadly speaking, has at its centre an epistemic notion of the ‘goodness’ and ‘badness’ of perceptual

¹² While Howard Robinson is one of the few theorists to discuss ‘The Intentionality of Sensation’ in any detail, in his Perception (167–72), his reconstruction of Anscombe’s argument is striking in eliding completely the first-person pronoun. This suggests he does not recognize that for Anscombe the question ‘What do you see?’ and its answer ‘I see x’ are essential to identifying the intentional object of sight. It suggests too that he does not appreciate that when a perceiver reports what they see, the description they give can be either intentional or material depending on their use of the verb ‘to see’.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

289

experience. A good experience involves acquaintance with that which can be represented, truly or falsely. A bad experience forecloses such acquaintance. Adopting Anscombe’s grammatical procedure, we can now allow that the question of whether a case falling under (3.P.) is a ‘bad’ case is sensitive to the actual circumstances in which the perceptual episode takes place, circumstances that include facts about the perceiver and her intentions. Similarly, the question of whether a case in which the intentional object is a material object is a ‘good’ one, can only be addressed if the circumstances of the perceptual episode are known. Indeed, as we will see, there may be good cases in which what you see does not exist, and bad ones in which it does. This is because, for Anscombe, the goodness or badness of perception is not exhausted by the epistemic distinction, veridical/non-veridical, but can accommodate a practical or ethical evaluation. We can approach this bold point—more work will be needed elsewhere—by considering a complaint against Anscombe made by Travis (2015), one that will have purchase with many contemporary philosophers of perception. Travis objects that ‘see’ is a success verb; as such there is no use of the kind described in 3.P. This objection is an opportunity to make explicit a feature of Anscombe’s account that might be read as a kind of proto-disjunctivism. We have stressed that verbs of sensation are intentional—that is, any intelligible use of these verbs implies an intentional object (even if the use itself is merely material). However, when we consider the language-game in which the question ‘What do you see?’ has its home, we should recognize that the point of that language-game—the point of our talk of what we see—is to communicate about our shared, publicly perceivable world. As such, the most usual use of a sensation verb is its material use. The purely intentional use of the verb ‘to see’ is one that, as we might put it, is a development from that more primitive language-game characterized by the question ‘What do you see?’ where the expected answer is a description of a joint object of perception. The idea that our use of the verb ‘to see’ contains these layers of sophistication is not a feature special to perception, but a feature of the idea of a practice. When a practice is complex—as human practices invariably are—we should also expect that a learner’s grasp of the practice to deepen over time as she comes to ‘get the hang of ’ the practice insofar as she needs to. And learners will also naturally differ in their competencies and interests. What one must grasp in order to join in a game of rugby in the park falls far short of the sort of the understanding of that game that is possessed by a professional player who has dedicated herself to participating in the sport, and who structures her life—her diet, exercise, sleep patterns, relationships, leisure—around its demands. Similarly, sensitivity to the structures of a linguistic practice can develop and deepen over time, along with one’s grasp of the concepts and words that shape those practices. We start by teaching a child the material use of the verb to see—a use the mastery of which is

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

290

     

the use of the intentional verb ‘to see’. Later, she will learn to talk about cases where things ‘seem’ or only ‘look’ a certain way, or cases where things do not look as they are; we may begin to offer comparisons between things ‘looking the same’ or ‘looking different’. Now she learns the merely intentional and merely material uses. If she becomes a painter or an ophthalmologist or a magician, her mastery of some areas of this practice will deepen. (If she becomes a philosopher, she will need to acquire terms with which to describe those specialisms: the term ‘object of sight’ will be among them.) How our concepts of sensation are elaborated and deepened, then, depends on our ‘getting the hang of ’ certain patterns of inference and enquiry, patterns, which are mediated and augmented in all sorts of ways by essentially perceptible artefacts, artefacts that are designed to be perceived and even contemplated, as well as by perceptual media of various degrees of technological complexity. Ophthalmologists test our eyes by asking if we can discern shapes against backgrounds and their locations. When I say ‘I see a red line to the left of a green one’ the doctor does not correct me: ‘No you don’t, the green is to the left of the red!’. We have critical practices around shared experience of artworks or designed artefacts or items of personal style—we may discuss and offer justification of what we can see and evaluate what things look like. A full philosophical description of the grammar of sensation will account for all of these forms and the practices that produce them and that they produce. As will now be clear, ‘asymmetry’ in application of the verb ‘to see’—the fact the purely intentional use of the verb is learnt after the material use—is not the asymmetry the disjunctivist had in mind. Further, Anscombe’s priority of material over merely intentional uses of the verb ‘to see’ does not attribute goodness and badness to individual perceptual experiences depending on whether they do or do not represent or present features of the environment. My report ‘I see a red line to the left of a green one’ need not be a bad case, even though there is no red line to the left of the green one—this is seen from the context. But it should now be clear too that there is nothing in 3.P, nor in this refusal to employ only an epistemic notion of ‘goodness’, that threatens the fact that, in the normal case, ‘see’ is a success verb after all. Travis seems to suppose that because, for Anscombe, the verb ‘to see’ can have purely intentional uses, this precludes a conception of perception whereby perception is, after all, ‘a way of providing acquaintance with that which can be represented, truly or falsely, as being such-and-such ways’ (2015, 62). Yet on Anscombe’s view, many sincere expressions of visual self-consciousness are neither intended to be nor should be treated as attempts to provide a material object of sight. And this is the case, even though the possibility of such uses depends on shared linguistic practice that begins from the material use of the verb. So, what should we say about such sincere expressions that are so intended (that is, that are attempts to give a material object of sight) but which miss their mark? Are they then bad? Again, Anscombe’s position draws us back to

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

291

the particular circumstances in which what is seen is seen—and allows for some manoeuvre. To bring this out, consider Travis on the experience of a person suffering an hallucination: Of course, someone who suffers hallucinations (a schizophrenic, say) may be said to ‘see’ things that are not there. But here, I think, scare-quotes are important. You may ‘see’ ghosts, or lions in the kitchen, when there are none. The problem then is that you are ‘seeing’ things which are not there to be seen; that is, not seeing anything. (2015, 58)

Of course, where a person who is hallucinating reports seeing lions in the kitchen and her intended use is material, it is true that there is no material object of sight— this is just what we mean when we speak, ordinarily, of an hallucination. But the grammatical framework provided by Anscombe allows us to say much more about the situation, about what may be going wrong or right, and about how we might respond in the circumstance. So as not to foreclose an investigation, let us dispense with the loaded term ‘hallucination’ and with the psychiatric diagnosis. Instead, picture a familiar case in which a young child with a rich imagination insists ‘I see lions in the kitchen’ (her imaginary pet lions perhaps) or ‘I see ghosts under my bed.’ On the grammatical approach to perception, the philosopher’s engagement with the case does not end with the diagnosis that the sensation verb has no application because no lions or ghosts are there to be seen. It does not end with the diagnosis that what we have here is the ‘bad case’—bad because it is on the wrong side of the disjunct that has phenomenally indistinguishable veridical perception on the other. Rather, we can go on to ask how this use of the verb ‘to see’ fits into the pattern of use that has at its centre cases which involve shared visual experience with others. This returns us too to the import of Anscombe’s interrogative mode. The parent might seriously query ‘And what do you see now?’—here deploying the intentional use. She might offer gentle reminders: there are no ghosts, you are not seeing ghosts—now with the material use in mind. And perhaps the child even might come to say ‘I am not seeing ghosts’ even while she ‘sees ghosts’, as a way of trying to reassure herself and bring herself back into a shared visual experience with others. Or, the parent may wish to encourage and foster her child’s imaginative capacities—she may provide her with paint and paper and ask her to paint the lions so that she might see them too. This brief discussion of purely intentional uses of the verb ‘to see’ shows up one way in which the disjunctivist’s purely epistemic use of ‘good’ and ‘bad’ cases can lead us away from attending to the richness of our perceptual language-game, and the concrete circumstances in which perceptual experience takes place. Note how very different things are in the ophthalmologist’s surgery to in the painter’s study

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

292

     

or the child’s nursery. These differences are part of the structure that is Anscombe’s topic. Wholly intentional uses can be fully accommodated in the weft and weave of mature human practice. We end, then, by reflecting on the ethical dimension of our perceptual practices, a dimension that can now be brought into view. What then could be a ‘good’ or ‘bad’ case, by Anscombe’s lights? Since this large question will take us too far from concluding, we close with only a sketch, one that brings Anscombe into conversation with her onetime interlocutor: Iris Murdoch. In ‘Modern Moral Philosophy’, G. E. M. Anscombe writes: It is not profitable for us at present to do moral philosophy; . . . it should be laid aside until we have an adequate philosophy of psychology, in which we are conspicuously lacking. (1958/2005, 169)

The question of how Anscombe’s account of intention provides the sort of philosophy of psychology she sees as necessary before work on ethics can begin has been much debated. This is natural given a modern tendency to take the subject matter of moral philosophy to be overt intentional action. However, it is well known that the Aristotelian ethics Anscombe advocates in that paper has at its centre the idea that the virtuous person will not just act differently but she will see differently. Her actions are good and, to borrow a way of framing things associated more with Iris Murdoch, her vision is just. But these two aspects of her character are not accidentally connected. Anscombe’s later work distinguishes theoretical from practical truth.¹³ Roughly, an action that possesses practical truth or, better, is practically true is one the goodness of which derives from, or is, inherited by the goodness of its ends.¹⁴ Seeing fits into this pattern. Where one’s ends are not good, by whatever objective standard one appeals to here, the episodes of seeing and of perceptual activity that are directed toward that end, however indirectly, are not good either. Murdoch’s moral psychology explores in great detail what might be called the fallen nature of our sight—we are biased and prejudiced; we overlook what is there to see; the descriptions under which we see what is there to be seen are often unjust; her treatment of the case of ‘M & D’ in ‘The Idea of Perfection’ is her central case.¹⁵ But she also makes room for the possibility of a change in, and ultimately, the perfectibility of vision, a task which she sees as endless.

¹³ See for example ‘Practical Truth’ (1993). ¹⁴ See the undated manuscript ‘Good and Bad Human Action’, first published in 2005. ¹⁵ This much-discussed case, which introduces into modern moral philosophy the idea of ‘moral perception’, involves a mother-in-law (M) learning to ‘look again’ at her daughter-in-law (D), whom she dislikes. Through ‘just and loving attention’, M comes to see D under new descriptions. No longer ‘vulgar but refreshingly simple, not undignified but spontaneous’ (Murdoch 1970, 17–23).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

     

293

What perfectibility of vision could amount to in Anscombe’s moral psychology, and what the possibility of perfectibility requires, is a difficult question that does not have a place here. What we can say is this. Sight is a capacity that belongs to a human animal living with others in a material world of institutions¹⁶ and practices, many of which, for Anscombe—even notoriously—are corrupt. Accordingly, for the Anscombian theorist, illusion cannot be thought of merely as matter of some introspectable this being unmoored from any worldly causes or things. Rather, it concerns the nature of the circumstances—material, natural, intersubjective, and epistemic—in which the perceiving animal finds herself and the conduciveness of those circumstances to human life going well; matters that, on the face of it, seem distant from the philosopher’s tug of war.¹⁷

References Anscombe, G.E.M. undated/2005. ‘Good and Bad Human Action’. In Mary Geach and Luke Gormally (eds.) [HLAE] Human Life, Action and Ethics. Exeter: Imprint Academics, 195–206. Anscombe, G.E.M. 1957/2000. Intention. Oxford: Basil Blackwell. Anscombe, G.E.M. 1981. ‘Introduction’. In [MPM] Metaphysics and the Philosophy of Mind: Collected Papers Volume II. Oxford: Blackwell, vii–x. Anscombe, G.E.M. 1958/2005. ‘Modern Moral Philosophy’. In Mary Geach and Luke Gormally (eds.) [HLAE] Human Life, Action and Ethics. Exeter: Imprint Academics, 169–94. Anscombe, G.E.M. 1993/2005. ‘Practical Truth’. In Mary Geach and Luke Gormally (eds.) [HLAE] Human Life, Action and Ethics. Exeter: Imprint Academics, 149–67. Anscombe, G.E.M. 1965/1981. ‘The Intentionality of Sensation: A Grammatical Feature’. In [MPM] Metaphysics and the Philosophy of Mind: Collected Papers Volume II. Oxford: Basil Blackwell, 3–20. Anscombe, G.E.M. 1976/1981b. ‘The Question of Linguistic Idealism’. In From Parmenides to Wittgenstein: Collected Philosophical Papers Volume I. Oxford: Basil Blackwell, 112–34. Anscombe, G.E.M. 1976/1981a. ‘The Subjectivity of Sensation’. In [MPM] Metaphysics and the Philosophy of Mind: Collected Papers Volume II. Oxford: Basil Blackwell, 44–56. Frey, J. and Frey, C. 2017. ‘Anscombe on the Analogical Unity of Intention in Perception and Action’. Analytic Philosophy 58(3): 202–47. Geach, P. 1957. Mental Acts: Their Content and Their Objects. Oxford: Routledge and Kegan Paul. ¹⁶ See Wiseman (2020). ¹⁷ Thank you to Heather Logue and Louise Richardson for their insightful comments on an earlier draft.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

294

     

Johnston, M. 2004. ‘The Obscure Object of Hallucination’. Philosophical Studies 120 (1): 113–83. Matthen, M. (ed.). 2015. Oxford Handbook of Philosophy of Perception. Oxford: Oxford University Press. McGinn, Marie. 1997. The Routledge Guidebook to Wittgenstein’s Philosophical Investigations. London: Routledge. Murdoch, I. 1970. ‘The Idea of Perfection’. In The Sovereignty of Good. London: Routledge and Kegan Paul. Poulvet, R. 2008. After Wittgenstein, St. Thomas. Translated by M. Sherwin. South Bend, IN: St Augustine’s Press. Robinson, H. 1994. Perception. New York: Routledge. Schwenkler, J. 2019. Anscombe’s Intention: A Guide. Oxford: Oxford University Press. Travis, C. 2015. ‘Intentionally Suffering’. In Michael Campbell and Michael O’Sullivan (eds.) Wittgenstein and Perception. London: Routledge. Wiseman, R. 2020. ‘Brute Facts and Human Affairs’. Royal Institute of Philosophy Supplement 87: 85–99. Wiseman, R. 2016. Guidebook to Anscombe’s Intention. London: Routledge. Wittgenstein, L. 1958. Philosophical Investigations. Translated by G.E.M. Anscombe. Second edition. Oxford: Basil Blackwell.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 20/3/2021, SPi

Index Note: ‘n’ indicates a footnote, with the number following ‘n’ indicating the footnote number when there is more than one footnote on the page. abilities situation-dependent 67–9 types of 68n8 abstracta 19, 133, 134, 135, 137, 139n12 abstract entities see abstracta accuracy, of perceptual experiences 139–42 acquaintance 1, 43, 116, 121, 289, 290 conscious relation of 50, 55 direct object 124 with external objects 116 with mind-independent objects 85 with peculiar entities 135 with things around us 8 actual experiences 261 admissible contents of perceptual experience 2, 12, 17, 19, 147, 148, 151, 152, 154 Advancement of Science, The (Kitcher) 41 adverbialism 1n1, 19, 109, 113, 122, 123, 138, 142, 217 affective phenomenology 185 affordances in perception 269 Allen, Keith 1, 5, 8, 9, 12, 13, 17, 19, 44, 46, 48, 48n5, 49, 52, 53, 58n13, 59n14, 127, 134, 199, 221n9, 224n13 analysis, of descriptions of a phenomenon 264–7 Anscombe, G. E. M. 6, 7, 13–4, 20, 24n1, 276–93 appearances 54, 55, 57 characterization of 58 Kant and 237, 240 sensory 195 Aristotle 30, 133, 279, 292 artists 272, 274 assertoric language 110 astronomic paradigms 30 audition 148, 150, 151, 188, 205, 208 auditory-visual multimodal illusion 155 Austin, J.L. 36, 52, 53–4, 72, 91, 117n19 awareness 43, 198 conscious 49, 58 non-representational 197, 198, 200 of our surroundings 251 property 132 quality/awareness gap 46, 47

relations to external objects 36 relation to abstract entities 135n9 of sense-data 127 unmediated 198, 199 visual 71, 76 bad cases 60, 63, 64, 65, 66, 288, 289 according to Anscombe 292 disjunctivist’s use of 291 uninstantiated qualities in 132 visual experiences in 71, 76 see also good cases Bayesian processing 195–6, 204 Bayne, Tim 94n6, 95n10, 148, 151, 154, 155, 174, 186, 238 behaviour, causal explanation of 80 belief 5, 23–4, 31, 115–17, 131, 181–2, 186, 188 Berkeley, George 9n10, 140, 142 Block, Ned 79n14, 90n5, 94n9, 154, bodily ownership 160, 166 ‘box-and-arrow’ diagrams 177 brain damage 185–6, 251–2 Brentano, Franz 279 Brewer, Bill 89, 90n4, 148, 155n6, 179n3, 197, 225 Briscoe, Robert 152, 154, 155n5 Brogaard, Berit 95n10, 148, 149, 151, 154, 155n5, 156n8, 225 Burge, Tyler 1, 14, 23–32, 37, 38, 44, 67n7, 80, 86, 87–8, 194, 197, 209, 210, 227, 230n25 Byrne, Alex 2n2, 45n2, 64n4, 95n10, 129n3, 131, 147–8, 174 ‘can,’ senses of 68, 69, 72 Campbell, John 1, 5, 14–5, 44, 47, 52–3, 56, 64n3, 90n4, 148 Capgras delusion 186 Cartesian Meditations (Husserl) 246 causal concepts 236, 237 causal dependence 86n1 Causal Exclusion Argument 49, 50 causal explanations 224 of behaviour 80 of psychological events 219

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 20/3/2021, SPi

296



Cavedon-Taylor, Dan 5, 12, 20, 149, 224, 269n7 CBA see cost–benefit analysis (CBA) Chalmers, David 5, 46, 90n5 Chisholm, Roderick 123n26, 138 Churchland, Paul 94n6, 133–4 Clark, Andy 174, 177, 178, 181, 196, 201 clinical disorders 154 cognitive capacities 79, 100, 226 cognitive hypotheses 188 cognitive phenomenology 101, 102, 185, 186 cognitive processing 204 cognitive psychology 24 colours 46–7 colour experience 44, 46, 120, 133 experiencing impossible colours 133–4 mind-independent 46 naïve realist theory of 44, 46 perception of 44 ‘Problem of Colour’ 45 teaching of 111–12 understanding of 98 common kind assumption 10, 113–14 common sense 18, 71, 78, 79, 262 explanatory practice 77, 81 understanding of colour 98 view of the world 263 conceptual negotiation 91–2, 95 concreta 134n8, 135 confrontation with objects 198 conscious awareness 49, 58 consciousness 31, 39–40, 88n3, 94n9, 262 hard problem of 5, 121. see also Problem of Consciousness higher-order theories of 94n8, 102n15 neural correlates of 122 perceptual experience as inner content of 8, 259–60, 267, 274 Consciousness Explained (Dennett) 39 conscious visual presentation 63, 76 Conservatism about admissible contents 152–4, 158, 159, 175, 176–7, 185 consistency of predictive processing framework with 186–7 content and phenomenal character 66n6 debate between Liberalism and 148–9, 151, 188 and low-level properties 174 motivations for 188 and the rubber hand illusion 160–1, 163, 165–6, 168 constitution 198, 200 of perception 225 constitutive explanations 227–9 constructivist theorists/theories 218, 224, 227

computational 218, 228 differences between ecological theories and 219 content of perceptual experience 12, 19, 132n6, 147–8, 174–7, 184–6 content-relationalism 136, 137, 138 cost–benefit analysis (CBA) 11, 12, 14, 19 Cotard delusion/syndrome 186, 247 Crane, Tim 122n25, 123n27, 130, 132n6, 136, 138, 217n3, 222n11, 225 Crisis of European Sciences (Husserl) 246 Critique of Pure Reason (Kant) 237, 242 crypto-mechanism 271 Darwin, Charles 37 data compression 178 deceptive experiences 24 see also hallucinations; illusions delusions 181n6 Capgras delusion 186 demystification of perceptual talk 7, 108, 109, 112–20 Dennett, Daniel 39, 89, 193, 238 Descartes, René 2n3, 26, 129, 260 describing experiences 259, 264 descriptive naïve realism 53–4 determinism 8, 59–60 De Vignemont, Frederique 152, 163, 166–7 direct objects 279, 280 concepts of 279, 280, 283 of perception 20, 136 disjunctivism 1, 2, 8, 91, 100, 108, 288–9 Burge and McDowell on 23–8 compatibility with vision science 23–8, 86 epistemic 64n3 metaphysical 64n3, 64n4, 113 moderate 64n4 naïve realism and 113–14 phenomenal 85 rejection of 103 scepticism about 87 Drayson, Zoe 11, 14, 15, 17, 20, 51n8, 72n11, 80, 218n4, 229n23,24 Dretske, Fred 129, 131–2, 174, 196, 197 Ducasse, C.J. 90n5, 138 ecological theories/theorists 218–19, 223, 224, 228 eidetic variation 239, 240, 247, 248, 251, 253 eidos see essence(s) Elgin, Catherine 41 eliminativism about common-sense concepts of perception 16, 18

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 20/3/2021, SPi

 about perceptual phenomenal character 101–2 about secondary qualities 139n13 empirical experience 236, 244 empirical phenomenology/ phenomenologists 236, 237 empirical psychology 24, 65, 147, 239, 251 Empiricism 262, 274 empiricist accounts of perception 234 epistemic role 5, 6, 7, 79, 215, 279 epistemology 15, 27, 31, 39n6, 149 essence(s) 234, 239, 250, 254, 259 of a given entity 77 in Merleau-Ponty’s phenomenology 264 as non-deducible from experience 266 of perceptual experience 7, 13 ethics Aristotelian 292 of perception 277 Euclidean geometry 79, 249 experiences 258, 274 multimodal 12 as states of perceivers 26 status of perceptual 259–62 unimodal 12 experiential monism 64, 65, 67 experiential naturalism 49, 50 experiential pluralism 64 moderate 65, 77 explanatory gap 45n2, 192–3 expressivism 110 externalism/externalist 71n11, 132n6, 192 Extreme Conservativism about common-sense concepts of perception 16, 18 facial perception 156, 157 Farkas, Katalin 86, 93 first-personal experiences 55, 118, 288 Fish, William (Bill) 5, 9, 10, 11, 14, 18, 19, 44, 47, 48, 50, 52, 80, 86, 87–90, 94n7, 101, 102, 112, 114n11, 116n16, 148, 149, 154–5, 197, 201–2, 217n3, 218n4, 227, 230n25, 277 flavour 152, 194, 205, 209 Fodor, Jerry 15, 193, 206, 229n22 folk psychology 18, 39n6 free energy 195 fundamental kinds of state 24–5, 77–8, 108 Galileo 30 genetic phenomenology 18, 247 geological time, Darwinian approach to puzzles surrounding 37 Gert, Joshua 1n1, 5, 6–7, 8n9, 9, 10, 12, 19, 48n5, 109n4, 120n22,23, 122n25, 138

297

gestalts 165–6, 264–5, 266 and the rubber hand illusion 165 Gibsonian perspective 194 Gibson, J.J. 24, 194, 218–19 global supervenience thesis 49–50 Godfrey-Smith, Peter 29–30, 39 Gomes, Anil 13, 58, 224n13 good cases 63, 64, 65, 289 visual experiences in 66–7 see also bad cases Gow, Laura 1n1, 9, 19, 48, 50, 136n10, 137n11, 139n13 grammatical approach, to perception 6, 20, 277, 291 grammatical category 279, 280, 281, 283, 288 gustation see taste hallucinations 2, 24, 48, 66, 91, 127, 291 contents of 137 and delusions in psychosis 181n6 eliminativism about the phenomenal character of 101 as indistinguishable from perceptions 10–11, 100 interfering with introspective capacities 102 Merleau-Ponty and 11, 261–2, 270 phenomenal character of 131 physicalist explanation of 130 and response-dependent properties 201–3 schizophrenic 262 total 1, 9–10 veridical 71, 73 views of representationalists on 85 high-level properties 2, 5, 12, 95n10, 188 and illusions 147, 149 and the rubber hand illusion 161, 168 and the ventriloquist effect 157 Hinton, J.M. 86, 91, 94n7 Hohwy, Jakob 177, 181, 195–6, 204 Hume, David 13, 235–45, 246–8, 250, 252, 256 Husserl, Edmund 238–40, 241, 243, 245, 249, 251, 264 IBE see Inference to the Best Explanation (IBE) ideas 236 resembling physical objects 140 illusions 2n2, 24, 48, 147, 183, 205, 293 auditory-visual multimodal 155 bodily ownership 160 Merleau-Ponty and 261 Müller-Lyer illusion 182, 183, 206 Rubber Hand Illusion. see Rubber Hand Illusion imagination 270–1, 291 impressions 236 incommensurability 29

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 20/3/2021, SPi

298



indistinguishability of experiences 24–5, 91, 108, 261, 291 Indistinguishability Principle 9, 86, 93, 94, 98, 99 see also subjective indiscriminability; subjective indistinguishability Inference to the Best Explanation (IBE) 11, 12, 17, 56, 58–9, 121, 216, 219, 223–4, 225 information-theoretic descriptions 123 Intellectualism 262 intentionalism see representationalism intentionality 121, 138, 279, 283 ‘Intentionality of Sensation: A Grammatical Feature, The’ (Anscombe) 283 intentional object 279, 284, 285, 287n11, 289 Intention (Anscombe) 276 internalism 132n6 introspection 12, 18, 79, 86–7, 112 infallibility of 93–4 interfering or disabling conditions for 100n14 phenomenal character and 93–6 introspective capacities 98, 100, 102 Jackson, Frank 44, 46n3, 50, 127, 191, 196, 199 Johnston, Mark 2n3, 79n14, 98, 116n18, 120n23, 131, 149, 287n11 judgement 175, 181, 183, 184–5, 236 Kalderon, Mark 44, 48, 52, 65n5, 155n6, 198, 200 Kant, Immanuel 13, 57, 58, 235–40, 242–5, 246–7, 250, 253, 256 Kitcher, Philip 41 knowledge argument 46n3 Kriegel, Uriah 86, 93, 94, 131, 134n8, 138 Kuhnian paradigms 18–19, 23, 28–30, 32, 33, 37, 38–9, 87–9 Kuhn, Thomas 23, 28, 29, 30, 32, 33, 38, 39, 87, 88 Lakatosian research programmes 33–7, 40, 89 Lakatos, Imre 33–5, 37, 39, 40, 89 language 109, 248 teaching of 111 language-games 278, 285, 289 Levine, Joseph 114n14, 192 Lévy-Bruhl, Lucien 249 Lewis, David 92, 209 Liberalism about admissible contents 19–20, 147, 148–9, 151, 153–5, 167, 168, 175–6, 182–4, 186, 188 empirical support for 163–6 and high-level properties 148 and the McGurk effect 156 and predictive processing framework 174, 188 and the rubber hand illusion 160–6 and the ventriloquist effect 157, 158

linguistic, competence 281–2 linguistic capacities 248 local supervenience principles 9, 10–11, 86n1 Locatelli, Roberta 9–10, 12, 18, 19 location, perception of 148, 157–9, 204, 207 Logical Investigations (Husserl) 238 Logue, Heather 2n2, 4n8, 5, 47, 56, 64n4, 101, 102n17, 108n3, 114n11, 116n16, 118–20, 129n3, 149, 197, 199, 225, 228 Lowe, E.J. 127, 216, 219, 221n10 low-level properties 2, 95n10, 148n1, 149, 154, 177, 184–5, 188 and illusions 147, 155n7 paradigm examples of 157 of proprioception 161 and the rubber hand illusion 162, 163, 167 and the ventriloquist effect 159 luminosity 98 Mac Cumhaill, Clare 6, 7, 13, 16, 18, 20, 24n1 Macpherson, Fiona 2, 66n6, 85, 149, 150, 161, 187n11 McDowell, John 1, 5, 14, 23–8, 29–32, 38, 64n3, 65n5, 77, 78n13, 87–8, 114n15, 149, 179n, 194, 197, 216, 218n4, 226, 227, 230n25 McGurk effect 155–6, 205, 207 Martin, M.G.F. 7, 8, 10, 25, 44, 45, 49, 52–3, 57n11, 64n3, 65n5, 78n13, 79, 86, 90n4, 94n8, 99n13, 101, 102n17, 108n3, 114n11,12,15, 124n28, 148, 149, 150, 160, 192, 194, 197, 198, 201–3, 225 matching view of accuracy assessments 139, 143 material objects 284, 286, 287–8, 289–91 Matthen, Mohan 150, 209, 276 Meditations (Descartes) 260 Mele, Alfred 68, 72 mental kinds experiential pluralism and 63–84 world-involving experiences as 77–82 mental states, nature of 9n10, 67n7, 98, 116, 227 Merleau-Ponty, M. 7, 8, 11, 13, 18, 20, 52, 54n9, 234–5, 236n2, 238, 240, 243, 245, 258–74 account of language 248 analysis of descriptions of a phenomenon 264–7 gestalt forms 264–5 and Husserlian phenomenology 247, 249 notion of essence 264 Objective Thought 262 phenomenology of perception 18, 54n9, 246, 250–6, 258–67 radical reflection 245–6

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 20/3/2021, SPi

 on speech 254 transcendental phenomenological reduction 262–4 see also Phenomenology of Perception (Merleau-Ponty) metaphysical truths, modal status of 220 metaphysics of mind 15, 231 Millikan, Ruth 70n10, 72, 132n6 ‘Modern Moral Philosophy’ (Anscombe) 292 Montague, Michelle 139, 186 motion 156, 272 Müller-Lyer illusion 182, 183, 206 multimodal perception 150–1, 187 additive vs. integrative 152 constitutive/generative 152–3 Murdoch, Iris 292 naïve realism 1, 2, 8, 12, 19, 20, 23–5, 35–7, 85–7, 92–3, 94, 98–102, 128, 148–9, 198, 217, 223, 225, 276 and the asymmetric discreditation argument 197–201 descriptive 53–4 and disjunctivism 91, 113–14 historical resilience of 52 intelligibility of alternatives to 60 metaphysical 51–3, 54–5, 116 neopragmatist critique of 112–20 perceptual experiences and 43–4 and physicalism 48–52 prediction error theory and 195–204 and the problem of consciousness 43–8 representationalism and 89–90, 129n3 transcendental 55–60 Nanay, Bence 4, 95n10, 148, 149, 154, 174 naturalism 112–95, 210, 215–33 see also physicalism neopragmatism 48n5, 107–12, 115–24 Newton, Isaac 26, 30, 225n16 Noë, Alva 180, 208 non-relationalism about perceptual experience 137–9, 142, 143 Noordhof, Paul 14, 15, 16, 17, 20, 193, 195, 198, 208 Nudds, Matthew 16, 150, 208–9, 223 O’Callaghan, Casey 150, 151, 152, 156, 177n2, 205–6 Objective Thought 262, 264, 274 object-relationalism about perceptual experience 136 objects of sight 6, 279–81, 283, 284 intentional and material 287 observational properties 1, 95, 96, 96n11

299

olfaction 36, 148, 150–2, 194, 204–5, 208 ontological approach to theorizing about perception 6, 277 openness of perceptual experience 8, 29, 88n2, 112, 114, 115, 116, 194, 222, 260 ordinary language philosophy 53, 285, 286 Oxford Realism 52, 57 painters, and visual perception 271 Papineau, David 131, 132n6, 124n28 paradigms, incommensurability of 29 Pautz, Adam 35–7, 50, 102n15, 131, 132, 135n9, 137 Peacocke, Christopher 96, 97 perceptual field 269, 270 perceptual processing 4, 50, 204 affect modulating 185 and cognitive processing 204 errors in 175–6 representations and 197 perceptual psychology 1, 25–6, 31, 79, 168, 218–19, 227 personal-level 4n8, 183n9, 194, 198, 204, 228–9, 231 see also subpersonal-level perspective drawings 273 phenomenal character, perceptual 4–6, 9–11, 20, 35, 63n2, 66, 79, 85–7, 89–90, 92, 93–6, 97–9, 100–2, 118–20, 129, 131–3, 154, 192–3, 197, 201, 217, 235 see also ‘what-it’s-like ‘talk phenomenal contrast, method of 5, 154–5 Phenomenal Principle 128–30, 131–2 phenomenologists 13, 236, 238 phenomenology affective 176, 185–6 cognitive 101–2, 154, 176, 185–6 meta-philosophical 51 perceptual 8n9, 12, 32, 43, 54, 58, 101, 114, 115, 139n13, 154, 176, 181, 184, 221n9 see also phenomenal character, perceptual; ‘what-it’s-like ‘talk as tradition/method 12, 54, 234–56, 258–9, 264, 274 Phenomenology of Perception (Merleau-Ponty) 54n9, 234, 245, 252, 253, 254, 258, 271 Phillips, Ian 4n5, 79n14, 82 Philosophical Investigations (Wittgenstein) 278 philosophy of mind 15, 130, 142, 149, 227, 229 philosophy of psychology 15, 17, 292 physicalism 19., 127–30, 139, 142–3, 191, 192–4, 198 naïve realism and 43, 46, 48–52

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 20/3/2021, SPi

300



physicalism (cont.) non-reductive 49 representationalism and 130–7 see also naturalism Planck, Max 39 Platonic conception of universals 133, 134 prediction error theory 195–204 see also predictive processing framework predictive processing framework 17, 20, 174, 177–9 and admissible contents 181–8 and ‘seeing as’ 179–81 see also prediction error theory Price, H. H. 44, 90n5, 108n2, 127, 277 Price, Huw 108n1, 110 Primitive Mythology (Lévy-Bruhl) 249 private language argument 278 Problem of Colour 45 Problem of Consciousness 43–6, 48, 50n7, 51, 52 see also consciousness, hard problem of Problem of Perception 221–2 see also puzzle of perception proprioception 160–1, 162, 163–6, 167 proximate cause principle 210 see also same cause same effect principle pseudo-presence 270–1 psychological verbs 278, 283 psychology cognitive 24 empirical 24, 65, 147, 239, 251 folk psychology 18, 39n6 perceptual 1, 25–6, 31, 79, 168, 218–19, 227 philosophy of 15, 17, 292 social 39 speculative 15 puzzle of perception 7, 108 see also Problem of Perception qualia 45, 46 Quine, W.V.O. 39n6, 222–3 recognitional dispositions 154, 176, 179 reference 5, 107–8, 109–10, 111, 124, 197, 202 reflection on experience 12–13, 79, 114, 235–6, 238, 245, 246–50, 254–6 see also introspection reflective capacities 247, 248, 250, 254, 255 relationalism see naïve realism representationalism 1, 2, 20, 23–4, 47, 85, 129, 136, 138, 143, 194, 215–17, 221–2, 224, 225, 279 externalist and internalist versions of 132 vs. naïve realism 2, 198, 215, 216, 217n4, 219, 276

as an ontological theory of perception 286 and physicalism 131–7 Revelation 98 Richardson, Louise 194, 205, 206 Robinson, Howard 86, 90n5, 127, 128, 276, 288n12 Romdenh-Romluc, Komarine 7, 8, 11, 12, 13, 20, 249, 251, 262n4, 267n5, 270n8 rubber hand illusion 5, 147, 155, 159–68 Russell, Bertrand 32, 90n5, 108n2, 116, 127 same cause same effect principle 86, 93, 98 see also proximate cause principle Schellenberg, Susanna 1, 5, 131, 135 schizophrenic hallucinations 262, 291 Schneider (brain-damaged World War I veteran) 251–2 scientism 193 seeing as 179–81 seemings 154 Sellars, Wilfrid 109n4, 109n6, 119, 229, 230 sensations 98, 107, 112, 117, 159–60, 162, 165–6, 167, 183, 187, 192, 236, 276–7 concepts of 278, 282, 290 verbs of 279, 281, 284, 289, 291 sense-datum theory 32, 90n5, 108, 127–8, 135, 140–1, 143, 217, 276, 281, 285–6 acquaintance and 116, 121 disjunctivism and 113 naïve realism and 43, 55, 136 problems with 128n2 senses, the 2, 16, 150–1, 204–9, 271–2 sensorimotor contingencies 180–1 Sethi, Umrao 2n3, 114n13 sexual harassment 249, 265–7 shape perception 120, 148, 271–2 Siegel, Susanna 3, 5, 86, 95n10, 139, 147–9, 153–4, 174, 176, 179, 181, 184 sight 200, 293 see also vision situation-dependent abilities 5, 12, 67–74, 71n11, 80–1 smell 35, 150, 151–2, 205, 208 see also olfaction Snowdon, Paul 64n3, 77 social psychology 39 sortal concepts 166, 167 Soteriou, Matthew 4n7, 64n3, 65n5, 90n4, 102, 197 sound perception 157–8, 188, 207–8 spectrum inversion 119 speculative psychology 15 speech 156, 157–8, 177n2, 254 Spence, Charles 151, 158–9 Spener, Maja 2n2, 5, 12, 18, 19, 94n6 Sterelny, Kim 191

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 20/3/2021, SPi

 Stoljar, Daniel 131, 134 Stoneham, Tom 50n6, 114n13 Strawson, P.F. 8, 59–60, 114n14 structural properties of experience 237–8, 239, 240, 243, 244, 248, 269 Structure of Scientific Revolutions, The (Kuhn) 39 subjective indiscriminability 202–44 see also indistinguishability of experiences; subjective indistinguishability subjective indistinguishability 10–11, 66, 97 see also indistinguishability of experiences; subjective indiscriminability subpersonal-level 183n9, 227–9, 230–1 superficiality constraint 10, 12, 87, 97–9, 100–3 synaesthesia 159, 271 tactual-proprioceptive referral 162, 163–6, 167 taste 35, 148, 152, 194, 208 Theory and Reality (Godfrey-Smith) 29 Thomist tradition 279 Thompson, Brad 133 total hallucinations 1, 9–11 touch 148, 150, 160–1, 167 transcendental arguments 12, 56, 57, 58, 59, 243 and inference to the best explanation 224n13 used by naïve realists 221n9 transcendental naïve realism 55–60 Transcendental-Phenomenological Reduction 258, 259, 262–4, 267, 273 transcendental phenomenology 237, 238, 268 transparency of experience 57n11, 137n11 Travis, Charles 52, 155n6, 197, 198–9, 201–2, 289, 290–1 truth 107–8, 110–11, 255, 292 Twin-Earth 116 Tye, Michael 1, 64n3, 90n5, 123n26, 129, 131, 132, 135, 137n11, 138, 148, 154, 174, 177, 196, 199

301

unconscious perception 4, 165, 230n25 unimodal perception 150–1, 153 universals 132–4 Valberg J.J. 90–1 Vatakis, Argiro 157–9 ‘veil of perception’ 128n2 ventriloquist effect 157–9, 207 verbs of sensation 279, 281, 284, 289, 291 veridical hallucinations 25n2, 71, 73, 129n4, 131 veridical perception 2, 66, 88n2, 108 veridical perceptual experience 1, 29, 48, 108, 113, 127–8, 129n3, 137, 217 vision 25, 117, 148, 150, 204, 218, 292–3 vision science 14–5, 80, 86 see also visual science visual field 259 visual illusions see illusions visual perception 47, 198, 218, 271–2 visual science 25 see also vision science visuocentrism 150 visuo-cutaneous stimulation 162, 163 Ward, Dave 13, 18, 20, 54n9, 56n10, 268n6 ‘what-it’s-like ‘talk 118–20 Wilkinson, Sam 17, 20, 177, 181 Williamson, Timothy 98, 223n12, 224, 225n16 Wiseman, Rachael 6, 7, 13, 16, 18, 20, 24n1, 282n9, 293n16 Wittgenstein, Ludwig 192–3, 276, 277–8, 279, 286 Wood, Carmita 249, 265–6, 267 world-involving experiences 63, 66–7 explanatory demand for 74–7 as a mental kind 77–82 Zettel (Wittgenstein) 192–5