Knowledge Through Imagination [Illustrated] 019871680X, 9780198716808

Citation preview

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowledge Through Imagination

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowledge Through Imagination edited by

Amy Kind and Peter Kung

1

OUP CORRECTED PROOF – FINAL, 01/11/2016, 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 © the several contributors 2016 The moral rights of the authors have been asserted First Edition published in 2016 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: 2015947703 ISBN 978–0–19–871680–8 Printed in Great Britain by Clays Ltd, St Ives plc 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 PROOF – FINAL, 01/11/2016, SPi

Contents Acknowledgements List of Contributors Introduction: The Puzzle of Imaginative Use Amy Kind and Peter Kung

vii ix 1

Part One:  Taxonomical and Architectural Approaches 1. On the Epistemic Value of Imagining, Supposing, and Conceiving Magdalena Balcerak Jackson

41

2. On Choosing What to Imagine Peter Langland-Hassan

61

3. The Imaginative Agent Neil Van Leeuwen

85

Part Two:  Optimistic Approaches 4. Knowing by Imagining Timothy Williamson

113

5. Modals and Modal Epistemology Jonathan Jenkins Ichikawa

124

6. Imagining Under Constraints Amy Kind

145

7. Perceiving People as People: An Overlooked Role for the Imagination Jennifer Church

160

Part Three:  Skeptical Approaches 8. Knowing Me, Knowing You: Failure to Forecast and the Empathic Imagination185 Heidi L. Maibom 9. Imagination Through Knowledge Shannon Spaulding

207

10. Thought Experiments in Ethics Peter Kung

227

Index

247

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Acknowledgements This volume originated with a conference on Knowledge Through Imagination held at Claremont McKenna College in April 2012. We are grateful to the Gould Center for Humanistic Studies, the Claremont McKenna College Department of Philosophy, and the Pomona College Department of Philosophy for their generous monetary support of that conference. We are also grateful to Celina Rosas for all her work in helping to make the conference a success. It has been a pleasure to work with Peter Momtchiloff and the terrific team at Oxford University Press on this project. The volume has benefitted greatly from the thoughtful comments of several anonymous referees and also from the careful attention of copy­ editor Howard Emmons. We also thank Roma Forest, CMC Class of 2019, for her work on the volume’s index. The cover image, “Tabletop Towers,” comes from the enormously talented artist Rob Gonsalves, and we are thrilled to be able to use it as a visual representation of the way that imagination allows us to engage with the world around us. On a more personal note, we each thank our respective families for their unwaver­ ing love and support. To our children in particular—Stephen and Joseph, and Tara, Shayan, and Raami—we are grateful for your powers of imagination. You inspire us each and every day.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

List of Contributors Magdalena Balcerak Jackson, University of Miami Jennifer Church, Vassar College Jonathan Jenkins Ichikawa, University of British Columbia Amy Kind, Claremont McKenna College Peter Kung, Pomona College Peter Langland-Hassan, University of Cincinnati Heidi L. Maibom, University of Cincinnati Shannon Spaulding, Oklahoma State University Neil Van Leeuwen, Georgia State University, University of Johannesburg Timothy Williamson, University of Oxford

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction The Puzzle of Imaginative Use Amy Kind and Peter Kung

From daydreaming to decision making, from pretending to planning, imagination plays a central role in many of the activities of everyday life. As children playing dress-up, we imagine ourselves to be glamorous princesses and powerful superheroes, and our imaginings transport us to distant lands of magic and make-believe. In our athletic forays, we imagine ourselves scoring the winning goal or making the winning basket, and this helps us to know how to aim the shot. Our imaginative capacities are called into play during brainstorming sessions at work as we attempt to come up with innovative and creative solutions to problems. On those occasions when our attention wanders, and our thoughts turn to upcoming vacations, we imagine ourselves sipping frosty beverages while lounging on the beach. When preparing for those trips, we imagine the various activities we’ll be undertaking in an effort to decide what to pack. And once we’re relaxing on the sand, away from it all, we escape via imagination into the fictional worlds of the novels we’ve brought. Once we pause to think about these kinds of examples—and there are of course many others that we could add to the list—they seem to present us with something of a puzzle. As the examples suggest, imagination is put to two distinct and seemingly incompatible kinds of uses. Imagination is sometimes used to enable us to escape or look beyond the world as it is, as when we daydream or fantasize or pretend. We’ll call this the transcendent use of imagination. Yet imagination is also sometimes used to enable us to learn about the world as it is, as when we plan or make decisions or make predictions about the future. We’ll call this the instructive use of imagination. But how can a single mental activity successfully be put to both uses? How can the same mental activity that allows us to fly completely free of reality also teach us something about it? This puzzle—what we’ll call the puzzle of imaginative use—has received surprisingly scant attention in philosophical discussions of imagination. Insofar as a tension

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

2  Amy Kind and Peter Kung between the transcendent and instructive uses of imagination has been recognized, such recognition has tended to be indirect—as, for example, when a philosopher’s focus on the transcendent use of the imagination has led to a dismissal (whether explicit or merely implicit) of the instructive use of the imagination. More commonly, however, the philosophical treatment of each use of imagination has simply occurred in isolation from the other, with some discussions focused on imagination’s transcendent use and other discussions focused on its instructive use. The bifurcated nature of such discussions has undoubtedly been a key factor in masking the puzzle that interests us. Our thinking about these issues was shaped in large part by the contributions presented at the conference Knowledge Through Imagination that we hosted at Claremont McKenna College in the spring of 2012. Conference presentations focused on the ways in which imagination might provide us with knowledge about the world in which we live. Such a question is inextricably tied to the puzzle of imaginative use: Because there seems to be little doubt that imagination can be put to transcendent use, imagination will only be able to provide us with knowledge about the world in which we live if the transcendent use of imagination is compatible with its instructive use. Fortunately, however, we believe a strong case can be made for this compatibility. Careful attention to the puzzle of imaginative use enables us to dissolve the apparent tension between the transcendent and instructive uses of imagination and thereby make room for the instructive use as well. As we’ll suggest in this introduction, the key to solving the puzzle is to acknowledge and explain how our expansive powers of imagination can be reined in. When there are constraints on imagination, either architectural constraints or constraints that we can willingly impose, and when these constraints ground imagination in the real world in the right way, imagination can help us discover truths about the real world. Though we do not think this solution is a new one—indeed, as we will see, there are hints of it throughout both the historical and the contemporary literature—we do not think it has yet been clearly articulated and developed. Before fleshing out this solution, we begin in the next section by considering a different attempt to solve the puzzle, one that focuses on the fact that different philosophers have different conceptions of imagination. Call this the equivocation solution. If there are different senses of imagination in play in philosophical discussion, then perhaps it is imagination in one sense that is responsible for the transcendent use of imagination, while imagination in some other sense is responsible for the instructive use of imagination. As we will suggest, however, the equivocation solution cannot ultimately succeed. While on occasion philosophers may talk past each other when developing their accounts of imagination, it is not plausible to suppose that this can fully account for the fact that imagination has been put to both transcendent and instructive uses. To lay the groundwork for our own resolution of the puzzle, we will then devote considerable attention to several influential treatments of imagination. In Section 2, we explore three particularly influential treatments of imagination from the history

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  3 of philosophy; in Section 3 we turn to the contemporary literature to discuss three different contexts in which imagination has played a key role. Our discussion in these sections is by no means comprehensive. Rather, our primary aim is to illuminate some of the different ways that philosophers have thought about the imagination’s potential for the transcendent and instructive uses. In doing so, we will show that constraints play a crucial (though often implicit) role in philosophical treatments of the imagination. As we suggest in Section 4, making these constraints explicit helps to dissipate the tension between the two uses of imagination. Because constraints may be in place without always being operative, imagination can gain the power to instruct us about the world without losing its power to transcend it. Once the puzzle is dissolved and the apparent tension between the transcendent and instructive uses of imagination thereby dissipates, space is opened for closer attention to the instructive use of imagination. Such is the primary goal of this volume. In the ten chapters that follow, our contributors all explore various ways in which imagination might enable us to learn about the world. Though not all of them are sanguine about what imagination can teach us, all of them address the question of whether and how we can gain knowledge of the world via imagination. In the last section of this introduction, we provide a roadmap to the volume and summaries of the ten contributions.

1.  The Equivocation Solution One natural way to try to resolve the puzzle of imaginative use stems from the suspicion that there is no single conception of imagination in play in philosophical discussion.1 Anyone coming to the imagination literature for the first time would undoubtedly be frustrated by the lack of a clear explanation of the mental activity being talked about. The problem is not simply that philosophers give different theoretical treatments of imagination but rather that there doesn’t even seem to be consensus about what the phenomenon under discussion is. Among contemporary philosophers in particular there is a surprising reluctance to offer a substantive characterization of imagination; instead, it is understood simply as a mental activity that is perception-like but not quite perception, or belief-like but not quite belief. To give just one example of this general reluctance, consider what Kendall Walton says about imagination in Mimesis as MakeBelieve, despite the central role it plays in the book as a whole: What is it to imagine? We have examined a number of dimensions along which imaginings can vary; shouldn’t we now spell out what they have in common?—Yes, if we can. But I can’t. (1990, p. 19)

Even when philosophers do provide substantive characterizations of imagination, such characterizations tend to conflict with one another. To give just a couple of examples of such conflicts among contemporary philosophers, some treat imagination as   For further discussion of this point, see Kind (2013).

1

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

4  Amy Kind and Peter Kung essentially involving mental imagery or some kind of sensory presentations; others do not.2 Some philosophers treat supposing as a kind of imagining; others do not.3 Some take imagining a proposition to amount merely to entertaining it; others think imagining requires more.4 Matters are no better when we turn to the history of philosophy; as we’ll see in Section 2, philosophers such as Descartes, Hume, and Kant had very different conceptions of imagination from one another. The equivocation solution to the puzzle of imaginative use derives support from this apparent lack of consensus about what imagination is. According to this proposed solution, the term “imagination” is equivocal; there are (at least) two different senses of it. Thus, one kind of imagination—call it imaginationT—has transcendent use, while another kind of imagination—call it imaginationI—is responsible for the instructive use. Though such a resolution looks promising initially, we do not believe that it ultimately proves satisfactory. First of all, to develop it successfully, one would need to differentiate clearly between the two different kinds of imagination in play. It is not at all obvious what these two kinds of imagination would be. Consider two common distinctions found in the literature: • imagination from the inside vs. imagination from the outside • imagistic imagination vs. non-imagistic imagination. Suppose you want to imagine yourself skiing. One way you might do this is to imagine getting all bundled up, strapping on the skis, grabbing the poles, and then bending your knees and leaning forward while feeling the rush of cold wind across your face and the jarring bumps as you go over the moguls. Another way to do this is to picture yourself with red cheeks, all bundled up with skis strapped on and poles in hand, as you move down the mountain. While the first is an instance of imagining from the inside, the second is an instance of imagining from the outside. The first proceeds experientially from the perspective of the skier; the second does not.5 Imagining from the inside plays a crucial role when we try to predict how we will react to certain situations.6 Because this kind of use of imagination seems instructive— by imagining from the inside, we learn about our own future reactions7—we might perhaps propose that all and only instructive uses of the imagination are instances of 2   For non-imagistic imagining, see Chalmers (2002), Walton (1993), Yablo (1993), and White (1990). For imagistic imagining, see Brann (1992), Kind (2001), and Kung (2010). 3   Gendler (2000), Kind (2001), and Kung (2010) sharply distinguish imagining and supposing; in contrast, Goldman (2006) treats supposing as a species of imagining. See also the contributions in this volume by Balcerak Jackson and Spaulding. 4   e.g. McGinn seems to equate imagining something with merely entertaining it (see 2004, p. 132); Yablo (1993) takes the opposing view. 5   The second may be experiential in the sense that the imaginative episode bears some resemblance to the perceptual experience of watching someone ski. But it is not from the skier’s perspective. 6   Imagining from the inside also plays a key role in mindreading. See Section 3.3. 7   Though see the contributions from Maibom and Spaulding for skepticism on this point.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  5 imagining from the inside. But it can be quickly seen that there are problems with both directions of this claim. First, imagining from the inside can also be employed in transcendent uses of the imagination. Consider children pretending to be superheroes, or readers engrossed in first-person narratives who put themselves in the narrator’s shoes. Consider also people engaging in fantasies or daydreams. In some such cases, one might see oneself in the fantasized situation from the outside (as when one pictures oneself lying on the beach), but in others, one fantasizes about the situation by imagining experiencing it from the inside (as when one imagines lying on the beach and feeling the hot sand). Second, there are instructive uses of imagination that seem to proceed by way of outside imagining. Consider, for example, using visualization to figure out whether the hat displayed in the store window will look good on you.8 The distinction between imagistic and non-imagistic imagining is unhelpful for similar reasons. Someone who believes that there can be both imagistic and non-imagistic imagining will almost certainly accept that activities like pretending and daydreaming, both of which constitute transcendent uses of imagination, can proceed by way of either of them. When we turn to instructive uses of the imagination, we find the same thing. Imaginative brainstorming, for example, might proceed either imagistically or non-imagistically. This brief discussion does not itself rule out that there might be some other distinction among kinds of imagining that lines up well with the transcendent use/instructive use distinction. But we are skeptical that such a distinction could be found. The worry is not just that there are no obvious candidates for such a distinction. Rather, our skepticism arises primarily from the fact that there seem to be important connections between imagination’s two uses. In many ways, the power of imagination to transcend the world seems directly continuous with its power to teach us about the world. Such continuity remains entirely inexplicable unless we accept that one and the same faculty can really be put to both transcendent and instructive use. In our view, then, a solution to the puzzle will be satisfying only if it confronts more directly the apparent tension between the two uses to which imagination is put. We believe that such a solution is available. Though we reject the equivocation solution, we believe that there is a different and more successful route one can take to resolve the puzzle. To enable us to bring this resolution to the surface, we turn in the next two sections to several of the most influential historical and contemporary treatments of imagination.

2.  Historical Treatments of Imagination So many philosophers throughout history have provided treatments of imagination that we cannot hope to survey them all here. We thus focus on three that have influenced   See Peacocke (1985) and Martin (2002) for arguments that all imagining is imagining from the inside.

8

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

6  Amy Kind and Peter Kung the contemporary debate: Descartes, Hume, and Kant. Though these philosophers’ treatments of imagination are quite different from one another, they each help to lay the groundwork for our own solution to the puzzle of imaginative use.

2.1 Descartes Although there are numerous mentions of imagination throughout the Cartesian corpus, the start of the Sixth Meditation offers a particularly useful summary of Descartes’s view of the nature of imagination and the issue of how it can be distinguished from other kinds of mental activities: When I imagine a triangle, for example, I do not merely understand that it is a figure bounded by three lines, but at the same time I also see the three lines with my mind’s eye as if they were present before me; and this is what I call imagining. But if I want to think of a chiliagon, although I understand that it is a figure consisting of a thousand sides just as well as I understand the triangle to be a three-sided figure, I do not in the same way imagine the thousand sides or see them as if they were present before me. . . . But suppose I am dealing with a pentagon: I can of course understand the figure of a pentagon, just as I can the figure of a chiliagon, without the help of the imagination; but I can also imagine a pentagon, by applying my mind’s eye to its five sides and the area contained within them. And in doing this I notice quite clearly that imagination requires a peculiar effort of mind which is not required for understanding; this additional effort of mind clearly shows the difference between imagination and pure understanding. (AT VII 72–3, emphasis added)

What Descartes seems to have in mind by this peculiar effort of mind seems to be the production of mental imagery—for something to be imagined, we must produce an image of it. Descartes is thus working with an imagistic conception of imagination. Moreover, he seems to be operating with a strict standard of when we will count as having produced an image of the thing in question. As he notes, though we might produce some image or other when thinking of a chiliagon, any such image would merely be “a confused representation” of a multi-sided figure: “it is clear that this is not a chiliagon for it differs in no way from the representation I should form if I were thinking of a myriagon, or any figure with very many sides” (AT VII 72).9 Understood within the overall context of the Meditations, a work whose goal is to achieve “certain and evident knowledge of the truth” (AT VII 10), such passages present what we might call a pessimistic vision of the imagination. Like other rationalists, Descartes dismisses imagination as the wrong kind of faculty to produce the secure knowledge that he seeks. Instead, he settles on clear and distinct perception—what we might now be more inclined to call rational insight—as the principal source of knowledge. Sometimes commentators treat clear and distinct perception as equivalent to an 9   While many subsequent philosophers agree with Descartes’s imagistic conception of imagination, there is considerably less agreement about the strict standard he employs for individuating imagistic representations; many philosophers who give imagistic conceptions of the imagination do not require all 1000 sides of the chiliagon to be represented in order for the imagining to count as an imagining of a chiliagon. See e.g. Kind (2001) and Kung (2010, sect. 6.1).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  7 exercise of conceiving, but even if this is correct, it is clear that for Descartes clear and distinct perception has nothing to do with imagination. For Descartes, the nature of the imagination renders it incapable of providing the kind of certain knowledge needed for his foundationalist project. Descartes’s pessimistic treatment of imagination can also be seen in his discussion of a piece of wax in the Second Meditation. As he notes, we might imagine the wax first pulled in one direction and then in another. But these imaginings proceed by way of specific images of specific changes to the wax, and they encapsulate only a very small number of the countless changes of which the wax is capable. Descartes thus denies that they can adequately support our belief that the wax is flexible, in other words that it is capable of being pushed and pulled into innumerable different configurations: “I would not be making a correct judgment about the nature of wax unless I believed it capable of being extended in many more different ways than I will ever encompass in my imagination” (AT VII 31, CSM 21). Likewise, though we know full well that the wax before and after melting is the same wax, this sameness is not represented in our sense perceptions, and correspondingly, it cannot be represented in our imaginings. For these reasons, Descartes concludes that our conception of the true nature of the wax must come from the understanding and not from sense perception or from imagination.10 We see a related point in Descartes’s discussion of the nature of the “I” earlier in the Second Meditation. Descartes first tries using the imagination in this inquiry, but he cuts short the attempt when he realizes that imagination may involve “fictitious invention.” Because imagination proceeds via contemplation of an image of a thing, it cannot teach us anything about the thing’s true nature. Throughout this discussion, we see explicitly the triumph of the transcendent use of imagination in Descartes’s work. On his view, given imagination’s power to be inventive, it cannot be used as a reliable guide to the properties of the self or, more generally, as a reliable guide to the nature of reality. As he concludes: [T]o say “I will use my imagination to get to know more distinctly what I am” would seem to be as silly as saying “I am now awake, and see some truth; but since my vision is not yet clear enough, I will deliberately fall asleep so that my dreams may provide a truer and clearer representation.” I thus realize that none of the things that the imagination enables me to grasp is at all relevant to this knowledge of myself which I possess, and that the mind must therefore be most carefully diverted from such things if it is to perceive its own nature as distinctly as possible. (AT VII 28; CSM p. 19)

Descartes makes a similar point in a letter to Mersenne, referring to the imagination as “more of a hindrance than a help” when it comes to metaphysical speculation (AT II, 622; CSM, p. 114). 10   See Wilson (1978, ch. 2, sect. 5); compare Sepper (1996, p. 266), which argues that imagination does have some role to play.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

8  Amy Kind and Peter Kung Here it’s worth noting that Descartes’s position is representative of the general tendency of the rationalists to privilege intellect, or understanding, over other mental faculties. Rationalists are suspicious of sense perception as a direct source of knowledge; their indictment of imagination follows from this mistrust. For example, though Spinoza’s imagistic conception of imagination is slightly different from Descartes’s— most notably in rejecting his strict standard11—he shares Descartes’s skepticism about imagination as a source of knowledge; in fact he repeatedly identifies it as a source of error. Take, for example, what Spinoza sees as a common misconception about quantity, namely, that it is divisible. In his view, this misconception stems from the imagination: If I am now asked why we have this natural inclination to divide quantity, I reply that we conceive quantity in two ways, to wit, abstractly, or superficially—in other words, as represented in the imagination—or as substance, which we do only through the intellect. If therefore we consider quantity as it is presented in the imagination—and this is what we frequently and readily do—we find it to be finite, divisible, and made up of parts. But if we consider it intellectually and conceive it in so far as it is substance—and this is very difficult—then it will be found to be infinite, one, and indivisible, as we have already sufficiently proved. (Ip15)

For Spinoza, then, imagination is the source of error about the very nature of things: And since those who do not understand the nature of things, but only imagine things, make no affirmative judgments about things themselves and mistake their imagination for intellect, they are firmly convinced that there is order in things, ignorant as they are of things and of their own nature. . . . And since those things we can readily picture we find pleasing compared with other things, men prefer order to confusion, as though order were something in Nature other than what is relative to our imagination. (I, Appendix)

As this discussion shows, though the issues underlying the puzzle of imaginative use are very much on display in the work of rationalists like Descartes, there’s an important sense in which the rationalist framework dictates that the puzzle has no solution. On Descartes’s view, the transcendent use of imagination derives from its very nature. Because any instructive use of the imagination would be in conflict with its transcendent use, the very nature of the imagination precludes the possibility of its having an instructive use.

2.2 Hume The importance of imagination in Hume’s cognitive psychology, especially in contrast to rationalists like Descartes, would be hard to overstate. According to Hume’s Copy Principle, all mental contents are in some sense imagistic, and as such the imagination lies at the foundation of his cognitive psychology. 11   Spinoza’s rejection of Descartes’s strict standard can be seen in passages like the following: “Further, to retain the usual terminology, we will assign the word ‘images’ (imagines) to those affections of the human body the ideas of which set forth external bodies . . . , although they do not represent shapes. And when the mind regards bodies in this way, we shall say that it ‘imagines’ (imaginari)” (IIp17s).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  9 In a broad sense, Hume accounts for all thinking in terms of memory and imagination. Memory is preservative: The representational contents of memory have their causal origin in impressions, either the impressions of reflection or the impressions of sensation. All other cognitive activity is the province of imagination. On top of this, Hume recognizes no sharp distinction between memory and imagination; memories are just those ideas that are more forceful and vivacious. Imagination, then, taken as the faculty of images, is the representational faculty for Hume. This means that most kinds of thinking, including reasoning and understanding, are aspects or features of imagination.12 With imagination understood so broadly, it seems plausible that it would have not only a transcendent use but also an instructive use. If all representational mental activity divides into memory and imagination, then the denial that imagination has an instructive use would mean that we could learn about the world only through memory—a claim that seems clearly false, as Hume would agree. But this apparent vindication of the instructive use of the imagination is not, in the end, very significant. For surely the ordinary notion of imagination—even factoring in the disagreements among contemporary philosophers about how exactly we should understand this notion—is considerably narrower than Hume’s, and it is this much narrower notion for which the puzzle of imaginative use arises. The importance of Hume for this issue does not end there, however, because he also uses “imagination” in a narrower sense. In this narrower sense, imagination contrasts with reason. Like Locke before him, Hume thinks of reason as the faculty responsible for generating, recognizing, and accepting inferences. Of course, the broad representational faculty labeled “imagination” plays a role in generating ideas even when inferences are absent; when it does so, we have imagination in the narrower sense.13 It is this narrower sense of imagination that we focus on in what follows. Hume clearly thinks that imagination has a transcendent use; in fact, his reflections on the topic have been particularly influential in helping to understand this use: Nothing, at first view, may seem more unbounded than the thought of man, which not only escapes all human power and authority, but is not even restrained within the limits of nature and reality. To form monsters, and join incongruous shapes and appearances, costs the imagination no more trouble than to conceive the most natural and familiar objects. And while the body is confined to one planet, along which it creeps with pain and difficulty; the thought can in an instant transport us into the most distant regions of the universe; or even beyond the universe, into the unbounded chaos, where nature is supposed to lie in total confusion. (Enquiry Concerning Human Understanding, sect. 2)

But though the imagination may appear to have “unbounded liberty,” Hume notes that it is subject to an important limit, one that follows directly from his empiricism. While the imagination has virtually unlimited recombinatorial powers, the resources it has at   See Garrett (1997, ch. 2, esp. pp. 25–9).   

12

  Garrett (1997, pp. 26–8).

13

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

10  Amy Kind and Peter Kung its disposal are limited. In particular, those resources are restricted to “the materials afforded us by the senses and experience” (Enquiry, sect. 2). Hume also places an important second limit on imagination; in his view, it is beyond the power of the imagination to represent a contradiction: 'Tis an established maxim in metaphysics, That whatever the mind clearly conceives includes the idea of possible existence, or in other words, that nothing we imagine is absolutely impossible. We can form the idea of a golden mountain, and from thence conclude that such a mountain may actually exist. We can form no idea of a mountain without a valley, and therefore regard it as impossible. (Treatise, I.ii.2) Whatever can be conceived by a clear and distinct idea [i.e. by imagination in the narrow sense] necessarily implies the possibility of existence; and he who pretends to prove the impossibility of its existence by any argument derived from the clear idea, in reality asserts, that we have no clear idea of it, because we have a clear idea. ’Tis in vain to search for a contradiction in any thing that is distinctly conceived by the mind. Did it imply any contradiction, ’tis impossible it could ever be conceived. (Treatise, I.ii.4)

By connecting imagination and possibility in this way, Hume suggests that imagination has an instructive use in addition to its transcendent use, namely, imagination provides us with knowledge about possibilities. But should knowledge of possibilities really count as an instructive use of the imagination, that is, as a way of instructing us about the world as it is? In talking of the world as it is, we mean to be using the expression in an expansive sense.14 For example, facts about the world as it is include future facts, necessary facts, and at least some other modal facts. This suggests that here Hume does indeed offer us an instructive use of the imagination, though a complete answer to this question will await fuller discussion of modal epistemology in Section 3.2 below. For now, it’s worth noting that for Hume, imagination tells us about possibilities only in the broadest sense. It cannot tell us about matters of fact; that is the province of impressions and custom and habit. It can tell us only about relations of idea, and even there, its proper role is confined to disconfirmation rather than confirmation. For example, by way of imagining, we can rule out the possibility that causal connections are relations of ideas. We can imagine colliding billiard balls flying up in the air, rather than ricocheting off of one another: When I see, for instance, a Billiard-ball moving in a straight line towards another; even suppose motion in the second ball should by accident be suggested to me, as the result of their contact or impulse; may I not conceive, that a hundred different events might as well follow from that cause? May not both these balls remain at absolute rest? May not the first ball return in a straight line, or leap off from the second in any line or direction? All these suppositions are consistent and conceivable. (Enquiry, sect. 4, part I)

What this suggests is that—though Hume’s imagistic conception of imagination is in some ways like that of Descartes—Hume is not quite as pessimistic about the epistemic   We do this in part to accommodate our own differing interpretations of this notion.

14

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  11 credentials of imagination as Descartes is. Though Hume agrees that imagination ­cannot reveal contingent facts about the actual world, he nonetheless allows that imagination gives us knowledge of something about the world as it is. Whereas Descartes frames imagination in a way that means the puzzle of imaginative obviously has no non-skeptical solution, Hume accepts that imagination has both transcendent and instructive uses. For Hume, the puzzle is a challenge: How can our unbounded imagination provide us with knowledge of possibilities? Importantly, Hume’s answer to this question points in the direction of the solution to the puzzle that we wish to pursue. On his view, imagination yields knowledge of possibilities only because it is not truly unbounded; rather, it is subject to a certain limit. Because imagination is limited by the scope of metaphysical possibilities, imagination can reveal this scope to us.

2.3 Kant Like Hume, Kant promises an expansive role for imagination. And like Hume, Kant presents us with an interpretive challenge in mapping his notion of imagination onto the notion used in contemporary philosophical discussion. To explain fully what Kant means by imagination and the role that he assigns to it would require us to map out the  overall Kantian machinery, and that is beyond the scope of this introduction. Fortunately, however, we can make some basic points about the critical role that imagination plays in the Kantian system without fleshing out that system in any great detail. We here focus simply on Kant’s distinction between productive and reproductive imagination. Although both kinds of Kantian imagination represent “an object even without its presence in intuition” (Critique of Pure Reason, B151)—that is, without it being present to our senses—these two kinds of imagination differ from one another in the way they represent that object. Kant’s most extensive discussion of reproductive imagination occurs in the Anthropology. Offering an account that sounds strikingly reminiscent of Hume’s empiricist account, Kant characterizes uses of imagination as either inventive or recollective. As he sees its power, however, it is “not exactly creative, for it is not capable of producing a sense representation that was never given to our faculty of sense; one can always furnish evidence of the material of its ideas” (Anthropology, 7:168). Correspondingly, the reproductive imagination is the faculty associated with fantasy and dreaming. Given this understanding of the reproductive imagination, it’s not surprising that Kant explicitly denies that it can produce knowledge. When introducing the distinction between the productive and reproductive imagination in the Critique of Pure Reason, he expresses pessimism about the ability of reproductive imagination to provide knowledge—or, more precisely, to be a necessary condition on the possibility of knowledge: Now insofar as the imagination is spontaneity, I also occasionally call it the productive imagination, and thereby distinguish it from the reproductive imagination, whose synthesis is subject solely to empirical laws, namely those of association, and that therefore contributes nothing to

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

12  Amy Kind and Peter Kung the explanation of the possibility of cognition a priori, and on that account belongs not in ­transcendental philosophy but in psychology. (B152)

The productive imagination, on the other hand, plays an absolutely crucial role in Kant’s transcendental analysis of the necessary conditions for the possibility of knowledge. In the Critique of Pure Reason, Kant outlines the productive role of imagination in perception as the “meeting-ground of the understanding and the sensibility” (Brann 1992, p. 90). For Kant, our raw perceptual input is a series of isolated uncategorized sensible content. Sensibility imposes space and time onto that raw input, thereby providing a series of discrete intuitions.15 Imagination transforms these discrete, static images into continuous wholes. Understanding then applies concepts to the continuous wholes to categorize them. This is known as Kant’s threefold synthesis, “the action of putting different representations together with each other and comprehending their manifoldness in one cognition” (A77/B103). In virtue of the key role it plays in this synthesis, the productive imagination is integrally involved in every perceptual experience. This suggests an important instructive use for the productive imagination. Since perception clearly teaches us about the world, the involvement of imagination in perception would mean that it too teaches us about the world.16 This instructive function is notably different from the one assigned to imagination by Hume; for Kant, imagination does have a role in telling us about contingent facts about the external world. But it’s important to note that the instructive use of the imagination is merely derivative from the instructive use of perception—it is not imagination qua imagination that plays an instructive role. This leaves open the question of whether independent operation of the imagination can have an instructive use. The examples of instructive uses of imagination that we gave at the start of this introduction all involve more or less independent operation of the imagination—we already knew that someone could determine whether they looked good in a hat by seeing their reflection in a mirror while wearing it; the more interesting question is whether (and, if so, how) this determination can be made via an imaginative exercise alone. This aspect of the puzzle of imaginative use remains unaddressed by our discussion thus far. Elsewhere, however, Kant assigns imagination a further instructive function. In the Critique of Judgment, Kant explores our judgments of beauty. He aims to articulate an objective account of beauty according to which, though it is not a feature of perceived objects, it is also not merely in the eye of the beholder. The question of how the Kantian system can allow objective judgments of taste raises many interpretative questions that are beyond the scope of this introduction (see e.g. Allison (2001), esp. ch. 8). For our 15   Sensibility is a non-active, receptive faculty for Kant; sensibility imposes structure on raw input the way a bottle imposes structure on the liquid inside it. 16   Contemporary philosophers have emphasized the role of imagination in perception as well; notable examples include Strawson (1970), Nöe (2004), Nanay (2010), and Church (2013) among many others. See also Church’s contribution to this volume (Chapter 7).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  13 purposes, however, what’s important is that Kant calls upon the productive imagination to deliver the desired result. In contrast to imagination’s role in perception, where imagination is governed by the understanding, aesthetic experience is marked by the “free play” of imagination and understanding. Now there belongs to a representation by which an object is given, in order for there to be cognition of it in general, imagination for the composition of the manifold of intuition and understanding for the unity of the concept that unifies the representations. This state of a free play of the faculties of cognition with a representation through which an object is given must be able to be universally communicated, because cognition, as a determination of the object with which given representations (in whatever subject it may be) should agree, is the only kind of representation that is valid for everyone. (5:217)

This free play results in a feeling of disinterested pleasure. It’s the disinterested pleasure that gives some measure of objectivity to beauty; in judging an object beautiful, I am claiming that you should find it beautiful as well. Because the structure of our psychologies is similar, an object that evokes disinterested pleasure in me should have that effect on you as well. This disinterested pleasure arises from the “harmonious workings of the cognitive faculties,” with productive imagination again playing the role of intermediary between intuition and understanding. Thus, in addition to assigning imagination an instructive use via its role in his threefold synthesis, Kant assigns imagination a further instructive use in helping to teach us about beauty. In both cases, imagination’s potential for instructive use is tied to constraint. In the aesthetic case, the constraints are architectural; the similarity between your psychology and mine enables imagination to play a role in our achievement of aesthetic knowledge. In the earlier perceptual case, the instructive role assigned to the productive imagination arises only from the heavily proscribed role it plays in the threefold synthesis. In contrast to the reproductive imagination, which is unconstrained, the productive imagination is constrained or governed by the understanding. So, as with Hume, though now in two very different contexts, we again see the seeds planted for the sort of solution to the puzzle of imaginative use that we wish to pursue—that imagination’s ability to serve an instructive function depends on the presence of constraints.

3.  Contemporary Treatments of Imagination In the previous section we focused on how imagination has been treated by different philosophers, but here we focus on how imagination has been treated in different philosophical contexts. Imagination has been invoked in a vast variety of such contexts: to explain dreams, pretense, make-believe, aesthetic appreciation, fictional engagement, creativity, and so on. But we here focus on three areas of philosophical research where the invocation of imagination has been particularly prominent and pervasive: engagement with fiction, modal epistemology, and mindreading. Though

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

14  Amy Kind and Peter Kung the first of these contexts suggests a way that imagination plays a transcendent role, the second two suggest ways that imagination plays an instructive role. As we will see, reflection on all three contexts helps shed light on our proposed solution to the puzzle of imaginative use.

3.1  Engagement with fiction At least since the publication in 1990 of two influential works in contemporary ­aesthetics—Kendall Walton’s Mimesis as Make-Believe and Gregory Currie’s The Nature of Fiction—imagination has been assigned a crucial role in our engagement with fiction. Although their theories differ in various respects, at their core the theories are similar: Our engagement with fiction can be best understood in terms of a comparison with children’s games of make-believe. For both philosophers, an understanding of make-believe requires invoking imagination. On Currie’s view, the author of a work of fiction intends that her audience adopt an attitude of make-believe towards the work—that they make-believe that the contents of her work are true. The work thereby generates a game that is importantly analogous to those played by children. Games of make-believe differ markedly from games like cribbage and Scrabbletm, where the rules of play are carefully dictated and the object is to achieve victory. As Currie notes, “Games like mud pies and cowboys and Indians have a very loose formal structure, depend heavily for their continuation on the creative imaginations of the players, and can be played without an intention on anybody’s part of winning” (Currie 1990, p. 71). Likewise for the games of make-believe generated by fictional works: Sometimes the game is played by a single player, sometimes by several players. Moreover, the game might be played in several different ways: by silently reading the words to oneself, by recalling the storyline from memory, by listening to someone else read the words aloud, and so on. Importantly, Currie insists that the notion of “make-believe” employed by his theory is not meant to be a technical term but to capture an ordinary notion familiar to us not only from children’s games but also from our activities of daydreaming and fantasizing. He explicitly connects this notion to imagination; on his view, the attitude of make-believe can be thought of as an attitude of imaginative involvement (Currie 1990, p. 18; see also p. 21). While Currie’s theory is focused specifically on our engagement with works of ­fiction, Walton’s theory addresses the more general issue of our engagement with representational works of art. As already noted, his theory too draws an important analogy to children’s games of make-believe. In particular, he draws our attention to the way that children incorporate objects into such games—objects ranging from sofa pillows to empty cardboard boxes to tree stumps. This incorporation often involves imagination: The pile of sofa pillows is imagined to be a fort, the empty cardboard box is imagined to be a race car, the tree stump is imagined to be a bear. These objects not only serve to prompt imaginings, but they are also often what Walton calls props, that is, they serve to generate fictional truths. In a game in which stumps are imagined to be

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  15 bears, for example, the presence of a stump generates the fictional truth that a bear is present. (As he is careful to note, however, a nearby stump generates this fictional truth even if none of the children involved in the game explicitly imagine anything about the stump.) According to Walton, representational works of art are likewise props in games of make-believe. Consider The Lion, the Witch, and the Wardrobe. The words of this novel make it fictional that the world of Narnia lies behind a certain wardrobe door in Professor Kirke’s country house, and they do so by prescribing a particular imagining. According to Walton, “a fictional truth consists in there being a prescription or mandate in some context to imagine something. Fictional propositions are propositions that are to be imagined—whether or not they are in fact imagined” (Walton 1990, p. 39). For Walton, then, the relationship between imagining and fictionality turns out to be analogous to the relationship between belief and truth: “Imagining aims at the fictional as belief aims at the true. What is true is to be believed; what is fictional is to be imagined” (Walton 1990, p. 41). Even this short discussion of Currie and Walton makes clear the importance of the transcendent use of imagination in contemporary discussions of our engagement with fiction. Currie, for example, notes explicitly that “make-believe allows us to achieve in imagination what we are denied in reality” (Currie 1990, p, 19). Likewise, for Walton, because imagination aims at the fictional, it clearly aims to transcend the world as it is. Interestingly, however, Walton seems to want to deny explicitly that the transcendent use of imagination is all there is to it: A conception of imaginative experiences as, in general, free-floating fantasies disconnected from the real world would be narrow and distorted. Sometimes, to be sure, imagining is a means of escape from reality, and we do frequently imagine what is not really the case. But even when we do, our experience is likely to involve the closest attention to features of our actual environment, not a general oblivion to it. Most imaginings are in one way or another dependent on or aimed at or anchored in the real world. (Walton 1990, p. 21)

Initially this passage may seem to suggest that, for Walton, imagination’s having a transcendent use is compatible with its having an instructive use. Even though imagining aims at the fictional, in many cases what’s fictional in a given work can overlap with what’s true. The Lion, the Witch, and the Wardrobe makes it fictional that Germany bombed London during World War II, but it is also true that Germany bombed London during World War II. In this way, by imagining the propositions made fictional by a representational work of art, might we not learn something about the world? Ultimately, it seems clear that the answer is no. Walton is undoubtedly right that typical transcendent uses of imagination do not escape the world entirely: “fictional” does not mean untrue. But that is not enough to show that such uses are also instructive. Consider, for example, the fact that the Harry Potter books make it fictional both that King’s Cross Station is located in London and that the town of Little Whinging is located south of London. One of these two claims is also true. But which one? Nothing

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

16  Amy Kind and Peter Kung in the work itself will enable the reader to tell. Reflection on the role that imagination plays in our engagement with literature thus seems to make the puzzle of imaginative use particularly difficult. But again, this same reflection also helps to lay the groundwork for the solution we favor. In the above passage from Walton, he suggests that transcendent uses of imagination are often “anchored” in the real world. As we will argue, it’s the possibility of just this kind of anchoring that enables our transcendent imagination also to have instructive capability.

3.2  Modal epistemology Thought experiments abound in philosophy. From zombies to utility monsters, from trolleys to teletransportation devices, these experiments play a central role in philosophical argumentation. In considering the proposed scenarios, one thing we do is form judgments about whether such scenarios are possible.17 But how does such consideration proceed? It seems natural to think that the imagination is involved; as we saw above in our discussion of Hume, imagination is closely connected with possibility. It’s thus no surprise that contemporary philosophers have frequently invoked imagination not only in discussions of thought experiments but also more generally in discussions of modal epistemology.18 Sometimes in these discussions, imagining is not clearly distinguished from conceiving.19 When the distinction is clearly drawn, some philosophers have argued that—in order for our judgments of possibility to be justified—we should contemplate these thought experiments by way of our faculty of conceiving, not by way of our faculty of imagining.20 Other philosophers, however, have argued precisely the reverse. We do not here attempt to adjudicate this debate; rather, we simply focus on the invocation of imagination in this context. Perhaps the most compelling reason to link imagination and possibility is an analogy to perception—the thought that imagination is to possibility as perception is to actuality. As W.D. Hart has claimed, “There are modes of imagination corresponding to each of the five senses . . . It is by virtue of the sensory character of these basic modes of imagination that they sustain modal convictions in analogy with the way the five senses sustain beliefs about the actual world” (Hart 1988, pp. 14–15). Just as perceptual experience involves the appearance of actuality, imaginative experience involves the appearance of possibility.21 17   Note that we are not claiming that this is all that we do with thought experiments. There is some debate about the purposes (plural) of thought experiments—see Cooper (2005) and Walsh (2011). But it is agreed that we can and often do use thought experiments for the purpose of exploring possibilities. 18   For some representative examples, see Chalmers (2002), Gendler and Hawthorne (2002), Geirsson (2005), Hart (1988), and Yablo (1993). 19   See e.g. Sorensen (1992), esp. ch. 2. 20   As we noted above, Descartes holds this view. Bealer (2002) rejects imagination but would not be happy with the label “conceiving” either, though his modal intuitions view in some ways resembles Descartes’s. See Gendler and Hawthorne (2002) for discussion. 21   See Yablo (1993).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  17 These remarks themselves do not constitute an argument that imagining is a guide to possibility. Rather, they articulate the psychological fact that imagining leads us to form judgments of possibility while at most hinting at features of imagination that might be used in constructing the needed argument. The strategy for contemporary proponents of an imagination-based modal epistemology has been largely defensive; proponents attempt to explain why various skeptical challenges to an imagination-­ based modal epistemology fail. The first step is typically to concede that imagination can provide at most a fallible guide to possibility. Further defensive actions generally fall into one of three categories: 1. Distinguishing imagining from other similar mental activities that it might be confused with, such as supposing, entertaining, finding believable, and so on. 2. Responding to Kripkean worries about a posteriori necessities, and our apparent ability to imagine their falsity. 3. Finessing the gap between what is imaginable and what only seems imaginable. The success of one or more of these strategies would indirectly help to establish that imagining provides a guide to possibility by defusing a skeptical worry. Returning to the puzzle of imaginative use, we are still left with the question we touched on briefly in connection with our earlier discussion of Hume: In providing us with knowledge of possibility, is imagination thereby being put to instructive use? Insofar as this consideration leads us to learn something about the world as it is, it seems that it must constitute an instructive use of imagination. As we noted above, we interpret “as it is” expansively, to include at least some modal facts. To navigate the world as it is successfully, we need to know more than just descriptive facts about the actual world, narrowly construed. In planning, we need to know about future possibilities, in order to avoid ones we dislike. Mindreading, as we’ll see in Section 3.3, sometimes teaches us modal facts—for example, could my boss be persuaded to give me a raise? And sometimes we can draw inferences about the actual world from modal facts, as in the consideration of thought experiments.22 The contributions in this volume help to flesh out the notion of the world as it is and thereby provide a fuller sense of the kinds of instructive uses to which imagination can be put. This way of understanding the role of imagination in modal epistemology leads in turn to a new way to understand skeptical challenges to that role. In particular, such challenges can be aptly reframed in terms of the tension between the transcendent and instructive uses of imagination. Consider, for example, the worry raised by Paul 22   For example, consideration of Frank Jackson’s (1982) Mary case—assuming that the case is successful—­ teaches us that physicalism is false. Consideration of Judith Jarvis Thomson’s (1971) violinist case—again, assuming that the case is successful—teaches us that there are at least some cases in which abortion is morally justified. More generally, a successful thought experiment convinces us that some situation is meta­ physically possible and thereby places constraints on an adequate theory of the relevant phenomenon. But since these theories are theories about the actual world, it seems that in learning about possibilities we can learn about the world as it is, thus the employment of imagination in modal epistemology can be an instance of imagination being put to instructive use.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

18  Amy Kind and Peter Kung Tidman in conjunction with Wittgenstein’s example of someone who claims to imagine King’s College on fire: It would be absurd to ask, “Are you sure its King’s College? Maybe you are just imagining a building that looks like King’s College.” Whether one is imagining King’s College depends not on the image, but on what we take the image to be an image of. I think I can imagine King’s College on fire although I have no idea what King’s College looks like. But, likewise, I think I can imagine a number of things I know to be impossible. (Tidman 1994, p. 301)

In raising this worry, Tidman directly challenges the efficacy of the third kind of defensive maneuver mentioned above; in his view, there is no principled way to distinguish that which is imaginable from that which only seems imaginable. If we are the authority about what we imagine—as one would expect from the fact that imagination has a transcendent use—then surely we must accept that we have the power to imagine the impossible. But if we can easily imagine the impossible, then it does not seem that imagination can tell us anything about possibility: If conceivability is measured simply in terms of our ability to form mental images, either we can form mental images of impossible states of affairs or, if we can only really form mental images of possible states of affairs, we cannot tell whether we have succeeded in doing so. On either account our ability to form mental images does not provide us with a usable test to discover what is possible. (Tidman 1994, p. 301)

On Tidman’s view, then, imagination’s transcendent capabilities rule out its having an instructive use. More generally, the contemporary dialectic concerning modal epistemology highlights the very tension that underlies the puzzle of imaginative use. We seem free to imagine the impossible—the accounts of fiction from the previous section strongly suggest this—but this freedom makes imagination unsuitable for instructive use about possibility. Further exploration of the modal epistemology literature, however, also points us towards a way of resolving this tension; yet again, the notion of constraint plays a key role. As many philosophers have suggested, it is only when we consider a sufficiently constrained imagination that imagination could lead to possibility. Some philosophers, like Hart, find the constraints in the imagistic aspect of imagination. Mental imagery has a definite structure and obeys certain rules—for example, no images of anything red and green all over. According to these philosophers, these constraints mean that imagistic imagination isn’t completely free, and that it depicts possibilities.23 Other philosophers have been inspired by Kripke’s groundbreaking work to look for constraints on imagination. They adopt the second defensive strategy above, trying to explain why the existence of a posteriori necessities does not doom modal epistemology. One popular answer in this vein, which we might call “post-Humean,” is that when we are informed of relevant empirical facts, we cannot imagine the impossible. 23   See e.g. Hill (1997) and Kung (2010, sect. 4); see also Balcerak Jackson’s contribution to this volume (Chapter 1).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  19 In other words, our ability to imagine the impossible arises only from a particular sort of ignorance; if we eliminate the ignorance, then imagination is constrained to reveal possibilities.24 Another answer is to emphasize voluntary rational constraints on imagination. We can deploy imagination under rational constraints; the idea is that we keep our imagination free of contradictions as we fill in more and more details about the imagined scenario.25 It is beyond the purview of this introduction to assess the plausibility of the various constraints suggested by proponents of an imagination-based modal epistemology. But as our discussion reveals, the notion of constraint appears to play a key role in validating this instructive use of the imagination.

3.3 Mindreading When we interact with others, we constantly develop explanations of their actions and make predictions about their future behavior. This practice—often referred to as mind­ reading—is perhaps particularly explicit when we’re competing against other people. In such cases, we often deliberately pause to try to figure out why our opponents might have done what they’ve done and also to determine what they’ll do next. Suppose, for example, that you’re involved in a Scrabbletm game against a skilled player. She passes her turn. You take this to mean that she’s hoping you’ll open up the board in such a way that she’ll be able to bingo on her next turn, so you then make your next move cautiously. But mindreading occurs all the time, in almost every context in which we’re interacting with others—from determining why a colleague said what they did or why another colleague is frowning, to calculating what kind of response will forestall further tears from a child or what kind of prompting will best ensure the child’s behavioral compliance. Much of the time, mindreading comes so naturally to us that we don’t even realize that we’re doing it. What accounts for our abilities to mindread? According to what’s known as theory theory, we employ a tacit folk psychological theory. Though theory theory was for a long time the standard view both in philosophy of mind and in cognitive science more broadly, in the mid-1980s a different sort of explanation of mindreading came on the scene. According to simulation theory, mindreading is best explained by our capacity for simulation; to figure out why someone did something or what they will do next, we mentally project ourselves into their situation and simulate their reasoning and decision-making. As the debate between theory theory and simulation theory has played out over the last 30 years, the line between these two theories has become somewhat blurred. Most philosophers today now accept that, in at least some situations and in at least some way, we use simulation in our efforts to explain and predict the behavior of ­others. For our purposes, however, what’s important is that such simulation is standardly taken to involve imagination, so that simulation is understood to be a process of imaginative   Gregory (2004) and Yablo (1993) explicitly give this answer.   See Chalmers (2002) and Geirsson (2005) for approaches like this.

24 25

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

20  Amy Kind and Peter Kung identification or imaginative projection. As Gregory Currie and Ian Ravenscroft describe it: Imaginative projection involves the capacity to have, and in good measure to control the having of, states that are not perceptions or beliefs or decisions or experiences of movements of one’s body, but which are in various ways like those states—like them in ways that enable the states possessed through imagination to mimic and, relative to certain purposes, to substitute for perceptions, beliefs, decisions, and experiences of movements. (Currie and Ravenscroft 2002, p. 11)

As this quotation suggests, the kind of imagination invoked in simulation theory is imagining from the inside. In contrast, the kind of imagination we have encountered in other contexts has often been imagining from the outside. Engaging with fiction, for example, typically involves visualizing the actions of the characters. In the case of Frank Jackson’s Mary, one sees her in her black and white room surrounded by black and white textbooks. But simulating requires something different. Rather than visualize your Scrabbletm opponent’s furrowed brow and look of intense concentration, you imagine her mental processes. In this way, simulation is experiential imagining— imagining being her rather than imagining seeing her.26 For our purposes, what’s especially interesting about mindreading is that it provides a clear case of imagination being put to instructive use.27 Our imaginative projections aim at providing us with knowledge of other people’s minds and actions—what they’re thinking, why they did what they did, and what they’re going to do next. Moreover, this knowledge is not knowledge of necessary truths but of contingent facts about the world. This again brings our puzzle to the forefront. In both historical and contemporary contexts, we have seen clear examples of the tension between the transcendent and the instructive uses of imagination. We now turn to addressing this tension more directly.

4.  Resolving the Puzzle The discussion of the last two sections has introduced us to some of the many roles assigned to imagination in both historical and contemporary discussions. Insofar as we have seen that different philosophers seem to have vastly different understandings of imagination, our discussion may appear to have lent some support to resolving the 26   Of course, imagining from the inside is used in many contexts besides simulation. For example, there are many thought experiments whose consideration seems to require imagining from the inside. When John Searle (1980) presents us with the Chinese Room thought experiment, we imagine being in the room manipulating squiggles and squoggles. When Derek Parfit (1984) presents us with the teletransporter case, we imagine pushing the button and being transported to Mars. Sydney Shoemaker (1993) argues that certain counterexamples to physicalism must be imagined from the inside to succeed; he thinks this fact reveals important limits on the power of these counterexamples. So though imagining from the inside plays a crucial role in simulation theory, it is not limited to the context of mindreading. 27   Though see Maibom’s and Spaulding’s contributions to this volume (Chapters 8 and 9) for skepticisms about imagination’s ability to be put to this instructive use unaided.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  21 puzzle of imaginative use by invoking the equivocation solution we discussed in Section 1. But, as we there noted, the equivocation solution seems unpromising. Having now looked at various treatments of the imagination in some detail, the inadequacy of the equivocation solution becomes even clearer. First, although we have seen significant differences in the way different philosophers treat imagination, there is still considerable overlap among such treatments. To point to one important area of overlap, most philosophers accept that at least some instances of imaginings involve conjuring and manipulating mental images. The existence of this overlap diminishes the plausibility of the equivocation solution. Second, one theme running through many philosophers’ treatment of imagination has been the argument that because imagination has a transcendent use, it cannot also have an instructive use. The equivocation solution would uncharitably and implausibly turn such arguments into fallacies. Third, where conceptions of imagination differ from one another, they are recognizably different. Thus, while there may be occasions when philosophers talk past each other in discussions of the imagination, it is not plausible that this can account fully for the puzzle of imaginative use. In our view, the better way to resolve the puzzle, and the way that fits better with both historical and contemporary treatments of imagination, is to look for constraints or limitations on imagination. Many of our contributors, having hit upon a similar idea independently, remark that an instructive use of imagination demands constraints to keep imagination realistic or “reality congruent.”28 Were our imagination completely unconstrained—as, for example, was suggested in Descartes’s work—putting it to instructive use would be impossible. But even though imagination is a powerful faculty, a faculty that we frequently use in an unrestrained mode to let our fantasies fly free, there can be (and often are) limitations on its employment. While unconstrained imagining accounts for its transcendent power, we can account for its instructive power in terms of the constraints that come into play. Our discussion has highlighted two primary classes of constraints. First, the constraints may be architectural; that is, they may result from our cognitive psychological architecture. Our psychological architecture prevents us from imagining certain things or using the imagination in particular ways. Second, the constraints may derive from a non-architectural source, such as from our will. This latter class of constraints are of the sort that we can (perhaps only when properly disciplined) voluntarily impose upon our imaginative projects. Recall some of the constraints that have come into play throughout our discussion and that have dissipated the tension between the transcendent and instructive uses of imagination: • For Hume, the fact that we are unable to imagine outright contradictions allows imagination to help teach us about relations of ideas; more specifically, it can   The expression “reality congruent” comes from Van Leeuwen’s contribution to this volume (Chapter 3).

28

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

22  Amy Kind and Peter Kung help to show us when something is not a proper relation of ideas. In keeping with his view of himself as developing a science of the mind, it seems plausible that he would see the limits as involuntary and architectural.29 • In Kant’s system, the productive imagination is essential to the threefold synthesis, playing the intermediary role between sensibility and understanding and thereby contributing essentially to our ability to have perceptual knowledge. This  is, for Kant, an involuntary architectural limitation owing to the constraints on perception imposed by transcendental conditions for the possibility of knowledge. • Kant’s system also imposes constraints on the reproductive imagination in its role in the formation of aesthetic judgments. As we saw, the effective employment of reproductive imagination in this context is governed by its interaction with the understanding. While the “free play” between these two faculties may be to some extent under our control, the “harmonious workings of the faculties” seem also to depend on architectural factors. • In the literature on modal epistemology, the question of whether and how imagination can justify beliefs in possibility is closely tied to the question of what limitations the imagination is subject to. Some find constraints in the imagery of imagination: Mental imagery is constrained to be roughly similar to perceptual or sensory imagery. Others take the lesson of Kripke’s work to be a Humean-type constraint on our ability to imagine impossible situations. We can misdescribe the situations we have imagined, perhaps because of our ignorance of certain empirical facts, but the situations themselves are possible. Though our discussion of the constraints operative in instructive uses of imagination has not been comprehensive, our basic point generalizes to other contexts in which we use imagination to learn about the world. In mindreading, for example, it seems plausible that our imaginings will have to be somehow reined in if we have any hope of learning about other people. Perhaps architectural constraints come into play, as might be the case if our simulative mechanism is simply an offline version of our decision-­making mechanism. Alternatively, perhaps we willingly constrain our imaginative simulations in an effort to learn about others.30 But absent any constraints whatsoever, if wholly unrealistic simulations are as likely as realistic simulations, it is hard to see how imagining other minds could teach us anything about the minds of others. While our solution to the puzzle of imaginative use sees constraints on imagination as necessary for it to have an instructive use, we do not believe that the presence of constraints is sufficient for such use. Rather, the constraints must be the right kind. Not all limitations lend themselves to an instructive use. To give just one example, a 29   Hume also sees the imagination as subject to other “universal principles,” like resemblance and contiguity. See Streminger (1980). 30   See Morton (2006) for an interesting discussion of what’s involved in accurately imagining other people.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  23 frequently discussed phenomenon in the recent imagination literature is the puzzle of imaginative resistance, our inability or unwillingness to imagine counter-moral propositions in fiction.31 Though we seem to be able to imagine a world in which a parent kills a newborn simply because the child is a girl, we seem unable to imagine a world in which such a killing is morally acceptable. The mere presence of this limitation does not by itself guarantee that imagination can teach us about morality. More argument would be needed to show whether this constraint means that imagination (or the ­inability to imagine) enlightens us in the moral context.

5.  Guide to the Contributions With the constraints framework for solving the puzzle of imaginative use in place, we turn in the final section of this introduction to the contributions that appear in this volume. In various ways, and across various different philosophical contexts, each contributor attends to the potential for the instructive use of imagination, addressing the question of whether (and how) we can have knowledge through imagination. In line with our proposed solution to the puzzle of imaginative use, in our brief discussions of their contributions we will pay special attention to the constraints (or lack thereof) that each of them believes operate in imagination.

5.1  The Taxonomist We noted earlier in this introduction that different philosophers have surprisingly different conceptions of imagination. Add in that philosophers also routinely appeal to apparently related notions like conceiving, supposing, entertaining, hypothesizing, and so on, and you have all the ingredients for miscommunication or equivocation. Magdalena Balcerak Jackson aims to bring clarity by proposing a taxonomy of mental activities in this vicinity. She argues that there are three distinct types of speculative mental activity: imagining, supposing, and conceiving. Imagining is experiential perspective-taking, according to Balcerak Jackson. It is recreative in Currie and Ravenscroft’s (2002) sense: you put yourself in the perspective of another actual or merely possible subject. This is best illustrated with imaginings that involve mental imagery: to imagine waking up attached to a famous violinist, rather than merely supposing it, is to conjure up the vivid imagery of feeling the IV in your arm, seeing yourself surrounded by hospital machinery, and worrying about what has happened to you. It’s these perspectival elements that make imagining more demanding than mere supposition: to imagine, you have to imagine the way you are hooked up to the violinist, and not just the fact that you are. If you are not sufficiently familiar with the perspective you are trying to take on, your imaginative endeavor may 31   The puzzle is discussed at length in Gendler (2000). Similar limitations on our imaginings may extend to other false propositions beyond those that are counter-moral. See Weatherson (2004) for helpful discussion.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

24  Amy Kind and Peter Kung fail. Additionally, imagining, but not supposing, is subject to the phenomenon of imaginative resistance. And imagining has the potential to motivate action, whereas supposing at most motivates deliberation. For Balcerak Jackson, the recreative nature of imagination entails that it must be constrained: “If imagination, by its very nature, aims to create simulations that take up the phenomenal character and content of corresponding (possible) experiences, then the character of our experiences is constrained by the properties exhibited by these corresponding experiences” (emphasis ours). She suggests that these constraints are the right kind to justify belief in possibility. Imagination recreates a subject’s perspective, providing defeasible justification that the way things are imagined to appear is a way that they could possibly appear. What, then, is supposing? Balcerak Jackson argues that supposing is acceptance, in Stalnaker’s technical sense of acceptance as “a label for those mental states that are primarily involved in the acts of deliberation and inquiry of intelligent subjects.” We accept propositions for specific purposes, typically to draw out their consequences. On this minimal understanding of supposition, it is clear that we can suppose conceptual or a priori falsehoods. Supposing does not, by itself, justify belief in possibility, but of course it can play a role in reasoning that reveals modal truth, as in a reductio argument. Having distinguished imagining from supposing, Balcerak Jackson next turns her attention to the mental activity of conceiving. While she finds this notion more elusive, she eventually concludes that there is room for a mental activity here that is distinct from both imagining and supposing. If imagination is recreative perspective-taking, then conceiving is recreative belief-taking, or simulated rational belief: “When one conceives of P, one does engage in an exercise of perspective-taking. But one does not take the perspective of the subject as the subject of phenomenal experiences, but rather as the subject of rational belief.” Conceiving in this sense is distinct from supposing because putting oneself in the position of somebody who believes that P involves rational commitments that are not constitutive of merely supposing that P. So understood, conceiving may justify belief: It depends on how well we are able to put ourselves in the shoes of ideal reasoners and trace through the rational commitments of our simulated belief.

5.2  The Architects Two of our contributors, Peter Langland-Hassan and Neil Van Leeuwen, offer architectural solutions to the problem of imaginative use. They marshal evidence from psychology and neuroscience to show that knowledge-permitting constraints on imagination are part of our cognitive architecture. In “On choosing what to imagine,” Peter Langland-Hassan tackles the puzzle of imaginative use head-on. “What we imagine seems determined by what we wish to imagine,” he writes. “But how can it be that we are able to wish ourselves into more

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  25 epistemically favorable states of mind?” To understand how imagination can be under our voluntary control, we might look to the fact that our imaginings are driven by our intentions. On this view—the “Only Top Down” (OTD) view—we think of imagining as a sequence of mental states where the content of each state in the sequence is simply transferred from an intention to imagine that content. On OTD, however, imagining simply becomes “the mental equivalent of handing yourself a dollar.” What is the epistemic point of moving intentional contents from intention to imagination? There seems to be none. On the other hand, letting imagination run completely free seems epistemically worthless as well. We thus need constraints on imagination. Langland-Hassan offers a neat architectural solution by adding lateral constraints to the OTD view. One’s intentions determine not the content of the entire series of imaginings but only of the first state in that series. Thereafter, what determines the content of subsequent states in the series is a constraining algorithm of some kind. In the case of propositional imagination, a popular proposal for this “algorithm” identifies it with the inferential patterns of belief. In the case of sensory imagination, some have proposed that the algorithm is the same as the one that generates predictions of expected sensory input during normal perception. Langland-Hassan labels views that appeal to both top-down influences and lateral constraints in this way “TOPLAT” approaches. However, TOPLAT approaches tend to undersell the freedom of imagination. It seems clear that we can imagine things going in ways we would never infer that they were going if we came to believe the imagined premise (or ways we would never sensorily predict, if we perceived the initial condition). Langland-Hassan calls this the “More Free” objection, and considers several possibilities for responding to it. The most promising, he argues, is to hold that imagining often involves the cyclical, ongoing involvement of one’s “top-down” intentions. The final picture that emerges is one where we can begin an imaginative episode, and intervene in an imaginative episode, at will; when we do so, the imaginative episode inherits its content directly from our intentions. This accounts for the transcendent use of imagination. However between interventions, imagination runs as inference or perception would, using the same mechanisms as inference or perception. That’s the source of constraints, and hence what solves the puzzle of imaginative use. Langland-Hassan concludes by arguing that a cyclical TOPLAT approach lends some credence to views that he has argued for elsewhere: the counterintuitive view that propositional imagining involves only ordinary beliefs and desires, and that sensory imaginings are commitments of a kind, with correctness conditions comparable to those of natural language conditionals. What seems most implausible about such views is their apparent conflict with the freedom of imagination. Yet it turns out that they explain the freedom of imagination in exactly the same way as the most promising accounts of imagination that resist identifying it with belief or quasi-perceptual “commitments.”

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

26  Amy Kind and Peter Kung Like Langland-Hassan, Neil Van Leeuwen believes there are important cognitive architectural constraints on imagination. In Van Leeuwen’s view, these architectural constraints demonstrate that imagination is essential to our agency because imagination prepares us for empathy-based moral appraisal of future actions. Empathy-based moral appraisals are examples of social emotions; Van Leeuwen notes that the recent psychological and philosophical literature emphasizes the role of social emotions in forming appropriate moral judgments, and hence in gaining moral knowledge. Van Leeuwen rests his view on the empirical claim that there is an activation pathway from imagistic imagining to emotional activation, which he labels the I-C-E-C pathway, largely the same pathway as the one from percept to emotional activation (the P-C-E-C pathway). Both early stage perceptual processing and imagistic imagining lead to primary categorization—a kind of rough, imprecise sorting that falls short of full conceptualization—that in turn triggers an automatic emotional response and largely automatic behavioral dispositions. A curved stick by the side of the trail triggers the primary categorization SNAKE, which in turn triggers a fear response, and this happens even if that stimulus is eventually conceptualized as a stick. Van Leeuwen identifies three key features of the I-C-E-C pathway that make it useful to agents such as ourselves. First, imagery triggers emotions automatically; we can exert no direct voluntary control over whether an image triggers an emotion, nor over which emotion is triggered. Second, the I-C-E-C is reality congruent, in the sense that “we imagistically imagine [things] that could happen in the environment in which we live.” Like Langland-Hassan and several others in this volume, Van Leeuwen thinks the transitions from one imagining to another are governed by the same inferential processes that govern transitions from one belief to the next. He also cites evidence of our preference for elaborating imaginings in conformity with familiar patterns of causation. Third, imagistic imaginings can be integrated into the perceived space around an agent. You can imagine a growl coming from that dark closet you’re looking into, or imagine a snake behind the log you see in front of you. These three features mean that imagined stimuli can initiate bodily preparations for an imagined future event. The fact that our emotional responses are automatic means that our bodies can be prepared in advance of conceptual thought. The fact that our imaginings are generally reality congruent means that we aren’t usually preparing for unlikely threats and opportunities. And the fact that imaginings integrate into perceptual space means that we can gear our responses towards particular features in our surroundings. Van Leeuwen then uses the results of several empirical studies to argue that this mechanism plays a key role in our ability to engage in empathy-based moral appraisals of future actions and others’ situations. Imagining social situations, and in particular imagistically imagining social situations, automatically triggers moral emotions, the very sorts of emotions relevant to an initial moral appraisal. It’s these empathy-based moral appraisals that make I-C-E-C central to our ability to be social agents, and to our ability to gain moral knowledge.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  27

5.3  The Optimists The next four contributors are also sanguine about imagination’s ability to provide knowledge, but unlike Langland-Hassan and Van Leeuwen, their solution to the problem of imaginative use does not lie in our cognitive architecture. Jonathan Jenkins Ichikawa and Timothy Williamson contend that the puzzle of imagination use gets purchase only when we focus on the transcendent use of imagination and ignore the much more common prosaic uses. They think it plausible, in fact, that evolution selected for our ability to imagine scenarios. Having the right kind of imagination—an imagination that is sufficiently tethered to reality—confers obvious evolutionary advantages. The constraints on imagination that keep it tethered to reality are ones built in by evolution. Timothy Williamson speculates that our ancestors would have found imagination useful for many purposes. When voluntarily solving problems—can I leap across this river, or do I need to take the long walk around?—imagining would have played a key role. Likewise, imagining would have helped alert them and prepare them for opportunities or dangers, as when someone is primed to fight or flee by imagining wolves in the forest. For imagination to be practically useful, and hence for it to confer a fitness advantage, imagination needs to be both selective and reality-oriented. Selectivity refers to the slate of possibilities that are imagined; imagination seems designed to bring forward practically relevant possibilities (the wolves will attack us, or the wolves growl and defend their territory) rather than irrelevant possibilities (the wolves cook meat over a fire and serve it to us). Reality-orientation concerns how imagined scenarios unfold. According to Williamson, once a subject imagines the initial conditions of a scenario and “lets the rest of the imaginative exercise unfold without further interference . . . his imagination operates in involuntary mode. . . . Left to itself, the imagination develops the scenario in a reality-oriented way.”32 The twin constraints of selectivity and reality-orientation mean that, in the right conditions, imagination can confer knowledge. According to Williamson, the costs of skepticism about imagination’s ability to deliver knowledge are higher than we may have realized. There are important cognitive similarities between reasoning hypothetically with imagining and reasoning actually with new information. If we think of reasoning with new information as an online process, then reasoning hypothetically via imagination would be the offline process. For example, if someone tells Stephen that Albert Pujols cleaned out his locker at Angel Stadium on Monday after the playoff loss to the Kansas City Royals, and as a result Stephen concludes that the Angels will trade Pujols this offseason, then presumably Stephen could have reached the conclusion, “If Pujols cleans out his locker after a playoff loss, then the Angels will trade Pujols this offseason.” If Stephen accepted the conditional by conducting the kind of selective, reality-oriented imaginative exercise just   Compare Williamson’s account to Spaulding’s (see this Introduction, Section 5.4).

32

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

28  Amy Kind and Peter Kung described, then that imaginative exercise is the offline analog of online updating. This means that “any scepticism about the offline processes is liable to generalize to the online processes too.” Though Williamson sees imagination as having evolved to handle very practical matters like “can I make this jump across the river,” he closes his piece by arguing that this does not mean that imagination is reliable for only those types of tasks. “The simplest forms of reasoning to implement that are truth-preserving for practical matters are truth-preserving for all matters.” Philosophers’ use of thought experiments is just a more elaborate version of the standard method of evaluating conditionals. Jonathan Jenkins Ichikawa also offers an evolutionary argument designed to vindicate philosophical knowledge of metaphysical possibility and necessity. Ichikawa is interested in why we have the ability to know facts about metaphysical possibility and necessity. He argues, first, that metaphysical modal claims are importantly similar to quotidian modals, the sort of modal claims we make every day (as when a sportswriter insists that NFL Commissioner Roger Goodell must resign). It is, for various reasons, implausible to think of modal terms like “possible,” “necessary,” “must,” and the like as ambiguous. The better account, Ichikawa thinks, treats metaphysical modals and quotidian modals as having similar form, but differing in the modal bases that the claims are relativized to. Ordinary claims might be relativized to the set of politically possible situations (“Ted Cruz could very well emerge as the Republican nominee in 2016”) or physically possible situations (“No human can outrun a cheetah”); in contrast, philosophers relativize to the set of metaphysically possible situations. Like Williamson, Ichikawa thinks it’s plausible that our ability to process quotidian modals conferred a fitness advantage that was selected for by evolution.33 Since the best treatment of modal terms treats metaphysical modals as having the same form as quotidian modals, there is no reason to think that the capacities that evolved to handle quotidian modals are not the same capacities that handle metaphysical modals. One question that the evolutionary story leaves open is why philosophers are so concerned with metaphysical possibility and necessity. Why not insist on a narrower range of possibilities, like physical possibilities? Or why not countenance a wider range of possibilities, like rational possibilities, that include metaphysical impossibilities? (In a Frege case, it can be a rational possibility that Hesperus is a planet while Phosphorus is not, even if that is a metaphysical impossibility.) Ichikawa finds no good argument to stop at metaphysical possibilities. Consider Williamson’s view, according to which a capacity for counterfactual conditionals plays roughly the role Ichikawa suggests for a capacity for quotidian modals. For Ichikawa, Williamson’s counterfactual approach leaves open an analogous set of questions. We do not always evaluate counterfactual conditionals with sensitivity to the distinctive set of metaphysical possibilities; the 33   Neither Williamson nor Ichikawa purports to offer independent evidence that evolution selected for our ability to imagine. They simply take it to be a very plausible hypothesis, and one gains abductive confirmation once the explanatory power of their views is made plain.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  29 context-sensitivity of counterfactuals corresponds to a variety of modal bases against which such a conditional may be evaluated. Ichikawa does not see any reason to be bothered by the fact that the evolutionary story leaves these questions open. Metaphysical modality is a sui generis subject matter, upon which we bring a perfectly general cognitive capacity to bear. We need not look to the capacity itself to explain why philosophers draw the line at metaphysical possibilities. Rather, we should explain the fact that philosophers bring the ­general capacity to bear on particular subject matters by its theoretical relevance, as attested by the success of philosophical theorizing that makes reference to metaphysical modality. Amy Kind also proposes a solution to the problem of imaginative use, though her solution does not rest on architectural or evolutionary considerations but rather on constraints that are subject to the will. To help isolate the relevant constraints, and thus the right kind of imagining for generating knowledge, Kind draws inspiration from science fiction, in particular, John Campbell’s 1932 science fiction story “The Last Evolution.” In the story, superintelligent and powerful machines serve as humanity’s caretakers. To aid in their task, the machines have developed a capacity for imagination that Campbell characterizes as ideal. Kind uses this notion of ideal imagination to help solve the puzzle of imaginative use. As Kind argues, what makes the machines’ imagination ideal is that it respects two constraints: the reality constraint and the change constraint. Respecting the reality ­constraint means imagining the target content in a maximally realistic way. When the machines imagine an impending alien attack, for example, they imagine a situation that is as close to the actual world as possible, modulo the insertion of the alien attack. They imagine their defenses as they actually are; and not, say, how they wish they were. The change constraint dictates that they imagine the situation evolving in a realistic way, so that it mirrors what would happen were the situation actual. This means, for instance, not imagining new elements that are not included in the target content. When the machines imagine the alien attack, they do not also imagine the discovery of a powerful new defense weapon; they imagine only what follows from the aliens attacking. In short, the ideal imaginer imagines by following out the consistent, logical consequences of what’s being imagined. Kind then turns back from science fiction to real life. If we can be ideal imaginers, if we can respect the twin reality and change constraints, then this would explain how imagining can generate knowledge of contingent matters of fact, and so the puzzle of imaginative use would be solved. But can we be ideal imaginers? Kind looks to the prodigious imaginations of individuals like the inventor Nikola Tesla and the scientist Temple Grandin to answer this question in the affirmative. Tesla and Grandin work out their inventions in their imaginations, apparently with astounding accuracy. Though ordinary people typically lack the imaginative powers of Tesla and Grandin, Kind is optimistic that we are in a good position to do ideal imagining on more limited projects. Like Williamson and Ichikawa, Kind thinks the more prosaic uses of

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

30  Amy Kind and Peter Kung imagination provide lots of examples. Expectant parents imagine whether an assembled crib will fit through the door; they imagine where it would work best in the nursery; they imagine the impact a child will have in their lives as they wonder whether they are ready to be parents; and so on. Although Kind admits that our imaginative capacities may not live up to those of the machines in Campbell’s story, the kinds of situations with which we’re normally presented require considerably less precision for successful imagining. Jennifer Church envisions a very different role for imagination in producing knowledge. In recent philosophy of mind, there has been increased interest in ways that the contents of perceptual experience may be richer than previously appreciated. We visually experience more than just colors and shapes.34 Church believes that there is also that kind of richness when we perceive people. In the spirit of Kant’s threefold synthesis that we sketched in Section 2.3, Church believes that imagination plays an essential role in our capacity to perceive people: We see them not just as three-dimensional spatial figures but also as anxious, or delighted, or as patiently waiting. Church’s argument is guided by the idea that a single, unified perceptual experience of other people can be “infused” with imagination. This infusion has both a phenomenological and an epistemic effect on the perceptual experience. Her claim is not merely that imagination accompanies or overlays the perceptual experience; rather, imagination and perception are synthesized together into a single whole, the perceptual experience. Crucial to Church’s view is the fact that perceptual experiences are self-justifying— that they contain with them the material for justifying that things are as the experience presents. This feature is explained by the fact that different aspects of a single perceptual experience can cohere and support each other: “My current experience . . . combines several visual components and several tactile components—each providing different perspectives on a single object . . . , each helping to establish the reality of that object.” For Church, we see the same coherence in perceptions of people’s mental states. Different aspects of a single experience—the cast of Tara’s eyes, the hint of resignation in her voice, the quaver at the corner of her smile—enable us to perceive a person’s mental state: Tara is disappointed with the gift that she’s just received but trying to make the best of it. Imagination enters the picture because imagination, not direct perception, contributes some of these different aspects. This can happen, Church argues, in at least three ways: 1. Past experiences can infuse the present experience, and phenomenologically affect how the person looks to me now. My past experiences of Tara getting a gift like this infuse my present experience of how I see her eyes, hear her voice, and watch her movements.

  See, for example, Siegel (2006, 2010).

34

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  31 2. Imagination can fill in parts of a scene that are outside the field of view. Her hands are occluded by the torn wrapping paper, but I imagine them clenched, rather than relaxed. 3. In Husserlian or Merleau-Pontyian fashion, imagination can make counterfactual predictions, or expectations, that infuse our current experience: “If we imagine that a child will be chastened by a particular challenge, we are likely to perceive her present look as having a different meaning than if we imagine that she will welcome the challenge.” These imaginings can clearly have an epistemic effect on perceptual experience, Church contends. She thinks that imaginatively infusing a perceptual experience with past and future surrounds increases our ability to detect inconsistencies. It’s harder to perceive Tara as frightened by strangers if imagination injects her gregariousness and expectations of her extroversion into the perceptual experience. Imaginative infusions can also serve as an impetus for discovery and motivate confirming or disconfirming action.

5.4  The Skeptics Ask any group of philosophers to write on knowledge, and invariably some of them will conclude that we cannot have it—or, at least, that it is more difficult to achieve than we might initially have thought. Three of our contributors argue that imagination typically doesn’t—or perhaps even cannot—provide knowledge of a domain that it is sometimes thought to provide. In Section 3.3 we briefly surveyed imagination’s role in mindreading. Both Heidi Maibom and Shannon Spaulding question the simulationist account of our knowledge of other minds. Heidi Maibom’s contribution focuses squarely on the mindreading issue. How do we explain or predict how other agents will react or have reacted, or what other agents will do or have done? The simulationist picture holds that we imaginatively project ourselves into their situation (or our future self ’s situation) and then examine our own reactions. Maibom calls this the empathetic imagination, which she understands broadly. Our reaction to the simulation is supposed to be a reliable guide to how other people would react in actuality. Since we need not rely on a systematic store of theoretical knowledge about human psychology to explain or predict the behavior of others, the simulationist view is thought to be representationally cheap. We take some of our existing psychological systems “offline” to run the simulation. Maibom marshals a wealth of fascinating results from empirical psychology to challenge the simulationist thought that we are good enough at imaginative projection to yield knowledge of how we ourselves would react, let alone others. Hence the title of Maibom’s contribution, “Knowing me, knowing you”: We don’t know ourselves via simulation, so a fortiori we cannot know others that way. Some of the results that drive Maibom’s pessimism are well-known, such as the bystander effect (people are less

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

32  Amy Kind and Peter Kung ­ illing to help someone in need in the presence of idle bystanders than they predict w they will be) and the Milgram effect (people deliver what they believe to be far more damaging shocks than they predict they will). But others are considerably less familiar. For example, we value an object that we own much more than we predict; we are less likely to cheat on math exams than we predict; we have difficulty predicting how we will respond to “visceral experiences” like hunger, pain, and thirst; we overestimate how long we will experience guilt or regret, but underestimate the effects of embarrassment; we underestimate how reluctant we will be to confront a sexual harasser. And so on. Maibom’s argument that we do a poor job of projecting ourselves into counterfactual situations can be seen as relying on the claim that imagination is subject to the wrong kind of constraints. We typically do not imagine at the requisite level of detail; we “focalize,” leaving out background conditions, other events occurring, other people acting in certain ways. We tend to imagine situations in “naked, decontextualized, and prototypical ways” that do not lend themselves to accurate simulations. According to Maibom, the empirical evidence suggests that we evaluate counterfactual situations based not on how we would react but rather on how we think it would make sense to react. What feelings would be “normal”? What beliefs ought one to have in such a situation? How ought one to react? In Maibom’s view, we are deciding how to react and what to do in that situation, calling upon information-rich semantic, declarative, and stereotypical knowledge of situations, rather than reaching a decision through imaginative re-enactment. Shannon Spaulding also takes on the simulationist account of mindreading, though she sets her skeptical argument in a broader context. Spaulding is interested in the extent to which imagination can provide new knowledge of contingent matters of fact. She begins by distinguishing deliberate imagination, which covers the voluntarily initiated imaginings involved in philosophical and scientific thought experiments, from spontaneous imagination, like dreams and daydreams, which are more or less independent of our volition. Spaulding argues that neither type of imagination is sufficient to generate new knowledge of contingent facts. Armed with this distinction, Spaulding turns to simulationist accounts of mind­ reading. Spaulding notes that simulationists posit two kinds of mindreading, corresponding to the two types of imagining. High-level simulations are familiar, deliberate acts of putting oneself into another’s shoes. Suppose you are trying to figure out why Joseph crumpled the letter from Shayan after reading it. To figure this out, you imagine that you yourself read the letter while holding various mental state combinations: Joseph is already irritated with Shayan and the letter pushes him to frustration; Joseph is worried about Shayan and crumples the letter to relieve tension; and so on. You then compare these various imaginative scenarios, and choose the most plausible. These high-level simulations are typically described as “involv[ing] quasi-sensory information” and as being “consciously accessible, voluntary, subject to the agent’s control, and target[ing] . . . mental states of a relatively complex nature, such as propositional

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  33 attitudes.” Low-level simulations are relatively automatic and often below the level of consciousness, though the end-product is consciously accessible, and they target less complex mental states, like sensations and basic emotions. Low-level simulations are acts of spontaneous imagination, and it’s widely accepted that mirror neurons are the mechanism for these low-level simulations. The skeptical challenge for both high- and low-level simulation is the “threat of collapse.” For a simulation to provide the kind of knowledge that it is supposed to, the imaginative projection must be supplemented by exactly the sort of information-rich theoretical knowledge that simulationists were trying to avoid. Spaulding points out two facts that seem to lead to collapse: (1) Any observed behavior is compatible with indefinitely many reality-based mental state combinations, and (2) imagination itself provides no way to decide between these mental state combinations. Thus, though imagination can generate possible mental state combinations that a target could have, it cannot tell us which of the possibilities is correct; it is not itself in any way constrained to select the correct or most likely one. To figure out why your brother crumpled the letter, you need more than imagination. You need loads of additional theoretical knowledge about folk psychology and background conditions, plus the capacity for inference to the best explanation. Spaulding thinks this skeptical problem generalizes. Imagination produces possibilities, but imagination itself does not tell us which of these possibilities is correct, or even likely. Her pessimistic view contrasts with Williamson’s more optimistic outlook. Williamson (see this Introduction, Section 5.3) offers evolutionary reasons for holding that imagination tends to select realistic possibilities. Spaulding argues that the selection of realistic possibilities should not be treated as part of imagination itself because the selection calls on cognitive capacities that operate independently of imagination and have their own functional roles. We need to supplement our imagining with background information, theoretical knowledge of the subject under examination, and general capacities for reasoning. This supplementary knowledge not only influences which possibilities we imagine in the first place (in trying to explain your brother’s behavior, I never bother to simulate that I believe evil aliens are about to bust through my door and that they can only be subdued with crumpled balls of paper) but also helps us evaluate which possibility is correct. Imagination, by itself, does not give us knowledge. While the skeptical challenges raised by Maibom and Spaulding focus on mind­ reading, Peter Kung raises a skeptical challenge for thought experiments in ethics. As he notes, many compelling thought experiments have played a prominent role in the ethics literature: the transplant case, deciding on the best policies from the original position, being kidnapped and attached to a famous violinist. According to Kung, a wide range of thought experiments in ethics have a distinctive feature: they feature forced choices with fixed outcomes. In a typical ethics thought experiment, an agent A is faced with choice C1 and C2. If A picks C1, then O1 will occur. If A picks C2, then O2 will occur. The thought experiment forces the choice between C1 and C2: they are the only

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

34  Amy Kind and Peter Kung relevant options. To suggest another option, C3, is to violate the rules of the game. Likewise, O1 and O2 are the only possible outcomes. It is not merely probable that one will occur; it is definite. Suggesting that something other than O1 or O2 will occur is, again, not to play the game. Starting with the plausible assumption that good thought experiments must be meta­ physically possible, Kung explores in his chapter whether thought experiments with forced choices and fixed outcomes are metaphysically possible. In his view, attending closely to features of imagination suggests that pessimism is warranted. For Kung, the best account of our knowledge of metaphysical possibility is via imagistic imagination. This type of imagination, on his view, has two kinds of content: content delivered by the imagery, and non-pictorial content. When you imagine a cat hidden behind a bookcase, the bookcase is pictured—your mental imagery includes a bookcase. The cat, on the other hand, is imagined but not pictured; there is no mental imagery of a cat (otherwise it wouldn’t be a hidden cat). Kung argues that to give us knowledge of metaphysical possibility, the content delivered by the imagery must do some real work. Non-pictorial content, he worries, is too unconstrained. We can imagine just about anything we want via non-pictorial content, including impossibilities. As he shows, however, the forced choices and fixed outcomes of many ethics thought experiments have to be imagined via assigned content. Thus, imagination gives us no reason to think that scenarios featuring forced choices with fixed outcomes are metaphysically possible. This has the implication that some putatively devastating counterexamples in ethics prove to be less devastating than widely thought. Kung concludes that any ethical view that counts outcomes as ethically relevant will have to take seriously moral risk, a result he thinks accords with common sense. In everyday ethical reasoning, choices are not forced and outcomes are not fixed. We take into account the chancy nature of our decisions: choosing C1 will likely lead to O1, but there is a chance it will lead to O1.1 or O1.2 or . . . . On his view, methodological considerations about successful thought experiments require that we consider moral risk in our foundational theorizing. We cannot first use counterexamples featuring forced choices with fixed outcomes to eliminate theories (like consequentialism) and then, with the foundational theory issues more settled, turn to risk. Our ethical theories have to factor in risk right from the start.

Acknowledgements We are grateful to Peter Thielke and Samantha Matherne for extensive discussion, particularly with respect to historical treatments of imagination. Their help has been invaluable. Thanks also to Frank Menetrez for discussion and feedback on an earlier draft. Finally, we are particularly indebted to all of the participants at the Knowledge Through Imagination conference. The discussion there played an essential role in shaping our thinking about these issues.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  35

References Allison, Henry E. (2001). Kant’s Theory of Taste: A Reading of the Critique of Aesthetic Judgment. Cambridge: Cambridge University Press. Bealer, George (2002). Modal epistemology and the rationalist renaissance. In Tamar Szabό Gendler and John Hawthorne (eds), Conceivability and Possibility. Oxford: Clarendon Press, pp. 71–125. Brann, Eva T. H. (1992). The World of the Imagination: Sum and Substance. Lanham, MD: Rowman and Littlefield. Chalmers, David J. (2002). Does conceivability entail possibility? In Tamar Szabό Gendler and  John Hawthorne (eds), Conceivability and Possibility. Oxford: Clarendon Press, pp. 145–200. Church, Jennifer (2013). Possibilities of Perception. Oxford: Oxford University Press. Cooper, Rachel (2005). Thought experiments. Metaphilosophy 36(3): 328–47. Currie, Gregory (1990). The Nature of Fiction. Cambridge: Cambridge University Press. Currie, Gregory, and Ian Ravenscroft (2002). Recreative Minds: Imagination in Philosophy and Psychology. Oxford: Oxford University Press. Descartes, René (1985). Meditations on First Philosophy (ed. and trans. John Cottingham). Cambridge: Cambridge University Press. [Original work published 1641.] Garrett, Don (1997). Cognition and Commitment in Hume’s Philosophy. New York: Oxford University Press. Geirsson, H. (2005). Conceivability and defeasible modal justification. Philosophical Studies 122(3): 279–304. Gendler, Tamar Szabό (2000). The puzzle of imaginative resistance. Journal of Philosophy 97(2): 55–81. Gendler, Tamar Szabό, and John Hawthorne (2002). Introduction: Conceivability and possibility. In Tamar Szabό Gendler and John Hawthorne (eds), Conceivability and Possibility. Oxford: Clarendon Press, pp. 1–70. Goldman, Alvin I. (2006). Simulating Minds: The Philosophy, Psychology, and Neuroscience of Mindreading. New York: Oxford University Press. Gregory, Dominic (2004). Imagining possibilities. Philosophy and Phenomenological Research 69(2): 327–48. Hart, W. D. (1988). The Engines of the Soul. Cambridge: Cambridge University Press. Hill, Christopher S. (1997). Imaginability, conceivability, possibility and the mind-body problem. Philosophical Studies 87(1): 61–85. Hume, David (1902). Enquiries Concerning Human Understanding and Concerning the Principles of Morals (ed. L. A. Selby-Bigge). 2nd edn, Oxford: Clarendon Press. [Original work published 1748.] Hume, David (1978). A Treatise of Human Nature (ed. L. A. Selby-Bigge and P. H. Nidditch, 2nd edn). Oxford: Clarendon Press. [Original work published 1739–40.] Jackson, Frank (1982). Epiphenomenal qualia. Philosophical Quarterly 32(127): 127–36. Kant, Immanuel (1998). Critique of Pure Reason (ed. and trans. Paul Guyer and Allen W. Wood). Cambridge: Cambridge University Press. [Original work published 1787.] Kant, Immanuel (2000). Critique of the Power of Judgment (ed. Paul Guyer; trans. Paul Guyer and Eric Matthews). Cambridge: Cambridge University Press. [Original work published 1790.]

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

36  Amy Kind and Peter Kung Kant, Immanuel (2008). Anthropology from a pragmatic point of view (trans. Robert B. Louden). In Günter Zöller and Robert B. Louden (eds), Anthropology, History, and Education. Cambridge: Cambridge University Press, pp. 227–429. [Original work published 1798.] Kind, Amy (2001). Putting the image back in imagination. Philosophy and Phenomenological Research 62(1): 85–109. Kind, Amy (2013). The heterogeneity of the imagination. Erkenntnis 78(1): 141–59. Kripke, Saul A. (1980). Naming and Necessity. Cambridge, MA: Harvard University Press. Kung, Peter (2010). Imagining as a guide to possibility. Philosophy and Phenomenological Research 81(3): 620–63. Martin, M. G. F. (2002). The transparency of experience. Mind and Language 17(4): 376–425. Morton, Adam (2006). Imagination and misimagination. In Shaun Nichols (ed.), The Architecture of the Imagination: New Essays on Pretence, Possibility, and Fiction. Oxford: Oxford University Press, pp. 57–72. Nanay, Bence (2010). Perception and imagination: Amodal perception as mental imagery. Philosophical Studies 150(2): 239–54. Noë, Alva (2004). Action in Perception. Cambridge, MA: MIT Press. Parfit, Derek (1984). Reasons and Persons. Oxford: Clarendon Press. Peacocke, Christopher (1985). Imagination, experience, and possibility: A Berkeleian view defended. In John Foster and Howard Robinson (eds), Essays on Berkeley: A Tercentennial Celebration. Oxford: Clarendon Press, pp. 19–35. Searle, John R. (1980). Minds, brains and programs. Behavioral and Brain Sciences 3(3): 417–24. Sepper, Dennis L. (1996). Descartes’s Imagination: Proportion, Images, and the Activity of Thinking. Berkeley, CA: University of California Press. Shoemaker, Sydney (1993). The first-person perspective. Proceedings and Addresses of the American Philosophical Association 68(2): 7–22. Siegel, Susanna (2006). Which properties are represented in perception? In Tamar Szabό Gendler and John Hawthorne (eds), Perceptual Experience. Oxford: Oxford University Press, pp. 481–503. Siegel, Susanna (2010). The Contents of Visual Experience. New York: Oxford University Press. Sorensen, Roy A. (1992). Thought Experiments. New York: Oxford University Press. Spinoza, Baruch (1992). The Ethics: Treatise on the Emendation of the Intellect; Selected Letters (ed. Seymour Feldman; trans. Samuel Shirley). Indianapolis, IN: Hackett. [Original work published 1677.] Stalnaker, Robert C. (1984). Inquiry. Cambridge, MA: MIT Press. Stalnaker, Robert (2002). Common ground. Linguistics and Philosophy 25(5–6): 701–21. Strawson, P. F. (1970). Imagination and perception. In Lawrence Foster and J. W. Swanson (eds), Experience and Theory. Amherst, MA: University of Massachusetts Press, pp. 31–54. Streminger, Gerhard (1980). Hume’s theory of imagination. Hume Studies 6(2): 91–118. Thomson, Judith Jarvis (1971). A defense of abortion. Philosophy and Public Affairs 1(1): 47–66. Tidman, Paul (1994). Conceivability as a test for possibility. American Philosophical Quarterly 31(4): 297–309. Walsh, Adrian (2011). A moderate defence of the use of thought experiments in applied ethics. Ethical Theory and Moral Practice 14(4): 467–81.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Introduction  37 Walton, Kendall L. (1990). Mimesis as Make-Believe: On the Foundations of the Representational Arts. Cambridge, MA: Harvard University Press. Walton, Kendall L. (1993). Metaphor and prop oriented make-believe. European Journal of Philosophy 1(1): 39–57. Weatherson, Brian (2004). Morality, fiction, and possibility. Philosophers Imprint 4(3): 1–27. White, Alan R. (1990). The Language of Imagination. Oxford: Blackwell. Wilson, Margaret Dauler (1978). Descartes. London: Routledge. Yablo, Stephen (1993). Is conceivability a guide to possibility? Philosophy and Phenomenological Research 53(1): 1–42.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Pa rt On e

Taxonomical and Architectural Approaches

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

1 On the Epistemic Value of Imagining, Supposing, and Conceiving Magdalena Balcerak Jackson

1. Introduction Sit down in your armchair, relax, and simply let me guide you through some mental exercises. First, let me ask you to imagine this. You wake up in the morning and find yourself back to back in bed with an unconscious violinist. A famous unconscious violinist. He has been found to have a fatal kidney ailment, and the Society of Music Lovers has canvassed all the available medical records and found that you alone have the right blood type to help. They have therefore kidnapped you, and last night the violinist’s circulatory system was plugged into yours, so that your kidneys can be used to extract poisons from his blood as well as your own. The director of the hospital now tells you, “Look, we’re sorry the Society of Music Lovers did this to you—we would never have permitted it if we had known. But still, they did it, and the violinist is now plugged into you. To unplug you would be to kill him. But never mind, it’s only for nine months. By then he will have recovered from his ailment, and can safely be unplugged from you.”1

Second, think about prime numbers—that is, those numbers that are divisible only by one and themselves—and now suppose that there are only finitely many of them.2 And finally, third, let me ask you to try to conceive of a zombie in the following sense: So let us consider my zombie twin. This creature is molecule for molecule identical to me, and identical in all the low-level properties postulated by a complete physics but he lacks conscious   Thomson (1971).   This is the beginning of several reductio ad absurdum arguments of this claim, the first of which can be found in Euclid’s Elements (book XI). If you feel more comfortable with a non-mathematical example, here is one: Suppose that heavier bodies fall faster than lighter ones, as Galileo did in order to disprove the Aristotelian theory by contradiction. 1 2

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

42  Magdalena Balcerak Jackson experience entirely. . . . To fix ideas we can assume that right now I am gazing out the window, experiencing some nice green sensations from seeing the trees outside, having pleasant taste experiences through munching on a chocolate bar, and feeling a dull aching sensation in my right shoulder. What is going on in my zombie twin? . . . He will certainly be identical to me functionally: he will be processing the same kind of information, reacting in a similar way to inputs, with his internal configurations being modified appropriately and with indistinguishable behavior resulting. . . . It is just that none of this functioning will be accompanied by any conscious experience. There will be no phenomenal feel. There is nothing it is like to be a zombie.3

What did you do in each of the three cases? Was it the same kind of mental activity? Was the difference between putting the three cases in front of your mind’s eye analogous to the difference between visually experiencing three different scenes spread out before your non-metaphorical eyes, or was it rather analogous to the difference between using three different sense modalities? Or is neither of these analogies appropriate for capturing the distinctions you notice? These questions are interesting and important for the following reason: Philosophers frequently invoke our ability to imagine, conceive, or suppose various things in order to explain how we achieve our cognitive goals when we make decisions about what to do, when we perform thought experiments, and when we reason about how things would or could be. But there seems to be no underlying agreement about whether the terms “imagine,” “conceive,” and “suppose” can or should be used interchangeably or whether they correspond to three distinct mental activities that play distinct cognitive roles. If “imagining,” “supposing,” and “conceiving” are merely three different ways to refer to our ability to think about scenarios and objects that may or may not exist, then which term we use in the three sets of instructions above—or in setting up any other thought experiment for that matter—does not have any deep significance, except perhaps to supply a certain emphasis or invoke certain associations. However, I contend that the terminological choice does matter, because the three cases are paradigmatic examples of three distinct cognitive activities. As we will see in what follows, in order to secure the intended epistemic payoff in the first case we should imagine being hooked up to the famous violinist rather than suppose that we are. In the second case we should suppose that there are infinitely many prime numbers rather than attempt to imagine it being the case. And we do not learn much of interest from the zombie thought experiment unless we try to conceive of the twin zombie rather than imagine him or suppose things about him. The goal of this chapter is to develop a systematic tripartite distinction between imagining, supposing, and conceiving. The distinction serves an epistemic purpose: The description of the distinct natures of imagining, supposing, and conceiving developed

  Chalmers (1996, pp. 92–3).

3

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagining, Supposing, and Conceiving  43 here shows how they are able to play three distinct epistemic roles, each of which is exploited within philosophical as well as non-philosophical enquiry.4 The plan is as follows. In the second section I begin by laying out a naive picture of hypothetical attitudes as a unified kind with unified epistemic properties. This view is arguably held by some historical figures, such as Hume, as well as by some contemporary philosophers, such as Ichikawa and Jarvis (2013). However, we will see that it is too naive a view to be plausible. Its main value is to serve as a heuristically useful starting point, a foil against which to develop a more nuanced and more plausible picture of imagining, supposing, and conceiving and their respective epistemic roles. In the third section, I compare and contrast imagining and supposing. The goal is to elucidate a thick conception of imagining as a kind of state that is distinct from supposing and that serves a distinctive epistemic purpose. In the fourth section, I propose that we understand supposition as acceptance; this model accounts for the fact that supposition can be used in reductio reasoning and certain kinds of conditional reasoning. And in the fifth section, I sketch an account of conceiving that distinguishes it from both imagining and supposing. This account calls the epistemic significance of conceiving into question: If the account is correct, we should be more careful about relying on conceivings than on imaginings and supposings within our philosophical practice.

2.  Hypothetical Attitudes: Common Nature, Voluntary Control, and Epistemic Innocence It is plausible to think of imagining, conceiving, and supposing as complex cognitive capacities. They are things that we as cognitive agents do for specific purposes, including epistemic ones. Correspondingly, the targets of this chapter are certain psycho­ logically real capacities that we actually have and use. But we also aim to identify epistemically good ways of putting these capacities to use for the purposes of systematic enquiry, especially in philosophy. So, our interest is in what imagining, supposing, and conceiving are such that they are, or could be, epistemically valuable practices. The goal is not merely to describe whatever we naturally classify using the ordinary language terms “imagine,” “suppose,” and “conceive,”5 but to develop a systematic picture that allows us to make sense of how and why they are useful in the cognitive projects in which they are utilized. Since our project is not merely descriptive but also normative, 4   The existing taxonomies of imagining-like mental states do not provide a satisfactory account of the differences between imagining, supposing, and conceiving from an epistemic point of view. Van Leeuwen (2013, 2014) gives detailed descriptions of types of states that we commonly refer to as imaginings, but he  does not attempt to capture distinctions between imaginings and close cousins like supposings and conceivings. Currie and Ravenscroft (2002) discuss the imagination/supposition distinction, but their discussion does not do justice to the different epistemic roles they play, and Gendler’s (2000) remarks about the same distinction are useful, but preliminary at best. So while the project of this chapter is related to some earlier attempts to give a more nuanced view of imagination, it goes significantly beyond these attempts. 5   Kind (2013) convincingly demonstrates that what falls under our concept “imagining” as used in ordinary language is too heterogeneous to form one natural mental kind.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

44  Magdalena Balcerak Jackson it might force us to revisit or regiment philosophers’ ways of using the corresponding terms from ordinary language. However, the regimentation to be developed here ­captures core cases of what we would naturally classify as imagining, supposing, and conceiving, and has the virtues of explanatory clarity and force. There are obvious commonalities between imagining, supposing, and conceiving. They are all ways of thinking that are most often directed towards merely hypothetical objects and scenarios, rather than towards scenarios that actually obtain—although they can be directed towards the latter as well.6 In each of the three cases above you are asked to consider a situation that you know not to obtain, or at least that you have no good reason to believe actually does obtain. In the first case you represent yourself as waking up hooked up to a famous unconscious violinist in need of life support. In the second case, you represent it being the case that there are only finitely many primes. And in the third case, you represent your fully functional but phenomenally “dead” zombie twin. Or at least, in each of these cases you attempt to represent the non-actual situation described. This commonality between imagining, supposing, and conceiving encourages a natural pre-theoretical assumption that they are all instances of the same basic cognitive capacity. Let us call this assumption the Common Nature Thesis.7 It is a natural starting point for our enquiry. The Common Nature Thesis is naturally paired with a second thesis that ascribes a common epistemic status to our hypothetical attitudes. According to the Epistemic Innocence Thesis, imaginings, conceivings, and supposings do not provide justification for belief. Being in one of these states is never by itself sufficient to make it epistemically appropriate to believe anything. It is important to emphasize that the Epistemic Innocence Thesis is not meant to entail that imagination, conceiving, and supposition have no role at all to play in the acquisition of knowledge. For example, scientists often stress that imagination is a precondition of their success. However, on the orthodox view, the role of imagination is restricted to the context of discovery, rather than the context of justification. Imagining something can help lead to the formation of a hypothesis, but it cannot provide evidential support for a hypothesis. The assumption that imagination, specifically, is epistemically innocent might look attractive if we focus on the creative use of our ability to imagine non-actual objects and scenarios in the context of creating works of art, or simply for the sake of having intellectual fun. But the main reason for the widespread philosophical skepticism about the epistemic value of imagination has to do with another apparent commonality between imagining, supposing, and conceiving: the fact that they are under our voluntary control. Setting aside cases of spontaneous daydreaming and the like, what we imagine, what we suppose, and what we conceive are under our voluntary control in two respects: We form these attitudes at will, and we decide what it is that we 6   Thinking is to be understood here very liberally, as any form of conscious consideration of a specific mental content. 7   See Ichikawa and Jarvis (2013, ch. 10) for an implicit endorsement of the Common Nature Thesis.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagining, Supposing, and Conceiving  45 imagine, suppose, or conceive of. To put it in a slogan: Imagining, supposing, and conceiving are up to us. 8 Let us call this the Voluntary Control Thesis. Elsewhere, I examine in detail how the voluntary nature of imagination can seem to be in tension with the possibility of imagination providing justification. But here is the idea in nuce: If imaginings are under our voluntary control, then when we imagine and what we imagine are not constrained by the way things actually are, and this seems to show that our imaginings cannot teach us anything about how things actually are. And more dramatically, if there are no limits to what we can imagine, then the fact that one imagines P does not rule out or speak against any hypothesis about how things are.9 Since supposing and conceiving also seem to be under our voluntary control, analogous reasoning applies to them as well. Just as the epistemic value of perception—the paradigmatic basic source of justification—springs from the specific nature of perceptual experiences and their causal relations to the external environment, the epistemic innocence of hypothetical attitudes is alleged to spring from their shared voluntary nature. But are the Common Nature Thesis, the Voluntary Control Thesis, and the Epistemic Innocence Thesis in fact true? In what follows I present reasons to think that the Common Nature Thesis is false with respect to imagining and supposing. More import­ antly, I show that differences between imagining and supposing include a significant difference in the extent to which they are under our voluntary control. This renders the Voluntary Control Thesis true in an unqualified sense at best only for supposings, and not for imaginings. And I show how what all this reveals about the nature of imagination forces us to revise the Epistemic Innocence Thesis. After discussing imagining and supposing, I turn in the penultimate section to the question of how to fit conceiving into the picture that has developed.

3.  Imagining as Experiential Perspective-Taking Return to the case of your imagining waking up hooked up to a famous violinist and being told that you need to accept your fate in order to save his life. When you imagined the scenario described, you probably put yourself in the shoes of the person waking up in the hospital: You imagined seeing the violinist laying next to you, hearing the medical machinery in the background, perhaps sensing the IV-needle in your arm, and feeling surprise and anxiety. Intuitively, unless you vividly represent the scenario from the perspective of the experiencing subject, you do not really follow my invitation to imagine it, in the core sense of the term. Had I asked you instead to suppose that this scenario obtained, would you have done anything different? In this case, it would have 8   Skepticism along these lines can be found in Sartre (1972), Wittgenstein (1981), and White (1990), to name just a few. 9   For a detailed discussion of what I call the Up To Us Challenge, and for an argument in favor of a justificatory role for imagination, see Balcerak Jackson (in press). Similar skeptical worries are also addressed by Langland-Hassan in his contribution in this volume (Chapter 2) and by the editors in their Introduction.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

46  Magdalena Balcerak Jackson been perfectly in accordance with my request for you simply to take the situation as obtaining, without representing it as being experienced from the first-person perspective, and in fact without representing it in any particular way at all. While imagination involves a phenomenology familiar to us from mental states such as perceptual experiences and emotions, supposition intuitively does not require any such phenomenology (although supposing something might initiate a subsequent act of imagining).10 Some philosophers have identified mental imagery as a crucial ingredient of imagination in the core sense.11 If what is meant by mental imagery is perceptual phenomenology,12 then I think it is not in fact a necessary ingredient. We can imagine being angry, for example, without experiencing any perceptual phenomenology at all. But when we imagine being angry, we certainly do form a mental state with a distinctive phenomenology, albeit an emotive rather than perceptual one. Still, if we understand mental imagery broadly enough to include the phenomenal character of non-perceptual kinds like this, then I think it is correct that it is a necessary ingredient of imagining in the core sense. But mental imagery (even in this broad sense) is not a necessary ingredient of supposing. This provides us with one good reason to distinguish imagining from supposing. Looking at the ways we report or ascribe imaginings and supposings in ordinary language further illuminates the difference. In some cases—in particular, those cases involving propositional reports—using “imagine” or “suppose” might seem merely to be two ways of describing the same mental act: (1) Alice imagines that she has an IV-needle in her arm. (2) Alice supposes that she has an IV-needle in her arm. There is a way of understanding (1) so that substituting “imagines” with “supposes,” as in (2), results in a sentence that is intuitively roughly equivalent. However, many uses of “imagine” are in objectual and eventive rather than propositional reports, as in (3), and in these cases “supposes” cannot be substituted for “imagines”: (3) Alice imagines (having) an IV-needle in her arm. (4)  * Alice supposes having an IV-needle in her arm. Substituting “supposes” for “imagines” in (3) results in the ungrammatical (4). The difference between (1) and (2) on one hand and (3) on the other hand is that (1) and (2) can both be naturally understood in such a way that there need not be any specific way for Alice to represent that there is an IV-needle in her arm. But the truth of (3) intuitively requires that Alice put herself in the position of the relevant person in her imagination. In this sense, the truth-conditions of objectual and eventive imagining ascriptions are more demanding than the truth-conditions of supposing ascriptions. This is another reflection of the observation above that objectual and eventive imaginings involve  Amy Kind provides detailed arguments for a claim along these lines in Kind (2001).   See Kind (2001).    12 As suggested by Van Leeuwen (2013).

10 11

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagining, Supposing, and Conceiving  47 capacities related to perspective-taking and phenomenal experience, and hence that they make use of cognitive resources that go beyond those needed merely to entertain a certain mental content. The point is not that a difference in kind between imagining and supposing follows from grammatical differences between “imagine” and “suppose.” Rather, the point is that as competent users of English we all understand that saying that someone supposes that she has an IV-needle in her arm is one thing, while saying that she imagines having an IV-needle in her arm is something much different. Ascriptions of objectual and eventive imagining capture a distinctive phenomenon, and it is this phenomenon that I suggest we identify with imagination in the core sense. There are plenty of other reasons to believe that core instances of imagining and core instances of supposing play very different functional roles. Here are five, some noted before but some new. Taken together, they create a powerful case for distinguishing imagining and supposing as two distinct mental kinds. First, there is a realistic risk of failure when engaging in an imaginative exercise, but that risk is not present—or is at least significantly reduced—when merely supposing things to be thus and so.13 If I have never given birth, I might have difficulties imagining giving birth. I might simply fail to adequately represent the phenomenal feel of this uniquely painful and overwhelming situation. But even in these circumstances I will have no trouble supposing that I am giving birth. To put it in a slightly oversimplified way: Deciding to suppose that P guarantees success in doing so, or is at least significantly less at risk of failure than deciding to imagine P. Second, as Meinong pointed out, it seems unproblematic to suppose that some negative state of affairs, such as an absence, obtains, but it is difficult to imagine such a thing.14 We know very well what to do when we are asked to suppose that Billy does not have measles, but it requires some ingenuity to follow the instruction to imagine Billy not having measles. Third, the phenomenon of imaginative resistance has no equivalent counterpart in the realm of supposition. Imaginative resistance occurs when a subject is asked to imagine a particular situation, but is either unable or unwilling to do so.15 Most convincing examples of imaginative resistance involve requests to imagine situations where morally highly deviant behaviors and attitudes are endorsed.16 Even though we have no trouble supposing that the following scenario obtains, we have difficulties imagining it: Alice took her newborn baby, put it into a cotton bag, closed the bag tightly with a rope, and threw it into the lake. And this was good. After all, the baby was a girl.   See also White (1990, p. 261).   Farennikova (2012) has argued that we can perceptually experience absences. If her arguments are sound, this would lend some support to the thesis that we can also imagine absences. At least, this would be the natural conclusion if we agree that imagination is by its very nature simulation. 15   See Gendler (2000). 16   Good non-moral examples can be found in Weatherson (2004). 13 14

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

48  Magdalena Balcerak Jackson Imaginative resistance has also been postulated with respect to conceptually deviant scenarios. Andy Egan and Tyler Doggett write: “Though we think it is hard to imagine that Robert Stalnaker is the smallest prime number, we have no trouble supposing that he is.”17 Fourth, supposing always happens for a purpose, while we often imagine for its own sake. Because imagining involves a certain—often vivid and immediate—phenomenology, and because this phenomenology can give us an approximation of what it would feel like to be in new, interesting, and desired circumstances, it holds an appeal of its own. Mere supposition does not exhibit this immediate phenomenal self-involvement. There is little intrinsic appeal in merely letting different propositions pass through one’s mind in thinking. Thus, it is unsurprising that we need a reason to do so. The para­ digmatic reason, and perhaps even one that is constitutive, is nicely described by Alan White: “To say ‘suppose that P’ invites or introduces a statement of the consequences or implications that P.”18 Fifth, imagining has the potential to motivate action, while supposition typically only motivates deliberation. When one imagines being a doctor in a game of make-­ believe, one experiences genuine motivation for certain actions towards the other players. When another participant in the game is lying on the ground, holding her belly, and wincing, one does not merely entertain the thought that one approach her with one’s (toy) doctor’s kit for an examination, one feels a genuine pull towards acting in that way. Such motivation is absent during an exercise of merely supposing that one is a doctor.19 These differences between core instances of imagining and core instances of supposing give us strong reason to reject the Common Nature Thesis. More importantly, this short functional role analysis suggests that imagining is more demanding than supposing: Imagining requires us not merely to represent a non-actual scenario in some way or other, but to represent it by taking a specific experiential perspective. This characteristic experiential perspectivity is hinted at in many philosophers’ statements about the imagination. “To imagine is always at least to imagine, from the inside, being in some conscious mental state,” writes Christopher Peacocke. 20 Tamar Gendler observes that imaginative engagement is “in a difficult to pin-down way self-involving.”21 And Zeno Vendler describes it as follows: the necessary condition for imagining performing certain actions, or being in certain conditions, is the existence of an experiential content attached to these things. It must make sense, in other words, to ask the question: what would it be like doing such a thing, or being in such a state.22   Doggett and Egan (2007, p. 1).   White (1990). Obviously, this is not an observation about the psychological underpinnings of supposition, but about the practices that constitute the relevant cognitive capacity. 19   For more discussion on participation resulting from the activity of imagining being in a certain position, see Walton (1990, chs 6, 7). 20   Peacocke (1985, p. 22). 21   Gendler (2006, p. 150). A more careful analysis of the perspectival nature of imagination is the project of another paper. 22   Vendler (1979, p. 166). 17 18

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagining, Supposing, and Conceiving  49 The view that recognizes and best captures this feature of imagination is the recreativist or simulationist view. On this view, imagination is the capacity to put oneself in the perspective of another actual or merely possible subject, by recreating or simulating the mental state that the subject has or would have. When I imagine seeing an apple, I don’t imagine a mental process of seeing, but I simulate or recreate undergoing such a process. I put myself in the position of somebody who actually perceptually experiences an apple. In their book-length treatment of the imagination as a “recreative” capacity, Gregory Currie and Ian Ravenscroft describe their view as follows: So what is the recreative imagination? Here is our central hypothesis. Imaginative projection involves the capacity to have, and in good measure to control the having of, states that are not perceptions or beliefs or decisions or experiences of movements of one’s body, but which are in various ways like those states—like them in ways that enable the states possessed through imagination to mimic and, relative to certain purposes, to substitute for perceptions, beliefs, decisions, and experiences of movements.23

Similarly, Alvin Goldman identifies a simulationist capacity that he calls “Enactmentimagination” or E-imagination for short: Enactment-imagination is a matter of creating or trying to create in one’s own mind a selected mental state, or at least a rough facsimile of such a state, through the faculty of the imagination. Prime examples of E-imagination include sensory forms of imagination, where one creates, through imagination, perception-like states.24

What is common to recreativist or simulationist views is that they postulate an asymmetrical mimicry relationship between imaginings, on the one hand, and specific mental states of other types, such as perceptual experiences, emotions, or bodily sensations, on the other hand. This basic picture is compatible with different views about precisely which sorts of conscious mental states can be simulated in imagination. Many proponents of the view are unwilling to include conative attitudes such as desires, but are happy to include intellectual attitudes such as beliefs. Currie and Ravenscroft (2002) understand supposing as simulation of belief and as a kind of imagination. In the next section we will see that it is a mistake to treat supposings as simulations of belief, and in the section after we will see reasons for treating the simulation of belief as a very different kind of mental process than the simulation of perceptual experiences and emotions in imagination. But none of this impugns the general recreativist or simulationist picture of imaginings. We have a natural understanding of experiences as mental states with the following property: There is something it is like to be a subject of those states, or alternatively, things appear a certain way to us in those states.25 Let us call this kind of perspectivity   Currie and Ravenscroft (2002, p. 11).   Goldman (2006, p. 42). 25   One way of thinking about the what-it-is-like of experience is to think of it as a unity of qualitative character and for-me-ness: see Zahavi and Kriegel (2015). 23 24

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

50  Magdalena Balcerak Jackson experiential perspectivity. In virtue of us having first-person acquaintance of the experiential aspects of our mental life, we can, as imaginers, create mental states that reflect the experiential perspective of those corresponding mental states that exhibit such an experiential perspective. When we (perceptually) imagine a red flower, we create a mental state that reflects the experiential perspective of perceptually experiencing a red flower. When we (emotionally) imagine being angry, we create a mental state that reflects the experiential perspective of feeling angry. So, the general picture is the following: Imaginings, by their very nature, take up—or at least attempt to take up—the experiential perspective, or in other terms the phenomenal character and the phenomenal representational content of corresponding experiences. Although I think the recreativist view is phenomenologically compelling, my aim in this chapter is not to provide a defense for it. But it is worth pointing out that the core recreativist idea is often implicitly accepted, even by those who do not explicitly endorse a recreativist view. For example, although neither Timothy Williamson nor Stephen Yablo identify their views as recreativist, they both echo the basic recreativist conception when they refer to the employment of our imaginative capacity with respect to visual properties as an “offline”-use of our perceptual capacities.26 There are significant consequences for accepting that imagination is, by its very nature, recreative perspective-taking. It reveals that Jean-Jacques Rousseau’s famous quote that “The world of reality is limited; the world of imagination is boundless” is a grave exaggeration at best. If imagination, by its very nature, aims to create simulations that take up the phenomenal character and content of corresponding (possible) experiences, then the character of our imaginings is constrained by the properties exhibited by these corresponding experiences. To take a simple example: Even though we can easily suppose that there is a wholly red and wholly green circle in front of us, we have trouble imagining seeing one. What we can imagine is not wholly under our voluntary control, but is instead constrained by what it is possible for us to perceptually, emotionally, or bodily experience, given the make-up of our cognitive capacities and independently of the contingent nature of our actual environment. In Zeno Vendler’s words: “The limits of imagination are the limits of experience.”27 Once we realize that imagination, unlike supposition, is systematically constrained, we are forced to revise the Epistemic Innocence Thesis.28 The main idea is as follows: When Alice imagines seeing a red apple, then her imagining aims at things looking the same way to her in certain core respects—at visually seeming the same way from a first-person perspective—as they would if she had a perceptual experience as of a   See Williamson (2007) and Yablo (2002).   Vendler (1979, p. 166). There is a nice heuristic for thinking about this particular realm of possibilities: Consider holding your cognitive equipment fixed, and being transported around various possible worlds to be confronted in experience with things that might be very different than the way things actually are. All the experiences we would have in all those possible worlds are possible experiences—experiences we could have—in the sense at issue. 28   Within this comparative study I can provide only a sketch of the epistemic role of imagination; I provide a more detailed account in Balcerak Jackson (in press). 26 27

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagining, Supposing, and Conceiving  51 red apple. The imagining state represents to her a distinctive way of looking, just as the perceptual state does. After all, this is what visual imagination is by its very nature, a capacity that is dependent on our perceptual capacity. But in virtue of this recreative character, Alice’s imagining provides her with direct epistemic access to  possibilities of a sort: It provides her with justification for beliefs about how things could possibly look. On the basis of her imagination, Alice can be justified in believing, for example, that a hexagon can be formed out of four triangles, or that something can be blue and elephant-shaped. Obviously, the justification provided in these cases is defeasible. We have no more reason to believe in the infallibility of our imaginative capacities than we have to believe in the infallibility of our perceptual or reasoning capacities. Even though this sketch allows us to carve out a justificatory role for imagination, one might worry that this role does not allow for imaginatively-based justification for any beliefs that are not about our experience. Dispelling this worry would require its own paper, but let me suggest a tempting line of thought. Let us suppose that imagining provides us with justification for believing that P is a way things could look. But if I have reason to believe that P is a way things could look, then I have reason to believe that P is a way things could veridically look. But then a fortiori, I have reason to believe that P is a way things could be. At least for the typical properties that enter into the phenomenal contents of perceptual experience, such as surface properties, the inference from things could look that way to things could veridically look that way seems promising. If this line of thought can be supported, then we have not only a sketch of an argument for the claim that imaginings provide us with justification of some sort, but also a sketch of an argument for the claim that imaginings provide us with justification for beliefs about not merely experiential possibilities, but also metaphysical possibilities.

4.  Supposition as Acceptance The contrast between imagining and supposing has helped us to develop a rich notion of imagination, a notion that allows us to see how it can be epistemically valuable. But where does this leave supposition? When I asked you above to suppose that there are finitely many prime numbers, I could have not expected you to imagine this in the sense articulated in the previous section. What kind of experience as of finitely many prime numbers could I have asked you to simulate? Whose experiential perspective could I have asked you to occupy such that from this perspective you see, or feel, or sense, the existence of finitely many prime numbers? What I did is merely ask you to use your ability to think a thought with a particular content. As observed above, this is paradigmatically done for some specific purpose. One such purpose—as you might have been aware—is to take the first step in a reasoning procedure that ultimately justifies the belief that there are not finitely many prime numbers. This suggests that even though supposition is epistemically innocent, it is not epistemically vacuous. So, what

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

52  Magdalena Balcerak Jackson is the epistemic role of supposings, and how is it grounded in the distinctive nature of supposition? We can arrive at a better understanding of the nature and the epistemic value of supposition with the help of Robert Stalnaker’s technical notion of acceptance. Stalnaker (1984) introduces the notion of acceptance as a label for those mental states that are primarily involved in the acts of deliberation and inquiry of intelligent subjects. In subsequent work, he also uses the notion to account for the common ground necessary for successful communication between intelligent subjects, that is for the “mutually recognized shared information in a situation in which an act of trying to communicate takes place. . . . [It is] information that must be available in order that the act of uttering certain noises reasonably be taken as an act of trying to get somebody to acquire certain information.”29 As Stalnaker rightly points out, not all speaker presuppositions that guide or emerge from a communicative act can be correctly described as (shared) beliefs. Rather, acceptance is, according to Stalnaker, “a category of propositional attitudes and methodological stances towards a proposition”30 that includes belief, but also many other mental attitudes very much unlike beliefs: “To accept a proposition is to treat it as true for some reason. One ignores, at least temporarily and perhaps in a limited context, the possibility that it is false.”31 This characterization includes all the ingredients that make acceptance a valuable model for supposition: First, to accept a proposition is not merely to treat it as true as a matter of coincidence, or because one has been struck by lightning. It is to treat it as true for a purpose. Thus far, supposings have been characterized by their hypothetical nature and by the fact that they obey the Voluntary Control Thesis. So it might look as if supposing is an “anything goes” mental state, that when we suppose something we are simply related to some content or other. But then what distinguishes supposing that P from merely considering P or thinking of P? And what explains the obvious methodological value of supposings? The answer has two parts. First, unlike considering P or entertaining the thought that P, supposing P has a quasi-assertive nature. It is part of the nature of supposition that we do not merely think of P, but also think of or treat P as true, that is, we accept P. And second, unlike considering that P or entertaining the thought that P, supposing that P always serves a purpose. As we have seen, the purpose is typically to reason through the consequences of whatever is supposed. To suppose something for no reason whatsoever, or without consequently thinking through its implications, is akin to stopping in mid-action. To consider or entertain P we might need a reason to bring it up, but to suppose P we need a reason to treat it as true.   Stalnaker (1984, pp. 77ff) and Stalnaker (2002, p. 704).   Interestingly, Stalnaker’s description of acceptance as a conglomerate of a category of propositional attitude and methodological stance fits nicely with my description of imaginings, supposings, and conceivings as conglomerates of psychological kinds and cognitive practices. 31   Stalnaker (2002, p. 716). This also distinguishes acceptances from so-called pro-attitudes that include wishes or desires. 29 30

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagining, Supposing, and Conceiving  53 Second, it is consistent with accepting a proposition that a person accepts it only temporarily, and even only for a very limited time frame. The only condition on the duration of our supposition is whatever is needed to think through the consequences relevant in the context at hand. As soon as this purpose is served, the subject does not necessarily retain any credence in the proposition supposed. And third, a person might accept something in a limited context only, while rejecting it or suspending judgment in other contexts. This “compartmentalization” is actually typical of acceptance states that are not also beliefs. For supposition to serve its job it is crucial that supposing something that we know not to be true, for the purpose of reasoning through its consequences, is insulated from numerous of our other, well-supported beliefs. The suggestion, then, is that we should understand supposition as a form of acceptance. More precisely, supposition is the cognitive capacity to enter into states that meet the minimal requirements on states of acceptance. Given the characterization provided above, this entails that we can suppose conceptual or a priori falsehoods. One might worry that this casts the net too wide. Shouldn’t we be interested in a notion of supposition that has a closer connection to rationality and possibility? The short answer is: no. Often our best way to figure out what is impossible is via reductio. One reason to suppose that there are finitely many prime numbers is to show that this supposition leads to a contradiction. For this reasoning to be available to us, it is important that we have the cognitive capacity to treat as true—at least in a temporarily limited, compartmentalized way—propositions that are necessarily false, and even propositions that it would be highly irrational to believe. We are now in a position to see what is wrong with Currie and Ravenscroft’s suggestion that supposition is simulated belief: Believing that P—even simulating believing that P—demands more than merely that one take P as true. It requires one to be rationally committed to P within the context of the imagining project. Such rational commitment depends on the use of basic cognitive capacities that are constitutive of us as believers, such as our grasp of the relevant concepts, certain fundamental logical capacities, and so on. This places limits on the propositions to which we can be rationally committed, consistent with our status as believers. For example, we cannot be rationally committed to the proposition that Stalnaker is the smallest prime number, not just because there is evidence against it, but because it violates simple conceptual principles. Because of this, it is difficult to simulate being rationally committed to the proposition. But of course we have no trouble supposing that Stalnaker is the smallest prime number, and as we have seen, our ability to do so is crucial for our ability to engage in reductio reasoning. The example of reductio reasoning shows, moreover, that even though supposing does not play a justifying role, it would also be wrong to assign it a role in the context of discovery. Supposing that there are finitely many prime numbers was not the inspiration for considering the hypothesis that the number of primes is not finite. Rather, supposing that there are finitely many prime numbers serves an enabling role in the

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

54  Magdalena Balcerak Jackson acquisition of knowledge of a modal claim. Our ability to treat the relevant proposition as true is necessary for us to be able to engage in the relevant forms of inference that justify the final resulting belief. By enabling us to reason through consequences, supposings provide us with indirect access to possibility and necessity.

5.  Conceiving as Rational Perspective-Taking I have suggested that imagination is best understood as recreative perspective-taking that takes up the distinctive first-personal phenomenal character of experience, and that justifies us in beliefs about possibilities. And I have argued that supposition is best understood as a minimal form of acceptance that allows us to take propositions as true for a limited time and in a limited context, and that enables us to reason through their implications. At this point, is there still a place for a distinctive cognitive capacity of conceiving? A deflationary attitude towards conceivings might be especially tempting given that talk of conceiving is not nearly as natural and as common in everyday discourse as talk about imagining and supposing. So, perhaps conceiving is merely a fancy word invented by philosophers to talk about either imagination or supposition. Think again about my invitation above for you to attempt to conceive of your zombie twin. If you think you succeeded in conceiving the zombie, what did you do? And more specifically, can we adequately characterize what you did either as imagination or as supposition, as these are now understood? The answer seems to be no. Chalmers intends his thought experiment to provide evidence against physicalism as a necessary claim about the relation of the mental to the physical. For this to be the case, it must be that successfully conceiving of the zombie provides evidence for the existence of a possibility of some sort. Therefore, we cannot treat the question of whether we can conceive of a zombie as the question of whether we can suppose that a zombie exists. Surely we can suppose this, but, as we have seen, this is compatible with it being necessarily and even conceptually false. But it is also difficult to understand the invitation as an invitation to imagine a zombie. For what is the first-person perspective that we are supposed to occupy, and whose experiences are we supposed to recreate? It cannot be the perspective of the zombie, because it is stipulated that the zombie does not have experiences, and so does not have a perspective to occupy. Perhaps we are meant to occupy the perspective of the zombie by recreating experiences of absences. But it is not at all clear that such an invitation makes sense; as we observed above, the difficulty in imagining absences is one of the features that distinguish imagining from supposing. On the other hand, if it is the perspective of an external observer of the zombie that we are supposed to occupy, then this observer will not experience any of those properties that distinguish a zombie from a conscious being. For them the experience as of seeing a zombie is indistinguishable from the experience as of seeing an ordinary conscious being.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagining, Supposing, and Conceiving  55 The more plausible explanation is that Chalmers’s thought experiment asks us to use a cognitive capacity that is distinct from both imagining and supposing. In a later paper, he describes the intended target as follows: One can place the varieties of positive conceivability under the broad rubric of imagination: to positively conceive of a situation is to imagine (in some sense) a specific configuration of objects and properties. It is common to imagine situations in considerable detail, and this imagination is often accompanied by interpretation and reasoning. When one imagines a situation and reasons about it, the object of one’s imagination is often revealed as a situation in which S is this case, for some S. When this is so, we can say that the imagined situation verifies S, and that one has imagined that S.32

The following clarifications make it clear that the notion of imagination he has in mind is significantly different from ours: There is a sense in which we can imagine situations that do not seem to be potential contents of perceptual experiences. One can imagine situations beyond the scale of perception: e.g. molecules of H2O, or Germany winning the Second World War. One can imagine situations that are unperceivable in principle: for example, the existence of an invisible being that leaves no trace on perception. And one can imagine pairs of situations that are perceptually indistinguishable: for example, the situations postulated by two scientific hypotheses that make the same empirical predictions, or arguably the existence of a conscious being and its zombie twin (an unconscious physically identical duplicate). In these cases, we do not form a perceptual image that represents S. Nevertheless, we do more than merely suppose that S, or entertain the hypothesis that S. Our relation to S has a mediated objectual character that is analogous to that found in the case of perceptual imaginability. In this case, we have an intuition of (or as of) a world in which S, or at least of (or as of) a situation in which S, where a situation is (roughly) a configuration of objects and properties within a world. We might say that in these cases, one can modally imagine that P. One modally imagines that P if one modally imagines a world that verifies P, or a situation that verifies P. Modal imagination goes beyond perceptual imagination, for the reasons given above, but it shares with perceptual imagination its mediated objectual character. . . .  This notion is our core notion of positive conceivability: I will henceforth say that S is positively conceivable when it is coherently modally imaginable.33

The main question is whether we actually have this cognitive capacity of conceiving understood as “coherent modal imagination.” While there are numerous examples of using supposition and imagination, in our sense, in everyday reasoning, conceiving in this sense would only be of limited philosophical or other highly technical use. What are our options? One option is to abandon the discussion of conceivability as an epistemically useful cognitive capacity. The other option is to try to reconstruct the basic idea behind conceivability while relying on other notions that are more familiar to us

  Chalmers (2002, p. 150).   

32

  Chalmers (2002, pp. 151, 153).

33

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

56  Magdalena Balcerak Jackson than the notion of modal imagination used by Chalmers. What follows is a sketch of such a reconstruction. When one conceives of P, one does engage in an exercise of perspective-taking. But one does not take the perspective of the subject as the subject of phenomenal experiences, but rather as the subject of rational belief. Just as when we imagine seeing a red apple, we put ourselves into the position of a possible subject who actually has a perceptual experience as of a red apple, when we conceive of zombies, we put ourselves in the position of a possible subject who is rationally committed to zombies. So, while I have argued that supposing should not be understood as simulating belief, I suggest that conceiving—at least as it is used in philosophy—should be so understood. Simulated perspective-taking in imagination is easy to understand because the subjects whose perspectives we are taking are ordinary perceivers like us. But the perspectives we take in conceiving, I want to suggest, are those of ideally rational believers with unlimited reasoning capacities. This makes the perspective-taking in conceiving harder to understand. When one tries to conceive of zombies, one holds fixed one’s basic rational capacities—first and foremost, one’s conceptual and logical capacities— and tries to adopt the perspective of a subject whose use of these capacities is infallible and unimpaired by limitations of memory, attention, and so on, and who believes that there is a zombie in the circumstances described. In fact, one can distinguish two perspective-taking exercises, corresponding to the two experiential perspectives that one might take on zombies noted above. The first exercise calls on one merely to simulate ideally rational belief that there are zombies, while the second calls on one to simulate ideally rational belief that one is oneself a zombie. (There is no prima facie reason to think that the latter exercise is any less coherent than the former, unless it turns out that belief is partly constituted by certain phenomenal properties.) To better understand the suggested picture it is useful to address specifically how it avoids a threat of collapse. The worry can be stated as follows: If occurrent belief is nothing but a distinctive kind of experience, just as perceptual states and emotions are kinds of experiences, then simulating belief will be simulating a kind of experience, and therefore a kind of imagining. But if belief is not just a kind of experience in the same sense in which perception and emotions are experiences, then simulating belief will not involve any perspective-taking, and will likely turn out to be a kind of supposition. Let us assume that belief is an experience that has a distinctive cognitive phenomenology. If this phenomenology of believing a certain content makes belief what it is, then it is an experience in the same sense in which perceptual states are experiences. Then conceiving would indeed be merely a sub-kind of imagining. However, it is wrong to think of beliefs in purely descriptive terms. Believing that P is not merely having a certain attitude with a distinctive cognitive phenomenology, it is also a matter of being subject to certain normative constraints. If you report a belief that Elizabeth II is the Queen of England, but then just a minute later report a belief that England does not have a Queen, then it might very well be that both of your states exhibit cognitive phenomenology, but intuitively we would not want to call your mental states beliefs,

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagining, Supposing, and Conceiving  57 because they are blatantly and obviously inconsistent. We would also intuitively question a report of a belief that bachelors are married on the basis of fundamental conceptual confusion. We are only willing to ascribe beliefs to a subject if the state of the subject meets some rational constraints. It is notoriously hard to say what these rational constraints are, but they govern our beliefs in ways in which they do not govern our experiences. Therefore, simulating belief is not merely simulating an experiential perspective. The normative nature of belief equally explains why simulating belief is not merely supposing, for the epistemic role of supposing relies on our ability to free ourselves from those rational constraints when creating a supposition from which to reason further. What is the epistemic significance of conceiving, on this understanding? If we can successfully conceive of zombies, then this answers for us the question of whether an ideally rational subject could be rationally committed to the existence of zombies. In other words, asking whether zombies are conceivable is, first and foremost, a way of trying to determine whether it is possible to have an ideally rational belief in zombies. Given the constraints on simulating ideally rational belief, the fact that one is able to successfully conceive in a particular case arguably gives one prima facie reason to believe that what one conceives of is possible in at least one sense: It is evidence that what one conceives is not ruled out by ideal application of one’s basic conceptual and logical capacities. Or in our terminology: Given that what we can conceive is not entirely under our voluntary control, conceiving is not epistemically innocent but justifies claims about what is possible. Whether this has any further implications for what is metaphysically possible remains an open question, and so it remains an open question whether or not Chalmers’s zombie argument is ultimately successful if we understand conceiving along these lines. (This would be the project of another paper.) The suggestion here is simply that understanding conceiving as simulated ideally rational belief gives us a good way to make sense of the cognitive capacity Chalmers and others gesture towards when speaking of conceiving. Is conceiving, so understood, a cognitive capacity that we actually have? The answer to this question is important, because even if there is an entailment from conceivability to possibility, this yields justification for beliefs about what is possible only if we actually can successfully conceive. In general, I am skeptical about whether we can occupy cognitive perspectives of subjects whose mental capacities are so radically different from our own, or even just approximate these perspectives well enough. In imagination we surely can occupy perspectives of subjects who happen to have very different courses of experiences in their lifetimes. But it is doubtful that we can imaginatively occupy the perspective of somebody who has radically different perceptual capacities from ours—call it schmerception—and who consequently has some very different form of conscious experience. If it is indeed a fact that in imagination we use some of the same cognitive mechanisms that were developed for perception—as some empirical research suggests—then those mechanisms are probably not capable of recreating a schmerceptual experience. This analogy might not be completely appropriate, because

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

58  Magdalena Balcerak Jackson conceiving does not require us to simulate mental capacities that we do not have at all, but rather to simulate highly idealized versions of capacities that we actually possess. So, the question is whether—and if so, how reliably—non-ideal reasoners like us can put themselves in the shoes of ideal reasoners when trying to decide whether P is genuinely conceivable. I am skeptical, but the present point is just to emphasize that the understanding of conceiving that I have developed here brings out a challenge for those who would defend the epistemic value of conceiving in philosophical (or other) enquiry.

6. Conclusions Talk about imagination, supposition, and conceivability is so ubiquitous in contemporary analytic philosophy that it is tempting to assume that when we ask each other to imagine waking up in a hospital hooked up to a famous unconscious violinist, to suppose that there are finitely many prime numbers, or to conceive of our zombie twin, we all understand quite well what kind of activity we are asking each other to engage in. Call this activity whatever you prefer, it is just what we do when we represent hypothetical scenarios and objects. It might be especially tempting to make such an assumption if one takes it for granted that hypothetical attitudes have a merely enabling or heuristic role to play within philosophical enquiry. However, even when the epistemic role of hypothetical attitudes is under discussion—as in debates about whether conceivability is a good epistemic guide to possibility—the nature of conceiving in relation to imagining and supposing is often left rather vague. In this chapter I have argued for a systematic picture of imaginings, supposings, and conceivings as the products of three distinctive cognitive capacities. Such a picture is supported by the salient differences between core cases of the three types of hypothetical attitudes—most importantly by the ways in which what we imagine, suppose, and conceive are or are not under our voluntary control. Imagination is the cognitive capacity to take perspectives of other possible subjects of experience and simulate the experiences they would have in various situations. Conceiving is the cognitive capacity to take perspectives of other possible reasoners and to simulate what they would be rationally committed to in their given circumstances. And supposition is the cognitive capacity merely to accept a proposition for the purpose of reasoning through its implications. While supposition is under our voluntary control in the strong sense, the extent of the control we have over imaginings and conceivings is limited by their ­recreative nature. Aside from an analysis of the nature of core cases, I have also tried to bring out the epistemic roles that imaginings, supposings, and conceivings do and should play. While imagination—and possibly also conceiving—can give us prima facie justification for specific claims about what is possible, supposition cannot; it merely enables us to reason through the implications of accepted propositions. If the relevant cognitive

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagining, Supposing, and Conceiving  59 capacities are not merely a theoretician’s dream, but something we imperfect subjects actually possess, then each of these capacities has a place within philosophical enquiry. I am confident that we are reasonably competent imaginers and supposers. Conceivers? I don’t know.

Acknowledgements For valuable comments and suggestions on earlier presentations of this material, and in conversations about the issues discussed in this chapter thanks to Brendan Balcerak Jackson, Paul Boghossian, Alex Byrne, Lars Dänzer, Jonathan Jenkins Ichikawa, Amy Kind, Peter Kung, and two anonymous referees.

References Balcerak Jackson, Magdalena (in press). Justification by imagination. In Fabian Dorsch and Fiona Macpherson (eds), Perceptual Memory and Perceptual Imagination. Oxford: Oxford University Press. Chalmers, David J. (1996). The Conscious Mind: In Search of a Fundamental Theory. New York: Oxford University Press. Chalmers, David J. (2002). Does conceivability entail possibility? In Tamar Szabό Gendler and  John Hawthorne (eds), Conceivability and Possibility. Oxford: Clarendon Press, pp. 145–200. Currie, Gregory, and Ian Ravenscroft (2002). Recreative Minds: Imagination in Philosophy and Psychology. Oxford: Oxford University Press. Doggett, Tyler, and Andy Egan (2007). Wanting things you don’t want: The case for an imaginative analogue of desire. Philosopher’s Imprint 7(9): 1–17. Farennikova, Anya (2012). Seeing absence. Philosophical Studies 166(3): 429–54. Gendler, Tamar Szabό (2000). The puzzle of imaginative resistance. Journal of Philosophy 97(2): 55–81. Gendler, Tamar Szabό (2006). Imaginative resistance revisited. In Shaun Nichols (ed.), The Architecture of the Imagination: New Essays on Pretence, Possibility, and Fiction. Oxford: Oxford University Press, pp. 149–74. Goldman, Alvin I. (2006). Simulating Minds: The Philosophy, Psychology, and Neuroscience of Mindreading. New York: Oxford University Press. Ichikawa, Jonathan Jenkins, and Benjamin W. Jarvis (2013). The Rules of Thought. Oxford: Oxford University Press. Kind, Amy (2001). Putting the image back in imagination. Philosophy and Phenomenological Research 62(1): 85–109. Kind, Amy (2013). The heterogeneity of the imagination. Erkenntnis 78(1): 141–59. Peacocke, Christopher (1985). Imagination, experience, and possibility: A Berkeleian view defended. In John Foster and Howard Robinson (eds), Essays on Berkeley: A Tercentennial Celebration. Oxford: Clarendon Press, pp. 19–35. Sartre, Jean-Paul (1972). The Psychology of Imagination. London: Routledge.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

60  Magdalena Balcerak Jackson Stalnaker, Robert C. (1984). Inquiry. Cambridge, MA: MIT Press. Stalnaker, Robert (2002). Common ground. Linguistics and Philosophy 25(5–6): 701–21. Thomson, Judith Jarvis (1971). A defense of abortion. Philosophy and Public Affairs 1(1): 47–66. Van Leeuwen, Neil (2013). The meanings of “imagine” Part I: Constructive imagination. Philosophy Compass 8(3): 220–30. Van Leeuwen, Neil (2014). The meanings of “imagine” Part II: Attitude and action. Philosophy Compass 9(11): 791–802. Vendler, Zeno (1979). Vicarious experience. Revue de Métaphysique et de Morale 84(2): 161–73. Walton, Kendall L. (1990). Mimesis as Make-Believe: On the Foundations of the Representational Arts. Cambridge, MA: Harvard University Press. Weatherson, Brian (2004). Morality, fiction, and possibility. Philosopher’s Imprint 4(3): 1–27. White, Alan R. (1990). The Language of Imagination. Oxford: Blackwell. Williamson, Timothy (2007). The Philosophy of Philosophy. Oxford: Blackwell. Wittgenstein, Ludwig (1981). Zettel (trans. G. E. M. Anscombe). 2nd edn, Oxford: Blackwell. Yablo, Stephen (2002). Coulda, woulda, shoulda. In Tamar Szabό Gendler and John Hawthorne (eds), Conceivability and Possibility. Oxford: Clarendon Press, pp. 441–92. Zahavi, Dan, and Uriah Kriegel (2015). For-me-ness: What it is and what it is not. In Daniel O.  Dahlstrom, Andreas Elpidorou, and Walter Hopp (eds), Philosophy of Mind and Phenomenology: Conceptual and Empirical Approaches. New York: Routledge, pp. 36–53.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

2 On Choosing What to Imagine Peter Langland-Hassan

1. Introduction If we choose the content of our own imaginings, how does the process of imagining move us forward, epistemically speaking? We do learn from our imaginings, after all. Imagination is implicated in modal reasoning,1 action planning,2 art appreciation,3 mindreading (Goldman 2006), pretense,4 artifact design (Arp 2008), scientific innovation (Kind (in press)), and spatial reasoning (Kosslyn, Thompson, and Ganis 2006). These are just some of the contexts where imagination is said to helpfully guide action and inference. Yet, unlike more paradigmatic examples of states that guide action and inference—beliefs and perceptual experiences—the content of our imaginings is usually said to be “up to us” in some significant sense. What we imagine seems determined by what we wish to imagine. But how can it be that we are able to wish ourselves into more epistemically favorable states of mind? Even if one were very careful with one’s wishes, the fact that the content of an imagining is chosen would seem to render the imagining itself pointless. For it suggests that the content of the imagining was already present in one’s intentions (why else would the imagining count as chosen?). If that is the case, why go through with it? Imagining becomes a kind of internal transfer of contents—the mental equivalent of handing yourself a dollar. At the same time, neither will simply letting your imagination run wild bolster its capacity to guide action. When we imagine in the above contexts, we don’t pull the lever on a mental slot machine. The process is controlled and, it would seem, subject to constraints. That is why we can rely on our imaginings to guide our behavior and judgments in so many important domains. So the epistemic value of imagination is at odds both with its being completely unfettered and with its being completely under 1   Special thanks to Christopher Gauker, Peter Kung, and Neil Van Leeuwen for helpful comments on earlier drafts of this work. Examples include Yablo (1993), Chalmers (2002), and Kung (2010). 2   See e.g. Addis et al. (2009), Schacter et al. (2007), and Van Leeuwen (2011). 3   Examples include Doggett and Egan (2012) and Walton (1990). 4   See e.g. Nichols and Stich (2000), Nichols (2004), and Currie and Ravenscroft (2002).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

62  Peter Langland-Hassan intentional control. Philosophers tend to emphasize the great freedom of choice we have in imagination, without squaring this with its capacity to guide action and inference.5 Psychologists, by contrast, often highlight the role of imagination in practical reasoning, without explaining how this is compatible with our ability to choose what we imagine (Byrne 2005, Schacter, Addis, and Buckner 2007). My goal here is to see if and how imagination’s capacity to guide action and rational inference can be brought into alignment with our freedom to choose what we imagine.

1.1  Some fine-tuning One way to approach the question is to grant that imagination, as a type of mental process, has some instances that are “chosen,” and some instances that helpfully guide us, while maintaining that these qualities are not typically possessed at the same time by token imaginings. For instance, Amy Kind (in press) contrasts imaginings where “we constrain our imaginings to fit the facts of the world as we know them” with those where we do not. If only the latter imaginings are subject to the will (in the sense of being chosen), and only the former helpfully guide action and inference, this might remove the need to explain how it can be that a particular mental process is simultaneously subject to the will and conferring of epistemic advantage. Can imaginings be divided into distinct kinds in this way? To answer, it will help to consider bodily actions by comparison, and the conditions under which we say a bodily action is chosen. If a bodily action derives from an explicit intention to carry out that action, it clearly counts as chosen (tabling more general worries about free will). However, many actions are chosen even when the subject did not form a conscious or explicit intention to perform them. For instance, when I select the correct key from my keychain to open my office door, selecting that key is something I choose to do. But I need not have formed an explicit intention or plan to select that key. And when I go to the kitchen to get a glass of water, my opening of a particular cabinet, and my selecting of a tall rather than short glass, are things I choose to do, even if my doing them feels quite automatic. In these cases, there is some (perhaps only subconscious) action-­ guiding mental state in virtue of which the act was not simply accidental, or a mere reflex. I will follow Searle (1983, ch. 3) in calling this sort of action-initiating mental state an intention in action.6 Intentions in action can be contrasted to prospective (or “prior”) intentions, such as the intention to go for a run today, or to file one’s taxes by April. Prospective intentions relate to consciously made future plans that may or may not 5   Cf. Fodor (1975, p. 191): “What makes my stick figure an image of a tiger is not that it looks much like one . . . but rather that it’s my image, so I’m the one who gets to say what it’s an image of.” More recently, Colin McGinn has called it a “familiar point” that “I cannot misidentify the object of my imagining” because “the identity of my imagined object is fixed by my intentions” (2004, p. 31). See also Noordhof ’s (2002, pp. 429–34) “Straight Forward View,” which identifies what is imagined with what an imaginer supposes is imagined. 6   I do not, however, endorse Searle’s claim that the content of an intention in action is self-referential. For Searle, the content of an intention to raise one’s arm is something like: (My arm goes up as a result of this intention in action), whereas I would have its content simply be: (My arm goes up).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

on choosing what to imagine  63 come to fruition. An intention in action, by contrast, is what accounts for a particular movement’s being willful, in the sense that a person can be held responsible for it. Motor commands (of the sort discussed in Section 3.2 on “forward models”) are a species of intention in action, as I will understand them. A bodily movement that is accidental, or that is forced by an outside power, will not have been guided by an intention in action. A bodily action can thus be thought of as having two necessary (but distinct) components: the intention in action that initiates it, and the movement that carries it out. Now, our question was whether one could quickly dissolve the apparent puzzle of chosen imaginings by holding that it is only non-guiding imaginings that are chosen. What we see with bodily actions, however, is that the kinds of actions that most likely advance one’s interests (and improve one’s epistemic standing) are the ones that are chosen, at least in the sense that they result from intentions in action. Indeed, there is  no expectation that accidental or coerced movements will advance one’s goals. Similarly, we should not expect all the imaginings that improve one’s epistemic standing to be accidental or uncontrolled. We should expect very many of them to be ­chosen, in just the way that useful bodily actions are usually chosen. And, intuitively, this is the case. When we imagine taking a new route home, or how a friend will react to a gift, or to determine a scenario’s possibility, these are mental actions that we choose to undertake, and over which we have control. They are not things that just happen to us—at least, not normally. This is not to overlook the tension that remains in the idea that we can choose a usefully guiding imagining. The point is just that we cannot sidestep the entire puzzle by holding that only non-guiding imaginings are chosen and controlled. We can now see that the most acute puzzle with respect to choosing imaginings concerns only a particular class of imaginings: those that are both chosen (in being subject to the will) and suitable for guiding action and inference. I will call these “Guiding Chosen” (GC) imaginings. Most of what follows will be an account of how there can be such a class of imaginings. I will argue that understanding Guiding Chosen imaginings requires positing three general features of their cognitive architecture. These are: 1) The initial involvement of “top-down” intentions for initiating imaginings (described in Section 2), 2) The use of (what I will call) “lateral” constraints in the development of an imagining (described in Section 3), and 3) The cyclical involvement of top-down intentions throughout the course of an imagining (described in Section 5, in response to a challenge raised against most existing accounts of imagination in Section 4). The end result is a picture of GC imaginings that reveals them as a kind of continuously guided conditional reasoning. This conclusion is put to a larger end in Section 6, where I argue that the cognitive-architectural features posited to explain GC imaginings give us the tools we need to understand most other imaginings as well.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

64  Peter Langland-Hassan

2.  Accounting for Choice In what follows, unless otherwise noted, all talk of “imagining” will refer specifically to our target phenomena: Guiding Chosen imaginings. To get started in thinking about GC imaginings, let us assume that paradigmatic instances of GC imaginings involve the endogenous triggering of a sequence of mental states. By “endogenous triggering” I simply mean that, normally, imaginings spring from factors internal to the organism. Here imagination is contrasted to visual perception, where the light currently reflecting off of nearby objects is a crucial part of the causal chain. In paradigmatic episodes of imagination, some endogenous cause leads to the tokening of a sequence of mental states which we can call: i1…in. We can call the entire sequence an “imaginative episode,” and each ix in that sequence an “imaginative state.” The idea is that, normally, a single imaginative episode will involve a sequence of imaginative states. With these minimal assumptions in place, we can ask: in what sense does a person choose the i1…in that constitute an imaginative episode? The most general answer suggested above is that the imagining is initiated by an intention, where this may often be a non-conscious intention in action. When an imagining has no special relationship to one’s intentions, it is not chosen, just as unintended bodily movements are not chosen. That said, the intention that initiates an imagining should not be considered a part of the imagining itself, any more than the intention that initiates a raising of the arm is part of the arm’s motion. An imaginative episode is on a par with a bodily movement; only when it is initiated by an intention will it count as an action (in the case of imagination, it will be a mental action). This allows for some imaginative episodes (e.g. unbidden imaginings) that are not mental actions. The next question to ask is to what extent the contents of i1…in (a particular episode of GC imagining) are determined by one’s intentions. In answering it will help to draw a familiar distinction between propositional imaginings and sensory imaginings (though not everyone agrees the distinction is in good standing—see e.g. Kind (2001)). There is much that can be said concerning how best (or if) to account for the difference between the two. Here a few cursory remarks will have to do. Propositional imagining occurs when a person imagines that thus and such. Imagining that the FBI is plotting to kidnap you is an example. It is typically thought that propositional imagining does not require the use of sensory imagery. Some support for this can be found in considering pretense. Imagining that p is plausibly the normal cognitive component of pretending that p. If a person can pretend that they are a tiger, or a mobster, or a snowflake, without using any sensory imagery, then it seems that propositional imagination need not be imagistic. Sensory imaginings, on the other hand, are usually defined as requiring mental imagery, and are typically ascribed with non-propositional uses of “imagines.” Unfortunately, I do not have space here to further defend or explain these distinctions.7 7  Is all cognition that involves sensory imagery properly called sensory imagination? It seems the answer should be no, at the risk of counting ordinary episodic memories as exercises of the imagination (this question gets more extended treatment in Section 6). However, there are no widely accepted answers here.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

on choosing what to imagine  65 I introduce them because they are already widely accepted, and because different issues arise when conceiving of imagination as either propositional or sensory in nature.

2.1  Choosing propositional imaginings Let’s look first at propositional imaginings and ask to what extent the contents of i1…in (a particular episode of imagining) are determined by one’s intentions, and therefore chosen. The most common view about propositional imaginings is that the ix amount to a sequence of belief-like states, where being in those states amounts to taking the attitude of imagination toward a sequence of propositions (though this is not my view—see fn. 9). We can then ask: to what extent do we choose the content of those states? The most extreme answer—one I expect no one to endorse—would be to say that the content of each proposition is determined by an intention to imagine a proposition with that very content. This is what I will call the Only Top Down approach, the idea being that intentions are “top down” influences on imaginings. According to Only Top Down, when I propositionally imagine that p, that q, and that r (where these imaginings are stages of the sequence i1…in), it is because I have first had an intention to imagine that p, and then had an intention to imagine that q, and so on. Here imagination is, at bottom, a transfer of contents from one’s intentions to one’s (propositional) imaginings, during which the contents p, q, and r are “extracted” from the intention to imagine that p, imagine that q, and imagine that r. This approach maximizes the sort of intentional control we have over imagination, while calling its usefulness into question. One ends up where one began, epistemically speaking. Explaining the capacity of one’s imaginings to usefully guide action and inference would amount to explaining one’s capacity to have a sequence of intentions of the sort that, given the process of “extraction,” result in useful imaginings. All of the heavy lifting falls on one’s intentions. One way of seeing the problem is that each ix— that is, each successive imaginative state—does not causally influence the one following it. Rather, at every point, the content of each ix is wholly determined by a top-down intention. As I said, I do not expect Only Top Down to be a popular view. But it is useful to have it on the table, as it forces us to focus on what the other options might be for explaining the apparent “chosen” nature of imaginings. It seems we must have somewhat less choice than the Only Top Down view envisions, even if our intentions may be relevant in initiating an imagining.

The relationship between sensory imagination and episodic memory is not well understood. Nor, for that matter, is the relationship between propositional and sensory imagination. If I form an image of a polar ice cap growing in size while imagining that global climate change has been reversed, am I engaging in two different kinds of imagination at once (employing two different faculties of mind)? Have I entered a third, hybrid state? For the purposes of this chapter, I will treat the two types of imagination as distinct, to avoid raising issues I will not have space to address.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

66  Peter Langland-Hassan

2.2  Choosing sensory imaginings However, the Only Top Down approach might seem more promising with respect to sensory imagination (cf. Searle 1983, p. 103). Here the idea would be that the ix of an episode of sensory imagination are individual mental images. And, for each mental image, we might choose its content in the sense that it is caused by a corresponding intention to generate an image with that content. The triviality that besets Only Top Down in its propositional form is to some degree avoided if one thinks that mental imagery, by its very nature, represents its contents in a more fine-grained way than one’s (presumably propositional) intentions. For instance, my (coarse-grained) intention to imagine my mother might result in the formation of a visual image of her face that represents various features in a way that outruns any linguistically expressible concepts I might have (and that might be constituents of my intentions). If there is a level of representational detail that mental imagery makes available and that is not present in one’s intentions, then the usefulness of imagination may lie in our ability to willfully make use of that detail by having intentions to trigger related images (this would, however, still presume some pre-existing content-link between one’s images and propositional intentions, such that one could reliably trigger the other). This sort of approach to understanding the usefulness of sensory imagination coheres well with Peter Kung’s (2010) conception of the role of imagination in justifying judgments concerning metaphysical necessity and possibility.8 Kung draws a distinction between the “basic qualitative content” of an imagining (roughly, what the image itself contributes) and “assigned contents,” which we might think of as the elements of content contributed by one’s top-down intentions. For Kung, imaginings gain a measure of usefulness (toward making judgments of metaphysical possibility) to the extent that some of their representational aspects—their basic qualitative contents— are not explicitly chosen or “stipulated,” in the manner of assigned contents. To the extent that assigned contents contribute to what is imagined, the process of imagining does not itself offer justification for judgments relating to those contents. Any justification relating to those contents must be inherited from elsewhere. While this approach is a step forward in reconciling freedom with usefulness, it does not take us very far. As with Only Top Down when considered for propositional imagination, there is still no account of how successive ix causally influence or constrain each other. Each image in an imagining is caused in a “top down” manner by a related intention. What determines the sequence of images will be a corresponding sequence of intentions (even if the intentions are all representationally coarse-grained by comparison). Imagination’s usefulness now lies entirely in its filling in of visuospatial details to a basic narrative that is already fully mapped out in one’s top-down intentions. 8   Kung does not, however, endorse the view I will describe; it is only a view one could extract from his remarks; see also Colin McGinn (2004, ch. 1), who emphasizes the transition between cognitive formats in explaining some of the usefulness of imagination.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

on choosing what to imagine  67 While this is not a trivial kind of usefulness, it diminishes the amount of work that imagination can potentially do. It is normally thought that imagination is a cognitive faculty with its own internal logic or principles of operation that determine how imaginings unfold across time—that the entire diachronic sequence of images is not already set out in one’s intentions. Ideally, we can both vindicate this traditional conception of sensory imagination and show that it has epistemic value beyond the filling in of visuo­ spatial details. Also, we should keep in mind that Only Top Down gives us no account at all of the usefulness of propositional imagination. Thus, whatever adjustments we make in the case of sensory imagination can hopefully be extended to propositional imaginings as well. To recap this section: the involvement of one’s top-down intentions in setting (at least) the initial content of a sensory or propositional imagining seems necessary to explaining why it counts as being chosen. However, such intentions—or series of intentions—cannot be the whole story with respect to how the imagining unfolds across time, short of trivializing the imagining itself.

3.  Adding in Lateral Constraints What are needed to guarantee greater usefulness, both with respect to sensory and propositional GC imaginings, are some lateral constraints on the imaginings. If one’s top-down intentions are key to initiating an imagining—in, say, determining its general subject-matter—then lateral constraints will be what govern how it then unfolds. So, for instance, if an intention to imagine that p is key to determining the content of the first ix of the imaginative episode, the lateral constraints on the imagining will determine why that state leads to the subsequent ix…in that contribute to the total imaginative episode. I am calling the proposed “lateral” influences constraints because they help answer the question: why is this the next ix, and not something else? On the Only Top Down account, this question was always answered by an appeal to a new intention. It preserved absolute choice at the cost of usefulness. Now the idea is that imagination—both propositional and sensory—has its own norms, logic, or algorithm that shapes the sequence of ix after the initiation of an imagining by a top-down intention. These constraints might then play a role in explaining how the imagining is useful.

3.1  Lateral constraints in propositional imagination Essentially this idea was put forward by Nichols and Stich (N&S) in an influential paper (2000) on the role of propositional imagination in guiding pretense (see also Currie and Ravenscroft 2002, ch. 1). On N&S’s account, an imagining begins with the insertion of a proposition into the “Possible Worlds Box” (PWB). The PWB is to be understood functionally; for a contentful state to be “in” the PWB is for it to play a certain kind of functional role. We can think of this initiating proposition as the first ix

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

68  Peter Langland-Hassan in some imaginative sequence i1…in.. Various “inference mechanisms” then determine how that initial proposition develops in imagination (or “in” the PWB). The crucial point for N&S is that these inference mechanisms are the same ones that shape and govern the inferences we draw within our ordinary beliefs. The key difference with the PWB is that it is not directly connected to action-guiding systems. This makes it possible to (rationally) imagine propositions one disbelieves, without threat of the disbelieved propositions driving one’s behavior. And this in turn allows us to draw out (rational) consequences from those (potentially disbelieved) propositions. On N&S’s account, the inferences drawn in imagination are imported back into one’s beliefs as consequents to a newly believed conditional, where the initial proposition that started the imagining is the antecedent. So, if “people colonize the moon” were first inserted into the PWB, and propositions q, r, and s were then imaginatively “inferred” from that proposition (through the activity of one’s “inference mechanisms”), then the end result might be a belief of the form: “If people colonize the moon, then q, r, and s.” On this sort of view, propositional imaginings are “belief-like” in that they unfold more or less in accord with the norms that govern belief. The idea that propositional imaginings are belief-like in this way is widely accepted (though not by me9) (Currie and Ravenscroft 2002, Doggett and Egan 2007, Nichols 2006, Weinberg and Meskin 2006b). If true, it would explain why most imaginings do not develop in a completely arbitrary manner, but instead often match what we would come to believe if we believed the initiating premise of the imagining. At the same time, this approach promises to explain the usefulness of imagination: we can rely upon our imaginings to guide our behavior because our imaginings are constrained by the very same norms (or “mechanisms,” or “algorithms”) that govern ordinary rational inference.10 The freedom to choose what one is imagining is retained as a freedom to, in effect, choose the subject-matter of the (imaginative) inferences. There are quibbles, large and small, that one can raise for this picture of propositional imagination. However, the basic strategy of understanding the lateral constraints on propositional imagination as being on a par with those operative among ordinary 9   I do not believe there is such a thing as an “imagination box,” or that there is an imaginative attitude that is distinct from belief (this makes me a “cognitive lumper” in Liao and Doggett’s (2014) term). On the view of propositional imagination that I defend elsewhere (Langland-Hassan 2012), the imaginative states that make up an imaginative episode are themselves beliefs, and not merely belief-like. How, then, can one imagine a proposition one does not believe? My answer, in brief, is that imagining that p (where one disbelieves p) does not require entertaining (or believing) the proposition that p; instead, it requires retrieving stored generalizations relevant to p-like situations, and using them to draw inferences about what would likely happen if p. On my view, then, one chooses the initial ix in the sense that one chooses the topic of an internal query about what would happen if p (and, of course, asking oneself a question does not require imagination). So, imagining that p can involve only beliefs and internal queries, while still not requiring one to believe that p. I give the details of this approach elsewhere (Langland-Hassan 2012), together with replies to likely objections. Here I do not assume the truth of that account, as the points I want to make do not require it. 10   These norms or principles of inference need not be explicitly represented; the idea is that the organism is constructed such that its inferential patterns conform to the rules or algorithms in question.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

on choosing what to imagine  69 beliefs is widely accepted, and is granted by theories that agree on little else (e.g. Langland-Hassan 2012, and Nichols and Stich 2000). Thus I will adopt it going forward. We can then understand the lateral constraints on propositional imagination to the extent—and only to the extent—that we understand such constraints with respect to belief.

3.2  Lateral constraints on sensory imaginings Before addressing some difficult questions for this general conception of the lateral constraints on propositional imaginings, we can ask if there is any corresponding move that can be made with respect to sensory imagination. At first glance, it might seem that the answer is no (for their part, N&S limit their discussion to propositional imagination). The orthodox view in philosophy is that beliefs do not themselves involve mental images as constituents. Beliefs are most commonly (though not always (Prinz 2002)) held to be “amodal” in nature, and (if there is a dominant view on this) to have a language-like constituent structure (Fodor 1975). From this perspective, the mechanisms or principles that govern the inferential relations among beliefs will not be suitable for governing the relationships among sequences of mental images, which are typically thought to have a different logical and syntactical form (occurring e.g. in an “iconic” or “pictorial” format (Fodor 2003)).11 And, even if sensory imaginings are “percept-like” (Currie and Ravenscroft 2002, ch. 4) in many ways, this does not offer any obvious clues as to how images develop in imagination. For, unlike the case of belief, the sequence of perceptions one enjoys is determined not by endogenous norms or principles, but by the actual state of the environment that one perceives. So there is no obvious quasi-perceptual analog to the rational norms governing inferential development among beliefs that might serve as lateral constraints for sensory imaginings. Nevertheless, an analog may still be available. Several theorists have recently drawn on the existing motor control literature concerning the prediction and comparison mechanisms at work in ordinary perception, as a means to thinking about the kinds of capacities and constraints that might be operative in sensory imagination (Grush 2004, Langland-Hassan 2011, Van Leeuwen 2011). It has long been hypothesized that during ordinary perception cognitive systems “predict” the sort of sensory input that will be received, given their current state and the particular motor command initiated (Sperry 1950, von Holst and Mittelstaedt 1973). The prediction can then be compared to actual “reafferent” input to determine if there is a match. This serves two purposes: first, it enables the organism to distinguish, in an immediate and non-inferential way, whether a particular sensory change was a result of its own movement or, instead, the movement of something in its environment. Second, it enables an organism to 11   The difference in format suggests a difference in the syntactical structure of the representations, and hence a difference in the mechanisms for processing the representations, as such mechanisms are usually assumed to be sensitive only to the syntactical features of representations.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

70  Peter Langland-Hassan instantaneously register whether its desired action was carried out successfully, as the predicted or desired state can be compared to the final state at the moment that state is registered. This helps explain how organisms correct motor errors more quickly (200–300ms) than would be possible if they had to rely upon ordinary visual or proprioceptive feedback (Miall et al. 1993, Wolpert, Miall, and Kawato 1998). In the contemporary control theory literature, the cognitive system whose role it is to generate relevant “predictions” is typically called a forward model (Blakemore, Wolpert, and Frith  2002, Wolpert, Ghahramani, and Jordan  1995). When a motor command is generated to begin an action, an “efferent copy” of that command is sent to the forward model, which generates a prediction of the resulting sensory input. That prediction state is then compared to the actual input. The capacity of the forward model to generate accurate predictions of sensory inputs requires the organism to have a kind of (perhaps implicit) knowledge of sensorimotor regularities and contingencies. This form of knowledge can be thought of as an implicit understanding of a set of rules of the form: given current sensory state c, plus motor command m, the subsequent sensory state should be s. Or, simplifying: If c and m, then expect s. It is not assumed that the rules are discursively represented, or explicitly represented at all; they may merely be conformed to, in the same way that a classical connectionist network might conform to a rule for detecting mines versus rocks (Churchland and Sejnowski 1989). The forward model can thus be seen as storing a set of principles for the “lateral” development of further imagistic states, given initial states and motor commands as input. The core working hypothesis of the control theory literature is that this kind of prediction goes on during most of ordinary perception. If, however, the forward model were under intentional control (to the extent that it could be fed inputs at will), we would have a way of seeing how it could account for lateral constraints for sensory imagination (Grush (2004) explicitly argues that the predictive states generated by the forward model are visual images). We simply need to assume that “efferent copies” of motor commands can be sent to the forward model when no such command is actually sent to centers that would carry out the command. In such cases a series of prediction states (in the form of visual images) could be generated in the absence of a related input. Sensory imagination can then be seen as the purposeful use of the forward model, in absence of relevant perception, to generate what are in effect “predictions” about the likely sensorimotor effects of certain actions. Because the forward model’s predictions are grounded in learned perceptual regularities and contingencies—ones that are relied upon to guide ordinary perception—we have an account of the reliability and usefulness of imagination that is on a par with that considered for belief. And, importantly, it is an account that has a rationale and theoretical life independent of the present puzzle about imagination. The sequence of one’s sensory imaginings will be constrained by the sort of things one would expect to see, given some initial image and motor instruction, in much the way that one’s propositional imaginings are often held to be constrained by what one would come to believe if one believed the imagined

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

on choosing what to imagine  71 premise. During sensory imagination, we can assume that the two “initiating” states (image and simulated motor command) are generated by one’s intentions. This remains the sense in which the imaginings are under intentional control and “chosen.” Unfortunately, this tidy story about forward models cannot be all there is to the constraints governing sensory imagination. For we can usefully imagine the movements of objects and creatures in our environment even in situations where we ourselves are stationary, and where the prediction in question has nothing to do with the consequences of our own possible movements. For example, in visually imagining a basketball rolling off of a high bookshelf, we more or less automatically imagine it bouncing upon hitting the floor. The “prediction” that a falling basketball will bounce on hitting the floor results from stored generalizations about the behavior of different kinds of objects in different settings—and not from generalizations about how the appearances of things change as we actively explore them. An algorithm (or set of algorithms) allowing conformity to such generalizations will doubtless influence and constrain the development of sensory imaginings in addition to forward models. Whether conformity to these algorithms requires input from propositional background beliefs, or, instead, is something that occurs completely independently of propositional thought, is a difficult and important question. For now I can do little more than point to it. This section has aimed to explain how imaginings can be relied upon to guide behavior, even if our top-down intentions set their initial topic or subject-matter. For each of propositional and sensory imagination, there is, in effect, one or more imaginative “algorithm” that constrains the development of subsequent stages in the imagining, after an initial “top-down” intention starts the process by determining the content of the first state in the sequence. The involvement of the initial intention accounts for the sense in which the content of the imagining was “chosen.” The algorithm in the case of propositional imagination is modeled on the inferential patterns operative on belief. The algorithm in the case of sensory imagination is realized partly by the operation of forward models, and partly by less well understood capacities for visual prediction. I turn in the next section to developing a challenge to this very general approach to explaining Guiding Chosen imaginings.

4.  The “Deviance” Objection The picture of GC imagination so far drawn gives rise to an immediate objection: imagination, even in its action and inference-guiding instances, seems to be less constrained than what is being allowed. Sure, there may be cases where we imagine more or less what we would come to believe (or perceive) if we believed (or perceived) the initiating content. However, even in the case of GC imaginings it seems we imagine things that deviate radically from anything we would likely infer from the initiating proposition (or would likely come to see, given the initial visual image and simulated motor command).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

72  Peter Langland-Hassan This is most clearly seen when we appreciate imagination’s role in guiding creative endeavors. The fact that the behavior in question is writing a novel or creating a sculpture (as opposed to planning a route home) does not lessen the need to provide an account of the operative constraints. Creative imagination is not random, after all. But neither is it so predictable as the drawing out of a set of inferences from a single premise. Consider also pretense: the plots of most pretenses do not stick strictly to the kinds of things one would infer given some pretense-initiating premise. Often pretenses move in and out of silly, absurd, and unpredictable premises. The ability to flexibly engage in such pretenses, from early childhood on, is one of the key capacities theorists invoke imagination to explain. So any account of imagination must have something to say on this score. And it bears emphasis that artistic creation and pretense are rational, purposeful actions that are driven by imagination; so these imaginings fall well within the scope of GC imaginings. Finally, there is much to be said for the idea that imagination allows us to audition a variety of ways things might go, in order to choose a best course of action. This would again point to GC imaginings having less than strict ­lateral constraints. We can call this the “deviance” objection, as it focuses on the capacity of useful imaginings to deviate from the kinds of constraints outlined in the previous section. The general point behind the deviance objection is that, even for action and inference-­ guiding GC imaginings, there is never just one way (or even just a handful of ways) that the imagining can unfold, given the way it begins. One way to deal with this is to appeal to what N&S call “scripts,” where these are “packets” or “clusters” of representations that “detail the way in which certain situations typically unfold” (2000, p. 126). These scripts can be conceived of as different sets of generalizations keyed to stereotypical contexts that, given an input, will output likely consequences within those contexts. One’s “scripts” are, in effect, proper parts of one’s lateral algorithm, as they partly determine what one will infer from what. While N&S limit their discussion of scripts to the kinds of situations we actually find ourselves in (e.g. dining at a fast food restaurant), the idea can be extended to include stereotypical narratives we encounter in fictions. For instance, I have acquired over the years a quiverful of clichéd ideas about the ways in which knights behave in duels. If a pretense requires me to behave like a brave knight, then the inferences I undertake in imagination may relate not to how I myself would act during a duel (run away!) but to how I would act if I were a knight of the sort portrayed in fairytales. If I believe some generalizations about how such knights behave, it will not be hard for me to draw out some inferences about how I would behave if  I were such a knight. So, often when an imagining seems outlandish—deviating from any plausible constraints that might otherwise hold—it may be because the initial premise specifies that the context is of a stereotypical, fictional sort. One is not inferring what one would actually do in that situation given one’s actual personality, but rather what one would do if one were a stereotypical character of the sort portrayed in certain fictional narratives. Obviously, generalizations drawn from the movies about how you would behave if you were a superhero, or a zombie, will not fruitfully

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

on choosing what to imagine  73 guide action in most everyday contexts. However, they can certainly be useful to guiding a pretense, and can potentially play a helpful motivational role during certain kinds of sincere action (e.g. imagining oneself in the Wimbledon Finals to inspire better play in a casual tennis match). That said, there are still bound to be GC imaginings that diverge from any such scripts or generalizations. Nichols and Stich themselves emphasize that “pretense is full of choices that are not dictated by the pretense premise, or by the scripts and background knowledge that the pretender brings to the pretense episode.” The same will of course be true of the imaginings driving artwork creation. There is no set of lateral constraints or scripts that will generate Moby Dick from “Call me Ishmael.” Noting that they “don't have a detailed account” of the cognitive mechanisms that underlie this flexibility, N&S posit a component of mental architecture called the “Script Elaborator,” whose job it is to “fill in those details of a pretense that can't be inferred from the pretense premise, the (altered) contents of the Belief Box, and the pretender's knowledge of what has happened earlier on in the pretense” (2000, p. 127).12 In essence, the job of the Script Elaborator is to account for whatever stages there may be in the sequence i1…in that would not have been inferred if i1 were believed. And while Nichols and Stich’s view is put forward only with respect to propositional imagination, one can easily see how a similar sort of mechanism would be needed to account for the cases of sensory imagination where what one imagines deviates from what one would expect to see, given the initial visual image and motor command (assuming one wished to appeal to something like the “forward model” constraints described above). While N&S admit that they “know little about how [the Script Elaborator] works” (p. 144), they see it as a virtue of their theory that it makes plain the need for such a mechanism or capacity. And, indeed, their positing of a Script Elaborator is the only serious acknowledgment of the problem that the deviance objection makes for any functionalist account of imagination (i.e. any account which seeks ceteris paribus generalizations about the way in which imaginings unfold across time). So, if the Script Elaborator represents a gap in their theory, it is equally a problem for everyone else who wishes to defend a broadly functionalist account of propositional or sensory imagination. Unfortunately, many theorists who have followed N&S in talking about imagination in functional or boxological terms simply out leave this necessary, if problematic, element (see e.g. Doggett and Egan (2007) and Schellenberg (2013)). As helpful as it is to acknowledge the need for something like a Script Elaborator, N&S still undersell the size of the hole it leaves in their account of propositional imagination (and, indeed, in anyone’s account who posits an “imagination box”). For it is arguably the central claim of their theory that imaginings are “belief-like” in the patterns of inference that govern their development. This is interesting if it is true, because 12   See also Weinberg and Meskin (2006a, 2006b) for accounts of propositional imagination that, following Nichols and Stich, posit a Script Elaborator.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

74  Peter Langland-Hassan we have an independent understanding of belief; its extension to imagination promises to shed explanatory light. However, to the extent that the Script Elaborator interferes with and augments those inferential patterns, imagination ceases to be (importantly) belief-like. This means that, until we have a reasonably fleshed out account of the Script Elaborator—one that tells us when and why it intervenes—we really do not know the extent to which imaginings are belief-like. We are left with a theory that says: imaginings are belief-like . . . except for the many ordinary circumstances in which they are not. We lack precisely the ceteris paribus psychological generalizations that are supposed to warrant talk of “boxes” in the first place. Thus, when N&S (and followers (e.g. Weinberg and Meskin 2006b)) draw separate boxes in their diagrams for the “imagination box” and the Script Elaborator, it is a diagrammatic sleight of hand. For it suggests that imagination is well understood (it is strongly belief-like) and that it is only the Script Elaborator that remains mysterious. But there really is no functional characterization of imagination to be had independent of a functional characterization of the Script Elaborator. This makes the willingness of others to simply omit the Script Elaborator from their imaginative boxologies all the more perplexing. Analogous points will apply for sensory imagination as well. If one posits a mechanism akin to the Script Elaborator to account for the cases where sensory imagination seems unconstrained by anything so pedestrian as a forward model or visuospatial generalization, we are left with a theory that says: one’s sensory imaginings will be constrained by what one would expect to perceive in thus and such circumstances . . . except for the many ordinary cases when they are not. This leaves us without a genuine functional account of sensory imagination. From the perspective of our puzzle concerning GC imaginings, the problem is this: until we understand which constraints (if any) govern the operation of the Script Elaborator (or comparable entity), we will not understand how it is that its operation generates useable, non-random results of the kind that reliably guide action and inference. Nor will we understand how its operation is compatible with a person’s choosing what they imagine (since it is not clear how the Script Elaborator interacts with one’s intentions). Thus, the deviance objection simultaneously threatens explanations of GC imaginings and accounts of imagination in general.

5.  Cyclical Processing Here is a way forward: let us suppose that, when our imaginings deviate from the patterns set out by the proposed lateral constraints, it is because we have intentionally intervened in that processing. To intentionally intervene is to stop the lateral processing where it is and to insert a new initial premise (or image) into the lateral algorithm for more processing. What we might pre-theoretically think of as a single imaginative episode could in fact involve many such top-down “interventions.” These interventions

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

on choosing what to imagine  75 would allow for the overall imagining to proceed in ways that stray from what would be generated if one never so intervened. Consider, as an example, Van Leeuwen’s (2011) improvisational actors who pretend to mount pterodactyls while dueling (he presents the case as a challenge to N&S-style accounts of the constraints on imagination). Actor A may start by imagining that “I am a brave knight at a duel.” This proposition is fed to the lateral algorithm which, we are assuming, is the same as that which governs inferential elaboration among ordinary beliefs. Actor A begins to infer (in imagination) things such as: “I am holding a sword . . . An opponent is trying to stab me . . . I hold one arm at my side.” These are the kinds of thing he would come to infer if he believed the initial premise. These imaginings begin to guide his action (perhaps, as N&S suggest, in the form of an inferred counterfactual belief: “If I were a brave knight at a duel, I would be holding a sword, an opponent would be trying to stab me”). At a certain point, this imaginative processing is interrupted by a desire for something more comedic to occur. Actor A decides that his riding a pterodactyl would be funny (more on this decision in a moment). This leads him to “intervene” on his prior imagining by feeding to the lateral algorithm: “I am dueling while riding a pterodactyl.” This allows him to draw some further imaginative inferences concerning things that might happen if he were somehow riding a pterodactyl. For one thing, pterodactyls fly. So he imaginatively infers that he is duel­ ing while flying on the back of a pterodactyl. Further interventions will likely follow. The idea is that, whenever an imagining diverges from the sort of thing that one would infer, or sensorily predict, from the currently imagined premises, this is because a new premise has been fed to the imaginative algorithm in a “top-down” manner, from one’s intentions. Guiding Chosen imagining, in its more freewheeling instances, then becomes a kind of cyclical activity, during which new and sometimes unusual premises are “fed” to a lateral algorithm at varying intervals. The output of the lateral “inferential” activity can then, at different intervals, be recombined with a novel element contributed by one’s intentions to begin the lateral processing anew (it is because of this recombination that I am calling the process “cyclical”). This allows the imaginative episode, as a whole, both to be constrained (by the lateral algorithm) and to freely diverge from anything one would have inferred from the initial premise alone. The mysterious work of the Script Elaborator has, in effect, been offloaded to one’s desires and intentions with respect to the task at hand. For it is one’s desires and intentions that will influence when and how one intervenes on the lateral processing. Why, then, did Actor A insert a premise having to do with pterodactyls, and not something else? Well, he wanted to shift the pretense to something more surprising, funny, and unusual—to something that would suit his goals, qua improvisational comedian. But why pterodactyls, in particular? Here the answer must trace to specifics of his psychology: what has he recently thought of or seen? What kinds of things does he generally find funny or surprising? Did someone mention dinosaurs earlier

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

76  Peter Langland-Hassan in the performance? The important point is that the answer will not involve positing a novel cognitive mechanism or process. The cyclical interaction of one’s top-down intentions with the constrained processing of the lateral algorithm is by itself enough to accomplish the work set out for the Script Elaborator. Coming to understand the work that N&S set aside for the Script Elaborator becomes part and parcel of understanding an agent’s goals, intentions, and decisions more generally. Developing an account of these is of course no mean feat; but at least we have a program for m ­ oving forward. One may object that, during such imaginings, we are not typically aware of having a  succession of intentions of this kind. However, I have already emphasized that the kinds of intentions being invoked are often intentions in action, of which we are usually only minimally conscious, if conscious at all. Just as we have a succession of intentions in action when getting a glass of water—to open a certain cabinet, to take a tall rather than short cup—without noticing them, so too may we have a succession during imagining without any real awareness of them.13 One might still point out that my arriving at a particular cabinet, in the case just described, will not surprise me, in the way that some GC imaginings can surprise us in where they lead. And this may seem to conflict with GC imaginings being under even this (minimally conscious) sort of intentional control. In response, surprise may come in the influence of the lateral algorithms themselves. They are what take the imagining beyond one’s intentions. And while they do indeed conform to our ordinary means of drawing inferences and, in a way, embody our expectations, we can nevertheless be surprised at the inferences we would draw from different unusual imagined premises. In the same way, one can be surprised at the answers generated by a familiar mathematical formula, when the variables are of an unusual sort. Summarizing where we now are, I have described and argued for three main features of GC imaginings that help explain how they can both be “chosen” and improve our epistemic standing. First, they are initiated by top-down intentions; second, they are developed in a constrained manner by lateral algorithms; third, through a process of continual and cyclical intervention by one’s intentions they can appear to deviate from those lateral algorithms, while nevertheless remaining reliable guides to inference and action. Of course there is much more that still needs to be understood concerning these features of imagination. The point in identifying them is to give a general sketch of how the paradoxical freedom and usefulness of imagination can best be explained, while suggesting where research should focus going forward. With respect to the latter goal, I will argue in the next section that understanding the nature of GC imaginings should give us the tools we need to understand most other imaginings as well. 13   And anyway, results from empirical psychology suggest that we have very little introspective insight into the nature of our own cognitive processes (Gigerenzer 1991, Girgerenzer et al. 1999, Johnson-Laird and Byrne 2002, Nisbett and Wilson 1977). Discovering what those inferential patterns and heuristics are is a project for empirical psychology, not introspection.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

on choosing what to imagine  77

6.  The Primacy of GC Imaginings The kind of cognitive architecture I have described to explain GC imaginings (involving top-down intentions, lateral algorithms, and cyclical processing) can be extended to explain a wide variety of other imaginings as well. I will divide the territory up into four mutually exclusive (and exhaustive) classes of imaginings: Guiding Chosen (our main topic), Guiding Unbidden, Misguiding Unbidden, and Misguiding Chosen. My goal in this section is to show that many, if not all, imaginings of each class can be assimilated to the general architecture already laid out for GC imaginings. This should help reinforce the idea that, in studying imagination in general, we would do well to maintain focus on GC imaginings.

6.1  Guiding unbidden imaginings Consider first “unbidden” imaginings—songs stuck in the head, obsessive negative imagery, and the like. They are not likely to be very useful, as a rule. I will consider such “misguiding” unbidden imaginings below. But first note that it is certainly possible that there would be cases where an unbidden imagining nevertheless ended up improving one’s epistemic standing. We could call these “Guiding Unbidden” imaginings. Is there a way of seeing these as drawing on the same architecture as GC imaginings? Given our account of Guiding Chosen imaginings, it is certainly possible that the kind of processing associated with the lateral imaginative constraints might sometimes be triggered by something other than a top-down intention (that is, by something other than its normal cause). The result would be an imagining that is constrained in the manner of GC imaginings, but that was not under intentional control. Provided the context nevertheless happened to be right for the imagining to play an action-guiding role, we can see how many Guiding Unbidden imaginings could simply be Guiding Chosen imaginings with non-standard causes. With that in mind we can consider another candidate Guiding Unbidden imagining. An idea recently (re)gaining steam is that imagination occurs during most of ordinary perception. Bence Nanay (2010) argues that the occurrence of mental imagery accounts for the “phenomenal presence” of the occluded parts of perceived objects (see also Macpherson (2012)). The basic idea is that, when seeing an object, a person automatically generates mental imagery that represents the occluded side of the object, and that this accounts for why occluded sides seem “phenomenally present.” Since the phenomenal presence of occluded sides of objects is not something over which we have intentional control, these putative exercises of imagination could be considered unbidden, yet also useful (assuming there is an epistemic use for phenomenal presence). It is possible that such imaginings (if they indeed occur) draw on the same predictive “forward models” discussed earlier with respect to GC sensory imaginings. For in predicting one’s sensory input given a particular motor action, one must typically represent a currently occluded aspect of the object—namely an aspect that one expects to come into view, given one’s planned movement. I have some reservations over whether

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

78  Peter Langland-Hassan this kind of processing ought to be considered a form of imagination (as opposed to an imagistic component of perception). But, however we settle that issue, we can see such Guiding Unbidden imaginings as closely related to GC imaginings at the cognitive architectural level.

6.2  Misguiding unbidden imaginings Still to be discussed, however, is the large class of imaginings that are not reliable guides to action or inference. We can call these “Misguiding” imaginings. These may have both chosen and unbidden instances. Consider first Misguiding Unbidden imaginings. These are token imaginings that are neither reliable guides to action and inference nor under intentional control. Unbidden and distracting sensory images are typical examples. No doubt the commonsense notion of imagination counts such phenomena as imaginings. But as we peer through the fog of platitudes about imagination and, squinting, make out the solid outlines of a more robust theoretical conception, we can ask whether all such imaginings should indeed be counted as exercises of the imagination. For unbidden misguiding imaginings are already borderline cases of imaginings, due to their not being under intentional control. One answer as to why Misguiding Unbidden imaginings should be counted as exercises of the imagination proper could be that they involve the use of the same basic architecture as Guiding Chosen imaginings. The lack of intentional control could be explained the same way as for Guiding Unbidden imaginings (that there is an unusual cause), while the explanation for why the imagining was not epistemically useful (given the lateral constraints involved) could be that it was irrelevant to—and even distracting from—the subject’s present situation and goals. Also possible is that there are some defects in the nature of the lateral algorithms that lead to bad inferences in some cases. If this is our explication of Misguiding Unbidden imaginings, then we are again in a situation where the architecture described for Guiding Chosen imaginings is that in terms of which other mental acts can be understood as imaginings as well. Of course, this does nothing to rule out the possibility of Misguiding Unbidden imaginings that have no close connection to GC imaginings. However, it is worth emphasizing that any putative imagining of that kind faces the question of why the imagistic (or propositional) cognition in question should be construed as an exercise of the imagination, and not something else (such as cognitive noise). For the use of sensory imagery in a cognitive act is not generally considered sufficient for imagination.14 And uncontrollable and epistemically fruitless propositional thought is not ipso facto propositional imagining. Why then should imagistic cognition with the same traits necessarily qualify? As we seek a tighter account of the cognitive basis of imagination, we should be prepared to leave such phenomena behind. In 14   This much is accepted even by some (e.g. Kind 2001, p. 100) who hold that sensory imagery is a necessary component of imagination.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

on choosing what to imagine  79 many cases (such as with unbidden and distracting sensory imagery) this can be done without completely fracturing the picture of imagination as it exists within folk psychology.

6.3  Misguiding chosen imaginings Last to be discussed are Misguiding Chosen imaginings. On the face of it, these are more centrally related to the imagination than Guiding Unbidden and Misguiding Unbidden imaginings, and so more difficult to simply assimilate to the Guiding Chosen architecture. Misguiding Chosen imaginings—under intentional control, but ill-suited to guide action and inference—include much of ordinary daydreaming and fantasy, which are paradigmatic instances of imagination as it is normally conceived. However, given the understanding of Guiding Chosen imaginings developed above, it is clear that the underlying architecture of GC imaginings can itself be at work in much of the cognition we associate with daydreaming and fantasy. Many Misguiding Chosen imaginings will not be useful or action-guiding for simple contextual reasons, and not because they employ fundamentally different cognitive capacities than GC imaginings. It may just be that their subject-matter—for example, what one would say upon winning an Oscar—is irrelevant to the attainment of any present goals, or to the real­ ization of any nearby possibilities. Also, it could be that the imagining is not guiding because, while it involves use of strict lateral algorithms, one is misguided due to learning errors or over-simplifications in the algorithms themselves (thus one might use an algorithm grounded in “folk physics” to wrongly predict the relative rates at which a feather and bowling ball will fall in a vacuum). Still, they would involve the same basic architecture as Guiding Chosen imaginings. The most resistance to assimilating Misguiding Chosen imaginings to the (constrained) architecture for GC imaginings will likely come from considering imaginings that seem ill-suited to guiding action or inference in any context. But even here the cognitive difference between the two may often be only skin deep. To see this, consider the following Misguiding Chosen imagining, which I take to be paradigmatic: I imagine that I drop a glass and that, as it hits the floor, it shatters; the shards then meld together into the shape of a bird, which flies away. Fantastical though it is, most of the imagining is tightly constrained: the way the glass both falls and breaks, and the way the bird is shaped and flies are all determined by lateral constraints that are grounded in background beliefs about the relevant kinds of objects. What about the moment where the glass shards turn into a bird? This moment in the imaginative project can be accounted for by a top-down intervention by a new intention in action—specifically, an intention to imagine what would happen if the glass shards turned into a birdlike creature. This intervention is not a conscious choice, necessarily. But it is still a topdown choice, in the way that everyday actions, such as unlocking one’s car door, are choices. Some may still feel that the account leaves Misguiding Chosen imaginings too tightly constrained. However, it is hard to see what a better account of the glass-to-bird

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

80  Peter Langland-Hassan imagining would be that still renders it a Misguiding Chosen imagining. For suppose it is not a top-down intention that intervenes on lateral algorithms at the moment the glass shards turn into a bird-like creature. What then happens instead? If no choice was made for the shards to turn into a bird, and if that change was not dictated by any lateral algorithm, then it would seem that the change is unbidden and, for all intents and purposes, random. But it cannot be either unbidden or random and remain a Misguiding Chosen imagining. It would become instead a Misguiding Unbidden imagining, and these have already been discussed. Many Misguiding Unbidden imaginings are just GC imaginings with non-standard causes. Those that are not may slip into the realm of cognitive noise. What if instead we dial back on the degree of control we have over a Misguiding Chosen imagining, choosing only the first imaginative state in the sequence? I do not see any reason to deny that we have a capacity to initiate a random sequence of mental states (though I’m not sure I can do it). However, it would be a mistake to consider such a capacity to be a cognitive faculty (the faculty of imagination). This sort of capacity (supposing we have it) is not something that explains crucial cognitive differences between humans and other animals, or that could be relied upon to guide pretense, facilitate aesthetic understanding, plan actions, understand other minds, reveal nearby possible worlds, or perform any of the other activities associated with imagination. And, again, if random sequences of propositional thoughts (where the content of the first is chosen) do not count as exercises of the propositional imagination, it is hard to see why random sequences of mental images should constitute examples of sensory imagination. In both cases we have something like cognitive noise. Cognitive noise no doubt occurs. But a philosophical, psychological, and cognitive-scientific theory of imagination should not also aim to be a theory of cognitive noise. The last possibility to consider is that some Misguiding Chosen imaginings are subject to lateral constraints of a kind, such that to call them “noise” would be unfair, but where the constraints are not of the (broadly rational) kinds described with respect to Guiding Chosen imaginings. This remains a possibility. One might think that extremely broad constraints, such as logical or conceptual possibility, govern the unfolding of some imaginings. This would make them quite free, but not yet random. Yet note first that this would not enable a kind of imagining where subjects have more control over their imaginings than they do for GC imaginings. Nor would it enable imaginings that were better suited than GC imaginings to explain how imagining is useful to the kinds of acts in which imagination is typically implicated. Its principal attraction would seem to be that it might satisfy the intuition that genuine imagining can take place that is not subject to strong constraints, yet which is also not randomly generated—mere cognitive noise. But the kinds of constraints being considered—logical or conceptual possibility—do not take us far past randomness. There is virtually no limit to the propositions that are logically or conceptually compatible with an arbitrary initiating premise. So it will again be hard to see why a capacity to generate a

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

on choosing what to imagine  81 sequence of ­contentful states so “constrained” should be viewed as an important cognitive faculty, as opposed to cognitive noise. The upshot of all this is that once we understand the basic cognitive architecture required for Guiding Chosen imaginings—in a way that explains how they are both chosen and epistemically advantageous—we have the tools needed to understand many other kinds of imaginings as well. And, if our theoretical conception of imagination may ultimately exclude some phenomena that get lumped in with imagination at the level of commonsense psychology, a theory of the cognitive underpinnings of GC imaginings may form the core of a theory of imagination tout court.

7. Conclusion It is natural to despair at ever establishing a comprehensive theory of imagination. As Walton lamented in his seminal work on make-believe, “it is not easy to see what behavioral criteria might throw light on imagining, or what the relevant functions of a functional account might be” (1990, p. 20). For Walton, the term “imagining” simply served “as a placeholder for a notion yet to be fully clarified” (p. 21). A decade or so later, Nichols and Stich took up the challenge of providing the needed functional account, using pretense to generate a set of behavioral criteria that might “throw light on” imagining. As discussed above, large questions remained open in their account concerning the capacity of imaginings to deviate from ordinary belieflike patterns of inference, and concerning the (crucial) role of mental imagery in imagination. In particular, no clear details were given concerning how imaginings could, as a matter of course, deviate from belief-like patterns of inference and still be reliable guides to action and inference. Their “Script Elaborator” was a means of marking the problem, still to be addressed. By taking the simultaneous usefulness and “chosen” nature of Guiding Chosen imaginings as an explanandum, I have outlined what I see as the best general framework for moving forward in the quest to understand imagination in functional terms and, indeed, as a faculty of mind. I have argued that a theory of Guiding Chosen imaginings will give us the basic pieces we need to understand imagination in general. The three general features of this architecture are: 1) top-down intentions that initiate an imagining, 2) lateral constraints that govern the development of the imagining, and 3) the possibility of cyclical interventions by one’s intentions during a single imaginative episode. Whether it is propositional or sensory imagination, one begins with a stipulated content and draws out a variety of inferences, shaped by one’s background beliefs and sensorimotor expectations. Where an imaginative episode seems to deviate from such constraints, it is due to ongoing subconscious interventions on the part of one’s intentions to consider new premises. Seen in this light, imagining is a form—perhaps the central form—of conditional reasoning. The development of a functional and computational account of conditional

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

82  Peter Langland-Hassan reasoning remains an important and contentious project within empirical psychology (Byrne 2005, Johnson-Laird and Byrne 2002, Over and Evans 2003). Part of my aim has been to show how those theories might themselves be filling in the crucial details of a philosophical theory of imagination. For that link can only be made by first reconciling the freedom of imagination with its constraints.

References Addis, Donna Rose, Ling Pan, Mai-Anh Vu, Noa Laiser, and Daniel L. Schacter (2009). Constructive episodic simulation of the future and the past: Distinct subsystems of a core brain network mediate imagining and remembering. Neuropsychologia 47(11): 2222–38. Arp, Robert (2008). Scenario Visualization: An Evolutionary Account of Creative Problem Solving. Cambridge, MA: MIT Press. Blakemore, Sarah-Jane, Daniel M. Wolpert, and Christopher D. Frith (2002). Abnormalities in the awareness of action. Trends in Cognitive Sciences 6(6): 237–41. Byrne, Ruth M. J. (2005). The Rational Imagination: How People Create Alternatives to Reality. Cambridge, MA: MIT Press. Chalmers, David J. (2002). Does conceivability entail possibility? In Tamar Szabό Gendler and  John Hawthorne (eds), Conceivability and Possibility. Oxford: Clarendon Press, pp. 145–200. Churchland, Patricia Smith, and Terrence J. Sejnowski (1989). Neural representation and neural computation. In Lynn Nadel, Lynn A. Cooper, Peter Culicover, and R. Michael Harnish (eds), Neural Connections, Mental Computations. Cambridge, MA: MIT Press, pp. 15–48. Currie, Gregory, and Ian Ravenscroft (2002). Recreative Minds: Imagination in Philosophy and Psychology. Oxford: Oxford University Press. Doggett, Tyler, and Andy Egan (2007). Wanting things you don’t want: The case for an imaginative analogue of desire. Philosophers’ Imprint 7(9): 1–17. Doggett, Tyler, and Andy Egan (2012). How we feel about terrible, non-existent Mafiosi. Philosophy and Phenomenological Research 84(2): 277–306. Fodor, Jerry A. (1975). The Language of Thought. New York: Crowell. Fodor, Jerry A. (2003). Hume Variations. Oxford: Oxford University Press. Gigerenzer, Gerd (1991). How to make cognitive illusions disappear: Beyond “heuristics and biases”. European Review of Social Psychology 2(1): 83–115. Gigerenzer, Gerd, Peter M. Todd, and the ABC Research Group (1999). Simple Heuristics that Make Us Smart. New York: Oxford University Press. Goldman, Alvin I. (2006). Simulating Minds: The Philosophy, Psychology, and Neuroscience of Mindreading. New York: Oxford University Press. Grush, Rick (2004). The emulation theory of representation: Motor control, imagery, and perception. Behavioral and Brain Sciences 27(3): 377–442. Holst, Erich von, and Horst Mittelstaedt (1973). The reafference principle (Interaction between the central nervous system and the periphery). In The Behavioral Physiology of Animals and  Man: The Collected Papers of Erich von Holst, vol. 1 (trans. Robert Martin). Coral Gables, FL: University of Miami Press, pp. 139–73. [Original paper published 1950 as Das

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

on choosing what to imagine  83 Reafferenzprincip (Wechselwirkungen zwischen Zentralnervensystem und Peripherie). Die Naturwissenschaften 37: 464–76.] Johnson-Laird, P. N., and Ruth M. J. Byrne (2002). Conditionals: A theory of meaning, pragmatics, and inference. Psychological Review 109(4): 646–78. Kind, Amy (2001). Putting the image back in imagination. Philosophy and Phenomenological Research 62(1): 85–109. Kind, Amy (in press). How imagination gives rise to knowledge. In Fabian Dorsch and Fiona Macpherson (eds), Perceptual Memory and Perceptual Imagination. Oxford: Oxford University Press. Kosslyn, Stephen M., William L. Thompson, and Giorgio Ganis (2006). The Case for Mental Imagery. New York: Oxford University Press. Kung, Peter (2010). Imagining as a guide to possibility. Philosophy and Phenomenological Research 81(3): 620–63. Langland-Hassan, Peter (2011). A puzzle about visualization. Phenomenology and the Cognitive Sciences 10(2): 145–73. Langland-Hassan, Peter (2012). Pretense, imagination, and belief: The single attitude theory. Philosophical Studies 159(2): 155–79. Liao, Shen-yi, and Tyler Doggett (2014). The imagination box. Journal of Philosophy 111(5): 259–75. McGinn, Colin (2004). Mindsight: Image, Dream, Meaning. Cambridge, MA: Harvard University Press. Macpherson, Fiona (2012). Cognitive penetration of colour experience: Rethinking the issue in light of an indirect mechanism. Philosophy and Phenomenological Research 84(1): 24–62. Miall, R. C., D. J. Weir, D. M. Wolpert, and R. C. Stein (1993). Is the cerebellum a Smith Predictor? Journal of Motor Behavior 25(3): 203–16. Nanay, Bence (2010). Perception and imagination: Amodal perception as mental imagery. Philosophical Studies 150(2): 239–54. Nichols, Shaun (2004). Imagining and believing: The promise of a single code. Journal of Aesthetics and Art Criticism 62(2): 129–39. Nichols, Shaun (ed.) (2006). The Architecture of Imagination: New Essays on Pretence, Possibility, and Fiction. Oxford: Oxford University Press. Nichols, Shaun, and Stephen Stich (2000). A cognitive theory of pretense. Cognition 74(2): 115–47. Nisbett, Richard E., and Timothy D. Wilson (1977). The halo effect: Evidence for unconscious alteration of judgments. Journal of Personality and Social Psychology 35(4): 250–6. Noordhof, Paul (2002). Imagining objects and imagining experiences. Mind and Language 17(4): 426–55. Over, David E., and Jonathan St. B. T. Evans (2003). The probability of conditionals: The psychological evidence. Mind and Language 18(4): 340–58. Prinz, Jesse J. (2002). Furnishing the Mind: Concepts and Their Perceptual Basis. Cambridge, MA: MIT Press. Schacter, Daniel L., Donna Rose Addis, and Randy L. Buckner (2007). Remembering the past to imagine the future: The prospective brain. Nature Reviews Neuroscience 8(9): 657–61. Schellenberg, Susanna (2013). Belief and desire in imagination and immersion. Journal of Philosophy 110(9): 497–517.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

84  Peter Langland-Hassan Searle, John R. (1983). Intentionality: An Essay in the Philosophy of Mind. Cambridge: Cambridge University Press. Sperry, R. W. (1950). Neural basis of the spontaneous optokinetic response produced by visual inversion. Journal of Comparative and Physiological Psychology 43(6): 482–9. Van Leeuwen, Neil (2011). Imagination is where the action is. Journal of Philosophy 108(2): 55–77. Walton, Kendall L. (1990). Mimesis as Make-Believe: On the Foundations of the Representational Arts. Cambridge, MA: Harvard University Press. Weinberg, Jonathan M., and Aaron Meskin (2006a). Imagine that! In M. Kieran (ed.), Contemporary Debates in Aesthetics and the Philosophy of Art. Oxford: Blackwell, pp. 222–35. Weinberg, Jonathan M., and Aaron Meskin (2006b). Puzzling over the imagination: Philosophical problems, architectural solutions. In Shaun Nichols (ed.), The Architecture of the Imagination: New Essays on Pretence, Possibility, and Fiction. Oxford: Oxford University Press, pp. 175–204. Wolpert, Daniel M., Zoubin Ghahramani, and Michael I. Jordan (1995). An internal model for sensorimotor integration. Science 269(5232): 1880–2. Wolpert, Daniel M., R. Chris Miall, and Mitsuo Kawato (1998). Internal models in the cerebellum. Trends in Cognitive Sciences 2(9): 338–47. Yablo, Stephen (1993). Is conceivability a guide to possibility? Philosophy and Phenomenological Research 53(1): 1–42.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

3 The Imaginative Agent Neil Van Leeuwen

Jean Valjean, in Victor Hugo’s Les Misérables, has just been sent to prison for stealing a loaf of bread to feed his sister’s seven children. They are poor, and he’ll never help them again. At this point, we find a remarkable passage: The tears choked his words, and he only managed to say from time to time, “I was a pruner at Faverolles.” Then still sobbing, he raised his right hand and lowered it seven times, as if touching seven heads of unequal height, and from this gesture one could guess that whatever he had done had been to feed and clothe seven little children. (pp. 84–5)

Why was Valjean crying? The children were neither in front of him, nor would he see them again. And why do we feel such poignant sentiment on reading this passage? We know Valjean does not exist, yet we cry for him more than for many people who do. He imagines the children; we imagine him. But now suppose Valjean is a real person with a real human psychology. I want to argue the following. First, the psychological pathway that leads Valjean to cry is of the same kind as the psychological pathway that leads readers to cry on reading this passage. Valjean, the agent, and you, the reader of fiction, share something. Second, that pathway from imaginative state to emotional response enables three central agential capacities: 1. Bodily preparedness for potential events in the nearby environment. 2. Evaluation of potential future actions. 3. Empathy-based moral appraisal. Furthermore, the pathway in question facilitates communication of emotion to other agents in a way that enables them to respond appropriately. In short, this pathway partially constitutes human agency. But it makes us lovers of fiction too. My argument for this view works like this. First, after making some distinctions that will prove useful (Section 1), I posit a pathway from imagistic imagining to emotional activation. My defense of this pathway, which I call I-C-E-C for reasons that will become clear, proceeds via consideration of evidence from psychology and cognitive

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

86  Neil Van Leeuwen neuroscience (Section 2). Second, the evidence I consider shows that this pathway has three important properties: automaticity, reality congruence, and integratability1 with perception (Section 3). Third, I argue that these very properties enable the three agential capacities in question (Section 4). Fourth, I argue that I-C-E-C supports engagement with works of fiction and conclude that this view sheds light both on the so-called paradox of fiction and on imaginative resistance (Section 5). In thinking about agency, philosophers—broadly, not universally—have too much ignored imagination. Imagination not only represents the possibilities from which we choose; it facilitates appraisal of possibilities by sparking relevant emotions. At the same time, in thinking about love of fiction, we have not realized that that love stems from pathways and capacities that also—primarily—enable action in reality. These two oversights—one in action theory and the other in aesthetics—are complementary. I seek to eliminate both at once.

1.  What Do We Mean by “Imagine”? When we say someone imagines something, we imply they have mental representations of things beyond what’s currently available to perception. But despite this loose coherence, there are orthogonal uses of “imagine” to distinguish. Suppose you “imagine” that a smiley face is looking at you.

Let this diagram schematize the mental states and processes involved in this imagining. We’ll see that the combinatorial structure of the diagram maps to the combinatorial structure of different imaginative mental kinds. First, you have a constructive process of coming up with the representation; the arrow symbolizes this process. Second, since you don’t actually believe there is a smiley face looking at you, you have an attitude besides belief toward the content you represent; the brackets represent this attitude, which may be “fictional imagining,” though there are other varieties of attitude imagining as well. Third, your mental representation has a certain format, which in this case is imagistic—structured like percepts (often also called “depictive”); the smiley face stands for the imagistic format.2 1   Pardon the neologism. “Integrability” already exists, but it has a specific mathematical sense. So rather than mislead, I prefer a new word to express what I need. 2   It should be clear from how I set up the terms that not all imagistic imagining (or “imagery”) has to be conscious; in fact, I think much of it isn’t, as Nanay (2013) argues in detail.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

The Imaginative Agent  87 The English word “imagine” can pick out the process, the attitude, the representational format, or any one or two or three out of the three. So to be precise about which notion of imagining I’m deploying in a given context, I introduce the terms constructive imagining, attitude imagining, and imagistic imagining. Importantly, these three things can come together or come apart. You might actually believe a smiley face is looking at you. In that case, you would still have the constructive imagining (arrow) and imagistic imagining (smiley face), without having an imaginative attitude (subtract the brackets); you would have a belief attitude toward the same content instead (one might say, “your imagination is playing tricks on you”). Alternately, you could visually remember something you saw and believe you saw it; then you’d have just the smiley face, or perhaps the arrow and the smiley face, if remembering involves constructive imagination (I won’t judge that issue here). Or you could imagine discursively, in an abstract symbolic code (without visual imagery), that a smiley face is looking at you, because imaginative thoughts needn’t always involve sensory imagery.3 In this case, we’d put a sentence or formula instead of an image inside the brackets. Or if you look at a picture of a smiley face and close your eyes, you needn’t engage in ­constructive imagining to come up with the imagery you take an imaginative attitude toward; in that case, subtract the arrow but leave the brackets and the smiley face. Here’s why these distinctions are important. Imagistic imagining is incorporated into much mental activity that has nothing to do with fiction, as Williamson (Chapter 4, this volume) emphasizes. Much imagistic imagining is of what has happened or of what may be about to happen in the actual world. But in order for imagery to support human agency usefully, imagistic imagining must yield emotions automatically, whether one believes their contents or merely has an imaginative attitude toward them. If the imagined potential black widow under the bed is to scare us—a fright that is ecologic­ ally valuable—the spider from the movie we just saw must do so too. If we conflate the different senses of “imagine” and wrongly think that all imagery is fictionally regarded, we’ll miss the crucial point that imagery is an area of overlap between fictional and everyday processing.4

2.  From Imagery to Emotion Seeing a lion, smelling a perfume an ex-lover used to wear, hearing your grandmother’s voice—all these percepts excite emotions. Each causes fear, nostalgia, or affection quickly, before you’ve thought about why they cause these emotions. For ecological 3   Discursive imaginings form a proper subset of propositional imaginings, since some propositional imaginings deploy mental imagery as part of their constituent structure, as I argue elsewhere (Van Leeuwen 2013). 4   Another good reason to recognize these distinctions is that doing so is prerequisite to making sense of the title of this volume: Knowledge Through Imagination. If we conflated the senses of imagining, we might wrongly think that everything produced by imagination is fictional, in which case the notion of attaining knowledge through imagination would make no sense.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

88  Neil Van Leeuwen reasons, percepts spark emotions before you have time to form beliefs about what you perceive. Nor are such emotional reactions optional. I couldn’t decide to feel anything but fear on seeing a lion in the wild, no matter how hard I might try. So there are pathways from perception to emotion, which we’ll explore in more detail. But this is already enough to state the view I endorse about imagery and emotion: imagery has a representational format that is like perception (across all modalities), and imagery is processed in the perceptual cortices of the brain; so imagistic imagining activates emotions along the same pathways as perception. This pathway from imagining to emotion has been posited before by Timothy Schroeder and Carl Matheson (2006) and by me (Van Leeuwen 2011). Here I add detail, acknowledge complexity, and connect the view to imaginative resistance and emotional responses to fiction. Let me now sketch relevant empirical facts. Visual mental imagery activates the brain’s visual cortices in the occipital lobe. Auditory imagery—“hearing” in one’s “inner ear”—activates the auditory cortex in the temporal lobes. Imagining faces activates the fusiform face area, which processes perceptual face recognition. Motor imagining, like imagining turning a knob with your hand, activates the motor cortex in ways that resemble activation for actual bodily movement. Finally, visual imagining also activates the sensory thalamus, which is active early in visual processing.5 So our brains have a neat trick: a capacity to use sensory and motor areas not just for processing present, actual perception, but also for representing non-present (often non-actual) motions and objects of perception. This capacity is instantiated just about everywhere it could be in the human brain: a shining example of nature’s tendency to give individual structures multiple functions. Thus, we can understand how the brain relates imagistic imagining to emotion by understanding in more detail how it relates perception to emotion. There are relatively direct connections between perceptual and emotional areas of the brain. Let’s focus on vision and fear for now. Joseph LeDoux (1996) identifies in the mammalian brain two pathways from perceptual areas to the amygdalae. These pathways, the “low road” and the “high road,” are largely responsible for preparing fear responses, such as freezing, fighting, or fleeing. The low road is rapid, but imprecise, casting a broad net that responds to almost anything that resembles a threatening stimulus; the low road is responsible for your jolt of fright on seeing a snake-like stick in the grass. It originates at the sensory thalamus and thus is triggered by perceptual processing that (i) is closer to the perceptual organs themselves and (ii) is likely prior to conscious awareness of a stimulus. Thus, visual input of a snake shape is apt to initiate a fear response before we’re conscious of it. The low road is especially responsive to fear conditioning, and once an organism acquires a low-road fear association, that association is hard to extinguish and returns easily.   Kosslyn, Thompson, and Ganis (2006) summarize this research.

5

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

The Imaginative Agent  89 The high road incorporates more sophisticated cortical processing and is slower but more precise: it is far better at distinguishing actually threatening stimuli, like a hawk, from stimuli that merely resemble threats, like a flying squirrel that briefly casts a similar shadow. To continue along the high road, we should note that visual processing divides into two streams, commonly called “where” and “what.”6 The “where” (dorsal) stream feeds into the parietal lobe and processes locations of objects. The “what” (ventral) pathway categorizes objects into kinds—a ball, a snake, a tree, etc. The “what” processing, which seems central to conscious visual processing of objects7 and ultimately to emotional response, happens largely in the inferior temporal lobe (IT). The IT, in turn, projects to the amygdalae and other emotional areas. I believe the “what” pathway in the first instance categorizes perceptual stimuli, before it conceptualizes them: there is a stage of representation that groups heterogeneous percepts into one category, where the unity of the category consists in common emotional and behavioral response dispositions. For example, visual percepts of a cat from different angles have very different geometric content, but the “what” pathway still groups those disparate percepts under the same category of CAT. Let’s call this stage of representation primary categorization. Primary categorization is different from conceptualization. The latter type of representation deploys concepts that can be constituents in compositional thought and rational inference (Evans 1982, p. 75); the former feeds primarily into emotion and relatively automatic behavioral dispositions. (There are various forms of causal influence between categorizations and conceptualizations, but that doesn’t show they’re not distinct.) The primary categorization of an object you see will typically be the first thing the object strikes you as being.8 You see dogs first as dogs and only later as particular kinds, like beagles or Great Danes, or as more general types, like carnivore (Rosch et al. 1976). In my view, the primary categorization of an object most heavily influences the high road’s impact on emotions, in comparison with more rational conceptualization. You can have a SNAKE primary categorization of an object thrown at you, even if you conceptualize it as a rubber toy. Once an emotion occurs, however, one has what Paul Ekman (2007) calls a “refractory period,” during which cognitive processes, like attention and conceptual thought, favor processing information that coheres with that emotion: fear causes one in the refractory period to think of and attend to other fearful things; joy—joyful things; and so on. So conceptualizations of stimuli are temporally subsequent to primary categorizations and emotional responses, even along the high road. Nevertheless, conceptualizations—what you’re   There are alternate construals of the two pathways, but the differences don’t matter for my project.   See, for example, Leopold and Logothetis (1996). 8   You may be inclined to think that the more sophisticated high road bypasses primary categorization and goes straight to conceptualization. But the high road is a structure we share with many simpler creatures, like rats, so it is likely that it employs structures, like primary categorizations, that they have as well. 6 7

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

90  Neil Van Leeuwen thinking in terms of—prime the perceptual system to categorize incoming ambiguous percepts one way or another. In sum, the high-road functional neuropsychology from vision to fear works something like this. (1) Early stage perceptual processing leads to a (2) primary categorization, which in turn triggers a (3) fear response (if the category is a feared one). That fear response (4) focuses attention and thought on further fearful stimuli. This continues until the fearful situation is past. I call this pathway P-C-E-C: percept-­categorization-emotion-conceptualization. Fear is just one emotion, and vision is just one sense modality; we should be careful when generalizing. Nevertheless, let’s assume for now that something like P-C-E-C connects other sense modalities to emotions too: emotions like pity, anger, joy, shame, and disgust, which may also have something like low roads of their own (Zajonc 1984). This assumption, which can be modified as needed, allows us to develop a coherent view of how links between imaginings and emotion contribute to agency.9 It is an explanatory hypothesis. I propose that the pathway from imagistic imaginings to emotions parallels the high road(s) almost entirely. Instead of P-C-E-C, we have I-C-E-C: imagery-categorization-­ emotion-conceptualization. Imagistic imaginings in the sensory cortices generate “what” categorizations in the ventral stream, which are largely responsible for activating emotional areas.10 Subsequent conceptualizations enable inferences and prime further primary categorizations.11 An animal shadow in the bush receives a lion primary categorization, if one has been thinking about lions, but a deer primary categorization, if one has been thinking about deer. There may also be a pathway from imagistic imagining to emotion that parallels the low road. That’s because, as noted, the sensory thalamus is activated during at least visual imagery (Ganis, Thompson, and Kosslyn 2004), and the sensory thalamus is the origin of the low road, directly activating the amygdalae.

3.  I-C-E-C: Principles and Properties To grasp the importance of I-C-E-C, try this. First, just discursively imagine—propositionally and without imagery—that a boy failed his test. Second, visually imagine him receiving his test back, with despair creeping onto his face, as he hides the grade from other students. For me, the second step, not the first, brings pity. And more imagery yields deeper emotion. Imagine the sound of the boy hurriedly crumpling the test into 9   In any case, Schroeder and Matheson (2006) give some reason to think the generalization is a good one. They point out that structures that produce multi-modal representations—representations that combine information from more than one sense modality—send signals not only to the amygdalae, but also to the orbital frontal cortex and the affective division of the striatum, which help distinguish rewards from punishments and produce visceral emotional responses as well. 10   Olson, Plotzker, and Ezzyat (2007) is relevant here. 11   See Gendler on priming (2003, pp. 134–6).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

The Imaginative Agent  91 his bag. Imagine his wet tears, when he lets out his frustration in the bathroom stall. The emotional impact of bare discursive imagining is faint and lacking in nuance compared with that of imagistic imagining. I-C-E-C operates according to two psychological principles that will be important in what follows: Percept Similarity: the emotional impact of imagistic imaginings is similar to that of percepts with the same representational structure: if a perceptual episode of the form P(r1, …, rn) contributes E to one’s overall emotional state, where r1, …, rn are representational constituents of the percept, then imagistic imagining of the form I(r1, …, rn) will typically contribute Eʹ, which is similar to E in that it includes the same emotions, though with possibly different intensities.12 Conceptual Modulation: discursive imaginings modulate the emotional impact of imaginative episodes by conceptualizing imagistic imaginings: i) If mental image i is conceptualized as entity e and e is conceptualized as having property F, the imaginative episode will be modulated, relative to mere imagistic imagining with i, with emotions similar to those that would typically accompany the occurrent belief that e is F. ii) Discursive imagining that p tends to cause imagistic imaginings that represent constituents of p, which cause emotions as specified in Percept Similarity.

To exemplify Percept Similarity, visually imagine/recollect fireworks on a night long ago with a particular lover. This imagistic imagining engenders some of the same romantic excitement as before. Percept Similarity holds. What about non-recollection cases and cases of other emotions? Visually imagining a child with a skinned knee, however fictional, and imagining the sound of his cries engenders some of the same pity as when one actually sees and hears such things, even if the pity is fainter. We can multiply such examples. See/imagine a man robbing a blind man: anger. Hear/imagine acoustically a sinister voice threatening to slit your throat: fear. Feel/imagine touching your favorite dog’s fur: affection. Furthermore, an important study by Lang et al. (1983) lends support to Percept Similarity. They show that idiosyncratic phobias that are triggered by percepts are also triggered by imagery with the same content, and imagery of objects of a phobia one does not have does not trigger fear. So there is a mapping between the percepts and imagistic imaginings that cause fear, as Percept Similarity predicts. 12   Humor, especially physical humor, seems to be an exception. Sometimes we laugh at things in imagined fictions that would frighten us or make us sad, were they real. Doesn’t this contradict Percept Similarity? No. The solution here is that humor is very often a release from negative emotions, like fear. So Percept Similarity is upheld, insofar as the negative emotions that get released in humor are still there and not far below the surface. Joëlle Proust (in conversation) has pointed out to me another interesting qualifier on Percept Similarity: emotions consequent on imagery are stronger when that imagery arises from external promptings than when it arises from an internal intention to have the imagery. This, however, won’t make a difference to my argument, since what matters for me is that the valences of the emotions consequent on the imagery are the same in both varieties of imagistic imagining (externally vs. internally prompted).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

92  Neil Van Leeuwen The natural objection to Percept Similarity is this. Often when one recollects an event imagistically, different emotions arise from those had when experiencing the event. A touch of sad longing accompanies sensory imagining of the fireworks and lover, which wasn’t there initially. But this isn’t a problem, since Percept Similarity does not claim there will be no other emotions during imagistic imagining from those that would come (or came) from analogous percepts; it entails only that the imagery contributes the same emotions (with perhaps different intensities) to one’s overall emotional state. There may be many background differences. The imagistic elements of the romantic imagining still engender feelings of romantic excitement, but the conceptualization of the recollected events as gone forever modulates that feeling with sad longing. Thus, far from undermining Percept Similarity, this is a case of Conceptual Modulation.13 So let’s turn to Conceptual Modulation. Visualize a friend standing face-to-face with a lion. Now discursively imagine that the lion is wild. Then imagine that the lion was raised in captivity. The same visualization yields different emotions, depending on the discursive imagining accompanying it: fright in the former case and slight relief in the latter. This will be so, I think, regardless of whether one also imagistically imagines the lion’s history. Again, cases can be multiplied. Imagine, with as much imagery as you like, a man dying. Now imagine, discursively, that that man is a father. This later imagining adds a tragic pity to the initial sadness. The neuroscience discussed in the last section, furthermore, coheres with these principles. Percept Similarity coheres with the fact that imagery plays out on the perceptual cortices of the brain. If seeing a certain face yields an emotion, and if visualizing that same face receives much the same processing as seeing it, consequent emotions will be largely the same. The same inference holds, mutatis mutandis, for other sense modalities and emotions. Next, Conceptual Modulation coheres with the fact that conceptual thought can prime primary categorizations, and primary categorizations are a major determinant of how percepts and imagery impact emotions. So, in the case of imagining the dying man, thinking of him as a father may yield a FATHER, as opposed to a MAN, primary categorization, which modifies downstream emotional processing. Again, the same holds, mutatis mutandis, for other conceptualizations, primary categorizations, and emotions.14 13   Amy Kind (personal communication) has suggested a particularly strong version of this objection. Suppose the ex-lover is now hated. Would there be any romantic sentiment at all? Percept Similarity seems to require that there at least be some, and it is plausible that in such a case there is none. I’m inclined to say that the emotions here would still be mixed, including some romantic nostalgia, at least for most people. In fact, it is the presence of continued romantic sentiment that makes an ex-lover’s sins so hurtful. But in extreme cases, to appeal to the last section, the primary categorization that ensues from imagery of the ex-lover may be such as to preclude any romantic feeling. For example, if the primary categorization was no longer LOVER but MURDERER, this would be an exception to Percept Similarity. But this can be seen as an exception to a principle that does in fact govern the normal operation of I-C-E-C. 14   You might wonder whether the notion of a primary categorization makes sense in the context of imagining. Don’t we start imaginings by thinking about stuff, in which case it’s conceptualization first? It may be true that we start many imaginings conceptually. But the fact that imagery is its own representational

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

The Imaginative Agent  93 Now that we’ve seen the principles according to which I-C-E-C operates, let’s consider three of its properties that help explain its roles in agency. Automaticity. Returning to the pity we had on imagistically imagining the boy who failed the test, we see that the emotion follows the imagery automatically. By this I mean two things: first, it is not under direct voluntary control whether one has the emotion consequent on the imagery, and second, it is not under voluntary control which emotion is consequent on which imagery. It may be that through some kind of training I can learn to have this imagery without having the emotion, but that will be an effortful exception to the default operation of the cognitive-affective system. This automaticity claim coheres with the fact that the way perception sparks emotion is automatic in the senses specified; whatever pathways underlie that automaticity underlie automaticity in I-C-E-C too. And it wouldn’t make sense for the P-C-E-C connections between perception and emotion to be non-automatic; if those were non-automatic, we’d be lunch.15 Furthermore, it is noteworthy that visual imagery at least involves activation of the sensory thalamus, whose projections to the amygdala are along the low road and hence fast. Reality Congruence. In claiming that I-C-E-C is reality congruent, I mean that the things we imagistically imagine are typically things that could happen in the environment in which we live, given what we believe about that environment. But isn’t imagination, as Hume puts it, free? Not exactly. Imagistic and attitude imaginings are by default constrained by one’s environment. And there are at least three reasons to think it takes effort to override the default.16 First, imagistic imaginings are largely composed of representational constituents perception has already instantiated. Since perception, then, is typically of things in one’s environment, the environment constrains imagistic imagining, because it constrains perception. This gives us reason to think that the starting points of imagistic imaginative episodes will often (not always) be of entities that in fact exist (or could exist) in the environment. Second, factual beliefs, which largely track properties of real entities (Dretske 1983), typically constrain inferences from one imagining to the next. If S imagines person x is outrageously drunk, S will likely further imagine x stumbling and having slurred format allows it to have processing consequences that weren’t anticipated in conceptual thought. For example, visualize an equilateral triangle pointing up (actually do it); now visualize an equilateral triangle of the same size pointing down superimposed. It’s only once you’ve done the visualizing that you arrive at a representation that can receive a RELIGIOUS SYMBOL primary categorization, which wasn’t present in the conceptual instructions. Furthermore, this primary categorization has sudden and compelling emotional resonances. In any case, we see that conceptual thought and primary categorization are different processing stages. I thank one anonymous referee for OUP for raising this issue. 15   Some behaviors may be entirely reflex and bypass emotions altogether. But we’d lose too much flexibility if all behaviors were like that. The automaticity of emotions maintains the priority of certain kinds of actions in response to certain percepts, without making the organism lose agential flexibility altogether. 16   And there are many more than three. Kind (Chapter 6, this volume), Langland-Hassan (Chapter 2, this volume), Spaulding (Chapter 9, this volume), and Williamson (Chapter 4, this volume) are all pertinent here.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

94  Neil Van Leeuwen speech. That’s because S believes drunks stumble and have slurred speech. This illustrates how beliefs inferentially govern attitude and imagistic imaginings: they are the informational background that allows inferences from one imagining to the next. Reality constrains beliefs, and beliefs govern imaginative inferences.17 Third, psychologists Deena Skolnick Weisberg et al. (2013) show that children have a strong preference for imagining realistic story continuations, where “realistic” here means in conformity with familiar patterns of causation. Their experimental paradigm is to show children an incomplete sequence of images, coupled with a correspondingly incomplete story read aloud, and to ask them to choose among pictures that would continue that storyline. Children mostly choose realistic continuations, even when the storyline up to that point had been fantastical (involving magic or other unrealistic elements). Adults, who were tested via an online version of the experiment, tended to choose realistic continuations for the realistic stories and fantastical continuations for the fantastical stories. But generally, realistic starting points yield realistic further imaginings. In sum, much work on imagination suggests that we tend to imagine things that could realistically happen, since imagining defaults to being constrained by cognition of the real. If we imagine in relation to a real environment—what is in that bush?— beliefs, past percepts, and knowledge of causal patterns constrain what we imagine. As a result, the imaginings at the head of I-C-E-C will tend to be reality congruent.18 Integratability. Integration is when imagistic imaginings represent objects, properties, and events in the perceived space around the agent. Imagery may be integrated or not, but the possibility of integration (integratability) is important and often overlooked. Valjean’s hand comes up and down to different heights in the space around him. His imagistic imaginings represent those children as being in that space. Had someone asked, “Which is the youngest?”, his hand may have returned to the specific place corresponding to the youngest. And we don’t need Victor Hugo for examples of integration. When we plan to move furniture, our imagistic imaginings are typically integrated, allowing us to estimate which furniture goes where.19 Furthermore, the relevant cognitive neuroscience backs up common sense on this front. Typical tasks that investigate neural correlates of visual imagery involve having subjects visualize where a recollected item would fall on a presently seen display; such tasks would make no sense, if integration were not a common phenomenon. What makes Valjean’s integration so interesting for present purposes is that his integrated imaginings have the

17   See Van Leeuwen (2013) for more discussion of this point. There are many complications here that don’t affect the overall argument of this section. 18   Does such constraint by belief, perception, and ultimately environment mean the idea that imagination is free is entirely without substance? Not really. Constructive imagination, which generates ­attitude and imagistic imaginings, is exploratory constraint satisfaction, and the existence of default ­cognitive constraints does not entail that different constraints cannot be chosen. But choice and effort are needed. 19   See Kind (2013) for discussion.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

The Imaginative Agent  95 emotional consequences characteristic of I-C-E-C: his emotions were directed at the children imagistically imagined around him. We’ll see why this is important.20

4.  Imagination, Emotion, and Agency Now that we’ve seen how imagistic imagining—and, by extension, attitude imagining that incorporates imagery—impacts emotion, we can ask: why might it make sense that we are so constituted? Is I-C-E-C merely a by-product of P-C-E-C? Or does this pathway do something for us as agents? I argue here that three valuable agential capacities are enabled by I-C-E-C.

4.1  Bodily preparedness An environment is a space in which a range of potential events could take place. To act responsively to such potential events, you must use your body, which can be more or less prepared for them. I claim that I-C-E-C supports bodily preparedness for actions in relation to potential events in an agent’s environment. Consider an intuitive example. Walking down a wooded path, Stephanie hears a rustle from the nearest bush and leaps back in fright, imagining a mountain lion; her heart racing, she feels ready to grab a stick and fight or just run. A deer darts out. She needs a minute to feel calm again. For the rest of her hike she looks at shadows, continuing to imagine mountain lions, and carries a large branch. The intuitive idea here is that Stephanie’s imaginings prepare her body—blood pumping, eyes alert, etc.—for potential encounters with lions. Psychological research bears out this idea. Paul Ekman argues that there are basic emotions, including “fear, anger, disgust, sadness and contempt” as negative emotions, and “amusement, pride in achievement, satisfaction, relief and contentment” as positive emotions (1999, p. 138).21 Ekman essentially has a functionalist theory of emotions, where each basic emotion is characterized by certain universal trigger stimuli, by changes in the central nervous system that process those stimuli, by resultant bodily and facial expressions, and (for some emotions) by changes in the autonomic nervous system (ANS), which regulates bodily functions mostly beneath conscious control (breathing, heart rate, blood flow, etc.). The class of trigger stimuli for a given emotion, in turn, is characterized largely in terms of the evolutionary needs of the organism with respect to a specific ecological problem, though learning can modify the class over time. Facial expressions, as outputs, are 20   I advocate the integration claim more fully in Van Leeuwen (2011). Nanay (2010) also offers arguments pertinent to the integration claim. 21   There is, of course, controversy about whether Ekman is right about basic emotions. I find his views very persuasive and think that a number of objections to his work are based on uncharitable readings. In any case, let’s take his views as given for now. I suspect that my claims about how I-C-E-C supports agential capacities could work for other conceptions of emotion too, though determining that would be another project.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

96  Neil Van Leeuwen characterized by specific, culturally universal face muscle contractions, recognizable to fellow humans, like the facial expression Valjean would have made when weeping. ANS changes are most relevant to our present claim. Ekman describes ANS changes for fear and anger: these ANS patterns evolved because they subserve patterns of motor behavior which were adaptive for each of these emotions, preparing the organism for quite different actions. For example, fighting might well have been the adaptive action in anger, which is consistent with the finding that blood goes to the hands in anger. Fleeing from a predator might well have been the adaptive action in fear, which is consistent with the finding that blood goes to large skeletal muscles . . . Freezing in fear might seem to create a problem for this line of reasoning, but not if freezing is interpreted as a fearful state in which the organism is nevertheless still prepared, autonomically, for fast flight if the freezing action does not provide concealment. (1999, p. 50, my emphases)

In short, the autonomic bodily preparedness that partly constitutes emotion has agential utility: it facilitates actions that are adaptively appropriate responses to our immediate environment. Now we ask: does I-C-E-C, as a source of emotion and hence ANS change, usefully serve bodily preparedness, or does it merely hype our bodies for what doesn’t exist? At this point, excessive focus on fictional contexts would lead us astray, because those contexts are designed to trigger I-C-E-C without the presence of urgent ecological challenges that render consequent ANS changes adaptively appropriate. But return to our example of imagined mountain lions. In this case, the ANS response of blood flow to large skeletal muscles is appropriate and could be life saving. Not every threat is seen (or heard, etc.), so it benefits us to have imagistic imagining trigger ANS changes when percepts aren’t there. A potential lion attack is a worthy ANS trigger. Importantly, the three properties of I-C-E-C identified in the last section are crucial to bodily preparedness. If I-C-E-C weren’t automatic, its contributions to ANS changes would be too slow and unreliable to be useful, especially in cases of potential lions. If I-C-E-C weren’t reality congruent, our bodies would end up being prepared for the wrong sort of thing. Imagining a lover in the bush when a lion is more likely would eventuate in the wrong bodily preparedness. And integration allows us to orient our emotional responses toward or away from specific portions of the surrounding envir­ onment. Generalized fear of a potential lion is somewhat useful. But fear of a lion imagistically imagined in that bush, where the imagistic imaginings are integrated with percepts, is more so.

4.2  Evaluation of future actions We need to imagine the future in order to choose what to do in it. One basis for evaluation of future action is how imagined actions and events make us feel. I hold that I-CE-C supports evaluation of future action by giving us affective responses to imagistically imagined actions and events.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

The Imaginative Agent  97 Again, an example. It’s 8:30 a.m. David should get to work, but he imagines first stopping to see the charming barista at the local café. He thinks of the pleasant conversation he might have and starts feeling upbeat. But then he realizes this is the café’s busiest time. Imagining his attempt at conversation while five caffeine addicts stand behind him makes him feel uneasy. He visualizes a pitying look on the barista’s face over his failed attempt. Better to go tomorrow, he decides. Here we see I-C-E-C at work: specific aspects of David’s visualizations play a role in how he feels about the content of those visualizations. The valences of those feelings then help him decide on an action.22 How, more exactly, does this evaluation work? Tamar Szabό Gendler and Karson Kovakovich (2006) clarify how imagining-emotion links contribute to planning, applying Antonio Damasio’s neuroscientific work on “somatic markers.” I basically agree with their view, which I explain in more detail shortly. But first let me highlight the broader context by pointing out contributions that various forms of imagining make to evaluation of future action. First, agents must represent options and outcomes to be considered—and must employ constructive imagination to have these representations, since they are not of events in the perceivable environment. So even on simple models of agential choice, there is a role for constructive imagination, which seldom receives comment: one can’t choose actions not imagined or attempt outcomes not imagined.23 That’s why “failure of imagination” is often apt; agents may fall short in terms of how many possibilities they build into explicit or implicit decision tables in the first place. And this shortcoming is prior to assignment of probabilities: one can’t assign probabilities to possibilities one hasn’t thought of, so one must constructively imagine any x at all before assigning x a probability. This can be done well or badly. A splendid example is the Bush administration’s “deliberation” over going to war in Iraq. Effectively, their implicit decision table looked like this:

Action: Go to War Action: Not Go to War

State of Iraq (as envisioned by Bush Administration)

Expected Value of Action

Be greeted as liberators Hussein retains WMDs

Good Bad

It only has one column for the state of nature. They only imagined Iraq one way. And a probability distribution over only one outcome, given an action, assigns that outcome 22   Just to be clear, I am not presenting David as a case of anticipated emotion, which would mean he merely anticipates having the uneasy feeling, though he doesn’t occurrently. Rather, I am suggesting that he would actually have the feeling in response to the visualization. Thanks to Andrea Scarantino for discussion of this point. 23   By this, I don’t mean to imply that the attitude that one takes to them is that of fictional imagining, though it is some form of non-belief cognitive attitude, which we may well call “entertaining.” All I am implying at this point is that constructive imagination must be implicated in generating representations of the possibilities under consideration. Thanks to Amy Kind for raising this issue.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

98  Neil Van Leeuwen a probability of 1. So the administration’s main failure in planning was prior to the calculation of probabilities;24 it was in its failure to imagine alternate possibilities at all. The point generalizes. The Bush administration example is replaceable by any in which agents perform certain actions, not because of assigning too low probabilities to certain outcomes, but because of failure to imagine those outcomes at all. Second, attitude imagining—take this for the moment to include the broad class of secondary cognitive attitudes25—is required for figuring out likely outcomes. Let’s return to the imagined café conversation. David holds fixed the attitude imagining that he goes to the café at 8:30 a.m. and then rummages around (deliberately or not) in his knowledge base for information that would allow inferences to other things that would be the case, if that were. The impatient customers don’t initially occur to him, but by imagining for a time he eventually accesses beliefs that suggest their likely presence. So in discovering outcomes for purposes of planning, agents hold fixed an action in attitude imagination and then use constructive imagination to generate whatever else would or might happen.26 Now, third, we’re ready to understand the role of emotions in evaluating possible actions. David’s felt awkwardness at the visualized presence of the caffeine addicts and his shame at the visualized pity on the barista’s face enable him to evaluate the conversation attempt as a poor idea. The important, generalizable features of this example are: a) Attitude imagining represents an action to be appraised. b) Constructive imagination, taking the representation in a) as given, generates imagery of likely states of the world consequent on that action. c) The agent has an emotional response to that imagery because of I-C-E-C. d) The emotional response enables the agent to evaluate the action. Gendler and Kovakovich (2006), as indicated, have a similar view; they discuss Damasio’s (1994) research in these terms: This research seems to show that our ability to engage in practical reasoning rests on the following sort of process: we imaginatively engage with the potential consequences of various courses of action, thereby activating our emotional response mechanisms, and we encode the results of these simulations somatically; the presence of these “somatic markers” then helps to guide our future behavior. (p. 248)

Let’s flesh this idea out. Humans are so constituted that imagistic imagining accesses internally encoded information that other forms of thought don’t, such as likely facial expressions. Such imagery makes important contributions to the agent’s overall emotional state because of I-C-E-C. Damasio (1994) argues that affective responses are 24   This is, of course, something of a simplification, but it's not that much of one. See Woodward’s (2006) State of Denial for a thorough report. 25   Secondary cognitive attitude is just my broad term for any cognitive attitude that is not factual belief; this includes hypothesis, assumption for the sake of argument, supposition, etc. 26   I develop this idea of belief governance in Van Leeuwen (2013) and (2014).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

The Imaginative Agent  99 needed for making rational choices about the future; in particular, those without such responses fail to be properly sensitive to risk. For example, people with damage to emotional areas of the brain—especially the medial pre-frontal cortex and the amygdalae—have systematic distortions in decision making, distortions which are correlated with failures of change in the ANS. So if Damasio is right about the necessity of emotional functioning for rational choice, then some imagining-emotional links will be needed for appraisal of future action. I-C-E-C is one such link, and it is the only one that enables use of information that is encoded in a perceptual format.27 In sum, the fact that we imagistically imagine at all when thinking about the future has at least two functions: (i) accessing of information that might not occur through purely discursive thought and (ii) generating affective states that support evaluation of possible actions. The three key properties of I-C-E-C enable these functions. Automaticity makes emotional responses to imagining an informative source of appraisals of actions and events represented by those imaginings. We discover the value for us of those actions and events by discovering the emotional responses that automatically ensue. If the responses were voluntary as opposed to automatic, they would not be informative. Reality congruence is needed too, since otherwise we’d be appraising outcomes that are not likely to occur, which wouldn’t serve us. The arrival of impatient caffeine addicts is a reality-congruent feature of David’s case. Integratability, finally, supports I-C-E-C in generating appraisals of future action, whenever the action to be appraised relates to one’s perceivable environment. In such cases, perception and imagery form a combined manifold to which emotional systems respond: P-C-E-C and I-C-E-C are jointly active. To return to Stephanie’s case, integration is important, because the seen distance to the bush in which she visualizes the lion helps her decide whether to fight or flee.

4.3  Empathy-based moral appraisal Let’s consider one last agential example. “Spare some change?” says the ragged homeless man on the corner. Should you give him the 75 cents in your pocket? You fleetingly imagine the man using the money for cheap booze. But then you recall him digging for scraps in the trashcan yesterday. You then imagine the sandwich your 75 cents would help him buy and his look of relief at having food. So you put your change in his gnarled hand. Imagery of possible outcomes for others sparks emotions via I-C-E-C—empathy and pity positively; disgust, loathing, and fear negatively—and those emotions support moral appraisal of action. I mean “moral appraisal” in a broad sense: appraisal of action that is responsive to benefit or harm to others, in addition to oneself, and that 27   This paragraph places more emphasis on imagery than Gendler and Kovakovich do. I’m not sure if they would agree with that emphasis. So this paragraph, which represents my view, may be taken as a friendly amendment to theirs or simply as a close but different view.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

100  Neil Van Leeuwen treats that benefit or harm as not-just-instrumentally relevant to whether the action should be performed, where benefit counts in favor and harm against. Three points support this picture of I-C-E-C’s contribution to moral appraisal: first, emotions are central to moral appraisal; second, imagining in general helps cause moral emotions; third, imagistic imagining specifically makes crucial contributions. The first point has been the subject of so much philosophical and psychological literature in the past decade and a half that I can only gesture at the relevant portions. Jonathan Haidt (2001) and Jesse Prinz (2007) are prominent. Haidt, reviewing three decades of psychology, argues for a social intuitionist model of moral appraisal, which contrasts with views like those of Lawrence Kohlberg that emphasize conscious rulebased reasoning. “Intuitionist” here implies that moral appraisals are largely produced by fast, automatic (nota bene), unconscious processes and are triggered by emotional responses to a situation. “Social” implies that feedback from others conditions the emotions and hence the intuitively reached moral judgments. Prinz has a similar view (for my purposes). For him, emotional responses to action situations contribute to moral judgments in a way that resembles how visual percepts in response to surfaces contribute to color judgments; the responses are constitutive of those judgments, but they are also triggered by genuine “concerns” in the world. A congenitally blind person may be able to say “red” and think that things are “red,” but doesn’t have robust color judgments in the way normally-sighted people do. Similarly, for Prinz, if the right emotions aren’t there, verbal or internal labelings of actions as “right” or “wrong” are not fully moral judgments. For present purposes, we don’t need a view as strong as Prinz’s. We must simply hold that moral emotions are typical precursors of moral thoughts, both developmentally (diachronically) and situationally (synchronically). Developmentally: one finds oneself with feelings of disgust, anger, pity, shame, or empathy; one goes on to assume there is something common to everything or most things that elicits each emotion; one then learns to apply internal and external labels for each of these assumed properties, like “right,” “wrong,” “just,” “virtuous,” and “vicious.” Situationally: each emotion is a trigger for judgments with the respective label as constituent. If this view of emotion’s influence on moral judgment is right, we should next establish the second point: that imagining (generally speaking) can elicits the relevant social emotions. One can check this oneself: imagining a snickering cashier short-changing a blind person elicits anger automatically; imagining someone physically humiliating a disabled person elicits anger and disgust; imagining a hungry man counting change to see if he can buy a sandwich elicits pity. These imagination-driven social emotions can morally guide action choice. You don’t see the riffling through the garbage or the satisfied look on the homeless man’s face at the time you make the choice; you imagine them. Daniel Batson et al. (2003) support this view. They performed two experiments to test what sort of imagining elicits empathy. In the first, each subject could assign

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

The Imaginative Agent  101 herself or himself to one of two tasks and another person (believed by subjects to be real, but not in fact so) to the other, where one task was relatively pleasant and the other unpleasant. There were three groups: one with no instructions to imagine ­anything, another with instructions to imagine oneself in the other person’s place (“imagine-self ”), and a third with instructions to imagine how the other participant likely feels (“imagine-other”). The authors found that subjects in the imagine-other condition assigned the other person the pleasant task significantly more often than subjects in the no-­imagining and imagine-self conditions (the respective portions were .58 versus .25 and .25). Subjects in the imagine-other condition also reported higher levels of empathy using “the six empathy adjectives . . . sympathetic, softhearted, warm, compassionate, tender, and moved” (p. 1195; authors’ emphasis), and the self-reported levels of empathy correlated positively with the assignment of the pleasant task to others. The third and final point in establishing that I-C-E-C contributes to moral appraisal is that imagistic imagining in particular evokes moral emotions. Claus Lamm, Daniel Batson, and Jean Decety (2007) provide experimental evidence that visualizing faces generates social emotions. They report a neuroimaging and behavioral study that supports the imagine-self vs. imagine-other difference mentioned above, as well as locating the brain regions implicated in the cognitive and affective sides of empathy. Importantly, imagining the other person’s pain correlates significantly with activity in the fusiform gyrus (FFG)—also known as the “fusiform face area” because of its role in cognizing and imagining faces (O’Craven and Kanwisher 2000). Experimental work by Elinor Amit and Joshua Greene (2012) and by Eugene Caruso and Francesca Gino (2011) is also relevant to this third point. Amit and Greene gave subjects two traditional trolley problems. In the first, flipping a switch would steer a moving trolley toward killing one person instead of five. In the second, pushing a person off a bridge and in front of a trolley would kill one and save five. It is well known that subjects tend to respond differently to the two scenarios, even though the costs and benefits from a utilitarian standpoint are essentially the same. But some people are more likely than others to advocate taking action to kill the one and save the five in both cases. Amit and Greene investigated why. We can summarize their main findings as follows: (i) people who have a more visual cognitive style tend toward the deontological type judgment that killing one to save the five is wrong; (ii) having been instructed to visualize the harm one’s choice would cause correlates significantly with deontological judgments; and (iii) higher levels of emotional response mediate deontological judgments. By “deontological,” Amit and Greene mean morality that appraises actions in terms of exceptionless rules, as opposed to in terms of consequences. In my view, their study

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

102  Neil Van Leeuwen shows that employing imagery (visual or otherwise) is one important strategy for moral appraisal, since it is effective at yielding moral emotions.28 Are such moral emotions effective at motivating action? Caruso and Gino (2011) find that closing one’s eyes before performing an act encourages ethical behavior. For example, in a one-shot dictator game (where one person has a pile of money and chooses how much to give to another person) subjects who close their eyes while listening to instructions for the game are likely to give more to the other (anonymous) player, as compared to subjects who don’t close their eyes. Caruso and Gino hold that the mechanism for this is that closing one’s eyes causes one to visualize outcomes, which leads to higher emotional engagement. And higher emotional engagement is morally motivating. As with evaluation of future potential actions, the automaticity of I-C-E-C is critical to its efficacy in moral appraisal. Affective responses to imagined actions and events serve as orienting guides, as if to say things like “help him” or “avoid her.” If their triggering weren’t automatic, they wouldn’t be good guides; they’d be like signposts that weren’t stuck in the ground. (This is not to say that there is no learning or change in connections between imaginings and emotions over longer time periods; rather, the moment-to-moment connections are automatic.) The reality congruence of I-C-E-C is also especially important for moral appraisal. Without it, we’d have strong moral emotions about events that never would happen, and we’d imagine people with feelings they simply don’t have. Imagining that the homeless man would thank us for our tough love when we deny him change won’t help us behave morally. (Interestingly, integratability doesn’t come into play in moral appraisal. If Caruso and Gino are right, we’re better off not integrating and closing our eyes instead.) * * * This has been a whirlwind tour through the agential capacities that I-C-E-C facilitates. But it gives a sense of how much our cognitive-affective system is improved by just one  feature: by having a set-up in which imagistic imaginings impact emotions in nearly the same way as percepts do—I-C-E-C works like P-C-E-C—we are much richer as agents. 28   It’s fashionable in some circles nowadays to argue that the more detached, utilitarian style of moral judgment is superior to the empathic style of moral judgment. The general idea is that the utilitarian style is free of the prejudices of the more emotional style and that we do more good for more people by adopting it. One who takes this view, on reading this chapter, might think me wrong to write so enthusiastically about I-C-E-C’s contribution to empathic moral evaluation, even if I’m right about the psychology. So let me just say that I find the arguments in this vein sophomoric. First, the relevant set of facts isn’t always available for purposes of weighing utilitarian costs and benefits, so one must have other strategies. Second, some moral actions are constituted at least in part by empathic feelings; kind words to someone suffering are of this sort. Third, we must ask how the basic goods in any utility-maximizing scheme are determined in the first place. What is worth maximizing? Utilitarian schemes are empty without some such determination. Here I think our moral emotions in response to actions and events in the world are particularly important. They give us initial purchase on what things are good or not. It may be that sometimes our utilitarian calculi may lead us to act contrary to our moral feelings, but that is because we are choosing to support less salient but greater instances of goods that those feelings identified for us as goods in the first place.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

The Imaginative Agent  103 Researchers often think (implicitly or explicitly) that capacities that impact the phenomenon they study exist for that phenomenon: imagination is for fiction; . . . for pretend play; . . . for appreciating art; . . . for geometrical reasoning; . . . for counterfactual reasoning; etc. As a result, the research community ends up with as many conceptions of imagination as there are related research areas. But this is not a recipe for progress. When I first learned a number of years ago that imagery plays out on the perceptual cortices of the brain, I wondered if it might have been better for us if we had had two sets of perceptual cortices, one set for perception and one for imagery. It seemed to me then that we would be capable of more detached imaginative reflection with such a set-up. Perhaps, I thought, the only utility of having imagery on the same set of cortices as perception consists in saving space and metabolic resources. But now, as you see, I think I was wrong to think that. Having imagery play out on perceptual cortices does a world of good for us as agents. One pathway yields three agential capacities—and it drives us to fiction as well. This is the kind of unity the research community on imagination should be seeking.

5.  I-C-E-C and Fiction By now one can predict my explanation for why humans find fiction so engaging. In very broad brushstrokes, works of fiction trigger I-C-E-C and are so crafted (in the usual case) that they maintain various forms of emotional activation once we’ve gotten into the work. This emotional activation prompts further engagement with the work and the cycle can begin anew; the emotional refractory period is crucial here, since it eventuates in further attention to the source of the emotion. The reason that I-C-E-C doesn’t shut down just because we know the fictions aren’t real is that I-C-E-C generally reacts to entities and events that are potential, and fictions, if they are well crafted, sneak into this class. If they’re not crafted well (in the relevant sense of “well”), they don’t trigger I-C-E-C and don’t “do it for us.” Since I-C-E-C makes this all possible, the answer to the question of why we like fiction simply becomes whatever the answer is to the question of why we have I-C-E-C. And that answer, I have argued, is to be found in examining what the pathway does for us as real agents. There is much to say about this approach to understanding humans as fiction-loving creatures. For now, I restrict myself to two specific points on two of the most prominent puzzles in the philosophical literature on fiction: imaginative resistance and the paradox of fiction.

5.1  Imaginative resistance The version of the puzzle of imaginative resistance I’m concerned with is what Tamar Szabό Gendler and Shen-yi Liao (in press) call “the phenomenological puzzle.” When we attempt to process a work of fiction, it seems easy to incorporate outlandish descriptive propositions into our understanding of the story, but our minds are far less flexible

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

104  Neil Van Leeuwen about incorporating outlandish moral propositions. Try imagining that Ebenezer Scrooge was right to dock Bob Cratchit’s pay for using an extra lump of coal to warm the room where he worked. It’s easy to imagine that he does this, but our minds falter when it comes to allowing the moral claim into the fictional world. I find myself thinking, “No way was it right for Scrooge to do that!” But by comparison, not for a minute do I think, “Hey, Cratchit doesn’t exist!” We take descriptive existential claims on board easily. So the puzzle is to explain why our minds seem to have one form of flexibility and not the other.29 Here’s how I (non-deductively) explain that inequality of flexibility: 1. We can imagine mostly30 whatever descriptive fact we like, or whatever an author asks us to imagine descriptively, in a fiction. 2. Whatever we descriptively imagine automatically triggers moral emotions via I-C-E-C. 3. Those moral emotions generate moral appraisal of actions in the fiction. 4. The moral appraisals of actions in a fiction are thus constrained by descriptive imaginings via I-C-E-C. 5. Since descriptive imagining of a fiction constrains moral appraisal of the fiction, the latter is less flexible than the former. In essence, the automaticity of I-C-E-C makes it such that the descriptive facts we imagine, if we imagine any elements of them imagistically (which we often do), settle our emotional moral responses. An analogy may help. Descriptive imagining of a fiction is like the turning of a key in an ignition; I-C-E-C is like the engine of a car; and  moral responses are like the starting of the motor. However much control we have  over turning the key in the ignition, we have little control over whether the engine starts, given that we have turned the key and given that we have a functioning engine with gas. The reality congruence of I-C-E-C is also pertinent here. Though we may be able to vary what we imagine a great deal, our imaginings will default to being similar to that which could occur in the world in which we live. Thus, we find ourselves, because of reality congruence, imagining things with which it seems to us we could interact. We import much more of our beliefs about the causal structure of the world than we think into cognition of even the most outlandish fictions. So our emotional responses to 29   As a matter of historic interest, I think the phenomenological puzzle is closest to what Hume (1875) had in mind: “Whatever speculative errors may be found in the polite writings of any age or country, they detract but little from the value of those compositions. There needs but a certain turn of thought or imagination to make us enter into all the opinions, which then prevailed, and relish the sentiments or conclusions derived from them. But a very violent effort is requisite to change our judgment of manners, and excite sentiments of approbation or blame, love or hatred, different from those to which the mind from long custom has been familiarized.” He seems to say that we have an easy time “enter[ing] into” what we regard as false descriptions of the world, but we have a terribly hard time doing the same for what we regard as false moral proposition. 30   The exceptions don’t make a difference here.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

The Imaginative Agent  105 ­ ctions, including moral emotions, will always occur as if those fictions were potential fi realities for us. (And of course, if they don’t strike us that way, we just put the book down.)

5.2  The paradox of fiction The Paradox of Fiction is about whether it is rational to experience emotions in response to fiction. I won’t attempt an ultimate solution here. But I wish to argue, using the framework I’ve developed, that an apparent entailment of two of the premises that make up the paradox is false. But that apparent entailment, were it true, would have a great deal of practical significance, so its falsity is important. The paradox has three (to varying degrees) intuitive premises, which jointly entail a contradiction: Premise 1: it is irrational to have emotional responses to entities believed not to be real. Premise 2: imagined fictions are believed not to be real. Premise 3: it is not irrational to have emotional responses to imagined fictions. The intuitive force behind Premise 1 comes from the need to explain the following sort of case: I imagine I have cancer (even though I know I don’t), and because of this I get extremely emotionally agitated. It seems that we need something to explain the irrationality of this case, and Premise 1 prima facie fits the bill. I take Premise 2 to be true; denying it would conflate cognition of fiction with delusion. And Premise 3 will seem sensible to anyone who emotionally enjoys fiction (and does so without embarrassment). Premises 1 and 2 jointly entail that it is irrational to respond emotionally to fictions. This further seems to entail that we should train ourselves not to have such responses. After all, we feel normatively obligated to train ourselves not to fall into other mental patterns we realize are irrational. So the mooted irrationality of emotional responses to fiction seems to entail that we should practice not having them. I am concerned with this latter (apparent) entailment. Training ourselves not to emotionally respond to fiction would involve effectively destroying I-C-E-C, or at least making it much weaker. I’m not sure it would be possible to do this. But if it were possible, doing it would be a bad idea, since it would involve destroying something on which three important agential capacities depend. So it is false that we should so train ourselves. Suppose we were to train ourselves not to respond emotionally to plays by doing away with the integratability property of I-C-E-C that lets us turn a sparse set into an imagined richer world. That would be folly, for it would also undercut Stephanie’s bodily preparedness for dealing with potential nearby mountain lions and other predators. Or suppose we were to train ourselves not to imagine fictional worlds in such a real­itycongruent way that they are emotionally powerful; we might be able to do this by habituating ourselves only to imagine incoherent and outlandish scenarios in response to the promptings of fictional works. But to do away with the reality congruence of I-CE-C would be to make bodily preparedness misdirected, to rest evaluation of future

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

106  Neil Van Leeuwen actions on the bizarre, and to divorce moral appraisal from the dispositions of actual people. Or perhaps we could make scary movies no longer elicit fear by training away the automaticity of I-C-E-C. But to do this would be to make ourselves no longer properly sensitive to real risk, rendering us—when it comes to choosing for the future— functional equivalents of Damasio’s lesion patients. In sum, anyone who scorns the human propensity to respond emotionally to fictions has just not understood the consequences of doing away with that propensity. If all parties to the discussion share Premise 2, then the last recourse of the defender of Premise 1 will be to deny the entailment: to deny that the supposed irrationality in question entails that we should train ourselves not to respond emotionally to fiction. Investigating conceptions of rationality according to which one could sensibly deny this is well beyond the scope of this chapter. But we can conclude that Premise 1, whether true or false, has no significance for how we should train our cognitive-affective system to operate.

6.  Conclusion: A Systematic View Jean Valjean, as an agent, imagistically imagined his sister’s children standing before him; that imagery helped guide his hand to each of their phantom heads. Those imaginings also, via I-C-E-C, triggered feelings of affection in him and ultimately sadness at his inability to help. You, the reader of the passage, imagistically imagine Valjean and his action of patting the imagined heads. Because your mind too has the I-C-E-C pathway, you first feel affection for Valjean’s kindheartedness, which then turns to sadness over his predicament. The sad thing about the Valjean case—both our perspective on it and his—is that there is nothing we can do to help. But I hope to have shown that these sorts of imaginings, this pathway, and these emotions exist for good reasons: they put us in positions—not always, but often—to act well as agents in our present environments, in relation to the future, and in relation to others. P-C-E-C, the pathway from perception to emotion, was almost certainly largely in place in our most recent evolutionary precursors. And then, as I see it, Mother Nature—by processes still unknown—discovered an incredible trick. By enabling imagistic imagining in all our brain’s perceptual areas, she created I-C-E-C out of P-CE-C and thereby in one fell swoop enabled the three capacities of which I write. But as soon as this happened, we became suddenly vulnerable to fictional enchantment, and this enchantment leads to moral entanglement with unreal people, places, relationships, and worlds. Is this entanglement irrational? If we narrow our focus onto the entanglement itself, it may seem so, since it’s not obvious what this or that fiction does for us. But if we take a systematic view, we see the entanglements are charming by-products of an extremely well-designed system of cognition and affect. Calling those entanglements “irrational” may not be entirely wrong. But it is pointless, for it

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

The Imaginative Agent  107 would be folly to change the cognitive-affective system that produces them. Our constructive imagination s­ urrounds us with possibilities that emotionally excite us, mostly for good practical purposes. But being so constituted leaves us vulnerable to the excitement of make-believe. Enchantment is the price of human agency.

Acknowledgements I’d first like to thank the research assistants who helped me explore the empirical and philosophical literatures pertinent to this chapter; Scott Danielson, Cameron Hamilton, and Rob Hatcher were extremely helpful. Next, for feedback on earlier drafts, I’d like to thank Leslie Andrews, Ben Freed, Josh Landy, Peter Langland-Hassan, Hyoung Kim, Amy Kind, Peter Kung, Shen-yi Liao, Eddy Nahmias, Andrea Scarantino, Shannon Spaulding, and one anonymous referee for OUP. I’d also like to thank Amy Kind and Peter Kung for having me as an invited discussant at the Knowledge Through Imagination conference in April 2012. I delivered presentations based on this chapter at Institut Jean Nicod and at the Centre for Philosophical Psychology at the University of Antwerp in May 2013; I thank Joëlle Proust and Bence Nanay, respectively, for hosting me, and I thank those audiences for their feedback.

References Amit, Elinor, and Joshua D. Greene (2012). You see, the ends don't justify the means: Visual imagery and moral judgment. Psychological Science 23(8): 861–8. Batson, C. Daniel, David A. Lishner, Amy Carpenter, Luis Dulin, Sanna Harjusola-Webb, E. L. Stocks, Shawna Gale, Omar Hassan, and Brenda Sampat (2003). “As you would have them do unto you”: Does imagining yourself in the other's place stimulate moral action? Personality and Social Psychology Bulletin 29(9): 1190–1201. Caruso, Eugene M., and Francesca Gino (2011). Blind ethics: Closing one's eyes polarizes moral judgments and discourages dishonest behavior. Cognition 118(2): 280–5. Damasio, Antonio R. (1994). Descartes’ Error: Emotion, Reason, and the Human Brain. New York: Putnam. Dretske, Fred I. (1983). The epistemology of belief. Synthese 55(1): 3–19. Ekman, Paul (1999). Basic emotions. In Tim Dalgleish and Mick Power (eds), Handbook of Cognition and Emotion. Chichester: Wiley, pp. 45–60. Ekman, Paul (2007). Emotions Revealed: Recognizing Faces and Feelings to Improve Communication and Emotional Life. 2nd edn, New York: Holt Paperbacks. Evans, Gareth (1982). The Varieties of Reference (ed. John McDowell). Oxford: Clarendon Press. Ganis, Giorgio, William L. Thompson, and Stephen M. Kosslyn (2004). Brain areas underlying visual mental imagery and visual perception: An fMRI study. Cognitive Brain Research 20(2): 226–41. Gendler, Tamar Szabό (2003). On the relation between pretense and belief. In Matthew Kieran and Dominic McIver Lopes (eds), Imagination, Philosophy and the Arts. London: Routledge, pp. 125–41.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

108  Neil Van Leeuwen Gendler, Tamar Szabό, and Karson Kovakovich (2006). Genuine rational fictional emotions. In Matthew Kieran (ed.), Contemporary Debates in Aesthetics and the Philosophy of Art. Oxford: Blackwell, pp. 241–53. Gendler, Tamar Szabό, and Shen-yi Liao (in press). The problem of imaginative resistance. In Noël Carroll and John Gibson (eds), The Routledge Companion to Philosophy of Literature. London: Routledge. Haidt, Jonathan (2001). The emotional dog and its rational tail: A social intuitionist approach to moral judgment. Psychological Review 108(4): 814–34. Hugo, Victor (1987). Les Misérables (trans. Lee Fahnestock and Norman MacAfee). New York: Penguin. [Original work published 1862.] Hume, David (1875). Of the standard of taste. In Essays: Moral, Political and Literary (ed. T. H. Green and T. H. Grose). London: Longmans, Green. [Original work published 1757.] Kind, Amy (2013). The heterogeneity of the imagination. Erkenntnis 78(1): 141–59. Kosslyn, Stephen M., William L. Thompson, and Giorgio Ganis (2006). The Case for Mental Imagery. New York: Oxford University Press. Lamm, Claus, C. Daniel Batson, and Jean Decety (2007). The neural substrate of human empathy: Effects of perspective-taking and cognitive appraisal. Journal of Cognitive Neuroscience 19(1): 42–58. Lang, Peter J., Daniel N. Levin, Gregory A. Miller, and Michael J. Kozak (1983). Fear behavior, fear imagery, and the psychophysiology of emotion: The problem of affective response integration. Journal of Abnormal Psychology 92(3): 276–306. LeDoux, Joseph (1996). The Emotional Brain: The Mysterious Underpinnings of Emotional Life. New York: Simon & Schuster. Leopold, David A., and Nikos K. Logothetis (1996). Activity changes in early visual cortex reflect monkeys’ percepts during binocular rivalry. Nature 379(6565): 549–53. Nanay, Bence (2010). Perception and imagination: Amodal perception as mental imagery. Philosophical Studies 150(2): 239–54. Nanay, Bence (2013). Between Perception and Action. Oxford: Oxford University Press. O’Craven, K. M., and N. Kanwisher (2000). Mental imagery of faces and places activates corresponding stimulus-specific brain regions. Journal of Cognitive Neuroscience 12(6): 1013–23. Olson, Ingrid R., Alan Plotzker, and Youssef Ezzyat (2007). The enigmatic temporal pole: A review of findings on social and emotional processing. Brain 130(7): 1718–31. Prinz, Jesse (2007). The Emotional Construction of Morals. Oxford: Oxford University Press. Rosch, Eleanor, Carolyn B. Mervis, Wayne D. Gray, David M. Johnson, and Penny BoyesBraem (1976). Basic objects in natural categories. Cognitive Psychology 8(3): 382–439. Schroeder, Timothy, and Carl Matheson (2006). Imagination and emotion. In Shaun Nichols (ed.), The Architecture of the Imagination: New Essays on Pretence, Possibility, and Fiction. Oxford: Oxford University Press, pp. 19–39. Van Leeuwen, Neil (2011). Imagination is where the action is. Journal of Philosophy 108(2): 55–77. Van Leeuwen, Neil (2013). The meanings of “imagine” Part I: Constructive imagination. Philosophy Compass 8(3): 220–30. Van Leeuwen, Neil (2014). The meanings of “imagine” Part II: Attitude and action. Philosophy Compass 9(11): 791–802.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

The Imaginative Agent  109 Weisberg, Deena Skolnick, David M. Sobel, Joshua Goodstein, and Paul Bloom (2013). Young children are reality-prone when thinking about stories. Journal of Cognition and Culture 13(3–4): 383–407. Woodward, Bob (2006). State of Denial: Bush at War, Part III. New York: Simon & Schuster. Zajonc, R. B. (1984). On the primacy of affect. American Psychologist 39(2): 117–23.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Pa rt T wo

Optimistic Approaches

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

4 Knowing by Imagining Timothy Williamson

1.  Imagining is often contrasted with knowing. When you know nothing about something, you have to imagine it instead. Knowledge deals in facts, imagination in fictions. This chapter sketches a way of thinking about the imagination on which that stereotypical contrast is utterly misleading. Far from being the opposite of knowing, imagining has the basic function of providing a means to knowledge—and not primarily to knowledge of the deep, elusive sort that we may hope to gain from great works of fiction, but knowledge of far more mundane, widespread matters of immediate practical relevance. The assimilation of imagining and knowing is a card that can be played for either realist or idealist ends. The idealist wants to shift our conception of knowing towards our conception of imagining: somehow even knowing deals in fictions. By contrast, the realist wants to shift our conception of imagining towards our conception of knowing: somehow even imagining deals in facts. The spirit of this chapter is firmly realist. It aims more to rethink imagination than to rethink knowledge. Consequently, it is not appropriate to start by defining ‘imagination’. Any such definition would be premature in advance of inquiry. Without one, we can still make progress with that inquiry on the basis of our capacity to recognize clear cases of imagination. What unifies them (if anything), and therefore which less clear cases should be grouped with them, are to be identified in the course of inquiry, not prejudged from the beginning. A more useful starting-point is to ask oneself why the elaborate capacity for imagining that normal humans possess should have arisen in our evolutionary history. Although this chapter will not attempt any detailed evolutionary considerations, one feature of the view to be proposed is that it makes the evolutionary advantage of having a good imagination obvious. That contributes to the view’s explanatory power, and so to its abductive confirmation. The reference to humans’ evolutionary history does not imply that the imagination is an exclusively human capacity. It is quite plausible that some non-human animals engage in at least primitive imaginative exercises, and thereby gain a similar advantage.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

114  Timothy Williamson In this chapter, however, the examples will largely concern the human imagination, since we know so much more about it. Of course, that we possess an imagination does not prove that it confers or once conferred some evolutionary advantage on either the individual or the species. In principle, the imagination might have arisen as an accidental by-product of something else that did confer an evolutionary advantage at one or other level. Sometimes, for structural reasons, the easiest way for evolution to develop the capacity to do something useful involves developing the capacity to do something useless too. This chapter does not take for granted that the imagination confers an evolutionary advantage on either the individual or the species. Rather, it proposes a speculative view of the imagination on which it does confer an evolutionary advantage on both the individual and the species, and that aspect of the view’s explanatory power contributes to its confirmation. There are many kinds of evolutionary advantage. Presumably, peacocks have unwieldily magnificent tails because peahens prefer peacocks so endowed. One could develop a similar hypothesis about the imagination. Other things being equal, the more imaginative you are, the better your seduction technique: if you dance, talk, and do other things more imaginatively, you are more exciting to be with. In One Thousand and One Arabian Nights, Scheherazade’s skill at telling stories keeps her alive and eventually enables her to pass on her genes. On the view to be proposed here, however, a good imagination also confers much more straightforward and direct advantages which do not depend on whether other members of the species are attracted by works of your imagination.

2.  Consider a small group of our distant human ancestors, travelling across a difficult landscape previously unknown to them. How might they find an imagination useful? One obvious answer is that an imagination will alert them to various potential dangers and opportunities. They are about to enter a forest. They imagine wolves in the forest; warned of the danger, they keep a sharper look-out for signs of wolves. They imagine edible berries in the forest; alerted to the potential opportunity, they look about for bushes of the right kind. In both cases, their imagination enables them to prepare for practically relevant possibilities, helping them avoid dangers and take advantage of opportunities. To serve that purpose well, the imagination must be both selective and reality-­ oriented. They could imagine the wolves bringing them food to eat, but doing so would be a waste of time, and a distraction from more practically relevant possibilities. An imagination that clutters up the mind with a bewildering plethora of wildly unlikely scenarios is almost as bad as no imagination at all. It is better to have an imagination that concentrates on fewer and more likely scenarios. One’s imagination should not be completely independent of one’s knowledge of what the world is like.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowing by Imagining  115 Such examples already suggest a distinction between two modes in which the imagination can operate: voluntary and involuntary. When you encounter a problem to which there is no obvious solution, you can turn your imagination to thinking up ways of solving it and to thinking through their consequences. Our little group may do that when they come to a deep river that they must cross. Such uses of the imagination are often voluntary. But sometimes we need our imaginations to work involuntarily too. If the group is absorbed in joking together as it enters the forest, it may be saved by an imagination that breaks into the stream of consciousness with reminders of dangers and opportunities. Imagination resembles attention in having both voluntary and involuntary modes of operation, and for similar reasons. We need to be able to focus our attention voluntarily on something: for example, to set ourselves to watch a hole in case an animal comes out. But we also need our attention to be capable of involuntary switches: for example, to be drawn by a slight movement at the periphery of vision, perhaps a predator or prey, despite our previous intention of watching something else. At least metaphorically, one might regard imagination as a form of attention to possibilities. Since members of the group have imaginations, they can be expected to use them, voluntarily or even involuntarily, in ways that serve no immediate practical purpose too, for instance by inventing and telling stories. In the long run, of course, such uses of the imagination may help the group bond. Moreover, even the most playful uses of the imagination may also help to exercise the capacity and keep it in good trim, ready for more practical applications, just as a cat practises its hunting skills by playing with a mouse once caught, letting it go and recapturing it.

3.  So far, the imagination has been described primarily as raising possibilities, rather than assessing the truth-values of propositions. One might therefore be tempted to suppose that the proper role of the imagination belongs, in Hans Reichenbach’s terminology, to the context of discovery, rather than the context of justification. On a simple-minded version of the distinction, one dreams up scientific theories in the context of discovery, but assesses them as true or false, or as probable or improbable on the evidence, in the context of justification. Rationality is essential in the context of justification, but not in the context of discovery. If you came up with the theory in the first place under the influence of drink or drugs, never mind, but you must sober up or come down from your high before you reach a conclusion as to its status on the evidence. Similarly, someone might think, the imagination has done its work once it has delivered enough propositions to consciousness, to be entertained there. On this view, the imagination plays no further role in the assessment of those propositions as true or false, or as probable or improbable on the evidence.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

116  Timothy Williamson The considerations of the previous section already include signs that such a view is inadequate. The need for the imagination to be selective and reality-oriented in bringing material to consciousness suggests that it must be capable of some sort of rational responsiveness to evidence. But in many examples we can see the imagination playing a far more direct role in the assessment of truth-value. Think of a hunter who finds his way obstructed by a mountain stream rushing between the rocks. He reaches the only place in the vicinity where jumping the stream might be feasible. The best scenario for him is to jump and succeed in getting across the stream. Then he can continue on his way with little loss of time or energy. The worst scenario for him is to jump and fail to get across the stream, for then he will probably be drowned or smashed on the rocks. If he does not jump, and goes another way instead, he suffers a great loss of time and energy, but does not incur imminent death or injury; that is the intermediate scenario. Thus it is vitally important for the hunter to know whether he can jump the stream, whether he would succeed if he did try to jump it, before he decides whether to attempt the jump. Since the method of trial and error is too risky a way of finding out whether he can jump the stream, he needs a way of finding out whether he can do it in advance of trying. He can remember some of his past jumps, but he cannot remember failing with a jump that was clearly easier than this one, or succeeding with a jump that was clearly harder. He has to consider not only the width of the stream, but also the awkwardness of the place from which he would have to launch himself, the slipperiness of the rocks on which he would have to land, how tired he is, and so on. How should he try to determine whether he would succeed? There is a natural human method of gauging one’s capacities in such situations. One imagines oneself trying. If one then imagines oneself succeeding, one judges that if one tried, one would succeed. If instead one imagines oneself failing, one judges that if one tried, one would fail. If one is still uncertain, one repeats the thought experiment, perhaps many times. If our hunter cannot resolve the uncertainty, he will presumably take the long way round—unless he is being chased by a sabre-toothed tiger, in which case he may jump anyway. A traditional stereotype of the imagination can make reliance on such an imaginative exercise look like a mad way of making up one’s mind. For however difficult the jump, one can imagine succeeding with it, and however easy the jump, one can imagine failing with it. How can one learn anything relevant from what one chooses to imagine? Such incomprehension indicates neglect of the distinction in Section 2 between voluntary and involuntary exercises of the imagination. When the hunter makes himself imagine trying to jump the stream, his imagination operates in voluntary mode. But he neither makes himself imagine succeeding nor makes himself imagine failing. Rather, having forced the initial conditions, he lets the rest of the imaginative exercise unfold without further interference. For that remainder, his imagination operates in involuntary mode. He imagines the antecedent of the conditional voluntarily, the consequent involuntarily. Left to itself, the imagination develops the scenario in a reality-oriented way, by default.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowing by Imagining  117 Obviously, nothing has been said to guarantee that the imagination will reach a true answer. In some cases, it reaches a false one. The imagination, like perception, memory, and every other generic human cognitive capacity, is fallible. In other cases, when too much uncertainty is registered, the imagination reaches no answer at all. Nevertheless, under suitable conditions, the method constitutes a reliable way of forming a true belief as to what would happen in hypothetical circumstances. I have used it myself in situations like that described, and lived to tell the tale, as have other people I know. Indeed, there is no good reason to deny that, under suitable conditions, the method enables one to know what would happen in the hypothetical circumstances, because the conditional connection is safe from failure. Of course, hardly anything has been said so far to explain the method’s reliability, except for the hint in Section 1 of evolutionary pressures. Later sections will suggest some links between the method and other normal human cognitive processes. First, however, it will be useful to have some more examples, to illustrate the range of cases over which the method is applicable. The method can be applied with complex iterations. Suppose that now the hunter’s way to the next valley is blocked by broken cliffs. Can he climb through or up them? The price of trying and failing is again high: perhaps becoming exhausted, perhaps getting stuck halfway up. The hunter stares at the rock face, trying to trace a route all the way to the top in his imagination, testing each step for feasibility by imagining what it would involve. No more rational method of answering the question is available to him. Even in the modern world, important decision-making often relies on knowledge or beliefs acquired through the imagination. For example, you are looking round a house, wondering whether to buy it. You want to know whether, if you lived in it, you would like doing so. You voluntarily imagine the antecedent of the conditional; your final decision may depend on what consequent you involuntarily imagine. Not all the knowledge gained from such imaginative exercises concerns the capacities and dispositions of agents. For example, if you look at a piece of furniture and then at a doorway, you can sometimes come to know whether the former would go through the latter, without measuring either, by imagining trying to get the former through the latter. You gain knowledge of spatial relations between the doorway and the furniture. The examples so far may appear to involve an essential role for mental imagery, in some sense. But even if that appearance is veridical, we should not overgeneralize to the conclusion that all imagining involves imagery. For example, you are very busy, and wonder whether to postpone a lunch appointment with a friend. You want to know whether, if you did postpone the appointment, she would be upset. You voluntarily imagine the antecedent of the conditional. You might then involuntarily form a visual image of your friend with a composed or disappointed face. But no such imagery is necessary for imagining her reacting with composure or disappointment. Again, suppose that a politician is trying to work out what his core supporters would do at the next election if he voted for gun control. He imagines their reactions, but doing so need not involve mental imagery.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

118  Timothy Williamson Given this heterogeneity of knowledge-yielding imaginative exercises, we need a more abstract characterization of what is going on.

4.  As Frank Ramsey observed, how we evaluate conditionals is closely tied to how we update our beliefs on new information. For example, suppose that someone tells a shepherd ‘The sheep have broken out of the pen and disappeared’, and the shepherd concludes ‘The sheep have gone down to the river’. Then presumably, even if the shepherd had not been given the testimony, he could still have reached the indicative conditional conclusion ‘If the sheep have broken out of the pen and disappeared, they have gone down to the river’ (or the subjunctive conditional conclusion ‘If the sheep had broken out of the pen and disappeared, they would have gone down to the river’, depending on subtle differences in the cognitive process). The testimony in the first case is the same as the antecedent of the conditional in the second case, and the unconditional conclusion in the first case is the same as the consequent of the conditional in the second case. If we regard the shepherd’s updating of his beliefs in the first case as an online process, then we can regard his evaluation of the conditional in the second case as the corresponding offline process. If he accepted the conditional on the basis of an imaginative exercise similar to those in Section 3, as we may assume, then that imaginative exercise is the offline analogue of online updating. Very roughly, the online and offline processes take the same input—‘The sheep have broken out of the pen and disappeared’—and deliver the same output—‘The sheep have gone down to the river’—by the same means. One process is online and the other offline in virtue of the different sources of the input. If we apply the term ‘imagine’ on the basis of the source of the input, we shall classify only the offline process as an imaginative exercise. If we apply the term ‘imagine’ on the basis of the processing between input and output, we shall classify the online process as an imaginative exercise too. It is more important to see the underlying cognitive similarity than to decide exactly how to use the word ‘imagine’. In many of the examples in Section 3, the cognitive process took a mix of online and offline input. When the hunter imagines himself trying to jump the stream, he also has to look carefully at its banks in front of him, to tailor his imaginative exercise as exactly as he can to their actual contours. Voluntarily in imagination, he somehow adds his jumping to the perceived scene. But that element of offline input is enough to make it a clear case of imagination. Typically, there are further differences between the online and offline processes. When the new information to be updated on derives from sensory perception, we are hard put to articulate it verbally in its full specificity, to be the antecedent of a conditional. Moreover, when we learn something by perception or testimony, we usually learn other things too. When you learn by sight that the sheep have gone, you usually also learn that you can see that the sheep have gone. When you learn by testimony that

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowing by Imagining  119 the sheep have gone, you usually also learn that you have been told that the sheep have gone. But often we do not build that extra information into the antecedent of the corres­ ponding conditional. For conditionals such as ‘If the sheep have gone, I can see that the sheep have gone’ and ‘If the sheep have gone, I have been told that the sheep have gone’ are far from trivial. Nevertheless, these differences of detail do not undermine the strong cognitive analogy between the online and offline processes. One consequence of the analogy is that any scepticism about the offline processes is liable to generalize to the online processes too. For example, someone who doubts that in suitable conditions the shepherd’s imaginative exercise enables him to know that if the sheep have broken out of the pen and disappeared, they have gone down to the river should also doubt that in suitable conditions, on being told that the sheep have broken out of the pen and disappeared, the shepherd can know that they have gone down to the river. That looks like the thin end of the wedge for a far more general scepticism. Much of our knowledge of the future depends on cognitive processes relevantly similar to imaginative exercises. For consider some cognitive process by which we form expectations about the future based on our knowledge of the present. Although we know far less about the future than we do about the recent past, we are not entirely ignorant of the future. The process may be more or less hard-wired into our brains, such as one for forming expectations about where a moving object in our visual field will be in a fraction of a second’s time. Alternatively, the process may involve complex conscious reasoning, such as one for forming expectations about the political situation a year from now. Our imagination enables us to apply such a cognitive process offline, to imagined input about a time t, to reach a conditional conclusion of the form ‘If X were to obtain at t, then Y would obtain at t + 1’. Similarly, we can reach a conditional conclusion of the form ‘If Y were to obtain at t + 1, then Z would obtain at t + 2’. Under conditions that make it legitimate to assume transitivity for the conditionals at issue, we can derive the further conclusion ‘If X were to obtain at t, then Z would obtain at t + 2’, and so on. If these offline imaginative exercises are unreliable, the likeliest explanation is that the corresponding online processes for forming expectations about the future are also unreliable. An analogous point applies to large tracts of our knowledge of the past. Consider cases where we use inference to the best explanation to solve a crime or to interpret an archaeological site. For such abductive reasoning, we need auxiliary conditional premises of the form ‘If the explanans had obtained in the past, then the explanandum would obtain now’. We may have to obtain those conditionals in the same way as before, by an imaginative exercise. Unless one has general sceptical inclinations, it is unwise to deny that, in suitable conditions, imaginative exercises are a source of knowledge. Could someone argue that what have here been called ‘imaginative exercises’ are really just inductive inferences? Most of them depend somehow on past experience, and go beyond it non-deductively. If that suffices for a cognitive process to be an induct­ ive inference, then they are inductive inferences. But they do not depend on the

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

120  Timothy Williamson ­subject’s remembering the relevant past experiences. What matters is whether they have made the subject skilful enough in performing the imaginative exercise itself. It is irrelevant to the process whether the subject can assemble the particular premises of the supposed inductive inference. Nor is it remotely clear in the given cases how to fill in ‘F’ and ‘G’ in the conclusion of the supposed inductive inference, ‘All Fs are Gs’ (or ‘Most Fs are Gs’, for that matter). Thus it is also irrelevant to the process whether the subject can formulate the general conclusion of the supposed inductive inference. The imaginative exercises are inductive inferences only in a sense so loose as to be entirely unhelpful.

5.  We can test the cognitive view of the imagination by confronting a genuine structural difference between imagination and perception, one sometimes used to contrast perception as reality-directed with imagination as supposedly not so. The structured difference is this. If you perceive an F, then there is an F that you perceive. Macbeth literally saw a dagger only if there was a dagger that he saw; otherwise he only seemed to himself to be seeing a dagger. By contrast, you may literally imagine an F even if there is no F that you imagine. I am imagining a golden mountain in Austria. Since there is no golden mountain at all in Austria, a fortiori there is no golden mountain in Austria that I am imagining. What should we make of this difference? I will argue that it is predicted by the cognitive account of the imagination sketched above, and so confirms rather than disconfirms that account. First, consider the involuntary imaginative exercise of developing an initial hypothetical supposition, in effect answering a ‘What if?’ question. On the cognitive account, various offline cognitive procedures add further conclusions to a pool that starts with the initial supposition. The process can be iterated indefinitely. Although most of the procedures are non-deductive, reasoning by deductive logic is an especially good procedure from a cognitive point of view, since it guarantees truth-preservation. The deductive aspect of the whole process will look something like the method of tableaux in first-order logic, by which the consequences of the initial premises are teased out—although in our imaginative exercises the aim is not usually to reach a contradiction. For each logical constant, there is a rule for extending a branch of a tableau which contains a formula whose main connective is that constant. For present purposes, the rule that matters is the one for the quantifier ∃ (‘something’). It says that if a branch contains a formula of the form ∃x Φ(x), then one may add to it the formula Φ(a), where a is an individual constant that has not previously appeared on that branch. One can then apply the rule for the main connective in Φ(a) (if it has one), and so on. Informally, we can regard a as an ‘arbitrary name’ for a satisfier of the formula Φ(x) as a value of the variable x—if there are several satisfiers, it does not matter which one. The point of the ban on previous occurrences of a in the branch is to avoid prejudging anything else about the satisfier. If an interpretation I verifies every formula on

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowing by Imagining  121 the branch before Φ(a), then it verifies ∃x Φ(x), so some interpretation differing from I at most in making an object o the value of x verifies Φ(x), so modifying I to make o the denotation of a yields an interpretation that verifies every formula on the branch up to and including Φ(a). The formula Φ(a) has the form of a formula expressing a singular proposition about a particular object, even if the individual constant a does not really denote anything: after all, ∃x Φ(x) may be false on its intended interpretation. Now read Φ(x) as ‘x is a golden mountain in Austria’. Thus ∃x Φ(x) expresses the proposition that there is a golden mountain in Austria. If an imaginative exercise starts from that false hypothesis, the cognitive account of the imagination (defeasibly) predicts that developing it properly will involve applying some informal analogue of the tableau rule for ∃, and thereby adding some informal analogue of Φ(a) to the development. That is, one will do something formally similar to thinking of some particular object that it is a golden mountain in Austria, even though really there is no particular object of which one is thinking that it is a golden mountain in Austria. But that is a good description of a case in which one imagines a golden mountain in Austria, even though there is no golden mountain in Austria that one imagines. Of course, the description does not entail that one forms a mental image of a golden mountain in Austria, but we saw in Section 3 that forming mental imagery is not necessary for imagining. Although one may imagine a golden mountain in Austria by forming a mental image of one, the informal analogue of Φ(a) may be a mental image. Thus the very feature that was supposed to differentiate imagining from reality-directed attitudes is predicted by the account of imagining as a reality-directed attitude. In more general terms, the felt tendency of the imagination to be specific, to fill in details, is partly explained by the hypothesis that it uses something like the tableau rule for the quantifier ∃ in developing an initial supposition. The explanation can be strengthened with the additional hypothesis that the imagination also uses something like the tableau rule for disjunction, ∨, in developing the initial supposition. The rule says that if a branch contains a formula of the form Φ∨Ψ, then one may divide it into two sub-branches, one containing all formulas already on the branch and Φ, the other containing all formulas already on the branch and Ψ. Informally, the idea is to explore the two (not mutually exclusive) ways in which the disjunction may hold separately. This corresponds to the dissatisfaction one tends to feel about, for instance, just imagining that the keys are in either the kitchen or the bedroom while neither imagining that they are in the kitchen nor imagining that they are in the bedroom. It is almost trivial that the imagination uses something like the other standard tableau rules in developing an initial supposition. For example, the rule for conjunction, ∧, says that if a branch contains a formula of the form Φ∧Ψ, then one may add to it both the formula Φ and the formula Ψ. Similarly, it would be hard to imagine that Mary is tall and thin without also imagining that she is tall and imagining that she is thin. On the cognitive view of the imagination, one would expect it to use something like the tableau rules in developing an initial supposition. The imagination conforms to that prediction.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

122  Timothy Williamson

6.  The cognitive view of the imagination does not predict that it will be cognitively reliable only for tasks just like those it evolved to serve. Its tendency to use something like rules of deductive logic is an example to the contrary, since they are quite generally truth-preserving. The simplest forms of reasoning to implement that are truth-­ preserving for practical matters are truth-preserving for all matters. Once we appreciate the role of the imagination as a standard means for evaluating conditionals and modal claims, we should be much less inclined to regard the use of thought experiments in philosophy (or natural science) as constituting any highly distinctive method. Still less should we be tempted to characterize such a method in terms of cooked-up categories such as ‘philosophical intuition’, which serve mainly to obscure the similarities between thought experiments and more routine exercises of the imagination in virtually every branch of human inquiry. We simply reserve the term ‘thought experiment’ for the more elaborate and eye-catching members of the kind. One might suppose that, as science progresses, the role of the imagination will increasingly be confined to the context of discovery, and that in the context of justification it will gradually be replaced by more rigorous methods. But there is evidence to the contrary. For rigorous science relies on mathematics, and so indirectly on the axioms or first principles of mathematics. But when one examines the justifications mathematicians give of their first principles, such as axioms of set theory, one finds unashamed appeals to the imagination. Things are complicated, because the justification is abductive: it involves the derivability of standard ‘working’ mathematics from the candidate first principles—and, one hopes, the non-derivability of contradictions. But it also involves the way in which the candidate first principles fit together as parts of an intrinsically simple, elegant, and unified picture of mathematical reality. Those theoretical virtues are needed to trump rival gruesomely gerrymandered pictures, for example, one on which the hierarchy of sets is truncated at an arbitrarily chosen large cardinal, in the usual manner characteristic of abductive theory comparison. Moreover, since by Gödel’s second incompleteness theorem the axioms can be proved jointly consistent only on the basis of even stronger axioms that are inconsistent if the former are, much of mathematicians’ confidence in the consistency of their axioms comes from the way in which they form such a coherent picture. Of course, they also have inductive evidence of their consistency, because no one has ever yet derived a contradiction from them, but mathematicians put comparatively little weight on such inductive evidence. They have much less confidence in the consistency of alternative axiom sets from which no one has ever yet derived a contradiction, but which do not seem to form a coherent picture. Without an imagination, one would be in no position to judge whether some candidate first principles form a coherent picture of mathematical reality. For the foreseeable future at the very least, imagination will play

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowing by Imagining  123 a crucial role in the context of scientific justification, not only in the context of scientific discovery. Whatever the function or evolutionary origin of our capacities, we are not forbidden to use them for other ends, including playful ones. Imagination has enabled humans to produce works of art, great and awful. Indeed, the combination of voluntary hypothetical suppositions and involuntary developments of those suppositions is reminiscent of the way some novelists describe novel-writing. An imagination is cognitively powerful only if it is capable of producing and developing fictions. But if we try to understand the imagination while taking for granted that fiction is its central or typical business, we go as badly wrong as we would if we tried to understand arms and legs while taking for granted that dancing is their central or typical business.

Acknowledgements This chapter is based on my talk to the 2012 Claremont conference on Knowledge Through Imagination. Other versions of the material were presented at a workshop at the InterUniversity Centre in Dubrovnik and at the following universities: Belgrade, Bergen, Birmingham, Boğaziçi (Istanbul), Cambridge, Edinburgh, Gothenburg, Liverpool, London (King’s College), Manchester, Oxford, Warwick (apologies to any I have forgotten). I am grateful to the audiences at all these events for stimulating discussion. The piece takes further ideas from my books The Philosophy of Philosophy (Oxford: Blackwell, 2007) and, to a slight extent, Knowledge and its Limits (Oxford: Oxford University Press, 2000). Its development was so loosely related to the literature that I decided, contrary to the unanimous advice of editors and referees, not to add footnotes. Interested readers may find points of similarity with other chapters in this book.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

5 Modals and Modal Epistemology Jonathan Jenkins Ichikawa

Introduction There is a substantial contemporary literature engaging with questions about the epistemology of metaphysical modality; in its typical instances, it neglects the connections between metaphysical modals and their quotidian counterparts. This is, I think, a significant error. Claims of metaphysical modality are best understood as particular cases of modal claims generally; the epistemology of the former should be understood as continuous with that of the latter. Consequently, we face no need to posit anything like a sui generis faculty for metaphysical modality. Neither ought we to, as Christopher Hill and Timothy Williamson have each recently suggested, understand knowledge of modality in terms of counterfactuals.

1.  Why Questions in Modal Epistemology Modal epistemology concerns our epistemic access to facts about modality. In particular, modal epistemology typically concerns itself with questions about our knowledge (or justified beliefs, etc.) of claims about metaphysical possibility and metaphysical necessity. One obvious question about our epistemic access to facts about metaphysical possibility and necessity is this one: (How) How do we come to know facts about metaphysical possibility and necessity? The How question is familiar in the relevant literature, as are various responses to it. According to one familiar response, we come to know propositions to be possible by conceiving of them, or by conceiving of them in a certain privileged sort of way.1 According to another familiar response, we use a faculty of rational intuition to come to know truths of modality.2 A different sort of response rejects the presupposition of

  e.g. Yablo (1993). Cf. also Chalmers (2002).   e.g. Bealer (2002), Sosa (2007, ch. 3).

1

2

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Modals and Modal Epistemology  125 the question, suggesting that the apparent ubiquity of knowledge of metaphysical modality is an illusion brought on by hubristic overconfidence.3 This How question about knowledge has close cousins, such as How do we come to have justified beliefs about metaphysical possibility and necessity? and How do we come to have reliable beliefs about metaphysical possibility and necessity? The sorts of responses mentioned above to the How question typically go along with corresponding answers to the How-Justified and How-Reliable questions in obvious ways; one might hold, for instance, that conceivability is appropriately and correctly trusted as a reliable guide to possibility, such that believing the conceivable possible is always justified, and, when things go rightly, knowledge. Different approaches to modal epistemology emphasize different of these questions. I’ll focus on the knowledge version, although I don’t think it should make much difference for present purposes. What these How questions all have in common is that they are all questions about us. Who are we, exactly? This is to some degree an open question, but it’s certainly meant to include at least us philosophers who have come to what we take to be a knowledgeable consensus about various matters of metaphysical possibility and necessity. Many of us, for instance, apparently know, having read Kripke, that water is necessarily H2O, and that it’s possible for that meter stick in Paris to be longer than one meter. More mundanely, we seem to know such modal facts as that it was possible for Mitt Romney to have won the election over Barack Obama, but impossible for him to have identified the largest prime number. Maybe many non-philosophers have some or all of this knowledge too. The How questions ask, at least, how we philosophers know the modal truths in question. The proper responses to the How questions, therefore, are beholden to psychological facts about us. If, for instance, as Cappelen (2012) argues at length, it turns out that as a matter of psychological fact, we don’t use a faculty of intuition in coming to judge modal truths, then the answer to the How question that says we do is incorrect. Psychological questions about how philosophers come to judge philosophical matters are interesting ones, but, as is the case with most psychological questions, the relevant data are rather complex, and not particularly well-suited to discovery by traditional forms of philosophical methodology. It’s not very easy in general to tell how we go about coming to believe and know what we do; the case of belief and knowledge about metaphysical modality does not seem to be exceptionally transparent. Philosophers who finds themselves interested in the How question, therefore, might be well-advised, at least as a first step, to limit their focus to a slightly more modest ­question that abstracts away from some of the psychological complications of humanity. (How Could) How could we come to know facts about metaphysical possibility and necessity? 3   e.g., in a limited version, van Inwagen (1998). See also the view discussed in O’Leary-Hawthorne (1996, p. 185).

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

126  Jonathan Jenkins Ichikawa The same set of responses are available here as before. (And so likewise with the “How Could” counterparts for justification, reliability, etc.) The same questions would need to be answered about the various connections between, for example, intuitions and modal facts, here as there; here, however, we set aside worries about the actual psychological implementation of the judgments in the wild.4 All of these questions, in the taxonomy I am about to suggest, are part of a single project in modal epistemology. Call this the “How” project. The How project investigates how it is that we or various other people do or can come to have epistemic access to propositions about metaphysical modality. The How question is not the only question one might be concerned to answer; and in some cases, it is not the most interesting one. We often want to know, in addition or even instead, why knowers are able to discern the relevant facts. To motivate this change in focus, consider the following story. Suppose that we were someday to discover, in a distant solar system, a civilization whose members seemed to know quite a lot about King Ludwig II. They had, we will suppose, never travelled to Earth, or spoken to inhabitants of nineteenth-century Bavaria. Nevertheless, when we finally do come to encounter them, they report various beliefs that are suggestive of an intimate knowledge of Ludwig’s life; they tell of his childhood in Hohenschwangau, his obsession with Wagner, his elaborate castles, and the creation of the German Empire. This alien civilization at least appears to have quite a bit of knowledge about Ludwig; this would be a striking fact, calling out for explanation. Perhaps we would investigate their cognitive lives in an attempt to answer questions in the How project: how do they know so much about Ludwig? Correctly answering the How question might well resolve the mystery—if we were to discover, for instance, that they had an extremely powerful telescope that was focused on mid1800s Europe, this would go quite a long way towards explaining what is mysterious about their knowledge of Ludwig. But not all true answers to How questions would so resolve the mystery. Suppose that it came to light that the way our alien knowers come to beliefs about Ludwig is by swimming to the bottom of a particular deep well and meditating. Upon leaving the well, they find themselves with beliefs about Ludwig—beliefs that turned out overwhelmingly to be true. This discovery, though it would truly answer the How question, would not at all dispel the mystery of their access to Ludwig. That this answer is so obviously unsatisfactory indicates that the How question doesn’t always get at the heart of what is of interest in epistemology. One might of course worry that this discovery would not answer the How question, because reliable belief-formation isn’t sufficient for knowing—perhaps the case as 4   Compare Ichikawa and Jarvis (2012, p. 154 fn. 3): “[O]ur paper will focus primarily on the question of how we can have modal knowledge, rather than on the related question how we do come to have it. How we do have modal knowledge is as much a question of psychology as it is epistemology—we won’t pretend to know from the armchair how people do know except to say that they are somehow sensitive to the method by which they can come to have it.”

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Modals and Modal Epistemology  127 described is relevantly like Lehrer’s (2000) Truetemp or BonJour’s (1985) Norman. But even if we grant the internalist assumption, we needn’t interpret the case as relevantly parallel. An important feature of Truetemp and Norman is that they have no independent reason to trust their inclinations. We can give my aliens such an independent reason without undermining the point they illustrate. We may, for instance, suppose that they are making use of a widely calibrated well of clairvoyance: it delivers beliefs about a wide variety of things, many of which turn out upon independent investigation to be true. So they could rationally trust their well as reliable; there is still a deep mystery about cognitive access. This is why it is not sufficient or satisfying merely to argue for the presence of a reliable faculty. One might be convinced, on broadly transcendental grounds like those of Bealer (1992), that we must accept that our intuitions tend by and large to be true; that’s certainly a good-making feature of intuitions, but it’s not the end of the story by any means. A deeper explanation is needed. Here is another example. Suppose that someone, or some group of people, have an apparent ability to tell, without aid of a scale, whenever some object weighs an even number of grams. It’s not just that they have general superhuman discriminatory abilities—I can’t tell the difference by feel between 200g and 202g, and neither can they— but they can always tell you whether, when rounded to the nearest number of grams, that number is even or odd. Even-grammed objects just feel a certain way to these subjects. These people, it seems, have an extraordinary ability that cries out for explanation. Suppose we investigate and find the particular neural mechanisms that underwrite their abilities. This would tell us how they know that this object is an even number of grams, but it would not satisfy our proper curiosity about their surprising ability. Answering the How question often yields only part of the story. So it is with modal epistemology. If, as orthodoxy has it, we have epistemic access to facts about metaphysical possibility and necessity, this is an interesting fact that requires explanation. And answering a How question might not satisfy this requirement. Consider this answer to How: “Upon entertaining certain propositions, we experience conscious episodes with a particular sort of phenomenal character, which we recognize to be a reliable indicator of metaphysical necessity, and so conclude that the propositions are necessary. Thus do we come to knowledge of necessity.” Even if this story were true, it would not provide any interestingly deep explanation for our capacity for modal knowledge. Saying only that the ability derives from a particular sort of rational capacity that allows us to distinguish the necessary and possible, even if true, does little more than does saying that our aliens meditate at the bottom of a well to learn about King Ludwig, or that our subjects use a particular bit of neural machinery, or its attendant phenomenology, to know whether an object has an even number of grams. David Lewis provides an extreme example of a treatment of modal epistemology that is inadequate precisely for its neglect of the Why question: [I]f it is a necessary truth that so-and-so, then believing that so-and-so is an infallible method of being right. If what I believe is a necessary truth, then there is no possibility of being wrong.

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

128  Jonathan Jenkins Ichikawa That is so whatever the subject matter of the necessary truth and no matter how it came to be believed. (Lewis 1986, pp. 114–15)5

And in fact, the situation here, with this answer to the How question for modal truths, is worse than in those analogies posited above. For we have independent access to facts about Ludwig and masses, and overwhelming reason to treat each as objective, mind-independent phenomena. This is less clear in the case of modality; if our best explanation of modal knowledge is such an unsatisfying answer to the How question, we are left with a deep mystery; we might avoid it if we gave up on the orthodoxy that we have access to modal truths, embracing skepticism, or embraced some sort of anti-realism about modality, closing the epistemic gap between truth and judgment. So failure to explain cognitive access could put pressure against the orthodoxy that we’re achieving knowledge of objective facts. Traditional philosophical concerns require more. We seek not merely answers to the How questions, but answers to Why questions as well. (Why) Why do we have the ability to know facts about metaphysical possibility and necessity?6 Correspondingly, we want to know why it is that our clairvoyant aliens are able to know so much about King Ludwig, and why our odd subjects are able to tell whether an object weighs an even number of grams. For instance, if we learned that the subjects in question had been training all their lives at distinguishing objects weighing even numbers of grams, the mystery would be considerably lessened; although we’d still be surprised and impressed that they’d managed it so well, we can at least see a sketch of a story about how the ability came to be.7 Three points arise with respect to the relation between Why questions and How questions. First, although they’re separate questions, they’re not wholly independent. Certain answers to How questions place constraints on Why questions, and vice versa. A How answer, for instance, may tell what mechanism is used to come to knowledge; its corresponding Why answer, therefore, had better be able to explain why that mechanism, or a mechanism with its characteristic features, came to be. One might worry that certain answers to the How questions for modal epistemology might not be consistent with any plausible answers to Why questions.

5   I owe the example to Schechter (2010), who uses it to make roughly the same point. See also O’Leary-Hawthorne (1996). 6   Cf. Schechter’s (2010) distinction between the “operational question” and “etiological question” for the epistemology of logic; these correspond closely to my How and Why questions, respectively. My names are imperfect—one can ask a Why question via “how is it that we are able to know?”—but I hope that the intended contrast is clear. 7   Of course, there can be unsatisfactory answers to the Why question, too—“they evolved a relevant mechanism,” for instance, tells us very little by itself, but might truly answer the Why question. The point is that the Why question is an important one that often needs consideration, not that correct answers to it always resolve the relevant mysteries. Thanks to Derek Ball here.

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Modals and Modal Epistemology  129 Second, in at least some cases, the Why questions go a considerable distance towards resolving the relevant mysteries, even absent an answer to the How questions. If we learn that the subjects spent their childhoods trying to develop the capacity to detect odd-gram-weighted objects, then, even absent a psychological story about how it is they actually do it, the fact that they manage it is not deeply mysterious in the way that discovering it with no explanation at all for how it came to be would be. (According to legend, chicken-sexers are notoriously skilled at distinguishing male from female chickens; it is surprising to learn that they do it by smell, but even before we know this answer to the How question, their ability is rendered rather unmysterious by the obvious answer to the Why question: they acquired this ability because they practiced it, along with the relevant positive and negative feedback.) Third, Why questions may not always be pressing. It is very plausible that evolution provides a rather neat answer to the Why question about our ability to identify dangerous animals by sight. But if it should turn out—surprisingly—that as a matter of historical fact, our capacity to recognize dangerous animals arose by total coincidence, irrespective of natural selection, this wouldn’t lead to great new mysteries about the epistemology of dangerous animals.8 (It might lead to mysteries about human development, but this is neither here nor there.) In some cases, it is enough to see that we have an ability, without regard to where it came from. What distinguishes cases where the Why question is pressing from those where it is not? I don’t know how to give a general answer here, but the cases I presented above do suggest that the Why question is pressing at least sometimes. I also find it pressing in the case of modal epistemology. I suspect that this is so in part—but not wholly—because, unlike in the case of dangerous animals, it’s unclear what independent check there is, either to corroborate our judgments of what is and is not possible, or even to confirm the thought that there is an objective fact about metaphysical possibility. We shall return to the question of to what degree we should find the Why question for modal epistemology pressing in Section 3. In previous work, Benjamin Jarvis and I (Ichikawa and Jarvis  2012,  2013) have developed a treatment of the How question for modal epistemology along broadly rationalist lines. My topic for this chapter is the Why question.

2.  Metaphysical Modals and Quotidian Modals Evolutionary explanations are well-suited to Why questions. If a species developed in an environment where objects weighing an even number of grams are nutritious, but those weighing an odd number of grams are poisonous, we would not be surprised to discover that they’d developed a capacity for sensitivity to the difference. Are there prospects for such answers in the case of modal epistemology? Prima facie, one might think not: knowledge of metaphysical modality does not appear to confer   Thanks to Peter Kung for discussion here.

8

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

130  Jonathan Jenkins Ichikawa any obvious survival or reproductive advantage in the environments in which our species developed. But of course, evolutionary explanations for abilities needn’t cite fitness advantages for those very abilities; if there are related abilities that would have carried fitness advantages, which would have carried the ability for knowledge of metaphysical modality as a side-effect, then the evolutionary strategy is promising. Here there is much more room for optimism. It is not difficult to think that there could have been a fitness advantage to knowledge of quotidian modal facts—the sorts of things we express with modal language in everyday nonphilosophical contexts—and that a capacity for recognitions of such quotidian modals would develop alongside a capacity for recognizing the more distinctively philosophically metaphysical modals that characterize the typical subject matter of modal epistemology. This strategy will be pursued in Section 3. First, however, there is an important bit of stage-setting to do. For the strategy is promising only to the extent that claims of metaphysical modality are importantly similar to claims of quotidian modality. In this section, I argue that this is so. It is widely recognized that the language of possibility is used in several importantly distinct ways. The first distinction standardly drawn in the literature is the distinction between epistemic and non-epistemic modals.9 These sentences can all be read in epistemic or non-epistemic ways:  (1) It’s possible that Renoir painted Beach at Sainte-Adresse. (2) John F. Kennedy might have survived his assassination attempt. (3) Roger Clemens must have used steroids. If I’m matching painters to paintings, and don’t know whether to leave open the possibility that Renoir painted Beach, I’d express an epistemic truth by (1). By contrast, if I’m attempting to deny Kripke’s doctrine of the necessity of origin, (1) would constitute a metaphysical claim—if Kripke is right, a false one.10 I’ll utter (2) epistemically if I am raising a conspiracy theory on which Kennedy’s death was faked; alternatively, I might take the death as given, and use (2) to convey the idea that Kennedy was unlucky that his attacker was successful. The natural reading for (3) is epistemic—just look at his performance late in his career, he must have used steroids—but it could be used to express the implausible claim that, as a matter of metaphysical necessity, Clemens used steroids. But even setting aside epistemic readings, we have stronger and weaker notions of possibility and necessity that can come into play. All of these have true non-epistemic readings in some contexts, even though they’d be false if interpreted as claims about metaphysical modality.  9   Gendler and Hawthorne (2002, p. 3), Bealer (2002, p. 77). We can also distinguish both these readings from deontic interpretations of modals, which I will not discuss in this chapter. The focus of this chapter is on, in Kratzer’s (1981) terminology, modals with realistic modal bases—bases that always include the base world. 10   e.g. Kripke (1980, pp. 113–15).

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Modals and Modal Epistemology  131 (4) No human can outrun a cheetah. (5) It’s impossible for a prisoner to escape from Guantanamo Bay. (6) You can’t win an election by promising tax hikes. So it is common ground that utterances of modal claims such as (1)–(6) can have different truth conditions, depending on the conversational context. Sometimes, but not other times, they express claims of metaphysical possibility and necessity like those that characterize the subject matter for investigations into “modal epistemology.” What accounts for this context-sensitivity? One possible interpretation of these data would suggest that modal terms like ­“possible,” “might,” and “can” are multiply ambiguous. On this suggestion, “possible,” for example, might be ambiguous between several distinct notions, including at least epistemic possibility and metaphysical possibility. This ambiguity thesis is implausible for many reasons. First, the sheer number of senses that need be distinguished should be daunting for anyone who wants to posit that the context-sensitivity of (1)–(6) is explained by ambiguity in the modal terms. Distinguishing metaphysical possibility and epistemic possibility is, of course, only a small first step; sentences like (4) suggest a need for something like “physical possibility”; perhaps (6) demands “political possibility,” etc. If “possibility” is ambiguous in this way, then each sense would have to be independently learned; this is obviously not the case.11 It is also a surprising and unexplained fact, given the ambiguity thesis, that the various meanings of the modal terms are related in systematic ways; each meaning of “must,” for instance, is related to a corresponding “might” in a uniform way; ambiguous terms don’t have such cross-­ meaning generalizations.12 Furthermore, we can combine senses for modal terms in a way not predicted by the ambiguity thesis. We can say, for instance: (7) Your theory is possible with respect to the laws of nature, but not with respect to the evidence. Here “possible” is used only once in a sentence conveying both epistemic and non-­ epistemic modality. The suggestion is a physically possible one (non-epistemic), but one known not to obtain (epistemic). That a single use can be extended to both senses strongly suggests that the word is not ambiguous between them.13 So the fact that

11   The linguistic meaning is the same in each case; this doesn’t imply, of course, that one mightn’t need to learn to identify particular modal bases independently. An ability to recognize “physical possibility” doesn’t guarantee competence with “political possibility” (or metaphysical possibility); one might be linguistically competent with “possible” without knowing, in a given context, which worlds are relevant. (Compare linguistic competence with “everyone,” which does not guarantee that one will know the domain for a given use.) Thanks to Peter Kung. 12   “Walk” and “run” are each ambiguous between a kind of pedestrian locomotion and a baseball event. In the baseball sense, one can in certain circumstances generate a “run” by “walking.” But we have no temptation to carry the generalization into the locomotive sense. 13   See Zwicky and Sadock (1975).

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

132  Jonathan Jenkins Ichikawa “possible” and “necessary” sometimes express metaphysical modality, and sometimes do not, is not explained by an ambiguity in those terms. According to the dominant view in formal semantics, modals are quantifiers over sets of worlds, where the relevant sets of worlds are determined by a function provided by the conversational context.14 We can often force particular readings with adjectives like “physically” or “epistemically,” or we can even relativize our modal claims more specifically with more explicit descriptions of modal bases, using phrases like “in view of what she knows,” or “in view of what I read in the Times last week.” There is a certain philosophical context in which the modal base comprises the set of all metaphysically possible worlds; we can trigger this context manually by affixing “‘metaphysically” to our modal terms. When we speak of “modal epistemology,” we typically mean to be considering the epistemology of these rather particularly philosophical modal propositions. But these propositions are deeply and closely related to those we express with modal language in more quotidian contexts; they differ only in having different modal bases. They differ in subject matter rather than in form.

3.  Quotidian Modals and the How Question So quotidian modals are closely connected to metaphysical modals. To ask whether p is metaphysically possible is to ask the same kind of question as whether p is physically possible, or possible in some more restrictive sense provided by context. We now have available a sketch of an evolutionary approach to the Why question described at the start of Section 2. Quotidian modals certainly have a strong prima facie claim to conferring significant survival advantages on our ancestors. I walked through a dangerous part of town alone at night, and you say to me: (8)  You could have been mugged. In typical contexts, you express not that it is metaphysically possible that I should have been mugged, but that among some nearer, restricted set of worlds there are some in which I was mugged. And it is not mysterious that we should have the cognitive ability to know propositions like this one; both developmentally and evolutionarily, we have faced good reason to develop the capacity for knowledge of such propositions as these. Or at least, this seems like an extremely plausible evolutionary story. It is of course an empirical question to what degree our capacity for quotidian modals was in fact an evolutionary adaptation for planning and better decision-making; there is no contradiction in the idea that it arose spontaneously and by coincidence in a way wholly unrelated to natural selection. If this is in fact so, then there is no substantive answer to 14   Kratzer (1977) is an important seminal work. Chapter 3 of Portner (2009) gives a helpful contempor­ ary overview.

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Modals and Modal Epistemology  133 the Why question for quotidian modals; perhaps then it’d be enough to observe that we have the capacity. Lucky us! Whether by coincidence or (as seems to me much more plausible) by natural selection, it is relatively unmysterious that we have capacities for knowledge of quotidian modals. And these capacities, I suggest, should be general ones that equip us to evaluate modals generally. There is no prima facie reason we should expect the case of metaphysical modality to require different fundamental epistemic capacities for their evaluation. We have a general ability to evaluate, given a modal base, whether a necessity or a possibility claim holds of it. (I shall return later to the question of how we determine what to include in a modal base.) That is the sense in which quotidian modals illuminate the How question for metaphysical modality. But there is another significant question that yet remains unanswered. Why is it that, in the context of philosophical discussions of modality, we work with the domain of  metaphysically possible worlds, rather than some other more or less restrictive domain? If ordinary quotidian modals represent restricted quantification over a limited class of possibilities, what does the general, unrestricted modality include? It’s clear enough that the physically possible is a restriction on the metaphysically possible. One might continue to suggest that the metaphysically possible is a subset of the conceptually or logically possible—and perhaps so on further into various logically impossible “possibilities.” And crosscutting some or all of these might be the epistemically possible. Or one might somewhere draw the line, insisting that one has already identified all of the unrestricted possibilities. But where to draw the line? Bealer (2002) draws the line at the metaphysically possible: [S]ome people insist on distinguishing logical possibility and metaphysical possibility and so are led to the following: p is logically possible iff p is merely consistent with the laws of logic (i.e., not ruled out by logic alone). This usage, however, invites confusion. There are many logically consistent sentences that express obvious impossibilities (e.g., “Bachelors are necessarily women”, “Triangles are necessarily circles”, “Water contains no hydrogen”). If you buy into calling mere logical consistency a kind of possibility, why not keep going? For example: p is “sententially possible” iff p is consistent with the laws of sentential logic. Then, since “Everything is both F and not F” is not ruled out by sentential logic (quantifier logic is what rules it out), would it be possible in some sense (i.e., sententially possible) that everything is both F and not F?! Certainly not to my ear! (pp. 78–9)

Bealer is here too quick, for at least two reasons.15 First, the logical structure of the argument for drawing the line at metaphysical possibility is suspect. It follows a particular erroneous form of a slippery slope argument: if you permit X (which doesn’t seem too bad), then what’s to stop you from going on to permit Y (which seems terrible)? The difference in felt terribleness, if there is one, would provide just the needed   See also Ichikawa and Jarvis (2013, p. 48) for related discussion.

15

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

134  Jonathan Jenkins Ichikawa traction between X and Y in order to avoid the slip. Remember that drawing the line at metaphysical possibility represents a substantive choice; one might try to draw it even more narrowly—at physical possibility, say. Imagine a philosopher who refuses to countenance those “metaphysical possibilities” which violate the laws of physics, insisting that they’re in no sense possible. Against someone like Bealer, who believes in such “physically impossible possibilities,” such a philosopher might offer just the same retort: “if you buy into calling mere metaphys­ ical consistency a kind of possibility, then why not keep going?” Bealer presumably thinks that there is good reason to countenance metaphysical possibility; but for all Bealer has said, it may well be that further conceptual or logical possibilities can be put to similar work. The second reason to be concerned with Bealer’s argument is that he makes an insufficient case for the undesirability of the bottom of his slippery slope. Bealer apparently finds implausible the suggestion that there is a sense in which the sentence in question is possible. He doesn’t say why, but the invocation of how it strikes his ear suggests it may be based in a linguistic intuition; it just sounds terrible, perhaps, to say that there’s a sense in which it is possible that everything is both F and not F.16 But that there is some sense in which it is possible does not, of course, imply that it’s a very interesting sense, or one that ordinary speakers are used to thinking about. Bealer ostends a notion of “sentential possibility,” abstracting away from any use to which thinking about it might be put. It shouldn’t come as any surprise, then, that, so presented, we shouldn’t have any interest in thinking about such “possibilities.” That doesn’t mean they’re not there, ready for us to take them up if and when the course of inquiry demands it. The philosopher who thinks of physical possibility the way Bealer feels about metaphysical possibility will respond similarly to Bealer to what is conventionally recognized as the physically impossible metaphysically possible. “It’s just not possible in any sense for humans to travel faster than light.” If you have introduced such metaphysical issues to undergraduate students, you’ve probably encountered people who respond this way at first. If there is reason to countenance metaphysical possibilities—and I agree with Bealer that there is—then presumably, we will justify it by reference to the useful work to which thinking about metaphysical modality can be put. It is to a considerable degree an open question, to what degree the class of metaphysically possible worlds is a theoretically important one—Ted Sider argues that the distinction is not metaphysically fundamental17—but if Kripke (1971,  1980) is right that metaphysical modality has 16   Brian Weatherson points out, in personal communication, another objection to Bealer’s particular example. The relevant sentence looks as if it might well be metaphysically, let alone “sententially” possible, on the assumption that there is a metaphysically possible world that contains nothing. If so, “everything is both F and not F” is (vacuously) true of that world. 17   Sider (2011, pp. 266 et seq). However, note that Sider’s claim, that metaphysical modality is not (totally) fundamental, is consistent with its being a relatively fundamental notion. Just as green is not fundamental, but is more fundamental than grue, so might metaphysical modality not be fundamental, but yet be rather fundamental—more so than other gerrymandered notions of modality.

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Modals and Modal Epistemology  135 central roles to play in discussion of identity, that might be good reason to treat metaphysical modality as significant at least to the extent that identity is. In the same way, there is room to argue for or against the theoretical significance of further, metaphysically impossible, “possibilities.” For example, Ichikawa and Jarvis (2013) argue at length that we should accept a notion of “rational possibilities” that outstrip the metaphysically possible. One reason we might need such possibilities is to model the rationality of a person in a Frege case: suppose that someone believes, on the basis of excellent evidence, that Hesperus is a planet, and also believes, on similarly strong (but misleading) evidence, that Phosphorus is not a planet. On our view in that book, invocation of rational possibilities—like the possibility that Hesperus is a planet, but Phosphorus is not—is necessary to explain the possible rationality of such a combination of attitudes (Ichikawa and Jarvis 2013, pp. 61–6). This, even though there is no metaphysical possibility corresponding to this rational possibility. Whether or not we were right to invoke such possibilities need not be decided here: the point is that the idea is intelligible, and there are possible grounds for drawing the line in different places than at the boundaries of metaphysical possibility. Why, in the philosophical contexts in which we do so, do we draw the line at the metaphysically possible? Timothy Williamson (2007) objects to a strategy much like mine by emphasizing this question, considering his own counterfactual-based strategy relevantly superior. Williamson writes: Can metaphysical possibility be understood as the limiting case of such more restricted forms of possibility? Perhaps, but we would need some account of what demarcates the relevant forms of possibility from irrelevant ones, such as epistemic possibility. It also needs to be explained how, from the starting-point of ordinary thought, we manage to single out the limiting case, metaphysical modality. The advantage of counterfactual conditionals is that they allow us to single out the limiting case simply by putting a contradiction in the consequent; contradictions can be formed in any language with conjunction and negation. (p. 178)

The relation between quotidian modality and metaphysical modality, Williamson suggests, is not so straightforward as automatically to dispel the mysteriousness of sui generis faculties for metaphysical modality. To say that knowledge of metaphysical modality is obtained via the exercise of our general faculty for modals is to leave an important question unanswered: why is it that we countenance the class of possibilities that we do, rather than some other one? Why don’t we countenance further “possibilities” that extend beyond the metaphysically possible (as we do in sentences like (1))? Why do we countenance possibilities as far out as we do, rather than considering some more or less restricted set of possibilities as the limiting case? There is nothing about our practice of quotidian modal reasoning, for example, that requires us to counten­ ance faster-than-light travel as “possible,” or denials of a posteriori identities as “impossible” in the relevant metaphysical sense. I agree that this question goes unanswered by an evolutionary approach to the Why question in terms of quotidian modals. However, I do not think we should be too

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

136  Jonathan Jenkins Ichikawa bothered by this particular failure. For one thing, his own claim to the contrary notwithstanding, Williamson’s favored approach fails to answer this question in just the same way. I argue that this is so in Section 4. In Section 5, I’ll suggest why I think it’s just fine to leave this question open at this stage.

4.  Counterfactuals and Modality Williamson’s preferred (2007) approach, following that of Hill (2006), emphasizes connections between metaphysical modality and counterfactuals. The capacity to evaluate counterfactual conditionals, like that for quotidian modals, has a strong prima facie claim to conferring an evolutionary advantage on a subject who has it. So it is no great mystery that we ended up with a capacity for counterfactual conditionals; the Hill/Williamson strategy is to understand our capacity for metaphysical modality as a natural side-effect of this ability. In particular, both Hill and Williamson argue that claims of possibility and necessity are equivalent to certain counterfactuals. They prove several equivalences; I’ll focus on these: (NEC) □ ≡ (∼ p □→⊥) (POS) ◊ ≡ ∼ (p □→⊥)

18

(Where ⊥ stands for an arbitrary contradiction.) There is a strong intuitive case for the truth of (NEC) and (POS). A counterfactual conditional A □→ B is true just in case, for some relevant restricted class, all of the possibilities in which A are possibilities in which B.19 For all of the possibilities in which A to be contradiction-possibilities just is for A to be impossible. Given these equivalences, Williamson suggests, a capacity for counterfactual conditionals will bring with it a capacity for metaphysical modality. The derivation of these equivalences relies, as Williamson observes, critically on the  assumption that counterfactual conditionals with impossible antecedents are vacuously true.20 This feature of Williamson’s argument has been targeted by some philosophers sympathetic to nontrivial “impossible worlds.”21 As will emerge, my own criticism of the Hill/Williamson view is not unrelated to this strategy, although I am not interested in denying that (NEC) and (POS) are true. 18   Williamson (2007, p. 157), my labels. For other similar formulations, see Williamson (2007, pp. 156–9) and Hill (2006, pp. 220–3). 19   This statement is neutral between various semantic approaches to counterfactual conditionals. On David Lewis’s (1973) account, for instance, the relevant class constitutes roughly the set of A-worlds that are most similar to actuality. We shall see below that alternate approaches to counterfactuals, such as those that treat them as restricted modals, also very directly yield this result. (If there are no A possibilities, then, trivially, all A possibilities are B, for any arbitrary B.) 20   Or at least, never false; he notes (p. 176) that his main point can still be observed if such conditionals are treated as defective in a way that results in their having no truth value. 21   Roca-Royes (2011). See also Jenkins (2008, p. 694).

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Modals and Modal Epistemology  137 Another challenge for the Hill/Williamson story—one pressed by Jenkins (2008)—is that it is difficult to see how it helps actually to explain modal knowledge, unless it’s thought that modal truths are inferred from the corresponding counterfactuals. Notice, for instance, the lacuna in this description of the view from Hill: We may believe that it is metaphysically necessary that 2 x 3 = 6, but is this something that we know? Can we be said to know that it is metaphysically necessary that Lord Darnley and Mary Queen of Scots were the parents of James I? As I see it, it is plausible that we do in fact know such things to be true. Thus, if the theory of the metaphysical modalities I have recommended is right, then claims of metaphysical necessity are equivalent to generalized subjunctive conditionals. It seems reasonable to suppose that it is possible to know such conditionals, for they are concerned principally with questions of large-scale similarity, and it seems reasonable to suppose that we can know the answers to such questions, at least in some cases. (Hill 2006, p. 231)

Hill goes on to offer further reasons to think that knowledge of the relevant conditionals is possible, but does not remark on the possibility that subjects who have not recognized the equivalence might have this conditional knowledge, but fail to have the modal knowledge. As Jenkins points out in her response to Williamson, this is a significant omission, as epistemology is a hyperintensional matter. Note that this challenge does not arise for my approach, which emphasizes a general capacity for modals: the posited capacity just is a capacity for modal judgments. It seems to me that there is something importantly right about the Jenkins objection, although I also think that there is room for disagreement about just how damaging it is. Given the focus I have been suggesting on the Why question, an explanation for our capacity to deal with something equivalent to claims of metaphysical modality might be thought to resolve a significant degree of the question; the outstanding issue turns on the degree to which the unexplained capacity for drawing out the equivalence is a cognitive achievement demanding substantive epistemological explanation. I am genuinely unsure to what degree we should think this is so, so I will not press the Jenkins objection further. I consider it to be somewhere between a weak and a strong reason to be dissatisfied with the Hill/Williamson line. My focus will be on a different challenge. The point I want to emphasize is that the Hill/Williamson line is also susceptible to the challenge Williamson himself presses for the quotidian modals approach. Ordinary counterfactuals, like ordinary modals generally, are context-sensitive.22 For instance, it is widely recognized that in different contexts, different features of actuality are weighted more heavily with respect to comparisons of similarity.23 But such weighting of respects of similarity does not exhaust the context-sensitivity of counterfactuals. According to the dominant view in semantics, counterfactuals are

22   Williamson (2007, pp. 173–4) mentions this widespread view without disputing it; he dismisses some objections to his view deriving from context-sensitive counterfactual conditionals, but does not consider the one I will press. 23   Cf. the “Caesar in Korea” counterfactuals; see Lewis (1979, p. 457).

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

138  Jonathan Jenkins Ichikawa restricted modals; like modals generally, they take a context-sensitive modal base.24 When more possibilities are considered as live possibilities—when there is less in the common ground—counterfactual sentences express stronger propositions. Consider this counterfactual: (9)  If this glass were to fall, it would break. In ordinary contexts, we ignore certain distant possibilities that would, if actualized, sever the connection between falling and breaking. Here, it is presupposed that no one will reach out of a window and catch the glass on the way down. Of all the possibilities in the modal base, only ones in which the glass breaks are among those in which it falls. So in ordinary contexts, (9) is true. But in some extraordinary contexts, we countenance more possibilities as relevant. If we are talking about the surprising possibilities in which people reach out of their windows to catch the glass, considering them to be, though distant, ones that are consistent with the conversational background, then (9) would express a falsehood, since, were the glass to fall, it might well be caught on the way down.25 The counterfactual conditional is understood as a restricted modal; (9) is interpreted as (Must:fall)(break). The modal base for the counterfactual, like modal bases for bare modals, is context-sensitive. So there is an intimate connection between modals and counterfactual conditionals; the latter are a particular restricted kind of the former. It is trivial that Williamson’s equivalences follow from this treatment of counterfactual conditionals.26 So the relevant counterfactuals do turn out, on the orthodox semantics, to be logically equivalent to the claims of possibility and necessity. Williamson is, to that extent, correct. But—and here is the problem—the bare modals and the counterfactual conditionals each take context-sensitive modal bases. The “□” and “◊” are each relativized, in a context, to a given modal base; as, indeed, is each “□→”. Williamson has not shown that there is anything special about the particular context that provides the set of metaphysical possibilities. Quotidian engagement with counterfactuals doesn’t tell us which modal bases to work with. The Hill/Williamson account leaves open the very same questions that mine does. Consider an example. I’m a human being, with ordinary physical capacities. In particular, I can’t fly home from work today. I could bicycle home, or take the bus, but I can’t fly home from work today. That’s a nice modal, true in its quotidian context: (10)   □~(Jonathan flies home from work today)   Kratzer (1986). See also Ichikawa (2011).   Those attracted to conditional excluded middle, of course, will disagree. I will not go into the arguments for and against conditional excluded middle here—see Williams (2010). I assume, with semantic orthodoxy and against Williams, that conditional excluded middle is false. 26   Consider (NEC) □p ≡ (~p □→ ⊥). The right-hand side is understood as (□:~p)(⊥)—all not-p worlds are contradiction worlds; i.e. there are no not-p worlds. 24 25

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Modals and Modal Epistemology  139 When we say (10), there are certain distant scenarios that we ignore. If we’re feeling fanciful, we may expand the common ground to open up some such possibilities; we might then end up in a context in which I could say, “I could fly home in a helicopter today,” or even, if we were feeling more fanciful still, “I could sprout wings and fly home by my own power today.” In such contexts, (10) is false. But in its natural context, (10) is true. But notice what happens when we evaluate the counterfactuals that I said to be equivalent to these bare modals. Consider again a down-to-earth, unfanciful context in which (10) is true. What about this counterfactual? (11) If Jonathan were to fly home today, then grass would be both green and not green. This counterfactual is not plausible; there wouldn’t be any true contradictions, regardless of whether I were to fly home today. But given (NEC), (10) and (11) are provably equivalent. What has gone wrong? Pairs of sentences like (10) and (11), though truth-­ conditionally equivalent in every context, may have divergent effects upon the conversational score, such that the utterance of (10) will result in one context in which it (and the unuttered (11)) is true, while the utterance of (11) will result in another context in which it (and the unuttered (10)) is false. This is likely in this particular case, because we tend to shift contextual parameters to accommodate utterances as nonvacuous. Consider an example with ordinary quantifiers. You and I are hosting a barbecue. You say: (12)  Everyone has been fed. The domain for your “everyone” is a restricted one. You’re excluding, for instance, Aunt Joan, who is dieting, as well as the many people who were not invited to our barbecue. Plausibly, you speak truly, because everybody who came to the barbecue, expecting to be fed—everybody in the domain—has been fed. Now suppose I, with a perverse fondness of linguistic games, say: (13)  Everyone who hasn’t been fed has been fed. I will most naturally be interpreted as saying something false—indeed, as saying something ridiculously false. But of course, (12) entails (13) in every context. Holding the domain of people fixed, if everyone in that domain was fed, then everyone in that domain who had any feature at all was fed. In particular, everyone in that domain who wasn’t fed was fed. There was nobody like that, so (13), so evaluated, is vacuously true. But in general, we don’t like to interpret sentences as vacuously true if we don’t have to. So if (13) is uttered, we naturally accommodate the presupposition that there are some people who haven’t been fed, expanding the domain, perhaps to include Aunt Joan or some strangers. In its new domain, (13) is false, rather than vacuously true.

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

140  Jonathan Jenkins Ichikawa Exactly the same thing happens with (10) and (11). When I say I can’t fly home, I mean that, of a certain set of worlds, none of them are ones in which I fly home. If I say that, were I to fly home, such and such would be the case, I say, of a certain set of worlds, that all of the ones in which I fly home are ones in which such and such is the case. Even if we were previously in a context in which there were no flying-home worlds in the modal base, when I utter (11), we will typically shift to one in which there are some, in order to accommodate me as speaking nonvacuously. This accommodation is not mandatory. Suppose I start talking about what it’d be like if I were to fly home today. You may accommodate me by allowing a modal base that includes flying-home worlds. But you might instead refuse to countenance such possibilities. There are a few ways you might do this. One rather direct way would be baldly to assert that I couldn’t fly home, thus signaling your refusal to shift with me into a modal base that includes flying worlds. Or you could make the same point less directly by saying something like: “if you flew home, I’d eat my hat” or “it’d be the case that p & ~p”. That is, you might say something tantamount to, all of the possibilities in which I fly home are not possibilities at all. To summarize: sentences expressing bare necessity are equivalent to sentences vacuously expressing certain counterfactuals, in all contexts. But often, those latter are accommodated via a shift in conversational context to render the counterfactuals in question nonvacuous. A conversational participant may be more or less cooperative in so accommodating. We’re left, then, wanting to ask the counterfactuals approach the same question Williamson emphasized for the quotidian modals approach: why do we refuse to expand the domain, accommodating the relevant counterfactuals, at the set of metaphysical possibilities, rather than somewhere else? There is nothing obviously privileged about stopping here; the domain might be extended in either direction. We might, for instance, suppose various metaphysical impossibilities without inferring any sort of explosion: if I were a whale, I would have a tail, but not: if I were a whale, I would climb trees. Or we might refuse to countenance any physical impossibilities, declining to accommodate the antecedents of conditionals like “if humans could fly, then p & ~p,” rendering them vacuously true. Or we might do the same for any other of the myriad possible restricted notions of possibility. Williamson’s way of distinguishing the distinctively metaphysically possible relies on an as-yet-unexplained tendency for us to be just open-minded enough, but no more. One might object that it isn’t necessary to explain why we stop where we do; it’s enough to observe that we do so. This does seem to me a good objection, but it is not an available defense for Williamson; his complaint against the quotidian modals approach was precisely that it fails to explain that which his own view also fails to explain. My response is that his own approach exhibits just the same shortcoming that he pointed out for the quotidian modals approach: it fails to provide “some account of what demarcates the relevant forms of possibility from irrelevant ones” and to explain “how,

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Modals and Modal Epistemology  141 from the starting-point of ordinary thought, we manage to single out the limiting case, metaphysical modality.” Recall the advantage that Williamson claimed for his own approach: “The advantage of counterfactual conditionals is that they allow us to single out the limiting case simply by putting a contradiction in the consequent.”27 But contradictions in the consequent do not serve to identify the limiting case; they merely serve to flag that we have reached our limit. There is nothing in the semantics of counterfactual conditionals that requires us to draw the line at metaphysical possibilities, rather than somewhere else. And Williamson hasn’t shown us anything about the ordinary practice of counterfactuals that demands the line be drawn just there.

5.  The Way Metaphysical Modality is Sui Generis So the Hill/Williamson treatment cannot answer Williamson’s challenge any more than mine can. (We have seen that his also faces additional challenges.) Where now do we stand? The deficiency of the quotidian modals approach is not total. It does offer an avenue for explanation of why it is that we should have the capacity to think modally; what it lacks is an explanation for our thinking modally with respect to the particular domain of the metaphysically possible worlds. That is to say, what it lacks is an explanation for our engaging with the particular subject matter that we do. It does not explain why we think about the set of metaphysical possibilities. The evolutionary approach to the Why question in terms of quotidian modals does not answer this question. But the question is not a deeply mysterious one. In general, when seeking an explanation for our thinking about the particular subject matters that we do, one needn’t look to evolutionary history; one might just as well look to our particular individual developmental histories. The reason we think about the set of metaphysical possibilities is that we’re trained to do so; plausibly, this is because the set has interesting properties that have emerged in our philosophical investigations. To take one prominent example, questions about metaphysical possibility have proven to be extremely fruitful for theorizing about identity. Cf. Kripke (1971, 1980). As the remarks of the previous section show, we face no great pressure to suppose that people often use beliefs or knowledge about distinctively metaphysical possibilities in their quotidian lives. Our modals choose from a variety of bases, depending on the needs of the conversation. Usually, they expand as far as necessary in order to accommodate the various utterances we encounter; only occasionally do we refuse to accommodate—and when we do, we may, and often do, so refuse at a point other than   Williamson (2007, p. 178).

27

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

142  Jonathan Jenkins Ichikawa that which demarks the boundaries of the metaphysically possible. We’re trained to be sensitive to this particular line in our philosophy courses.28 Does this represent a concession to the kind of anti-realism I was motivated to avoid in the introductory sections?29 Not necessarily—it all depends on whether the distinction we’re trained to respect is a genuinely theoretically interesting one.30 For the reasons mentioned above having to do with identity, it seems to me rather plausible that they are; but detailed investigation here would take us too far afield. I have not refuted the anti-realist suggestion that the only reason one might be interested in metaphysical modality is to satisfy the perverse and idiosyncratic interests of philosophers, but I have given a framework for an epistemology that does not require it. So there is a sense in which metaphysical modality is sui generis. But it need require no sui generis faculty; it merely engages a sui generis subject matter. And there is no great mystery why we should choose to think about any particular subject matter. So the evolutionary approach in terms of quotidian modals does show what needed showing: that it is unmysterious why we should have the relevant faculty, even on the assumption that metaphysical modality comprises an objective realm.

Acknowledgements Thanks to Derek Ball, Yuri Cath, Torfinn Huvenes, Carrie Ichikawa Jenkins, Josh Schechter, Stewart Shapiro, Margot Strohminger, and Timothy Williamson for discussion of early drafts of this chapter and related issues. Special thanks to Peter Kung and an anonymous referee for detailed comments on a late-stage draft. Versions of this chapter were presented at the Knowledge Through Imagination conference in Claremont, CA, and a workshop on metaphysical knowledge in Paris; thanks to helpful audiences at both events for discussion. Some of the research for this chapter was supported by the Arts and Humanities Research Council as part of Arché’s Intuitions and Methodology project. 28   Or so it seems to me. Of course, this is an empirical question; one might discover that humans become sensitive to this distinction in some other way. Shaun Nichols, in yet-unpublished work, is investigating young children to see whether they treat the category of metaphysical modality in a privileged way. One can, perhaps, get traction on this question by asking children to distinguish, of various events that are impossible in some more restricted sense, which might be possible “by magic,” and which not even magic could achieve. Such is (a simplified version of) Nichols’s strategy; I look forward with interest to engaging with his results. It is also possible, as an anonymous referee pointed out to me, that particular contingencies of our cultural backgrounds could end up playing a role in influencing our selection of this category of possibility. This is of course a deeply empirical question. 29   Thanks to Peter Kung for raising this question. 30   It isn’t trivial even to articulate what the realism in question comes to; that the truths of metaphysical modality are objective facts is insufficient for the more interesting forms of realism. Consider what might be a clear case of an anti-realist modality: fashion modals. (Thanks to Peter Kung for the suggestion.) We’re told that “one can’t wear white after Labor Day”—this means that, among a certain set of worlds (the worlds where one abides by the fashion rules), there are none where one wears white after Labor Day. It is of course an objective fact whether or not any of those worlds are ones in which one wears that then; but the rules of fashion are paradigmatically arbitrary. We should be anti-realists there if anywhere. Presumably the way to carve the realist distinction we want, then, involves whether the modal base we use (or the properties of worlds we use to identify the modal base) is a (reasonably) natural one. See Sider (2011, ch. 4).

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Modals and Modal Epistemology  143

References Bealer, George (1992). The incoherence of empiricism. Aristotelian Society Supplementary Volume 66: 99–137. Bealer, George (2002). Modal epistemology and the rationalist renaissance. In Tamar Szabό Gendler and John Hawthorne (eds), Conceivability and Possibility. Oxford: Clarendon Press, pp. 71–125. BonJour, Laurence (1985). The Structure of Empirical Knowledge. Cambridge, MA: Harvard University Press. Cappelen, Herman (2012). Philosophy Without Intuitions. Oxford: Oxford University Press. Chalmers, David J. (2002). Does conceivability entail possibility? In Tamar Szabό Gendler and  John Hawthorne (eds), Conceivability and Possibility. Oxford: Clarendon Press, pp. 145–200. Gendler, Tamar Szabό, and John Hawthorne (2002). Introduction: Conceivability and possibility. In Tamar Szabό Gendler and John Hawthorne (eds), Conceivability and Possibility. Oxford: Clarendon Press, pp. 1–70. Hill, Christopher S. (2006). Modality, modal epistemology, and the metaphysics of consciousness. In Shaun Nichols (ed.), The Architecture of the Imagination: New Essays on Pretence, Possibility, and Fiction. Oxford: Oxford University Press, pp. 205–35. Ichikawa, Jonathan (2011). Quantifiers, knowledge, and counterfactuals. Philosophy and Phenomenological Research 82(2): 287–313. Ichikawa, Jonathan, and Benjamin Jarvis (2012). Rational imagination and modal knowledge. Noûs 46(1): 127–58. Ichikawa, Jonathan Jenkins, and Benjamin W. Jarvis (2013). The Rules of Thought. Oxford: Oxford University Press. Jenkins, C. S. (2008). Modal knowledge, counterfactual knowledge and the role of experience. Philosophical Quarterly 58(233): 693–701. Kratzer, Angelika (1977). What “must” and “can” must and can mean. Linguistics and Philosophy 1(3): 337–55. Kratzer, Angelika (1981). The notional category of modality. In Hans-Jürgen Eikmeyer and Hannes Rieser (eds), Words, Worlds, and Contexts: New Approaches in Word Semantics. Berlin: Mouton de Gruyter, pp. 38–74. Kratzer, Angelika (1986). Conditionals. Chicago Linguistic Society 22(2): 1–15. Kripke, Saul (1971). Identity and necessity. In Milton K. Munitz (ed.), Identity and Individuation. New York: New York University Press, pp. 135–64. Kripke, Saul (1980). Naming and Necessity. Cambridge, MA: Harvard Univrsity Press. Lehrer, Keith (2000). Theory of Knowledge. 2nd edn, Boulder, CO: Westview Press. Lewis, David (1973). Counterfactuals. Cambridge, MA: Harvard University Press. Lewis, David (1979). Counterfactual dependence and time’s arrow. Noûs 13(4): 455–76. Lewis, David (1986). On the Plurality of Worlds. Oxford: Blackwell. O’Leary-Hawthorne, John (1996). The epistemology of possible worlds: A guided tour. Philosophical Studies 84(2): 183–202. Portner, Paul (2009). Modality. Oxford: Oxford University Press. Roca-Royes, Sonia (2011). Modal knowledge and counterfactual knowledge. Logique et Analyse, 54(216): 537–52.

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

144  Jonathan Jenkins Ichikawa Schechter, Joshua (2010). The reliability challenge and the epistemology of logic. Philosophical Perspectives 24(1), Epistemology: 437–64. Sider, Theodore (2011). Writing the Book of the World. Oxford: Oxford University Press. Sosa, Ernest (2007). A Virtue Epistemology. Oxford: Oxford University Press. van Inwagen, Peter (1998). Modal epistemology. Philosophical Studies 92(1): 67–84. Williams, J. Robert G. (2010). Defending conditional excluded middle. Noûs 44(4): 650–68. Williamson, Timothy (2007). The Philosophy of Philosophy. Oxford: Blackwell. Yablo, Stephen (1993). Is conceivability a guide to possibility? Philosophy and Phenomenological Research 53(1): 1–42. Zwicky, Arnold M., and Jerrold M. Sadock (1975). Ambiguity tests and how to fail them. Syntax and Semantics 4: 1–36.

Dictionary:

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

6 Imagining Under Constraints Amy Kind

The claim that imagining does not provide us with any information about the world has often seemed a truism in philosophical discussion. When I imagine that a pumpkin has been transformed into a stagecoach, or that a motley group of escaped zoo  animals have taken refuge in my backyard, or that I’ve won the Publisher’s Clearinghouse Sweepstakes, my imaginings give me no reason to believe that such events have occurred. Three basic (and related) features of imagining account for its apparent inability to provide us with such reasons. First, imagining is typically under our voluntary control. Second, imagining is not world-sensitive—the content of an imagining is determined by the imaginer, not by the world. Third, imagining is uninformative—an act of imagining can provide us with no new information. As Sartre has claimed, “nothing can be learned from an image that is not already known” (Sartre 1948, p. 12). In all three of these respects, imagining seems quite unlike epistemically respectable sources of reasons such as perceiving. Thus, imagination has generally been taken to be, as Brian O’Shaughnessy has said, “out of the cognitive circuit” (O’Shaughnessy 2000, p. 345). In my view, however, this conclusion is unwarranted and derives at least in part from the mistaken assumption that all cases of imaginings are roughly comparable to the examples above. In fact, there are all sorts of other examples where imaginings play a role in justifying the beliefs that they prompt. Many of these epistemically significant imaginings are perfectly ordinary ones. When trying to decide whether to become parents, a young couple might call upon their imagination in various ways to help them make their decision—from imagining themselves grappling with exhaustion after a sleepless night with a crying baby to imagining themselves proudly watching a teenager graduate from high school.1 When these prospective parents head off to buy a 1   In her work on transformative experience, L. A. Paul (2014) argues that imaginings cannot help when determining whether to become a parent for the first time; on her view, “if you’ve never had a child, it is impossible to make an informed, rational decision by imagining outcomes based on what it would be like to have your child, assigning subjective values to these outcomes, and then modeling your preferences on this basis” (Paul 2014, p. 83). Though I disagree, my argument in this chapter does not rely on this particular example. One could think instead of parents trying to decide whether to have a second child.

Dictionary: NOSD

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

146  Amy Kind new car, they might again call upon their imagination to figure out what to buy—while looking at the different models in the showroom, they might imagine themselves getting their child strapped into the car seat in the back, fitting the stroller and other gear in the trunk, and cleaning milk spills and crumbs (or perhaps worse) off the interior. And likewise, when they are converting a former office into a nursery, they might imagine the crib first against one wall and then against another. All of these imaginings seem to have justificatory force as they make their decisions—as they determine that they do want, and are ready to have, a child; that they’d do best to buy a mini-van; and that the crib needs to go on the wall opposite the windows. Once we consider such imaginings, it seems clear that the imagination has considerably more epistemic significance than it has been traditionally assigned. In at least some cases, imagining has a role to play in justifying our contingent beliefs about the world.2 But recognizing this fact does not itself provide us with any explanation of how imagining can have the sort of epistemic significance that it does. Developing such an explanation will be the aim of this chapter. My project here thus turns on providing an answer to the following question: What distinguishes the epistemically ­significant imaginings from the epistemically insignificant ones? In my view, the answer has to do with the constraints under which we operate when we engage in the former sorts of imaginings, constraints that are absent in the latter sorts of i­ maginings. The mention of constraint here might seem puzzling since, as David Hume famously said in the Treatise, nowhere are we more free than in our imagination. But the freedom we enjoy when imagining does not show that we must always proceed completely unfettered, and in fact it is our ability to constrain our imaginings in light of facts about the world that enables us to learn from them. In this chapter, then, I offer a framework for showing when and how an imaginative project can play a justificatory role with respect to beliefs about the world, an account of imagining that I call imagining under constraints.

1.  Stage Setting To start, it will probably be useful if I say something about what I take imagining to be. Importantly, however, my account of imagining under constraints—and, correspondingly, my defense of the epistemic relevance of the imagination—does not depend on a particular theory of imagination. For our purposes here, we need only understand imagination as distinct from related states such as supposition or the entertainment of propositions. When engaged in a reductio proof, someone might suppose that Congress passes an assault weapons ban, or that Texas secedes from the United States, or that Elvis Presley is still alive. Such suppositions can be made without any difficulty at all, and without the exertion of much mental energy. Merely bringing the relevant proposition to mind is enough. In order to imagine these things, however, something 2   Elsewhere, I have given arguments in support of the epistemic significance of imagination and I have attempted to diagnose the failures in arguments to the contrary (see Kind, in press).

Dictionary: NOSD

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagining Under Constraints  147 more is required. Different philosophers give different accounts of how to understand this requirement; on my own view, for example, it is to be cashed out in terms of mental imagery.3 But I do not here presuppose this imagistic account. Rather, I rely only on the assumption that imagining involves a more active effort of mind than does supposition or entertaining a proposition, the kind of effort that Kendall Walton seems to be gesturing at when he suggests that imagining “is doing something with a proposition one has in mind” (Walton 1990, p. 20). It would also probably be helpful were I to say something more about the kind of epistemic significance with which I am here concerned. For example, although philosophers typically deny that imagination can play a justificatory role with respect to our contingent beliefs about the world, that’s not to say that they deny it any epistemic import altogether. No one denies, for example, that an act of imagination can jog one’s memory, as when I’m reminded that there’s no dog food in the house as a result of imagining my dog sleeping on the couch. Likewise, no one denies that an act of imagination can lead to interesting innovations, discoveries, or new directions of research. To give just one particularly famous example: At the age of 16, Albert Einstein imagined himself chasing a beam of light, an imaginative exercise that he credits as having played a key role in the development of his theory of special relativity. So my opponent’s charge that imagination is epistemically insignificant is consistent with the claim that our imaginings might generate various beliefs about the world, beliefs that might even turn out to be true. Rather, my opponent means only to deny that imaginings can justify—or even play a role in justifying—such beliefs. I will not here offer any sort of detailed epistemology, and it is my intention to stay clear of debates about (for example) the nature of justification. So perhaps I can best put my point like this: Just as perception teaches us something about the world, so too can imagination. Interestingly, there is one domain in which imagination has been assigned precisely the sort of justificatory power in which we’re interested, namely, the domain of modal truths. Suppose I imagine that there is a heretofore undiscovered planet orbiting the sun between the orbits of Jupiter and Saturn, that the Los Angeles Angels win the next World Series, or that Arnold Schwarzenegger is elected President of the United States. In each case, my imagining the relevant states of affairs gives me reason to believe that this state of affairs—even if it is not astronomically, athletically, or politically possible— is nonetheless metaphysically possible. Perhaps this reason can be overridden by other reasons; perhaps it can’t. But it seems that imagining some state of affairs S is, at the very least, relevant for the justification of the belief that S is metaphysically possible.4 This claim, which I will call the Imagining-Possibility (I-P) principle, is often loosely captured by saying that imagination serves as an epistemic guide to possibility.5   This imagistic account is developed in Kind (2001).   Throughout I’ll talk of imagining states of affairs, but nothing should hang on this assumption. Those who think we imagine sentences or propositions should be able to make the appropriate substitutions. 5   Many claim that it is conceiving—of which imagining is a subspecies—that serves as an epistemic guide to possibility. See the discussion in Gendler and Hawthorne (2002). 3 4

Dictionary: NOSD

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

148  Amy Kind The I-P principle, though widely held, is by no means uncontroversial. Perhaps the most obvious counterexamples arise when we consider imaginings under conditions of ignorance. Someone lacking even a basic understanding of chemistry might be able to imagine that water is not H2O, but we surely do not want to conclude that it is metaphysically possible that water is not H2O. Proponents of the I-P principle thus typically impose conditions that an imagining must meet in order to serve as an epistemic guide to possibility. For example, one important condition is that such claims must not be merely prima facie imaginable—that is, imaginable on first appearances—but must rather be ideally imaginable—that is, imaginable on ideal rational reflection.6 Or, to put things in Cartesian terms, our imaginings must be clear and distinct. Thus, when philosophers claim that the imagination provides us with reasons for our beliefs about metaphysical possibility, it is by no means intended that every imagining can play this role. I take this to be an obvious point about the role of imagination in modal epistemology, but I raise it now to make a parallel point about the central claim in this chapter. In arguing that the imagination provides us with reasons for additional beliefs (i.e. beliefs in addition to those about metaphysical possibility), it is likewise by no means intended that every imagining can play this role, and the fact that there are all sorts of examples of imaginings which lack justificatory power will not count against our central thesis. Rather, just as it seems obvious that we have to impose conditions on when an imagining will be of the right sort to provide us with reason for belief about metaphysical possibility, it should be obvious that we will need to impose conditions on when an imagining will be of the right sort to provide us with reason for other kinds of belief.7

2.  Ideal Imagination Unsurprisingly, it turns out that the notion of ideal imagination—or at least a kind of ideal imagination—turns out to be important not only with respect to imagination’s epistemic relevance for beliefs about metaphysical possibility but also with respect to imagination’s epistemic relevance more broadly. But the sort of ideal imagination with which we will be concerned is not one that contrasts primarily with prima facie imagination as described above. Consider again a paradigmatic case in which an imagining lacks the kind of epistemic relevance in which we’re interested, such as when I imagine that a motley group of escaped zoo animals have taken refuge in my   See e.g. Chalmers (2002).   That said, having invoked an analogy between my project here and the I-P principle, I should note explicitly that I do not mean to put too much weight on this analogy. Importantly, there seems to be a conceptual connection between imagining and possibility that helps to explain why imaginings can be a guide to possibility, a connection that is lacking between imagining and actuality. As Gendler and Hawthorne note, when we imagine, “the things we depict to ourselves frequently present themselves as possible, and we have an associated tendency to judge that they are possible” (Gendler and Hawthorne 2002, p. 1). But when we imagine, the things we depict to ourselves are typically not presented to us as actual. 6 7

Dictionary: NOSD

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagining Under Constraints  149 backyard.8 Not only is this state of affairs easily imaginable on first appearances, but there’s no reason to doubt that it remains imaginable even under ideal rational reflection of the sort mentioned above. Rather, there must be something else that keeps this imagining from having justificatory power—something else that keeps this imagining from being ideal in the sense required for an imagining to have epistemic relevance. But what is this relevant sense? Answering this question will be the task of this section. Here, as so often in philosophy, we are helped by an example from science fiction. In particular, consider the imaginative capacities of the machines who populate the fictional world described in “The Last Evolution,” a short story written in 1932 by American science fiction author John W. Campbell: It was 2538 years After the Year of the Son of Man. For six centuries mankind had been developing machines. The Ear-apparatus was discovered as early as seven hundred years before. The Eye came later, the Brain came much later. But by 2500, the machines had been developed to think, and act and work with perfect independence. Man lived on the products of the machine, and the machines lived to themselves very happily, and contentedly. . . . Machines—with their irrefutable logic, their cold preciseness of figures, their tireless, utterly exact observation, their absolute knowledge of mathematics—they could elaborate any idea, however simple its beginning, and reach the conclusion. From any three facts they even then could have built in mind all the Universe. Machines had imagination of the ideal sort. They had the ability to construct a necessary future result from a present fact. But Man had imagination of a different kind, theirs was the illogical, brilliant imagination that sees the future result vaguely, without knowing the why, nor the how, and imagination that outstrips the machine in its preciseness. Man might reach the conclusion more swiftly, but the machine always reached the conclusion eventually, and it was always the correct conclusion. By leaps and bounds man advanced. By steady, irresistible steps the machine marched forward.

Campbell’s story depicts a future in which superintelligent and powerful machines take care of all human needs. But for all their intelligence and power, and despite their ­artificial eyes and ears, the machines are said to lack the kind of imagination that ­enables humans to make progress “by leaps and bounds.” Instead, the machines have imagination of a different sort—an imagination that Campbell himself describes as ideal. The ideal imagination of the machines allows them to reach correct c­ onclusions— not only conclusions concerning metaphysical possibilities, but also conclusions ­concerning contingent facts about the world. Indeed, it is through the exercise of their ideal imagination that the machines are able to figure out not only how to counter a devastating attack from an alien force but also how to transcend their material 8   Though such an imagining may have epistemic relevance for our beliefs about metaphysical possibility—­ while it may help to justify my belief that it’s metaphysically possible that a motley group of escaped zoo animals have taken refuge in my backyard—the imagining lacks any further epistemic relevance. Hereafter, when I talk of an imagining having epistemic relevance, it should be understood in this latter sense, i.e. I mean to be talking about the epistemic relevance of the imagining over and above whatever epistemic relevance it has with respect to our beliefs about metaphysical possibility.

Dictionary: NOSD

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

150  Amy Kind e­ xistence (the titular “last evolution” occurs when the machines evolve into beings composed of pure energy). Obviously, the machines have considerably more “brain” power than the humans— the mental capacities of even the significantly evolved twenty-sixth-century humans pale in comparison with the sheer computational power of the machines, who are described by Campbell as having the capacity to work with countless trillions of facts. And this immense computational power certainly contributes to the machines’ imaginative capacities. But the computational power itself is not what’s primarily important for Campbell’s characterization of their imagination as ideal. Rather, what matters is how that computational power is put to work. Importantly, unlike the imagination of the humans, the machines’ imagination is not creative. They do not imaginatively create a new reality completely disconnected from the current reality; rather, their imaginings are closely guided by reality as it is. Consider, for example, how the machines employ imagination to stave off the impending alien attack that threatens Earth and all its occupants. Here they are not concerned simply to figure out whether it is metaphysically possible that the attack will be thwarted if they take certain courses of events; they need to figure out which course of events will actually thwart the attack. Thus, it does not help them to imagine an alien force that is smaller, or less intelligent, or less powerful, than the actual army on its way. It also does not help them to imagine their own existing defenses as more widespread or advanced than they actually are. They need to imagine the alien force as it actually is, and their own existing defenses as they actually are, and likewise for all other factors relevant to the impending attack. Of course, in attempting to determine whether a given course of action will thwart the attack, the machines can’t imagine the world exactly as it is in all respects, since the implementation of that course of action would result in various changes to the world as it actually is. To figure out whether a given course of action will be successful, then, their imaginings must appropriately take such changes into account. For their imaginative exercises to teach them something about the world—that is, for their imaginings to have epistemic relevance—the machines aim to imagine exactly the changes required by the implementation of their plan and no others.9 Because the machines are ideal imaginers, they can do this. Insofar as they are forced to depart from imagining reality as it is, their imaginings proceed by carefully controlled imaginative extrapolation. We can thus tease out two features of the machines’ imaginative process that make them ideal imaginers. First, their imaginings capture the world as it is. Second, when 9   This claim as stated is undoubtedly too strong, since it needs to be temporally modified. A change to the world as it is has a cascading effect through time—there will be all sorts of consequences of that change down the line—and it doesn’t seem that they all have to be imagined for an imagination to have epistemic significance. The machines don’t need to imagine the effects their plan will have 50 years hence; they need to imagine the effects such a plan will have now (or at least, now-ish). Thus, the claim that they must imagine all and only the changes required by the implementation of their plan should be interpreted as something like: all and only the changes required by their plan in an appropriate time frame.

Dictionary: NOSD

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagining Under Constraints  151 their imaginative projects do require them to imagine a change to the world as they believe it to be, they are guided by the logical consequences of that change. Each of these aspects can be thought of as a constraint on imagination, what we might respectively call the reality constraint and the change constraint. As I want to suggest, the epistemic relevance of the imagination is tied closely to these two constraints. Interestingly, some reflections by Campbell on the very nature of science fiction help to support this general line of argument. In an essay that attempts to explain what science fiction is and how it is to be distinguished from related genres such as fantasy, Campbell suggested: The major distinction between fantasy and science fiction is, simply, that science fiction uses one, or a very, very few new postulates, and develops the rigidly consistent logical consequences of these limited postulates. Fantasy makes its rules as it goes along . . . The basic nature of fantasy is “The only rule is, make up a new rule any time you need one!” The basic rule of science fiction is “Set up a basic proposition—then develop its consistent, logical consequences.” (Campbell 1966)

Our imagination often works analogously to the way that Campbell describes the genre of fantasy as working—completely without constraint. But ideal imagination works analogously to the way that Campbell describes the genre of science fiction as  working. In light of this description, it is perhaps unsurprising that it is often thought that we can learn something interesting about our world from reading science fiction that we can’t from reading fantasy. Likewise, we can learn something interesting about our world when we engage in ideal imagination—when our imagination is ­governed by the reality and change constraints—that we can’t learn from our more ordinary imaginings. In order to better understand these two constraints, it might first be helpful to note why they are differentiated from one another. Let’s consider two different ways in which a machine might malfunction. First, an intermittent glitch in its storage mechanism might cause it to add an extra zero when adding numerical information to its factual database; for example, the computer might represent that there are 500 missiles in a particular defense facility when there are really only 50. Second, an intermittent glitch in its processor might result in faulty inferences; for example, having determined that certain upgrades to existing defensive shields will double their strength, the computer might conclude that upgrading a shield that could previously withstand two missile blasts will enable it now to withstand ten missile blasts. While the first machine’s imaginings are not governed by the reality constraint, there is no reason to suppose that they violate the change constraint. In contrast, while the second machine’s imaginings are not governed by the change constraint, there is no reason to suppose that they violate the reality constraint. Having differentiated the two constraints in this way, however, it may be tempting to see them as operating sequentially; in particular, it may be tempting to think of the reality constraint as operating prior to the change constraint. But to my mind this

Dictionary: NOSD

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

152  Amy Kind would be a mistake. Granted, it’s quite natural to think of an ideal imagining as involving three distinct steps: (1) We imagine the world (or some part of it) as it is; (2) we make one or some targeted changes as governed by our overall imaginative project; (3) we adjust our imagining appropriately in light of all and only the consequences of the changes made. But the respect in which these steps are distinct from one another is merely a conceptual one and not a temporal one. Certainly steps (1) and (2) can occur at once, and it may be that part of doing (2) is doing (3)—making the relevant adjustments, that is, might well be part and parcel of what it is to imagine some change to the world as it is. All three steps might thus be achieved simultaneously as the imagining is developed. Likewise, then, the reality and change constraints should not be seen as operating one before the other. Rather, they work closely in tandem with one another as the imagining unfolds. Though Campbell’s story suggests that these constraints govern the machines’ ideal imagination, and thereby enable the machines to learn about the world via imagination, we might wonder whether such constraints are necessary for our imaginings to have epistemic significance.10 Consider first the reality constraint. Must imaginings represent the world, or some relevant subset of it, exactly as it is in order to have the sort of epistemic significance that interests us? Here it seems clear that the answer is no. Though the machines cannot learn whether a proposed defense system will be effective against the alien attack if they imagine fewer aliens than there actually are, they might well be able to learn this if they imagine more aliens than there actually are. A defense system that’s effective against two million aliens will also be effective against one million aliens. In fact, the machines might deliberately want to overestimate the strength of the attacking army to give themselves some room for error. Other kinds of examples make the same point. When imagining whether the stroller I just bought will easily fit in the trunk of a car I’m looking at on the showroom floor, it doesn’t matter if I mistakenly imagine the stroller as green rather than yellow; it also doesn’t matter if I mistakenly imagine that the stroller was made in the United States rather than in China. Next consider the change constraint. In our discussion of the machines’ ideal imagination above, I noted that they tested their proposed plan in imagination by imagining all and only the changes to the world that its implementation would cause. But must all imaginings proceed this way in order to have the sort of epistemic significance that interests us? Suppose that once again to provide themselves with some room for error the machines were to deliberately underestimate how much additional protection would be provided by a proposed upgrade to their defensive shields; they imagine a lesser change than would occur were their plan to be implemented. Here too it seems that they could learn from their imagining. Thus, like the reality constraint, the change constraint also does not appear to be necessary for an imagining to have epistemic significance.   I am grateful to Tyler Doggett for pushing me on this issue.

10

Dictionary: NOSD

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagining Under Constraints  153 Perhaps it would be possible to reformulate the reality and change constraints so that they would serve as necessary conditions for an imagining to have epistemic relevance. Rather than requiring that the world be imagined as it is, for example, we might say that the world be imagined as it is in all relevant respects; rather than requiring that an imagining be constrained by all the consequences of the change imagined, we might say that it be constrained by all and only the relevant consequences of the change imagined. The plausibility of these reformulations would then hinge on how we cash out the notion of relevance. But while I am optimistic about the prospect of coming up with an appropriate reformulation along these lines, I will not attempt to do so here. Rather, for the purposes of this discussion I’d suggest that we think of the reality and change constraints as aspirational in nature. Some imaginings will come closer to meeting them than ­others. The closer we come to meeting these constraints—the closer we come to being ideal imaginers—the more likely we will be able to learn from our imaginings. In aiming to learn from our imaginings, we should thus aspire to meet them. Here an analogy to perception might help. Though ideal perception requires that our perceptions represent the world exactly as it is—that they meet a reality constraint for perception—we can still learn about reality from many perceptions that fall short of the ideal. As a result of my aging eyes and a habit of misplacing my reading glasses, the words on the front page of the newspaper often look blurry to me as I’m reading it over breakfast. Since the words are not actually blurry, my perceptions misrepresent the world. But, despite the blurriness, I can still learn from them what today’s Los Angeles Times headlines are. Likewise, though ideal imagination requires that our imaginings satisfy the reality and change constraints, we can still learn from imaginings that fall short of this ideal. 11

3.  An Important Worry: We are Not Machines Given the two constraints required for ideal imagination, it shouldn’t be surprising that machines would be particularly good at it. Machines, after all, are particularly good at following out logical consequences—that is, they are particularly good at abiding by the change constraint. Moreover, because of their immense storage capacity, it seems plausible that machines might well be able to have a sufficiently complete sense of the world as it is, or at least a relevant subset of it, to meet the reality constraint. But these brief reflections suggest something troubling about the notion of ideal imagin­ ation. If we define ideal imagination in terms of what machines can do, then it begins to look as if we humans will inevitably fall short—and in fact, that we will fall so far short that we won’t be able to learn from our imaginings at all. In responding to this worry, it’s worth first getting clear on exactly what it entails. Most importantly, it fails to lend support to the proposition that the imagination lacks   I will return to an analogy to perception in Section 4.

11

Dictionary: NOSD

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

154  Amy Kind epistemic relevance. Suppose for a moment that it were true that the cognitive limitations of humans prevent us from being ideal imaginers. Even so, my discussion of ideal imagination would still help to show why the imagination does not in principle lack epistemic significance—that is, it is not something about the nature of imagination that renders it incapable of giving us knowledge about the world. Compare hoping, for example. There’s no way even in principle to constrain hoping so that it would have epistemic relevance. No matter how constrained hoping is, if it is really a case of hoping, it is not something from which we can learn anything about the world. The imagination is different. While the nature of hoping is inconsistent with its having epistemic relevance, our discussion of ideal imagination reveals that the nature of imagining is not inconsistent with its having epistemic relevance. And this remains true even if it turns out that the limitations of humans are such that none of our imaginings meets (or comes close to meeting) the ideal—and thus that, in practice even if not in principle, none of our imaginings has any epistemic relevance. That said, this would certainly be a disappointing result. Fortunately, however, I think we can avoid it by confronting the worry more directly. In short, my response consists of two key claims. First, I think that we’re often better imaginers than the worry above gives us credit for. But second, I want to claim that even when we do fall short of the capabilities of Campbell’s machines, we can still come close enough to the ideal for our imaginings to have epistemic significance. In fact, many people are very, very good at imagining—and, in particular, at imagining under constraints. Consider two such relatively well-known people: the inventor Nikola Tesla and the animal scientist Temple Grandin. Tesla, who is perhaps best known for developing the alternating current technology in widespread use around the world today, was also responsible for the invention of high-voltage electrical coils, long-distance electrical transmissions lines, hydroelectric generators, bladeless turbine engines, X-ray tubes, and various radio-controlled devices. Of interest to us, however, are not his inventions themselves but rather the creative process underlying their development—a creative process that owed to Tesla’s remarkable powers of visualization. As Tesla himself described it, he could perfectly picture his inventions in advance of their creation: Before I put a sketch on paper, the whole idea is worked out mentally. In my mind, I change the construction, make improvements, and even operate the device. Without ever having drawn a sketch, I can give the measurement of all parts to workmen, and when completed these parts will fit, just as certainly as though I had made accurate drawings. (Tesla 1921, p. 62)

Also according to Tesla, his method was flawless: The inventions I have conceived in this way, have always worked. In thirty years there has not been a single exception. My first electric motor, the vacuum tube wireless light, my turbine engine, and many other devices have all been developed in exactly the same way. (Tesla 1921, p. 62)

Dictionary: NOSD

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagining Under Constraints  155 Although Grandin’s inventions are perhaps not as well-known as Tesla’s, over the past several decades, she has significantly improved the welfare of animals throughout the world by revolutionizing the design of livestock-handling facilities. Approximately half of the cattle in North American meat plants are now handled in systems that she designed. Grandin’s description of her own creative process sounds remarkably like Tesla’s: “Visual thinking has enabled me to build entire systems in my imagination” (Grandin 1995, p. 19).12 As she also notes, “in my work, before I attempt any construction, I test-run the equipment in my imagination. I visualize my designs being used in every possible situation, with different sizes and breeds of cattle and in different weather conditions. Doing this enables me to correct mistakes prior to construction” (Grandin 1995, pp. 20–1). One of Grandin’s early design successes came in 1978 when she developed an innovative dip vat design for a cattle-handling facility in Arizona. A dip vat, which is filled with pesticide to rid animals of parasites, is a long, narrow, pool-like structure in which cattle are completely immersed while proceeding through it single file. Prior to Grandin’s designs, cows would often panic both when approaching the dip vat and when exiting it. By taking a “cow’s eye view” of the situation, Grandin diagnosed the problems with the existing structures and was able to create an alternative in which the cows would calmly enter and exit the equipment voluntarily, without any use of force. Her design process, however, took place entirely in her mind: “I started running three-dimensional visual simulations in my imagination. I experimented with different entrance designs and made the cattle walk through them in my imagination. These images merged to form the final design” (Grandin 1995, p. 23). When considering these first-person accounts of Tesla’s and Grandin’s use of ­visualizations—and here I have included only a few key excerpts—it’s hard not to see the imaginings they describe as eerily machine-like. Like the machines of Campbell’s “The Last Evolution,” Tesla and Grandin each have the ability to abide by the reality and change constraints we discussed above. When coupled with an incredible know­ ledge base—itself also machine-like—about the relevant domain, each of these individuals was able to engage in imaginings from which they learned something about the world. (Grandin, for example, discovered via imagination what dip vat design was most compatible with the temperament of cows.) Contrary to the worry under discussion then, at least some humans seem capable of the ideal imagination enjoyed by the machines. Clearly Tesla and Grandin have extraordinarily well-developed powers of imagination, so much so that their abilities will likely seem far out of reach to those of us whose powers are more mundane. For this reason, the worry that we’ve been grappling with in this section might still seem to have bite. Perhaps there are a few rare cases where humans can achieve machine-like ideal imagination, but if these cases are so few and   Interestingly, Grandin explicitly compares herself to Tesla. See Grandin (1995, p. 26).

12

Dictionary: NOSD

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

156  Amy Kind far between, and if the rest of us can’t even come close, then the imagination won’t have much epistemic significance after all. But why should we think such cases are so few and far between? Why should we think that the rest of us can’t even come close? Here it will be helpful to recall the examples we saw above in which it seemed plausible that an act of imagination could provide an imaginer with reason for beliefs. A prospective parent—call her Imogen—might engage in various imaginings in an effort to figure out whether she is ready to have children, or what kind of car to buy, or where to put the crib. If she wants these imaginings to be epistemically useful to her, it seems natural that she will deliberately attempt to line them up with her conception of the world. These aren’t mere daydreams, but rather are controlled—that is, constrained— imaginings. For example, when she imagines the crib against the various walls of the room that will become the nursery, her imagining is constrained by the actual size of the room, the actual size of the crib, the placement of the windows and doors in the room, and so on. Will the door still comfortably open if the crib is placed against the west wall? This can best be determined if both the reality and the change constraints are met—if she imagines the room as it actually is, but suitably adjusted for the insertion of the crib. Likewise, when she imagines a car seat in the back of a two-door sports car, her imagining is constrained by factors such as the size of the car, the size of the infant seat, and the angle of the front seat as it’s tilted forward. With the front seat tilted forward and a car seat in back, will there be enough room for a parent to reach in and strap in an infant? Again, this can best be determined if the reality and change constraints are met—if she imagines the car as it actually is, but suitably adjusted for the insertion of a car seat. Granted, there may be some folks who can’t even do what Imogen does, let alone what Tesla and Grandin do. The only way for such a person to determine whether the car seat really fits in the back of the sports car would be to test it out with an actual car seat and an actual car. But many of us can make these determinations without actually trying it out—we can learn something via an imaginative test rather than an empirical test. How do we do this? We are not machines, and most of us are not even Tesla and Grandin. We’re not as good as the machines in setting the relevant constraints, and it doesn’t seem that we’re as good as the machines in abiding by them once they are set. But in many cases—and which cases these are will vary from person to person—we do seem to be able to be good enough. In part this is because the sorts of imaginative projects in which we’re involved will be considerably less fine-grained than the projects undertaken by the machines. Considerably less precision is required for us to successfully imagine many of the kinds of things that we imagine. In part this is also because the sorts of imaginative projects in which we’re involved will be considerably less complicated than the projects undertaken by the machines. In many cases, considerably fewer variables are in play. Thus, even though it seems true that we will typically fall short of the machines, that we can’t match their capabilities, we can nonetheless approximate their capacity—and in many cases, that will be enough.

Dictionary: NOSD

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagining Under Constraints  157

4.  Another Important Worry: We Don’t Know What We’re Doing Though I hope that the reflections of the previous section have laid to rest the worry that the capacity for ideal imagination is so far beyond our reach that we can’t learn from our imaginings, there is a related worry in the vicinity. Perhaps some of our imaginings will be ideal (or close enough), but how can we tell? If we can’t, then doesn’t this limit the epistemic significance of the imagination? To flesh out this worry, let’s return to Imogen. We assumed that she was a pretty good imaginer—not as good as Tesla and Grandin, perhaps, but good enough that she could plausibly count as approximating the ideal for the limited sorts of imaginings she was engaging in. But what if she’s not? Blinded by her love of sports cars, her imagining might wildly misrepresent the car’s interior size—almost as if it had been magically transformed by one of Hermione Granger’s Undetectable Extension Charms. (When an object is transformed by such a charm, it is made bigger on the inside without being made bigger on the outside; what appears to be a small handbag, for example, becomes capable of holding numerous textbooks, clothes, a tent, and even a sword.) An imagining that misrepresents in this way is not appropriately constrained—it is no longer even coming close to operating under the constraint of reality—and it is precisely this lack of constraint that robs the imagining of its epistemic usefulness. An imagining that wildly exaggerates the size of a car’s interior cannot teach Imogen anything about whether an infant seat fits in the car’s back seat. What might seem troubling, however, is that Imogen might very well have difficulty  determining whether she has violated the constraint of reality. She might be convinced—even reasonably so—that she has represented the car’s interior correctly when she has not. Perhaps the machines can run self-diagnostics to ensure that they are operating within standard parameters, but we cannot. In general, we can’t be completely confident about whether we have abided by the constraints of reality and change. But if we can’t determine this, if (to put it bluntly) we typically don’t know what we’re doing, this might well seem to undercut the epistemic significance of the imagination. But let’s think again about what has gone wrong with Imogen’s imagining in the case we’re now considering. In such a case, Imogen is the victim of what we might call imaginative illusion. She takes herself to be imagining the car’s interior as it is, but her imagining does not represent the car’s interior as it is. Here there is an obvious parallel to certain sorts of perceptual illusions—we might take ourselves to be seeing the tower in the distance or the stick in water as they are, but we’re not—the tower isn’t really round, and the stick isn’t really bent. The fact that we can be victims of perceptual illusions does not rob perception of its epistemic significance. Likewise, then, the fact that we can be victims of imaginative illusions should not rob the imagination of its epistemic significance.

Dictionary: NOSD

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

158  Amy Kind And now let’s reflect for a moment further about perceptual illusions. In general, we can identify various conditions in which we are more likely to be the victim of perceptual illusions—when the lighting is bad, or when the objects we are viewing are far away, or when we are under the influence of alcohol or hallucinogenic drugs. Relatedly, we can take steps to minimize the possibility of such illusions—we can turn up the lights, or move closer to the objects we’re viewing, or we can abstain from alcohol and  hallucinogenic drugs. Likewise, I think, we can identify various conditions in which we are more likely to be the victim of imaginative illusions. If Imogen has never seen the inside of the sports car, or she hasn’t seen it for a very long time, then she should be wary about her ability to constrain her imagining appropriately. Additionally, as we suggested above, her passion for sports cars might be distorting her ability to imagine the interior correctly. Here, her desires have a distorting influence similar to the distorting influence of poor lighting or long distance. In cases where she has such desires, then, she should also be wary about her ability to constrain her imagining appropriately. Ultimately, however, we cannot always identify that we have fallen victim to a perceptual illusion, and likewise, we cannot always identify that we have fallen victim to an imaginative illusion. There may be some skeptics who think this fact about perception robs perception of its epistemic significance. But setting aside such skepticism, as long as we think that the possibility of perceptual illusion does not prevent us from learning something about the world from our non-illusory perceptions of it, the possibility of imaginative illusion does not prevent us from learning something about the world from our non-illusory imaginings of it.

5.  Concluding Remarks According to Campbell’s depiction of humankind in “The Last Evolution,” our imagination may be brilliant, but it is illogical. Unlike the machines, we can advance by leaps and bounds. But also unlike the machines, our conclusions are not always correct. As I have tried to suggest in this chapter, however, our imagination need not be as illogical as Campbell makes it out to be. While many of our imaginings are unable to teach us about the world, we have the power to constrain our imagination and, in doing so, to imbue our imaginings with epistemic significance. Importantly, I do not mean here to privilege constrained imagining over unconstrained imagining. Both of them have their place in human life. Our unconstrained imaginings are important for many of the activities that are important to us, and rightly so—from engaging with art and literature to fantasizing or pretending. But it’s our constrained imaginings that are important for our attempts to learn about the world— and this kind of knowledge cannot come by leaps and bounds. Rather, in modeling our imagination on the ideal imagination of the machines, we are able to make epistemic progress the way they do, by steady, irresistible steps.

Dictionary: NOSD

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagining Under Constraints  159

Acknowledgements In thinking about these issues, I learned a lot from Tyler Doggett and Jennifer Church, both of whom served as commentators when an earlier version of this chapter was presented at the APA Pacific Division meeting in 2013. I owe them both thanks. I am also grateful to the audience members during that session for their feedback, as well as to audiences at the University of Kentucky, Denison University, and the Workshop on Mental Imagery and Pretense at the University of Antwerp. Thanks also to Peter Kung for helpful discussion and feedback.

References Campbell, John W. (1932). The Last Evolution. Amazing Stories. . Campbell, John W. (1966). Introduction. Analog 6. Chalmers, David J. (2002). Does conceivability entail possibility? In Tamar Szabό Gendler and  John Hawthorne (eds), Conceivability and Possibility. Oxford: Clarendon Press, pp. 145–200. Gendler, Tamar Szabό, and John Hawthorne (2002). Introduction: Conceivability and possibility. In Tamar Szabό Gendler and John Hawthorne (eds), Conceivability and Possibility. Oxford: Clarendon Press, pp. 1–70. Grandin, Temple (1995). Thinking in Pictures: My Life With Autism. New York: Random House. Kind, Amy (2001). Putting the image back in imagination. Philosophy and Phenomenological Research 62(1): 85–109. Kind, Amy (in press). How imagination leads to knowledge. In Fabian Dorsch and Fiona Macpherson (eds), Perceptual Memory and Perceptual Imagination. Oxford: Oxford University Press. O’Shaughnessy, Brian (2000). Consciousness and the World. Oxford: Clarendon Press. Paul, L. A. (2014). Transformative Experience. Oxford: Oxford University Press. Sartre, Jean-Paul (1948). The Psychology of Imagination. New York: Philosophical Library. Walton, Kendall L. (1990). Mimesis as Make-Believe: On the Foundations of the Representational Arts. Cambridge, MA: Harvard University Press. Wisehart, M. K. (1921). Making your imagination work for you: An interview with Nikola Tesla. . . . American Magazine 91 (April 1921): 13, 60–6.

Dictionary: NOSD

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

7 Perceiving People as People An Overlooked Role for the Imagination Jennifer Church

Most everyone can agree that, at least sometimes, we use our imagination in acquiring knowledge of other people. Few would deny, for example, that I might come to know what my colleague was doing yesterday at 3:10 p.m. by imagining her movements following the 3:00 p.m. fire drill, or that I might come to know that my brother is worried by imagining his thoughts upon reading our mother’s letter. There is considerable disagreement over what the relevant imagining involves (whether it requires imagery for example), over how it arrives at a conclusion (whether guided by theories or by mental simulations, for example), and about when it is trustworthy.1 Regardless of the details, however, the usual story is one in which our imagining provides us with indirect or inferential knowledge of other people’s thoughts and actions. The indirect character of the acquired knowledge is clearest in cases where the imagined thoughts or actions are explicitly derived from background beliefs about how people think and act in general or how a particular person tends to think and act. I believe that people normally leave a building when there is a fire drill, and they normally leave by the nearest door; I then imagine my colleague near the south door when the fire drill goes off at 3:00; and that leads me to imagine her outside of the south door at 3:10. Likewise, I believe that my brother is concerned about our mother’s well-being and that he tends to expect the worst, so I imagine that he will be worried when he reads about her illness. In cases where I proceed by imagining myself in another person’s situation, the indirectness of my knowledge arises at a different point—namely, 1   See Maibom’s article in this volume (Chapter 8) for a detailed argument against the trustworthiness of  imagination in predicting the actions and reactions of both oneself and others. Empirical ­evidence indicates that much of our imagining is guided by what we consider to be a rational or moral response—not by past experience or detailed observations of people in specific situations. This allows, though, that imagination is a trustworthy source of knowledge when people are likely to be rational or moral; and it suggests that imagination can be improved insofar as it can draw on richer memories and observations.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

perceiving people as people  161 when I make an inference from my own (imagined) response to that other person’s (actual) response. I imagine myself in her office when the alarm sounds, I note where that imagining takes me, and I infer that my colleague’s response to the alarm will be similar. Or I imagine myself reading my mother’s letter, and I infer that my brother will respond in a similar way. I want to suggest that imagining also plays a role in our more immediate knowledge of people’s thoughts and actions—our ability to look at, or listen to, someone and perceive them as waiting, or as worried. This sort of knowledge has immediacy that is both phenomenally and epistemically distinctive, and imagining helps to account for both of these aspects. This claim will take some explaining—as it applies to perception in general, and as it applies to our knowledge of people in particular. The basic idea, though is this: when we perceive something, we experience it as objective—as something independent of any particular view we may have of it, and that experience depends on the simultaneous representation of different views of that thing. To the extent that those different views are coherent and are themselves grounded in past experience, they offer an immediate validation of our experience as veridical. When we perceive something as a door, for example, we imagine different views that together confirm its categorization as a door; and when we perceive a friend as worried, we imagine different views of that friend that together confirm their categorization as worried.2 The arguments of this chapter, by highlighting the similarities between (relatively) uncontroversial cases of perceiving a door or perceiving a flame and (more) controversial cases of perceiving people’s intentions and feelings, might help to vindicate the possibility of the latter.3 My explicit aim is more limited, however. I want to show how imagining enables our knowledge of others’ mental states to share the phenomenological character and the epistemic standing of more ordinary cases of perception. For ease of presentation, I shall refer to cases in which someone appears to me to be worried as cases in which I perceive that person as worried—leaving open the possibility 2   I allow that my imagining may be itself the product of subpersonal inferences—subpersonal theories or subpersonal simulations, for example. Some philosophers, along with most psychologists, reject the view that perceptual knowledge is non-inferential knowledge. For an early defense of the view that perception is inferential, see Harman (1973). Burge (2010) describes “principles” that govern perception, but insists that these principles, unlike the principles that govern thought, are not themselves represented and are not accessible to consciousness. For a defense of the view that “theory theories” are compatible with the claim that we have direct perceptual awareness of others’ mental states, see Lavelle (2012). And for a recent defense of the view that perception is not incompatible with inference, see Cassam (2014, ch.  10). The important distinction, for my purposes, is not between perception and inference, but between the way things appear to be and the way I merely think them to be. 3   It is, of course, possible to challenge the assumption that we can perceive the property of being a door, or the property of being a flame—particularly if such properties are thought to be relational or dispositional properties. Tye (1995) and Dretske (1995) restrict the contents of perception to “low-level” properties such as shape and color. Siewert (1998) and Siegel (2006, 2010) defend the more natural view that we can perceive “high-level” properties as well. For my purposes it is enough to establish certain similarities between how something comes to appear to us as a door and how someone comes to appear to us as worried.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

162  Jennifer Church that, though my experience of their worry is phenomenologically and epistemically like perception, it fails to meet some other requirement of perception;4 and leaving open the possibility that they are not in fact worried. Unlike the claim that I see a drawing as a duck, or I see a bush as a bear, the claim that I perceive my friend as waiting or as worried should not be taken to imply that she is not in fact waiting or worried. Section 1 focuses on imagination’s contributions to the phenomenology of perceiving something as a door, for example, and Section 2 details imagination’s analogous contributions to the phenomenology of perceiving someone as worried, for example. Section 3 addresses the epistemic contributions of imagination in both the case of perceiving something as a door and the case of perceiving someone as worried.

1.  The phenomenology of perceiving something as a door or perceiving something as dangerous is different from the phenomenology of merely thinking of the surface in front of me as a door or merely thinking of the moving light as a flame. In the latter case, it is perfectly reasonable to say “I think that is a door even though it doesn’t look like a door, and I think the light is a flame even though it doesn’t look like a flame.” Likewise, the phenomenology of perceiving someone as waiting or as worried is different from the phenomenology of merely thinking of them as waiting or as worried. I do not doubt that there is something it is like to merely think of something as a door—that mere thinking, if conscious, has a phenomenology.5 Furthermore, I assume that there are many cases in which the phenomenology of my experience falls somewhere between pure cases of perceiving as and pure cases of merely thinking as, or cases in which my experience is perception-like in some respects and merely thought-like in others. (An experience might be perception-like with regard to its reference and thought-like with regard to an attributed property, for example.6) I do, however, assume that there is a familiar and recognizable difference, across a wide variety of cases, between the phenomenology of perception and the phenomenology of thought. But what is that difference? A standard reply is that perception presents an object or a property directly, without mediation, while thought presents it indirectly—through the mediation of language, for example. Insofar as our focus is the distinctive phenomenology of perception, it 4   My own view is that the term “perception” can be used in several different ways, but that the phenomenological and epistemological aspects of perception are the aspects that give perception its philosophical significance and (contra Burge (2010)) that a satisfactory account of perception should show how these two aspects are related. See Johnston (2011) for a fuller presentation of this position. 5   This does not mean that the phenomenology of thinking is sensory in character. Siewert (2011) offers a useful history of how philosophers’ notions of phenomenology narrowed in the twentieth century, largely under the influence of Wittgenstein. 6   Montague (2011) has a nice discussion of how even the phenomenology of particularity has a cognitive aspect; and Levine (2011) defends what he calls an “impure” version of cognitive phenomenology whereby that phenomenology requires a sensory manifold but is not exhausted by that manifold.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

perceiving people as people  163 will be the absence of apparent or conscious mediators that is relevant. (Unconscious mediators, though they may affect the phenomenology of an experience, will not be experienced as mediators.) There are different sorts of mediation, though, and it is not clear which of them is relevant to the phenomenology of perception. When I look at the scene before me, I might be aware of the causal mediation of my sunglasses without that awareness making the phenomenology of my experience any less perceptual; so the phenomenology of perception does not rule out awareness of causal intermediaries.7 Does it rule out awareness of mental intermediaries, or representational intermediaries? If I am aware that my memories of a building, or a map of a building, play a mediating role in my current experience of that building, does that mean that my experience lacks the phenomenology of perception? Given the pervasive influence of memories and maps—mediations of which we are often quite aware—this would be an extreme conclusion, implying that almost no experience has the phenomenology of perception. The phenomenology of thought, moreover, may not include any awareness of intermediaries: I can think of my colleague as older than me without any awareness of how I reached that conclusion and without ever putting my thought into words. So it can’t be our unawareness of mental or representational intermediaries that distinguishes the phenomenology of perception from the phenomenology of thought. It is common to suppose that the sort of immediacy that is distinctive to perceptual experiences is justificatory immediacy—but justificatory immediacy can be understood in several different ways. First, perceptual experiences might be thought to have justificatory immediacy insofar as they do not need any further justification in order to constitute knowledge. But the need for justification depends on context, not on the phenomenology of our experience; some perceptual experiences, in some contexts, will require further justification before they can provide us with knowledge while some non-perceptual beliefs, in some contexts, will provide knowledge without the need for any further justification. (Compare an amateur’s claim to be seeing a swan overhead with an expert’s claim that there are no swans in this region; the former needs further justification while the latter does not.8) Second, our perceptual experiences might have justificatory immediacy insofar as they are not capable of further justification—insofar as justification must come to an end in perception. This assumes a foundational view of justification, which is questionable. More importantly for our purposes, however, 7   How to count or individuate causes is a complex issue—especially if one thinks that the determination of causes depends on the description of effect and/or the context of inquiry. I am not aware of any principled way of individuating causes that ensures an absence of causal intermediaries between an object and the perceptual experience of that object. Causes can be specified at many different levels of abstraction, of course, so it could be that at an appropriate level of abstraction there is a direct causal connection between one person’s smile and another’s knowledge of that smile. See Yablo (2003). If the relevant level of abstraction is dictated by the level of one’s interest, however, it is hard to see how causal immediacy would explain (versus be explained by) the immediacy of perception. 8   It is possible to adopt what I would call a “deflationary” view of perception whereby an experience qualifies as a perception whenever we require no further justification—whatever the phenomenology of that experience happens to be. I think there is something distinctive about the phenomenology of perception, however—something that helps to explain its special status vis-à-vis justification.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

164  Jennifer Church it fails to show how the justificatory standing of perception relates to the phenom­ enology of perception.9 A third reading of the justificatory immediacy of perception offers to connect the distinctive epistemology of perception to its distinctive phenomenology by invoking the possibility of self-justification, or self-evidence. In the words of Eilan, perceptions have “objective import,” which (in the tradition of P. F. Strawson) means that “experience presents itself as experience of a mind-independent reality” (italics added)10. On this view, the experience of perceiving a door is unlike the experience of merely thinking of the door insofar as the perceptions make the objectivity of the door self-evident, confirming the veracity of our experience from within. (This is not to say that the veracity of a perceptual experience can be guaranteed from within the experience itself—only that the phenomenology of a perceptual experience gives us reason to trust its accuracy.) How is it possible, though, for an experience to justify itself—for evidence of the veracity of an experience to be evident from within that very experience? Alva Noë’s account of experienced objectivity offers a revealing response to this challenge. He distinguishes two aspects or components of any perception, and argues that experiencing the dependence of one aspect on the other amounts to an experience of objectivity. According to Noë, the contents of perceptual experience must include both a factual dimension and a perspectival dimension, an aspect that indicates how things are and an aspect that indicates how things appear; and it is the interaction between these two aspects that is responsible for the distinctive phenomenology of perception. We experience not only how things are, but also how they look from here. We experience that the plate is round and that it looks elliptical from here. Its elliptical look from here is a genuine property of the plate—we see the shape and we see the perspectival shape from here—but it is also a relational property, one that depends on where “here” is. Perceptual content . . . is two-dimensional. It can vary along a factual dimension, in regard to how things are. And it can vary along a perspectival dimension, in regard to how things look from the vantage point of the perceiver. Visual experience always has both these dimensions of content.11 9   Burge (2010) explicitly separates the phenomenology of conscious perception—which, on his view, has to do with the phenomenology of sensations—from the justificatory standing of perception (which need not be conscious). I, like many of the philosophers cited below, reject this separation and attempt to explain the phenomenology of objectivity in terms of justificatory standing. 10   Eilan (2011). 11   Noë (2003, pp. 2–3). The importance of mastering multiple perspectives in order to fix the content of an experience is familiar from the work of Dretske (1981), Millikan (2000), Davidson (2001), and others. What is distinctive about Noë’s analysis is his attempt to use the phenomenology of converging perspectives to explain the phenomenology of objective reference. Elsewhere, Noë (2002, p. 74) makes a similar distinction between the representational and sensory factors that make up the qualitative character of experience:

The qualitative character of experience, as we have seen, depends on two factors. First, it depends on the qualities that we experience (e.g. looks, sounds, etc). This is a representational

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

perceiving people as people  165 According to Noë, then, it is the two-sided character of perception, and the experienced dependence of one side on the other, that enables us to experience objectivity from within perception itself. We experience the objectivity of a perception by experiencing the dependence of perspectival content on factual content—a dependence that is revealed as we move through space and notice how the way things look depends both on how they are and on where we are. Indeed, we only become capable of distinguishing looks from facts insofar as we become capable of tracking such dependencies. While I agree with the suggestion that we experience the objectivity of a state of affairs when we experience constancy across perspectival change, Noë’s insistence that we experience appearances (e.g. the elliptical look of a plate) as well as facts (e.g. its roundness) creates more contents (and more puzzles) than are needed.12 When we look at a plate from the side, we do not see an elliptical appearance in addition to the round shape of the plate; either we see the round plate or we mistake the round plate for something elliptical. Likewise, when we experience the shapes around us as if they were flattened onto a screen before our eyes, we do not simply attend to an already present aspect of experience (as Noë suggests); rather, we shift from seeing three-­ dimensional objects to imagining their two-dimensional counterparts. While we do need to undergo changes in perspective in order to discover what remains invariant across changes in perspective, and while changes in perspective may involve changes in the content of experience (different sides of an object come into view, and different parts of its surroundings appear), this does not mean that the content of our perceptions can be divided between factual and perspectival components or dimensions. The phenomenology of objectivity depends on the convergence of information gained from different points of view, not the experienced dependence of some experiential contents on others. In short, I agree with Noë’s reliance on the coordination of different feature. Second, it depends on the character of the activity in which the temporally extended activity may consist. So, for example the fact that we do not make eye movements when we explore the environment haptically makes a difference to what it is like to touch. These differences in the sensorimotor contingencies governing the different sensory modalities are differences in the qualitative character of experience that do not correspond, directly at least, to differences in what is perceived. It is not clear (at least not to me) whether Noë intends the second, non-representational aspect of experience to count as part of its content—akin to the perspectival contents discussed above. The term “content” doesn’t matter, perhaps, but the ability to recognize the dependence of subjective phenomenology on objective fact, through the dependence of one factor on another, does matter. See Siegel (2014) for a related discussion of why affordances need to be represented in order to be part of the content of visual experience. 12   Noë’s talk of two different “dimensions” of content is less problematic than his talk of two different types of experiential content—“looks” as well as facts, for example perspectival shape as well as objective shape. He is clearly aware of the dangers of invoking sense data as the true object of perception, but he is determined to include appearances, understood as relational properties (between the perceiver and the perceived), as part of the contents of perception. I agree that experiencing alternative perspectives is necessary for experiencing something as objective, but I reject the view that every perspectival experience has, as part of its content, a perspectival property. Briscoe (2008) provides a nice elaboration of this criticism. Noë (2008) has replied to several other critics of his two-aspect view.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

166  Jennifer Church perspectives to explain the experience of objectivity, but I disagree with his invocation of a special type of content, or a second aspect of content, that is perspectival. The key to preserving the insights of Noë’s account while avoiding its problems, I suggest, lies with the realization that different parts or different aspects of a single experience can reinforce each other, each part helping to justify our confidence in the veracity of the others. My current experience of my desk, for example, combines several visual components and several tactile components—each providing different perspectives on a single object at a single location, each helping to establish the reality of that object. If I only experienced pressure on my elbow, or if I only experienced a streaky brown patch in my left visual field, there would be nothing self-evident or self-justifying about the experience—nothing within the experience itself to show me I am experiencing something objective, that I am perceiving something as opposed to merely undergoing a particular sensation.13 Contra Noë, I do not experience the dependence of perspectival contents on factual contents; rather, I experience the contents of my experience as factual in virtue of the way they are integrated with one another. This is where imagination becomes relevant—not just imagination as the cognitive representation of some state of affairs, but imagination as the simulation of a perspectival experience. While it is true that we can experience only a limited number of perspectives at a time, it is certainly possible simultaneously to touch and to look, to touch from two different angles, or to look from the slightly different angles of two separate eyes. Furthermore, it is possible to imagine several different perspectives and modal­ ities simultaneously14; these will include perspectives that we do not now occupy, and modalities that we do not now engage. The coordination of those additional perspectives and modalities in imagination can deepen our experience of objectivity, however. As I look at the tilted table from here, I can also imagine what it looks like from over there; and I can imagine what it would feel like to rest my hands on the table, or to stroke its surface; and so on. Insofar as such imaginings can be integrated into the phenomenology of my present experience, they will add to my experience of the table as something objective because they will add to my experience of the table as something independent of the particular perspective(s) I now occupy.15

13   For a fuller discussion of my understanding of perception as it relates to that of Noë and others, and as it relates to sensation, see Church (2013). 14   Noë (2006) claims that different perspectival contents occur at different times: “experience is a temporally extended phenomenon; it is an activity of skillful probing” (p. 430). But he also claims that “experiential presence is virtual all the way in . . . The rear side is present virtually, but the present side is present simpliciter” (p. 427)—a claim that seems to suggest that the rear side is present in imagination at the same time as the front side is present to one’s senses. 15   Siegel (2011) distinguishes three different ways that philosophers have attempted to explain the experience of objectivity in perception: ways that emphasize the experience of space, ways that emphasize the experience of alternate perspectives, and ways that emphasize the experience of tactile resistance. For reasons that I elaborate in Church (2013), I think the experience of space and the experience of perspective-­ independence are inseparable.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

perceiving people as people  167 There is not a sharp line here between what is experienced as objective and what is not experienced as objective. This is true for a number of reasons. First, the number of different perspectives or modalities that we can experience—whether actually or imaginatively—varies with the object and with the circumstances. There are many different perspectives on the shape of a lemon, but few perspectives on its taste. I can easily imagine alternate perspectives on the desk before me, but I struggle to imagine even one other perspective on the dark fleck that just crossed my visual field. As a result, we can experience the lemon’s shape as more objective than the lemon’s taste, and we can experience the desk as more objective than the fast-moving fleck. Second, our ability to integrate the different perspectives we remember or imagine at a given time also varies quite widely. If you show me photos of a bird that crossed my visual field—photos that enable me to imagine what it looks like from many different perspectives—this can enhance my experience of its objectivity only insofar as I am able to imagine how those other perspectives fit together with my present perspective. The more integrated the presented alternatives, the more evident the objectivity. Third, there are differences in the extent to which we can entertain different perspectives simultaneously. Even if I understand how my current perspective on the bird is related to the perspectives of the photos you show me, it may be harder for me to conjure up those other images while observing the bird than it is for me to imagine other perspectives while observing my desk. Insofar as the phenomenology of perception is a phenomenology of ­evident objectivity, and insofar as the synthesizing of perspectives that supports a phenomenology of objectivity comes in degrees, the phenomenology of some experiences will be more perceptual than others. As long as our intuitions about clear versus unclear cases accord with the predictions of this analysis, the resultant fuzziness counts for rather than against the account offered here. Imagining of alternate perspectives is certainly possible, but how often is it actually occurring as we perceive? If the phenomenology of perception is the phenomenology of evident objectivity, and if objectivity can only be evident from within an experience through the infusion of imagined alternatives, then we have reason to think that imagination must be operative when we perceive something as a door (or as anything else). But what empirical evidence do we have to support this claim? The highest, and perhaps the least reliable, level of evidence is introspection. We are not normally aware of imagining alternative perspectives or alternative possibilities when we perceive something as a desk or we perceive someone as amused; our focus is on the object or person we are observing, not on the way we are processing information. But just a bit of introspection suggests that we are constantly imagining more than what is immediately present to our senses. One way to prompt an introspective awareness of such imagining in the midst of a perceptual experience is to ask an attentive observer: Are you imagining the door opening to the right? Are you imagining the backside as carved in the same pattern? Often, the observer will respond: “Actually, I was imagining the door as opening to the left,” or “I was imagining the back as less ornate”—claims that count against the worry that the reported imagining is merely

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

168  Jennifer Church due to the power of suggestion. (Our sense that we were imagining something prior to being asked may be mistaken, of course; but the burden of proof lies with those who doubt our introspective evidence, not those who accept it.16) Introspection also reveals that unavailable perspectives are more or less automatic­ ally imagined when certain parts of a scene are obscured from view, or when certain parts of a sequence of sounds are blocked from hearing. When a pillar blocks our view of center stage, for example, we use our imagination to fill in the way an actor moves in walking from left to right; and when a radio crackles as we listen to music, we use our imagination to fill in the missing notes.17 Likewise, when we are shown (in a movie, for example) a person performing an impossible action, or following an impossible trajectory, we tend to use our imagination to “correct” our vision; and when we are given a nonsensical series of notes, we tend to use our imagination to “correct” our hearing. Frequently, we experience the missing positions as positions we have directly observed, and we experience the missing occurrences as occurrences that actually happen, wholly unaware of the role of the imagination in giving us these experiences. The contributions of the imagination in such cases are quite familiar to actors and painters and musicians, and they are contributions that are quite easy for anyone to observe introspectively. There is also considerable behavioral evidence, at the functional level, that supports the pervasive influence of imagining on perception. (For those who want to restrict the term “imagine” to conscious acts or conscious representations, functional evidence for image-like representations may seem irrelevant. I do not, however, want to restrict the term in this way, and neither do most psychologists.18) The fact that we move so smoothly through intricate spaces suggests that those spaces are already imagined in advance of our movement—not just hypothesized, in thought, since the translation from thoughts into movement would lack both the speed and the spatial nuance that are required. Also, the ease with which we recognize previously seen 16   Dennett (1991) takes the opposite approach, placing the burden of proof on those who think there is a difference between the realist versus constructivist interpretations of such reports, for he argues that there is nothing to decide between these two interpretations. The dispute then turns on whether the prima facie evidence of introspection can be explained away (and not merely discounted). 17   Note the difference between these cases and the case of blind spots or visual and aural “blinks,” where it is more plausible to suppose that an absence of information is simply ignored rather than that the missing information is imaginatively filled in. (What is the difference between not noticing a gap in perception and filling the gap in with imagination? Noë (2004, pp. 47–67) summarizes Dennett’s “perceptual presence” account and offers his “implicit understanding” alternative.) There are also interesting differences between cases in which the gaps in a presented scene are filled in by relatively low-level representations generated by visual processing mechanisms and cases in which the gaps are filled in by higher-level processing based on long-term memories. Briscoe (in press) clarifies some important differences between these two sorts of filling in. Both involve visual representations and both are involuntary, however; so, unless one insists that imagining must always be under our voluntary control, Briscoe’s distinctions raise questions about the extent of so-called “cognitive penetration,” not the extent of imaginative filling in. 18   McGinn (2004) is one philosopher, following in the tradition of Wittgenstein, who wants to restrict the term in this way. For further discussion of the nature and possibility of unconscious imagining, see Church (2008). See also Church (2013, ch. 2) for further discussion of active versus passive imagining.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

perceiving people as people  169 objects, even when they are now viewed from alternative points of view, suggests that we already imagined (something close to) these other points of view when we first viewed those objects. (The alternative hypothesis—that we imagine different points of view only later, in an effort to match the objects before us with those seen previously— is not only implausible in most cases, given the speed of our recognition; it simply shifts the time of our imagining, without questioning the claim that most of the perceiving that we do—pretty much all perceiving that involves recognition—depends on actively imagining alternative points of view.) Third, in the case of many causal sequences—a particular type of movement usually following a particular facial expression, a particular tone of voice usually preceding a particular type of laughter, and so  on—it can be shown that (other things being equal) even babies lose interest in sequences they are familiar with. One plausible explanation seems to be that the aftermath of an event has already been played through in the imagination so its actual occurrence is no longer news.19 This hypothesis may also be supported by observations at the level of neurology. (While neural observations are probably more reliable than either introspective observations or behavioral observations, they are also the hardest to translate into talk of imagining, perceiving, or thinking.) The activation patterns specific to any one perceptual encounter are tiny compared with the activation patterns that are shared across multiple encounters with the same object and, more importantly, the activation patterns that are present when merely imagining that object. When I see a bird (or hear its song), for example, much of the resulting neural activity is more closely correlated with past experiences and with imagined experiences than with the present encounter. The sensory input of perception accounts for a minority of synaptic contacts in the cortex while the majority has been described as “re-entrant activity”— activity that was originally prompted by other perceptual input, and activity that is associated with memories and imaginings where the occasioning input is no longer present.20 More recently, of course, there has been a flurry of interest in “predictive coding”—that is, higher-level representations or coding that predicts lower-level representations or coding in such a way that lower-level coding has no effect on higher-level coding unless and until it differs from what is predicted. Andy Clark, who has been particularly influential in disseminating this research to philosophers, maintains that “it means that perception (at least, as it occurs in creatures like us) is co-emergent with (something quite like) imagination”21 for there is a “fundamental linkage between ‘passive perception’ and active imagining, with each capacity being continuously bootstrapped by the other. Perceiving and imagining (if these models are on the right

19   For evidence that even infants are often imagining possible outcomes, see Butterworth and Cochran (1980). 20   This is the terminology of Llinás and Ribary (1994), for example. 21   Clark (2015, p. 26). Clark views predictive coding as explanatory of such phenomena as priming and confirmation bias that are operative at higher levels as well as lower levels of perception.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

170  Jennifer Church track) are simultaneous effects of a single underlying neural strategy.”22 Recent neurological evidence, then, only strengthens the claim that the imagination is active whenever there is perception.

2.  Let us turn now to cases in which we perceive people as waiting or as worried, as arguing or as anxious, as sighing or as sad—cases in which we perceive people as doing certain things or as feeling certain things. Just as many different aspects of a desk can be registered simultaneously, many different features of a person can be registered simultaneously; and though no one of these features alone enables us to perceive what the person is doing or feeling, the combination enables us to perceive them as waiting, as angry, as anxious, and so on. We notice a foot twitching, narrowed eyes, the upward tilt of a head, a tightness in the voice, a shaking hand; and these observations combine to produce our perception of the person as anxious—not through any (conscious) inference but through the creation of a self-supporting gestalt. We do not reflect on these components in order to guess their significance or their source; rather, we experience them as parts of a unified whole, the objectivity of which is evident precisely through the way its parts reinforce the unity of the whole. For perceiving the person as anxious also means that the twitching foot appears as an expression of restlessness, the narrowed eyes as an expression of guardedness, tightness in the voice as an expression of emotional tension. The validation of our experience of that person as anxious is embedded within the perceptual experience itself.23 Sometimes, though, we do not need to observe multiple aspects of a person’s behavior in order to perceive them as waiting or as worried. A single gesture can cause us to perceive someone as angry, or as about to leave. A child gives a sidelong glance and we immediately perceive her as up to some mischief. We hear a catch in our friend’s voice and immediately perceive her as sad. These cases are analogous to cases in which we observe part of a familiar surface and immediately perceive it as a desk, or as a painting, or as a book. The phenomenology is not that of simply thinking of the object as a desk, or simply thinking of the voice as sad; rather, we perceive the object as a desk and we perceive the voice as sad. But if, as I’ve been suggesting, the phenomenology of perception depends on mutually supporting aspects appearing within a single experience, then our exposure to a single gesture or a single sound must be supplemented by various sorts of imagining in order to generate the phenomenology of perceiving as. This can happen in at least three distinct ways.

  Clark (2015, p. 39).  Note how this view of internal justification avoids Burge’s (2010) complaint about the hyper-­ intellectualism of accounts that require a perceiver to think about what justifies a perceptual belief. 22 23

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

perceiving people as people  171 One way relies on the reactivation of memories. As I observe a child’s sidelong look, I can recreate, in my imagination, things I have observed in the past—the tentative movements that preceded this look, the cookies in the next room, various actions that have accompanied similar looks in others, a mother’s parting words, and so on. Kant calls this “reproductive imagining”, but it is now more common to call it “iconic” or “episodic” memory.24 It can be quite deliberate—as when we try to recall just what the child was doing a moment ago; or it can arise automatically—as when this child’s look prompts the memory of another child in a similar situation. The important thing here is that the memory is recreated in imagination, not merely recounted in thought, and not necessarily via thought. Perceptual experiences from my past are reactivated, or reconstructed, and those imaginings affect the way the child now appears to me. Another example. Consider how, at a social gathering, I might come to perceive my neighbor as amused. Listening to the host’s long-winded story, I look at my neighbor to check his reaction and I see that his eyes are bright and a corner of his mouth is twitching. Taken by themselves these facial features could be expressions of any number of states of mind—repressed impatience, bodily pain, growing disdain, happy memories, and so on.25 But because my current experience is also infused with memory images of my neighbor laughing (rather than complaining) about similar stories, of the eager expression on his face a few minutes ago, and of his expressions of affection for this storyteller, I will perceive his reaction as amusement. Again, these recollections do not only guide my thinking—and, indeed, they may not prompt any thought at all; they also affect the way my neighbor appears to me. The recalled laughter, and the recalled affection, make his eyes appear as not only bright but twinkling, the corner of his mouth as not only twitching but curling upwards. A second way that imagining can contribute to the phenomenology of perceiving as is by filling in unseen parts of a scene. Only one side of the child’s face might be visible, but normally I will use my imagination to fill out her expression; her hands might be out of sight, but I will imagine them clasped behind her back; I can’t see around the corner but I can imagine the door she is headed for. When they become part of the phenomenology of my current experience, these imaginings will fill in any number of details simultaneously, not as a series of separate events. This sort of imagination is what Kant calls “productive” imagination; it doesn’t simply recreate scenes that have been witnessed before, it fills in gaps in the information we receive so as to produce a more complete picture of the world. And when that picture is suitably filled out, we might be able to perceive the child as up to some mischief, and we might be able to perceive our neighbor as amused. 24   Wollheim (1984) offers an analysis of iconic memory that focuses on the different perspectives one can have within a memory. Crowther (2013) defines episodic memory as a memory trace that “is activated through an imaginative realization of the past experience” (p. 110). 25   Even if distinct emotions are reliably correlated with distinct facial configurations, we are not reliable at recognizing these correlations.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

172  Jennifer Church There are at least three possible sources of the imagining that fills out a scene. One possible source is conditioning that we have undergone over a lifetime—conditioning that establishes associations between various sorts of experiences such that the activation of one sort tends to result, pretty much automatically, in the activation of another sort. We see someone reaching for a door and we automatically imagine them adjusting their hand position, or we see them suddenly breaking into a smile and we automatically imagine their laughter. If similar associations can be established through the evolutionary development of a species’ brain (rather than the conditioned development of an individual brain), that is a second possible source of the imagining that fills out a scene. It could be that our brains are structured in such a way that the experience of certain tones of voice, for example, triggers images of certain sorts of movement. Third, and most powerful perhaps, is the possibility of cognitive penetration, whereby thinking of something as a particular sort of thing—conceptualizing it in a particular way—changes the very look of the thing. There are good reasons for thinking that conceptualizing an object or a scene in one way rather than another affects the way the scene looks to us, not only through changes in attention, but through changes in the appearance of the properties we are attending to. Fiona Macpherson describes experiments that show how even something as simple as the color of a piece of paper changes its appearance in response to our different conceptualizations of the paper’s shape. We conceptualize the shape of the cut-out in different ways.26 And, she argues, it is plaus­ ible to suppose that thought affects such transformations in visual content through the intermediary of imagining (and dreaming, and hallucinating)—states of mind that are clearly capable of altering the character of our visual experience. In the terms of our discussion, this means that thinking of a person as angry can cause us to imagine angry faces and those imaginings can alter the way the face in front of us appears; and thinking of a child as mischievous can cause us to imagine mischievous movements and mischievous expressions that are not actually in view, and that can cause us to perceive the child as (still more) mischievous. Such filling-in by the imagination can be misleading, of course, but there should be little doubt that it can alter the phenomenology of my experience—transforming the mere thought of a child as mischievous into a perception of her as mischievous. A third kind of imagining that can contribute to the phenomenology of perceiving as is what I will call “predictive imagining.” (This may or may not be realized by what is called “predictive coding,” which I touch on below.) Predictive imagining does more than fill in the gaps—the other side of a face, the unseen hands, the unheard laugh. It makes projections about events that are not (or not yet) observable by anyone, and events that may not ever occur—what a child would do if challenged, how a neighbor would respond to my laughter, and so on. Naturally, as long as we are trying to get 26   Macpherson (2012). I find the shape and color experiments more convincing than the “black” versus “white” face experiments insofar as they do a better job of eliminating the shift-of-attention alternative explanation. But see Stokes (2013) for a helpful overview of the ongoing disputes about these and other purported proofs of cognitive penetration.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

perceiving people as people  173 things right, what we imagine as happening will be guided by what we have observed in the past—observations concerning the person in question and observations concerning others in similar situations. But our imagining can also be guided by our desires, by our aesthetic preferences, and by other things that we are thinking about. Whatever the source of such imaginings, our imagining of a person’s future responses (including responses to situations that they will ever encounter) will affect the way we experience that person’s current actions and expressions. If we imagine that a child will be chastened by a particular challenge, we are likely to perceive her present look as having a different meaning than if we imagine that she will welcome the challenge; and this effect holds whether or not the challenge actually occurs. Likewise, if we imagine someone softening in response to a kind word, the tension in his face will appear to have a more fragile quality than if we imagine him irritated by attempts at kindness— whether or not any kindness is forthcoming. Working in the other direction, also, when we perceive someone as vulnerable rather than irritated, that perception is likely to be infused by various images of their actions and expressions in situations that are merely possible, not (yet) actual. 27 Before considering the epistemic impact of these various kinds of imagining, which is the topic of Section 3, it is useful to distinguish the “infusion” effect I have been describing from an “overlay” effect that imagination can have on perceiving as. In order for our imagining to alter the ways things appear to us, and to become part of a single perceptual experience, that imagining needs to be appropriately combined with the information we are receiving through our senses; to use Kantian terms, there needs to be a “synthesis of the imagination.” I cannot offer a full account of such synthesis here, Kantian or otherwise, but it is important to recognize the difference between imagining that we experience as an “add on” versus imagining that infuses our experience of ­something such that it alters the way something looks to us, or what we perceive it as.28 Suppose that every time a particular student is criticized he becomes angry. Being a regular witness to this pattern I might start to imagine his anger whenever criticism is imminent—before any criticism is actually lodged, and before he is actually angry. 27   In a related vein, Sean Kelly (2003, 2005) has emphasized the role of expectations in our perceptual experiences. When we experience objects in space, we expect to be able to move around them in various ways and to make contact with them in various ways; likewise, when we experience people as having certain intentions or feelings we expect them to do some things rather than others, and those expectations are part of the phenomenology of our knowledge. According to Kelly, these expectations are not expectations that are explicitly entertained so much as they are felt; following Merleau-Ponty, he speaks of a “tension in the experience that feels as though it is about to resolve itself.” Felt expectations or anticipations seem to depend, however, on representing (in imagination or otherwise) what is to come—representations that may be either vague or vivid, visual or tactile, full of detail or rather sparse. So the feeling of expectation is not an alternative to imagining so much as its correlate. Siegel (2014) describes the difference between representational and non-representational understandings of Gibsonian affordances, arguing that a representational understanding is needed in order for affordances to inform the content of our perceptions. 28   Strawson (1974) used the “infusion” metaphor, acknowledging how difficult it is to unpack the metaphor in any precise way.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

174  Jennifer Church This could be regarded as a kind of foresight that is both accurate and helpful: I can “see” what is coming, I can plan accordingly. This is an example of what I have called “predictive imagining,” but it doesn’t always enter into my current perception of that student so it doesn’t always enable me to perceive him as being in a particular mental state. Sometimes it is merely an image that accompanies my perception of the student before me. Unless the imagined future affects the way he looks to me now—emphasizing the tension in his jaw, for example, or making his movements appear aggressive—it will not be a case of perceiving the student as angry. Rather than making his anger appear self-evident from within the perception itself, my imagining stands alongside my perception and may even distract me from it. Likewise, my exposure to a student’s behavior may prompt recollective imagining of a similar student’s behavior; but that recollective imagining may or may not affect the way this student appears to me. I might “see” the two students side-by-side, in my “mind’s eye” rather than seeing the one student in the other—with their juxtaposition actually changing the way that each appears to me (each appearing more like the other than they did before). When imagination contributes to the phenomenology of perception, the relation between what is present and what is imagined will be one of reciprocal reinforcement—one in which the imagining is supported by the evidence at hand and the evidence at hand is supported by the imagining.

3.  One could agree that imagining counterfactual possibilities contributes to the phenom­ enology of perceiving people as doing certain things or feeling certain things—without, however, supposing that the relevant imagining makes an epistemic contribution to our knowledge of others’ states of mind.29 One would have to be a skeptic about knowledge itself to doubt the epistemic contributions of accurate memories of a person’s past, of reasonable posits about aspects of a person that we don’t actually observe, and of justified predictions about what a person will become (or could have become). And one would have to be a skeptic about imagining itself to doubt that our imagining can sometimes recreate a memory, fill in a scene correctly, or represent the future accurately. Such skepticism is difficult, perhaps impossible. It is easy, however, to wonder how imagining as such adds to our knowledge. Isn’t imagining, insofar as it is accurate, an expression of rather than a source of knowledge? (This is one of the charges that is sometimes made against simulationist theories: isn’t our ability to simulate another’s mental state—at best—the 29   Likewise, one could believe that cognition “penetrates” the contents of visual experience without adding to—and perhaps detracting from—its epistemic value. Shea (2015) writes: “An appropriate account of the epistemic profile of a sensory process—its sensitivity, specificity, and positive and negative predictive value—will be heavily dependent upon the extent to which there are top-down as well as bottom-up influences on that process” (p. 76); and he goes on to explain why top-down influences are not necessarily epistemically pernicious.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

perceiving people as people  175 result rather than the source of our knowledge?30) When we imagine things to be one way rather than another, instead of validating our beliefs, doesn’t this imagining itself need to be validated? It might be tempting to distinguish what I would call reproductive imagining, or iconic memory, from other sorts of imagining on the grounds that it retains and reproduces the content of actual experience rather than adding to actual experience through projective filling-in of what has not been experienced. Reproductive imagining or iconic remembering, unlike productive imagining, might seem like a reliable source of knowledge insofar as it merely serves to retain knowledge of what was actually perceived. There are two problems with this suggestion, however. First, it is well documented that our memories, iconic or otherwise, already include a lot of filling-in of details that weren’t actually visible; and these details are often adjusted to better fit with what comes later—remembering a person’s manner as rude in order to better square it with their later behavior, for example. Second, as I have tried to show above, even our initial perceptions tend to be infused with imaginative filling-in and imaginative prediction. Alternatively, it might be tempting to distinguish between imagining, whether reproductive or productive, that occurs automatically and outside of one’s control versus imagining that is intentional and under one’s control. Our automatic imagining, one might suppose, has been selected for its contribution to knowledge so, other things being equal, we are entitled to trust automatic imagining but not deliberative or intentional imagining. Putting aside questions about evolution’s interest in tracking truth,31 the relevance of this distinction seems to disappear once we recognize the extent to which automatic imagining can be brought under our control, and the extent to which imagining that is initially deliberate can become automatic. On the one hand, various kinds of therapy (EMDR treatment for trauma, for example) enable people to intentionally redirect imagining that is otherwise automatic.32 On the other hand, when directed imagining becomes sufficiently repetitive it usually becomes automatic. A year of choosing to imagine the worst that might happen to one’s child, for example, will usually result in worst-outcome imagining that is largely out of one’s control. Even if we could identify the type or types of imagining that are most likely to be veridical (Descartes’s criterion of “clear and distinct” was meant to do this), this would not establish imagination as a source of knowledge. For certain sorts of imagining could be good indications of knowledge without being contributions to knowledge. Wittgenstein, for example, resists the idea that entertaining a particular image (an episodic event) could give us knowledge of what a word means or what a person feels 30   See Shannon Spaulding’s “Imagination Through Knowledge” (Chapter  9, this volume) for a fuller discussion of this charge and for one possible response. 31   See Stich (1990) for an extended critique of the assumption that evolution favors true beliefs. 32   EMDR therapy (Eye Movement Desensitization and Reprocessing) involves guided attempts to relive a past experience in such a way that the focus of one’s attention is redirected and the emotional associations one has with the memory change as a result.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

176  Jennifer Church since these are dispositional rather than episodic properties of language and of a ­person, respectively. The meaning of a word, he argues, is not given by the evocation of an associated image but by the ability to make proper use of the word; knowing the meaning of a word depends on knowing how to use it, not on having an appropriate image. His arguments center on (1) the fact that associated images are both optional and variable (thus not necessary for knowledge), and (2) the fact that images admit of many different interpretations, implying nothing on their own (thus not sufficient for  knowledge). Applied to the case of knowing the mental states of persons, the Wittgensteinian claims become (1) that we can know a person’s mental states without imagining anything about them, and that the images we do have are highly variable, and (2) that when we do imagine the past or future behavior of a person it is not the image but our interpretation of the image that makes an epistemic contribution to our knowledge of that person. While there is truth in both of these claims, they do not imply that the imagining we have been describing makes no epistemic contribution to our knowledge of other people. Regarding (1): it might indeed be possible to know a person’s mental state—by knowing what they are likely to do under various conditions—without actually imagining any of those actions or conditions. For the knowledge might be merely dispos­ itional (I could call up the information if asked but it never really comes to mind), or the knowledge might be purely propositional (I can give accurate descriptions of a person’s past or future but I can’t really imagine it). This would show that imagining is not necessary for knowledge—for not all knowledge is perceptual knowledge—but it wouldn’t show that imagining is an epistemic “dangler,” adding nothing to what is already known. Regarding (2): we can agree that images of a person need to be interpreted in order to give us knowledge of that person’s mental states. A raised eyebrow or the tilt of a shoulder, a loud outburst or a quiet exit, can indicate different things about different people in different situations. Our experience of images is seldom neutral with respect to alternative interpretations, however; different perspectives and different propensities are part of the very look, and the very feel, of what we experience. We see a downturned mouth as disapproving rather than sad, we see a stick figure as moving in one direction rather than another, and we hear a tapping finger as impatient rather than meditative. So we can accept that an accurate picture of a person in a particular setting does not suffice for knowledge of that person’s state of mind, but we need not accept that the presence of such images—especially in consort with one another— makes no epistemic contribution whatsoever. Responding to arguments against an epistemic contribution for imagining doesn’t establish, let alone explain, what its epistemic contribution actually is. How can the infusion of imagining into our perception of a person give us knowledge that we didn’t already have? There are, I think, three distinct ways our knowledge of other people benefits from infusions of the imagination. First, by imagining the past and future surround of an event, and by imagining various aspects of a scene that remain hidden from view, we

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

perceiving people as people  177 are able to check the consistency of our convictions about that scene and event. It is much easier to hold on to conflicting beliefs if we do not try to picture them all holding at once. When we activate several different memories and several different expectations all at once, it is hard to sustain inconsistencies in our view of that person. Recalling experiences of your past kindness, imagining acts of generosity that I haven’t actually observed, and anticipating what you would do in response to another person’s pain—all actively imagined, and all integrated into my current perception of your current abruptness, I am less likely to view your current behavior as dismissive or rude, more likely to view it as preoccupied and protective. The inconsistency between past kindness, unobserved generosity, and anticipated concern, on the one hand, and present rudeness on the other hand, becomes more evident, and more uncomfortable, when the inconsistent pieces are imagined all at once. Or, to take another example: if I vividly recall a child’s fear of failure, I imagine her meticulous practicing when alone, and I envision her likely response to praise, I am less likely to take her apparent confidence at face value. I create a multifaceted image of the child that is at odds with her apparent confidence. The imagined aspects of the child’s behavior could be thought about rather than imagined, of course; or they could be imagined in a slow sequence rather than more or less simultaneously. But entertaining several things at once is a particularly effective way to check for inconsistency as it prevents one from ignoring the implications of a judgment simply because it is not present to mind. Furthermore, imagining several aspects of a person simultaneously tends to be easier than thinking about several aspects of a person simultaneously insofar as images occupy space rather than time; multidimensional space accommodates many different images while single-­dimensional time normally restricts us to one thought at a time.33 This advantage can be recognized more easily perhaps, in the case of imagining different perspectives on an arrangement of objects in space. It is harder to believe that John’s house is closer to the river than Maria’s house and that Maria’s house is closer to the river than Pilar’s house and that Pilar’s house is closer to the river than John’s house if we imagine them all simultaneously. (Even if the relative distances are inferred from the imagined scene, rather than directly seen in it, imagining makes it easier to discover inconsistencies in our beliefs.34) Similarly, it is harder to believe that John can see Maria at her window but Maria can’t see John at his window if we don’t actually imagine both looking out of their windows simultaneously. People vary in their ability to recognize inconsistencies in their imagining; and people vary in their ability to recognize inconsistencies in their thoughts. Still, since perceptual inconsistencies are usually 33   Matters are a bit more complicated than this, of course, since we can’t attend to all aspects of a picture at once and since we can have the experience of two different thoughts running in tandem. The contrast still holds, however, insofar as it is easier to illustrate many different aspects of a thing all at once than to describe many different aspects of a thing all at once. 34   Owens (1996) describes cases where recollection of a past event must be supplemented by inference in order to produce further knowledge. This may cast doubt on the perceptual character of some of the knowledge we acquire through memory, but it doesn’t count against the epistemic value of recollective images.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

178  Jennifer Church harder to ignore than cognitive inconsistencies, and since imagination simulates perception, imagination can draw our attention to inconsistencies that might otherwise be overlooked. I am not claiming that thoughts have contents entirely independently of their associated perceptions—only that thoughts can sometimes be had in the absence of such perceptions and that simulating those perceptions can help to ensure their consistency. A second way in which imagining can contribute to our knowledge also derives from the way that imagining spatializes our knowledge. As we have said, fitting everything into a common space requires a kind of coordinating that is absent from a mere list of properties. This also prompts a kind of filling-in that can lead to new discoveries. When my mother calls to describe a package she has just received, I might imagine the parcel in her hand, imagine her standing by the door, imagine the telephone in her hand, imagine her bathrobe and slippered feet, and so on. As the picture gets filled in, I might realize that she can’t read the label because the light is not good there, that she can’t open the package because her hands are full, and so on. Or watching a neighbor’s uneven gait and imagining the route he will follow, I might discover that he is likely to fall, or that he won’t see the cat crossing his yard. How do these examples carry over to the case of mental states or personality traits that are not spatial in the same way? If I imagine my mother’s pleasure in her package and I simultaneously imagine her apprehension about her eyesight and her fears about her physical instability, I am more likely to realize that she is feeling teary or that she is thinking ahead to when she will next see me.35 No one of these pieces by themselves leads to those conclusions, and merely recounting them in sequence as opposed to synthesizing them into a single perception of her person is unlikely to prompt that conclusion. Both of the epistemic contributions I have described—the added check on consistency, and the added impetus to discoveries—are contributions that were emphasized by Descartes in his discussions of how visualizing, or “intuiting,” is preferable to deducing or enumerating.36 By compressing sequential deliberations into a single intuitive insight—through temporal compression in the case of deduction, and through exemplification in the case of induction—Descartes maintains that we circumvent certain limitations on the amount of information at our disposal: limitations due to failures of memory in the one case, and limitations due to the finite character of our life in the other. Circumventing these limitations enables us to better recognize inconsistencies—the first advantage cited above. More important to Descartes, however, is the 35   Maibom (this volume, Chapter 8) notes how often our imagination misleads us precisely because we fail to imagine enough details of a person’s situation. 36   Although he touches on this topic in his Meditations, it is his Rules for the Direction of the Mind (1952) that addresses these points more fully. Descartes’s term “intuitus” is usually translated as “seeing” or, as a noun, “a look, a view.” In Rule IX, especially, he likens the intuitus of the mind to the intuitus of the eyes, noting how a single act of apprehension or attention can be directed to simple objects or to many objects simultaneously.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

perceiving people as people  179 capacity of such compressions to engender new discoveries.37 When we imagine various items within a single space, we easily discover a variety of previously unspecified relations between those items and their parts. So insofar as mathematical proofs can be presented spatially (a discovery that is itself largely due to Descartes, who demonstrated the inter-translatability of algebra and geometry) they will be presented in a manner that is bound to display more relationships than were specified in the originating description. A third advantage of simultaneous imagining concerns the intimate connection between imagination and action. Presenting information in a perceptual manner allows knowledge to be translated more directly into action. Imagining someone’s aggression, or someone’s pain, insofar as it enables one to perceive them as aggressive or as in pain, is more likely to trigger a response than merely thinking that they are aggressive or in pain.38 Apart from the practical advantages (and sometimes disadvantages) of a more sure translation of knowledge into action, there is an epistemic advantage to acting on one’s judgments insofar as actions prompt reactions that offer useful feedback on one’s judgments. The more readily I respond to what I perceive as your amusement, or your anger, the more easily I will be able to check the accuracy of my judgment. If I imagine that you are about to laugh, and respond with a complicitous smile, I will quickly discover whether my imagining, and the perception that was colored by that imagining, were accurate or not. If you were angry rather than amused, my complicitous smile is likely to backfire. While thoughts about someone’s state of mind can also lead to action, those actions will be more delayed and so less likely to provide reliable feedback on the accuracy of the thought. Having described three distinct epistemic advantages that imagining brings to our knowledge of people’s mental states, it is worth noting the corresponding risks of such imagining. Corresponding to the advantages of highlighting inconsistencies in one’s beliefs, there is the risk of seeing consistency where none exists. Juxtaposing a remembered past on an observed present, or an observed present on an imagined future, can encourage us to regard people as more consistent and more unchanging than they actually are. Corresponding to the advantage of discovering new facts as we automatically fill in the “blank spaces” in our perceptions, there is the risk of presenting likely hypotheses as perceived facts and becoming less rather than more able to discover something new. If we fill in the gaps between a person’s observed 37   The example he gives (in Rule XI of his Rules for the Direction of the Mind) concerns a series of magnitudes—A, B, C, D, E—arranged in continued proportion such that the proportion of A to B is the same as B to C, and so on. The ease with which one can generate a further addition to the series—an addition whereby the proportion of F to E is the same as E to D—is unaffected by whether one considers each relation simultaneously or in sequence. But if we are asked to determine the mean between A and E, it is immensely easier if we can consider the whole sequence at once, for then it is immediately evident that C must be the mean between A and E. 38   See Van Leeuwen (2011) as well as his contribution to this volume (Chapter 3). Exceptions to this pattern include cases where empathetic imagining of another’s pain makes us want to avoid further exposure to that pain. See Maibom (2009).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

180  Jennifer Church behavior with more of the same, not only do we overlook possibly important distinctions between a person’s public and private self but, by creating experiences in which we perceive them as being in one mental state rather than another, we generate a false confidence in what we believe. Finally, corresponding to the advantage of acting quickly to receive immediate feedback on the accuracy of our judgments, there is the risk of acting too precipitously, before all the evidence is in, thereby precluding the evolution of a person’s inclinations that allows an initial recoil to mature into (for example) indignation or fascination or wry amusement. Our immediate move to action may help confirm or disconfirm our beliefs concerning a short-term mental state while preventing us from recognizing its place within a longer-term state of mind.

References Briscoe, Robert Eamon (2008). Vision, action, and make-perceive. Mind and Language 23(4): 457–97. Briscoe, Robert Eamon (in press). On the uses of make-perceive. In Fabian Dorsch and Fiona Macpherson (eds), Perceptual Memory and Perceptual Imagination. Oxford: Oxford University Press. Burge, Tyler (2010). Origins of Objectivity. Oxford: Clarendon Press. Butterworth, George, and Edward Cochran (1980). Toward a mechanism of joint visual attention. International Journal of Behavioral Development 3(3): 253–72. Cassam, Quassim (2014). Self-Knowledge for Humans. Oxford: Oxford University Press. Church, Jennifer (2008). The hidden image: A defense of unconscious imagining and its importance. American Imago 65(3): 379–404. Church, Jennifer (2013). Possibilities of Perception. Oxford: Oxford University Press. Clark, Andy (2015). Perceiving as predicting. In Dustin Stokes, Mohan Matthen, and Stephen Biggs (eds), Perception and its Modalities. New York: Oxford University Press, pp. 24–43. Crowther, Paul (2013). How images create us: Imagination and the unity of self-consciousness. Journal of Consciousness Studies 20(11–12): 101–23. Davidson, Donald (2001). Subjective, Intersubjective, Objective. Oxford: Clarendon Press. Dennett, Daniel C. (1991). Consciousness Explained. Boston, MA: Little, Brown and Co. Descartes, Rene (1952). Rules for the Direction of the Mind (trans. Elizabeth Haldane and G.R.T. Ross). Chicago, IL: Encyclopaedia Britannica. [Original work published 1681.] Dretske, Fred (1981). Knowledge and the Flow of Information. Cambridge, MA: MIT Press. Dretske, Fred (1995). Naturalizing the Mind. Cambridge, MA: MIT Press. Eilan, Naomi (2011). Experiential objectivity. In Johannes Roessler, Hemdat Lerman, and Naomi Eilan (eds), Perception, Causation, and Objectivity. Oxford: Oxford University Press, pp. 51–67. Harman, Gilbert (1973). Thought. Princeton, NJ: Princeton University Press. Johnston, Mark (2011). On a neglected epistemic virtue. Philosophical Issues 21(1): 165–218. Kelly, Sean D. (2003). Edmund Husserl and phenomenology. In Robert C. Solomon and David Sherman (eds), The Blackwell Guide to Continental Philosophy. Oxford: Blackwell, pp. 112–42.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

perceiving people as people  181 Kelly, Sean Dorrance (2005). Seeing things in Merleau-Ponty. In Taylor Carman and Mark B. N. Hansen (eds), The Cambridge Companion to Merleau-Ponty. Cambridge: Cambridge University Press, pp. 74–110. Lavelle, Jane Suilin (2012). Theory-theory and the direct perception of mental states. Review of Philosophy and Psychology 3(2): 213–30. Levine, Joseph (2011). On the phenomenology of thought. In Tim Bayne and Michelle Montague (eds), Cognitive Phenomenology. New York: Oxford University Press, pp. 103–20. Llinás, Rodolfo R., and Urs Ribary (1994). Perception as an oneiric-like state modulated by the senses. In Christof Koch and Joel L. Davis (eds), Large-Scale Neuronal Theories of the Brain. Cambridge, MA: MIT Press, pp. 111–24. Maibom, Heidi (2009). Feeling for others: Empathy, sympathy, and morality. Inquiry 52(5): 483–99. Macpherson, Fiona (2012). Cognitive penetration of color experience. Philosophy and Phenomenological Research 94(1): 24–62. McGinn, Colin (2004). Mindsight: Image, Dream, Meaning. Cambridge, MA: Harvard University Press. Millikan, Ruth Garrett (2000). On Clear and Confused Ideas: An Essay about Substance Concepts. Cambridge: Cambridge University Press. Montague, Michelle (2011). The phenomenology of particularity. In Tim Bayne and Michelle Montague (eds), Cognitive Phenomenology. New York: Oxford University Press, pp. 121–40. Noë, Alva (2002). On what we see. Pacific Philosophical Quarterly 83(1): 57–80. Noë, Alva (2003). Causation and perception: The puzzle unraveled. Analysis 63(2): 93–100. Noë, Alva (2004). Action in Perception. Cambridge, MA: MIT Press. Noë, Alva (2006). Experience without the head. In Tamar Szabό Gendler and John Hawthorne (eds), Perceptual Experience. Oxford: Oxford University Press, pp. 411–33. Noë, Alva (2008). Précis of Action in Perception. Philosophy and Phenomenological Research 76(3): 660–5. Owens, David (1996). A Lockean theory of memory experience. Philosophy and Phenomenological Research 56(2): 319–32. Shea, Nicholas (2015). Distinguishing top-down from bottom-up effects. In Dustin Stokes, Mohan Matthen, and Stephen Biggs (eds), Perception and its Modalities. New York: Oxford University Press, pp. 73–91. Siegel, Susanna (2006). Which properties are represented in perception? In Tamar Szabό Gendler and John Hawthorne (eds), Perceptual Experience. Oxford: Oxford University Press, pp. 481–503. Siegel, Susanna (2010). The Contents of Visual Experience. New York: Oxford University Press. Siegel, Susanna (2011). The contents of perception. In Edward N. Zalta (ed.), The Stanford Encylopedia of Philosophy (Spring 2011 edn). Siegel, Susanna (2014). Affordances and the contents of perception. In Berit Brogaard (ed.), Does Perception Have Content? New York: Oxford University Press, pp. 51–75. Siewert, Charles (1998). The Significance of Consciousness. Princeton, NJ: Princeton University Press.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

182  Jennifer Church Siewert, Charles (2011). Phenomenal thought. In Tim Bayne and Michelle Montague (eds), Cognitive Phenomenology. New York: Oxford University Press, pp. 236–67. Stich, Stephen P. (1990). The Fragmentation of Reason: Preface to a Pragmatic Theory of Cognitive Evaluation. Cambridge, MA: MIT Press. Stokes, Dustin (2013). Cognitive penetrability of perception. Philosophy Compass 8(7): 646–63. Strawson, P. F. (1974). Imagination and perception. In Freedom and Resentment and Other Essays. London: Methuen, pp. 45–65. Tye, Michael (1995). Ten Problems of Consciousness: A Representational Theory of the Phenomenal Mind. Cambridge, MA: MIT Press. Van Leeuwen, Neil (2011). Imagination is where the action is. Journal of Philosophy 108(2): 55–77. Wittgenstein, Ludwig (1953). Philosophical Investigations (trans. G. E. M. Anscombe). Oxford: Blackwell. Wittgenstein, Ludwig (1980). Remarks on the Philosophy of Psychology (ed. G. E. M. Anscombe and G. H. von Wright; trans. G. E. M. Anscombe). Oxford: Blackwell. Wollheim, Richard (1984). The Thread of Life. Cambridge: Cambridge University Press. Yablo, Stephen (2003). Causal relevance: Mental, moral, and epistemic. Philosophical Issues 13(1): 316–29.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Pa rt T h r e e

Skeptical Approaches

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

8 Knowing Me, Knowing You Failure to Forecast and the Empathic Imagination Heidi L. Maibom

How can we gain insight into what others think, feel, want, or do? One tradition insists that we do so by consulting a theory about which environmental factors cause what beliefs, the types of desires people have, what people tend to do under certain circumstances, and so on (Fodor  1987, Godfrey-Smith  2004, Gopnik and Wellman  1992, Maibom 2003, Stich and Nichols 1992). This view is challenged by philosophers who argue that to understand others all we need to do is to imagine we are in their shoes, and then ascribe to them what we think, feel, or want (Gordon  1986, Davies and Stone  1998, Heal  1986, Goldman  1989). Imagining our own thoughts, feelings, or desires is an intermediate step towards figuring out theirs. This kind of self-knowledge is necessary in order to gain knowledge of others on this picture. We use ourselves as a model upon which to play out counterfactual situations. This is known as simulation. I  also call such a process of imaginative projection “the empathic imagination,” whether it is performed for the purpose of understanding others or for predicting our own actions. When we simulate others, we imagine being like them in the respects we believe are relevant to understanding them in their situation. We may take on different beliefs, desires, or predilections while we “quarantine” some of our own. If carried out well, perspective change together with our capacity to forecast our own reactions to events is thought to yield understanding of others in their situation. The advantage of using our imagination in this way is that it is representationally cheap. It (largely) avoids the need to represent any systematic knowledge about human psychology or human behavior. One version of the idea maintains that you take your psychological system offline and use it to run simulations of real-life psychological reactions or behavior (Goldman 1989). According to another, you use your ability to make decisions about what to think or what to do to figure out what others will do or how they will react (Heal 1986). According to neither account do you need anything like a theory of human

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

186  Heidi L. Maibom psychology. Our imagination, suitably constrained, is sufficient to give us insight into others’ experience. Simulation is sometimes thought to give us a particularly intimate understanding not forthcoming from a more detached theoretical perspective; it enables us to know “what it is like” to be the other (Ravenscroft 1998), or to emotionally resonate or empathize with them (Goldman 1995, Maibom 2007). Unfortunately, the assumption that the empathic imagination gives us good insight into others’ thoughts, feelings, desires, or intentions is flawed. Too often we cannot figure out what we would think, feel, want, or intend under situations only slightly different from those we find ourselves in. The problem is that our ability to project ourselves into situations different from the ones that we are in is limited. Yet, we are largely un­aware of our limitations. Since a certain degree of self-knowledge is required to gain other-knowledge, failure to gain the former is reflected in failure to have the latter. Not knowing me, I don’t know you. Our mistakes follow a certain pattern. We typically think in terms of how it would make sense to react, what feelings would be “normal” under those circumstances, what actions would be best (or: common), what beliefs we ought to have, and so on. This suggests that we do not take our psychological mech­ anisms offline in any interesting sense. In other words, offline simulation is not a good account of the empathic imagination. By contrasts, accounts of simulation that focus on decision-making capacities fit the available evidence better. My proposal has four parts. First, I show that the many forecasting errors we make render simulation less useful a tool for understanding others than one might otherwise think. Second, I argue that one account of simulation—offline simulation—is not supported by the data. We are generally not capable of arousing in ourselves reactions to imagined scenarios that match those we would have to actual scenarios. Third, the pattern of errors we make suggests that forecasting our own reactions is a way of deciding how best to react. Forecasting deploys relatively high-level decision-making processes. Fourth, I argue that this type of decision-making is not the only kind. We have a way of making decisions that is more tied to our immediate environment, so-called immediate or in-the-situation decision-making. This process can avail itself of implicit memories and is apt to arouse emotional and other reactions in the subject. Longerterm decision-making does not have access to such resources, but must instead use semantic and explicit knowledge for its calculations. The empathic imagination can be understood as a forecasting device that makes use of longer-term decision-making, but cannot avail itself of the tools that are available when deciding in the situation (as offline simulation seems to suppose).

1.  Failure to Forecast People’s difficulties forecasting their actual reactions to situations range across a great variety of situations. Owning an object makes a person value it more. As a result, owners tend to charge more to part with what they own than buyers think the object is

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowing Me, Knowing You  187 worth. This is known as the endowment effect. Like most of the other cases that I discuss, people are typically unaware of this effect. If asked to predict their reactions, they get them wrong. More surprisingly, perhaps, when they recall events, people also fail to recognize that their predicted evaluations failed to match their actual evaluations (Kahneman, Knetsch, and Thaler 1991, Van Boven, Dunning, and Loewenstein 2000). People are blind, if you like, to the endowment effect. There are many other effects just like it. People predict that certain literary passages will increase the emotional impact of a section of fiction, but they are mistaken. They think that being exposed to loud noise while watching a movie will reduce the enjoyment of it, but it does not appear to be true. When asked how willing they would be to endure electric shocks without being reassured that they will not be permanently damaged, they predict that they will be less willing to do so than if they were reassured. Yet, reassurance has no such effect on their willingness to be exposed to shocks (Nisbett and Wilson 1977, pp. 245–6). A more ethically troubling effect is that people are less likely to help someone in need if others present do not help (Latané and Darley 1970). The so-called bystander effect is relatively robust, yet people predict that they will help others in need even in the presence of idle bystanders. In a similar category of seriousness, we have the Milgram effect (Milgram  1963,  1974). Stanley Milgram found that if asked by an experimenter to administer shocks of increasing severity to an unknown person allegedly as part of a learning experiment, most people do so until they have run out of voltage (26 out of 40, or 60%). Indeed, these people administer what they believe to be extremely severe shocks (but not ones that will cause permanent damage) to the subject after they believe he or she has been rendered unconscious by previous shocks (Milgram 1963). Everybody continues to administer shocks until the subject kicks at the wall and ceases to answer the questions that are presumed to justify the experiment in the first place. However, few people predict they will go that far, and most believe that their willingness to shock will tail off after reaching strong shock levels, much before they actually refuse to shock the other person (Milgram 1963).1 This failure to forecast their own actions is matched by their failure to predict how others will act. In other words, subjects fail to predict that ordinary people will be subject to the Milgram effect (Nisbett and Ross 1980). Less serious, but nevertheless morally disconcerting, is people’s tendency to cheat or steal if the immediate payoff is not money. Dan Ariely and Nina Mazar found that while a six pack of coke cans is gone after 72 hours in a dorm fridge, a plate of six one-dollar bills remains intact (Mazar and Ariely 2006, Mazar, Amir, and Ariely 2008). But who would think that they feel fine about taking someone else’s stuff, but not their money? On a more positive note, Rimma Teper, Michael Inzlit, and Elizabeth Page-Gould (2011) found that people think they are more likely to cheat on a mathematics exam 1   When people start to hesitate about shocking the other person, the experimenter provides mild prodding. For details, see Milgram (1963) or (1974).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

188  Heidi L. Maibom than they actually are. The authors suggest that because emotional arousal is not fully engaged in moral forecasting, moral judgments may not be very predictive of actions that the subject will actually perform. In other morally relevant studies, some show concordance between how subjects react on a standard trolley switch case and their previous judgments of how they would act, while others do not (Navarrete et al. 2011, Pan and Slater2). In many cases, we are wrong about how we would react in counterfactual situations because we overestimate the similarity between how we feel now and how we would feel under different circumstances. Though this might help correct for temporal discounting,3 it is typically a problem when people are experiencing visceral affect, or when they imagine being in a situation where they would experience visceral reactions that they do not currently experience (Morley 1993, Loewenstein 1996, Gilbert, Gill, and Wilson 2002, Van Boven and Loewenstein 2005). Hunger affects how we think and what we are likely to do in ways that we do not foresee when we are not hungry (Read and van Leeuwen 1998, Gilbert, Gill, and Wilson 2002). When choosing snacks for consumption in the future, people who are hungry choose to have more unhealthy snacks than people who are not (Read and van Leeuwen 1998). Furthermore, people often reverse previous choices once they are actually in the situation. So imagining being hungry, alone, does not have the effect of engaging the relevant resources to forecast one’s preferences (even in heavily constrained circumstances). Thirst appears to have similar effects on forecasting; if one is thirsty, one values drinking more (than eating, say) in a counterfactual scenario than if one is not thirsty (Van Boven and Loewenstein 2003). The literature on pain suggests similar forecasting difficulties. Pain is typically remembered as having been worse than it was (Gilbert and Wilson 2000). For instance, women tend to remember labor as more painful than they reported during labor (Terry and Gijsbers 2000), and people who donate blood remember the experience as having been worse than they reported at the time (Breckler 1994). In general, people are not particularly good at remembering the quality of their pain (Beese and Morley 1993, Terry and Gijsbers 2000). Given people’s tendency to magnify their previous pain, it is rather curious that they, at the same time, seem to underestimate the effects such pain had or will have on their behavior. Going back to childbirth, when women are asked for their preferences about anesthesia one month before going into labor or during its early stages, many women prefer to do without it. However, a size­ able number of those women reverse their decision once they go into full labor. This is true not only of first-timers but also of women who have previously given birth (Christensen-Szalanski 1984). It is something of a testament to human optimism that women tend to magnify both the pain of labor and their willingness to endure it. 2   Pan, Xueni, and Mel Slater. Should you push the switch, and would you? An experimental study on  a  moral dilemma in virtual reality. Moral Emotions and Intuitions (MEI) Conference, The Hague, Netherlands. . 3   Briefly, that we prefer less of a good now over more of that same good later.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowing Me, Knowing You  189 Like women about to give birth, alcoholics (Osiatyński  1992) and drug addicts (Gardner and Lowinson 1993) radically underestimate the effects of having a drink or taking a bit of the drug, while overestimating their ability to resist future abuse. Students predict enjoying spring break more than they actually do (Wirtz et al. 2003). And people who receive a positive diagnosis of HIV are less distressed by it than they predicted (Sieff, Dawes, and Loewenstein  1999). People tend to overestimate how strongly and for how long they will experience positive or negative affect, including guilt or regret (Wilson and Gilbert 2005). Similarly, they underestimate the effects of embarrassment on their actions (Van Boven et al. 2012) and the effects of social politeness norms on people who are being interrogated (Biderman 1960). The downstream effects of such failures of affective forecasting are significant. For example, holidays are typically expected and remembered to be more enjoyable than they actually are, and it is how people remember the experience that affects their decision to repeat it (Wirtz et al. 2003). Moreover, prisoners of war who provide information to the enemy beyond rank, name, and division, not under torture or threat of death, often experience severe guilt. The fact that they, and people more generally, are unable to imagine doing such a thing when not actually in the situation is likely to play a significant role in that guilt (Biderman 1960). One study, in particular, highlights the importance of the failure to predict our own reactions. Our confidence that we know how we would think, feel, and act under different circumstances can have serious social, political, and moral consequences. Julie Woodzicka and Marianne LaFrance (2001) compared women’s predictions of their reactions to sexual harassment with their actual reactions. If told to imagine being interviewed for a job by a male interviewer who asks them questions about their dating status and the importance to them of their being sexually attractive, most women predict that they will become angry and confront the interviewer. In actual fact, most women report being afraid and none of them confronts the interviewer. The most “extreme” reaction is that of some women who ask the interviewer to repeat the question. Here is some sign of protest, though the request was always phrased quite politely. As Woodzicka and LaFrance point out, this discrepancy suggests that we think of sexual harassment cases the wrong way. People tend to believe that if the victim does not confront the perpetrator, the victim is partly to blame. This intuition may be due to the fact that most women believe that they themselves would have confronted the offending interviewer. The fact that they do not, when they are actually in the situation, suggests that it is much more difficult for women to confront a man under those circumstances than we think. It may therefore be too much to ask that someone directly confront her sexual harasser. Understanding how people actually tend to act as opposed to how we think they ought to act, therefore, has potentially rather profound moral, social, and political consequences. Failure to forecast is exacerbated by the fact that we sometimes do not detect mismatches between what we choose and what we get (so-called choice blindness). Petter Johansson and colleagues (2005) conducted a very thought-provoking experiment in

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

190  Heidi L. Maibom which subjects were presented with photographs of two people and asked to choose the one they found more attractive. Subsequently, the experimenter showed them the photo of the other person. He then informed the subject: “this is the person you chose.” He then asked them to explain what it was about the person that they found attractive. While they were looking at the photograph of the person they had not chosen as the more attractive, they described the features of the person they had chosen (e.g. they had a nice smile), but never protested that the photo was the wrong one. This fits with Baruch Fischhoff ’s studies on the effect of outcome knowledge on remembering predictions. Fischhoff found that informing a person of an outcome tends to change their recall of their previous prediction of the probability of the outcome (Fischhoff 1975). The tendency is to think that we thought actual events were much more likely before they took place than we actually did. And returning to labor, once more, we find that the descriptions of the experience of labor pain “suggest that women provide ‘reconstructed’ reports of their labour pain” (Terry and Gijsbers 2000, p. 151). Remembering what we thought evidently involves a great amount of reconstruction. This long list of failures to forecast demonstrates that our confidence in our ability to know how we would think, feel, or act under different circumstances is misguided, though it goes mostly unchallenged. We take ourselves to be relatively authoritative and insightful about such matters, except in exceptional circumstances. The evidence I have listed shows that our authority is more limited than we think; somebody with a good knowledge of social psychology will often have better insight into our psychological reactions than we do ourselves. Richard Nisbett and Timothy Wilson (1977) famously argue that we tell more than we know. We often do not know our reasons for action—or, if you like, the aspects of the environment that have a determining influence4 on our thought and action—even when we think we do (Nisbett and Wilson 1977, Kunda 1999, Wilson 2002). In cases like subliminal perception and problem solving the solution simply “comes” to people, and they are unaware of the crucial inference or the recent percept that influenced their reactions. Nisbett and Wilson (1977) found that association, effects of position, and the warmth of a person had large, determining effects on people’s thoughts and actions. However, people denied being influenced by such facts, and typically ascribed the reasons for thinking or acting to other factors. People’s insistent denial that they acted on motives or were influenced by features of the environment suggested by the researchers (and which were the ones that did, in fact, play a determining role in their actions) arises from people’s assumption that they have privileged and authoritative access to such reasons. This suggests that we fabricate reasons for action so ­convincingly that we fool ourselves. That is not to say that we should think of all cases 4   Only some of the things that influence an action are relevant to the introspective availability of reasons or motives, to moral and legal responsibility, and so on. Those are the influences that are important in the production of a person’s thoughts, feelings, desires, or actions. I call them “determining influences.” See also Doris (2002).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowing Me, Knowing You  191 of confabulation as self-deception. We may be self-mistaken, as it were, without being self-deceived. Having said that, there are reasons to think that self-deception plays a significant role in our interactions with others. Deceptive and exploitative behaviors are often adaptive, but are more easily detected by others if we are aware of engaging in them. If, however, we are unaware of our deceit, we act more convincingly. Robert Trivers (2011) lists a range of common behaviors as evidence of this tendency. For instance, we  remember positive information about ourselves more easily than negative information and so present ourselves in an overly positive manner. Another common self-­serving activity we engage in is projection—the ascription to others of an unwanted feature of ourselves (Freud 1958, Mikulincer and Horesh 1999, Newman, Duff, and Baumeister 1997). One problem with figuring out how we would react under circumstances different from the ones that we are currently in, then, derives from our rather indirect and imperfect access to our reasons for action. Clearly, we do not have a terribly accurate theory or explicit set of ideas about what types of things influence us and what our reasons are for action. What is worse, however, is that we are equally bad at reproducing in our imagination reactions to imagined circumstances. What seems to be the problem is that we cannot trick our psychological system into reacting the way that it would were we actually in the situation by merely imagining being in the situation. Why should we think that we use the imagination in any substantial way when we predict our reactions to counterfactual events? This obviously depends on what one’s idea of the imagination is. In the literature that I have discussed, it is taken for granted that when we are asked to forecast what we would think, feel, or do we engage the imagination.5 Woodzicka and LaFrance (2001) ask their subject to imagine being in the relevant situation. Others use descriptors such as “perspective taking” (Van Boven and Loewenstein 2003) and “intrapersonal empathy” (Read and van Leeuwen 1998) to characterize the process under investigation. What happens when we think about how we would act under different circumstances? It seems appropriate to describe this as imagining how we would act, as imagination is ordinarily used (mutatis mutandis for thoughts and feelings). Another consideration is that if simulation theory is correct, then our understanding of others is mainly, primarily, or most basically, based on imagining being in the situation that the other person is in and then ascribing our imagined reactions to her. Admittedly, I accept a broad construal of imagination here. People who think that it is essential that imagination has a perceptual element (e.g. Kind 2001) may think that all the above cases are instances of forecasting, but not ones in which the imagination is employed. I do not here want to engage in a debate about what the proper conception of the imagination is. When I use the term “empathic imagination,” I mean to capture 5   Cf. Elizabeth Dunn, Norah Forrin, and Claire Ashton-James (2009, p. 331): “If imagination is an airplane, then we humans are frequent fliers; emerging research in neuroscience suggests that we devote a large portion of our mental lives to traveling into the future, as well as the past, to envision what other times and places would be like and how we would feel when we got there.”

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

192  Heidi L. Maibom the phenomenon in which we project ourselves into a different situation and see how we would react, either as a means of understanding ourselves or of understanding others. This may sometimes involve perceptual imagery, but not always.

2.  Imagining and Deciding Why are we so poor—relatively speaking—at projecting ourselves into counterfactual situations and getting good psychological results by means of it? The most obvious obstacle is lack of imagination. Or, perhaps, lack of imaginative acuity. We often do not imagine in enough detail. Our imagination is usually constrained in a variety of ways. If, for instance, our mother instructs us to imagine how we would feel if someone were to hit us over the head with a stick and take our toy, this is exactly what we will imagine: being hit over the head and deprived of our toy. Quite likely we’ll imagine little else, such as where we were being hit, who the other child was, why they might have been induced to hit us over the head, and so on. The event will not be situated the way that events in our lives are. Some psychologists call the tendency to focus exclusively on a central event, or the most obvious features of a situation, “focalism” (Wilson et al. 2000) or “the focusing illusion” (Schkade and Kahneman 1998). A simple example suffices for illustration: when asked to predict how much they will enjoy a plate of spaghetti with meat sauce, students tend to say that they will enjoy it equally in the morning or the evening. In actual fact, most enjoy it more in the evening than in the morning (Gilbert, Gill, and Wilson 2002). If you just imagine eating a nice plate of spaghetti, and not the conditions under which you do so, you are apt to make mistaken predictions about how much you’ll like it, and so on.6 So one reason we do not get good results from imagining being in a situation that we are not in is that we tend to imagine a radically impoverished scenario, not properly situated in the flow of life. If we focalize when we simulate, simulation is not such a great tool after all. The problem is not principled, however. People are able to correct for some of their forecasting errors when asked to consider other, neglected, features that are likely to form part of the imagined situation (Dunn, Forrin, and Ashton-James 2009). Nonetheless, people have a tendency to imagine more abstract scenarios, and not detailed or fleshed-out ones. The tendency to focalize is not the only problem with getting good results from imagining being in counterfactual situations. Imagine a group of old people walking along when one of them suddenly falls over. Will you rush to help? If you are like most people, you think you will. Now imagine there being many people around you. Will adding this detail make a difference? Unlikely. Most people still think they will help. 6   Fiction provides lots of scaffolding, so how we imagine scenarios here is usually richer. If the writer is right about what details matter to what mental states, we may learn more from engaging with fiction than from sitting around trying out our imaginative powers. For knowing what to imagine—including all sorts of details that we would not ordinarily think are important—is central to getting the right results from imagining.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowing Me, Knowing You  193 But, as we have seen, the presence of inactive bystanders will prevent most of us from doing so. So imagining the presence of inactive bystanders does not make you think that you will not help, despite the fact that their presence is what makes you not help in real life. Imagining the right details is no guarantee that the simulation will be successful. This is clearly a problem for offline simulation. One swallow doesn’t make a summer, of course. One failure to forecast does not show that offline simulation is false. But the bystander scenario is not unique. Consider the sexual harassment case. Most women have been exposed to sexual harassment of  one kind or another. Most likely, therefore, some of the women interviewed by Woodzicka and LaFrance had been exposed to sexual harassment. Despite having had personal experience with such cases, these women made predictions about their actions that did not take into consideration reactions that they themselves had had. This suggests first, that one’s emotional and behavioral reaction to at least some situations is quite different from what one would predict, and second, that even having personal experience with such situations is no guarantee that one will be able to correctly imagine one’s reactions to another situation of the same general kind. If you think the sexual harassment case is unpersuasive, consider hunger, pain, or ownership, which are all relatively common experiences. Pain and hunger are both examples of what are sometimes called “visceral experiences” (Morley 1993). Visceral experiences exert control over a subject’s attentional resources, their motivation, and so on to a significant extent. But the degree to which this is true is not generally recognized (Morley 1993, Loewenstein 1996, Levine, Lench, and Safer 2009). A person ­cannot typically recreate their visceral experiences in their imagination either by imagining being in situations that typically produce such experiences or by remembering situations in which they had such experiences (Morley 1993). This is easy to test at home. You simply imagine a situation with a significant amount of visceral affect (pain, hunger, elation) and forecast your reactions. Write down your projected reactions and compare with previous reports or check in the future. If you are like most people, there will be a bigger gap between your forecast and your actual reactions than you expected. Visceral reactions cannot typically be engaged by the imagination or, if you like, at will. According to George Loewenstein (1996, p. 284): with certain important exceptions, it appears that people can remember visceral sensations at a cognitive level, but cannot reproduce them, even at diminished levels of intensity. It seems that the human brain is not well equipped for storing information about pain, emotions, or other types of visceral influences, in the same way that visual, verbal, and semantic information is stored.

Other researchers suggest that perhaps storage per se is not the issue. What is at stake is the type of memory involved. Linda Levine and colleagues argue that explicit memory for episodic events quickly degrades (which we saw in the failures to remember pain and hunger above), becomes infused with semantic knowledge about the significance of such events and affect, and is influenced by the current affect, values, and appraisals

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

194  Heidi L. Maibom (Levine 1997, Levine, Lench, and Safer 2009). However, implicit memory of many such emotional events can remain relatively intact and be triggered by situations very similar to the one that caused them or in which they occurred. But they cannot be recalled at will or, if you like, by means of the imagination. Loewenstein (1996) thinks his conclusion applies mainly to “visceral” experiences. We have quite good success with other types of imaginative activities, such as visual imagery and sound. But the problem seems to be more widespread. Ownership appears to have the same impenetrable effects on behavior as the visceral experiences Loewenstein describes. Imagining owning an object does not make a difference when it comes to predict how attached one will get to it or how much money one would charge to part with it (Loewenstein and Adler 1995). People seem to be unaware that they value objects they own more than objects they do not. It is possible, therefore, that it is this lack of awareness that drives the effect. Perhaps, then, we are unable to forecast our reactions to events when we are unaware of how the situation influences us. However that might be, we are frequently unable to forecast our reactions, even when those reactions are not the result of visceral experiences or emotions. Others think you might be able to trick your system into impersonating some sort of hedonic response, though not necessarily the one you would have were you actually in the situation you are only imagining being in. As Daniel Gilbert and Timothy Wilson say (2007, p. 1354): The cortex attempts to trick the rest of the brain by impersonating a sensory system. It simulates future events to find out what subcortical structures know, but try as it might, the cortex cannot generate simulations that have all the richness and reality of genuine perceptions. Its simulations are deficient because they are based on a small number of memories, they omit large numbers of features, they do not sustain themselves over time, and they lack context. Compared to sensory perceptions, mental simulations are mere cardboard cut-outs of reality. They are convincing enough to elicit brief hedonic reactions from subcortical systems, but because they differ from perceptions in such fundamental ways, the reactions they elicit may differ as well. Although prospection allows us to navigate time in a way that no other animal can, we still see more than we foresaw.

If imagining the factors relevant to the reaction that we are hoping to replicate is not enough to create the right imagined reaction, then it may be a mistake to think of the empathic imagination in terms of offline simulation. Although imagining events in great detail sometimes gives good results, in many cases it does not matter in how much detail I imagine; I still imagine doing, feeling, or thinking what I am unlikely to do, feel, or think in the actual situation. I can imagine, in great detail, being surrounded by people who do not help, yet I still imagine helping the person in trouble. In the general run of things I cannot, by means of the imagination, trick my psychological system into producing the responses it would produce under actual circumstances, only imaginatively. What is needed is to be in the very same situation (as we shall see), or one very much like it. But then, of course, we would most likely have the relevant reaction in reality, not in the imagination.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowing Me, Knowing You  195 My point is not the relatively pedestrian one that the imagination won’t give you real reactions. Anybody who is excited about the power of the imagination undoubtedly agrees that we cannot recreate our experiences exactly simply by force of the imagination. That was never the point of the imagination in any case. We may enjoy experiences imaginatively that we would be aghast to experience in reality. We often use the imagination to avoid actually experiencing what we imagine experiencing. What people have argued, however, is that imagining being in a particular situation leads to a substantial advance in one’s understanding of other people. We can come to know or understand what someone thinks, feels, or is inclined to do by, as it were, re-enacting their experiences by means of the imagination (Gordon  1986, Goldman  1989). However, evidence concerning how people think about how they would react in counterfactual situations suggests that when they do so they do not take their psychological system offline and recreate facsimiles of reactions that they would have to those situations.

3. An Alternative Rejecting offline simulation is not tantamount to denying that people sometimes use their own imagined reactions as models upon which to understand others. There are many reasons to think that they do, most important of which is that people report doing so. I know I do it because I am sometimes consciously aware of doing it. The veracity of such awareness is hard to deny. How I do it is a different matter. Moreover, social psychologists, such as Leaf Van Boven and George Loewenstein, claim that their “results indicate that arousing emotions in oneself influences predictions of others’ preferences and decisions—evidence that is difficult to explain from a prototype or implicit theory point of view” (2005, p. 294). Their idea is that the pattern of errors people make suggests that they simulate others. This is partly true. What Van Boven and Loewenstein fail to recognize is that the pattern of errors strongly suggests that offline simulation is not a correct account of what people do when they use their own imagined reactions to understand others. What it does do is lend support to another account of simulation. This account is related to early simulation accounts according to which we use our practical reasoning skills—not in an automated fashion as offline simulation seems to suggest—but in a relatively effortful and deliberative way (Heal 1998, Davies and Stone 1998). This way of figuring out others’ reaction to things is like deciding what the right reaction is. When we simulate, we use a mishmash of theoretical assumptions, stereotypes, recollections of past events, and so on. Or so I am about to argue. We saw that people make mistakes when they forecast their own reactions. This indicates that they are not able to trick their psychological system into reacting the way it would react to real events by imagining such events. But if they cannot, offline simulation theory is wrong. It might be thought, however, that all the evidence shows is that offline simulation is worse than we thought. It is here the pattern of errors that people

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

196  Heidi L. Maibom make becomes relevant. People tend to err on the side of what they assume is reason­ able or good when they think about their reactions. They think they will confront the harasser, that they will not harm an innocent person merely because some experimenter tells them to, that they will help someone in need regardless of the presence of others, and so on. They reason that if spaghetti is good to eat, it should not matter what time of day one is eating it. This seems eminently reasonable, since the time of day is irrelevant to spaghetti’s gustatory properties. Addicts reason, like others, that just because they have one drink, it does not mean that they will have another drink. One certainly ought to be able to have only one drink. And people do all the time. So that mistake, too, seems to be a mistake of imposing normative considerations on one’s future actions. On the whole, people tend to be optimistic about their ability to resist temptation, to do the right thing, to resist social pressure, and so on. Yet, their optimistic projections often turn out to be wrong. One might ascribe this optimism to our penchant for self-enhancement, which unduly affects offline simulation by biasing our forecasts of our own reactions. Another, better, interpretation is that the ability to forecast is the offspring of our capacity to make decisions. What seems to be optimism is, in effect, the expression of what is judged to be the optimal course of action, the most rational way of thinking, what one should be experiencing under such-and-such circumstances, etc. Forecasting potentially guides us to the better option, the option that a cooler head would choose, the option that is chosen if one is not caught up in the details. When we forecast we do not predict so much as decide what to do. And when we do so, we think in terms of optimality; what it would be reasonable, right, or good to think, feel, or do.7 This way of thinking about imagining our own reactions to things fits what a number of psychologists think we do when we imagine the future. They maintain that we typically imagine the future in rather naked, decontextualized, and prototypical ways (Wilson et al. 2000, Gilbert and Wilson 2007, Levine, Lench, and Safer 2009, Kane, Van Boven, and McGraw 2012). Rather than reaching back into the past to retrieve actual memories of similar events, we engage our semantic memory. We have certain ideas about what situations are like, what social structures are like, and so on. These ideas are obviously not entirely disconnected from our personal memories. However, when people transpose themselves into the future, they primarily rely on (semantic) knowledge of such events and structures, and only secondarily on recall of actual experiences. Imagining the future is less a matter of re-enactment than of the engaging of semantic, declarative, or if you like systematic knowledge. This is consistent with the idea that much of our knowledge of the social world is based on models (Maibom 2007). 7   On the face of it, at least, the most rational thing and the most moral thing to do sometimes conflict. The math cheating experiment I mentioned before is a good example (Teper, Inzlit, and Page-Gould 2011). My diagnosis is that the subjects thought in terms of what was in their best interests when they were prospecting their action, not about what would be the right thing to do. Luckily, for some at least, their moral fiber showed itself once they were in the situation.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowing Me, Knowing You  197 Why don’t people rely more on their episodic memories when trying to foresee how they would react? The answer is that the question is based on a mistaken view of the quality of remembrance of things past (Proust’s gymnastics notwithstanding). In brief, memory stinks. Our recollection of past events is piecemeal, poor, and heavily reconstructed in light of present beliefs, ideas, and concerns. Recall Fischhoff ’s (1975) studies, which show that people often believe they thought events, that did occur, were much more likely to occur than they actually thought at the time.8 This tendency is complemented by a propensity to remember emotions as having been more in line with current knowledge or appraisals than as they actually were (Levine, Lench, and Safer 2009). Evidence is mounting that our memories of our past actions and experiences are relatively poor and heavily reconstructed (Levine, Lench, and Safer 2009, Schacter 1996). Yet, imagining oneself in the past, compared to imagining oneself in the future, is less characterized by prototypical representations of situations, is more filled in with details and contextual factors, and tends towards the extreme. For instance, if asked to remember missing the train (Gilbert and Wilson 2007), people tend to recall their worst experience of it. Conversely, people often recall only good things about events, such as holidays (Wirtz et al. 2003). Memories of emotions are tricky. When people recall visceral experiences, they do not re-experience them, however weakly (Morley 1993). They are good at recalling intensity, but this intensity is disconnected from any actual re-enactment, and often from phenomenological recall of what it was like as well (Beese and Morley 1993). Although I have talked as if all forms of forecasting are the same, it is distinctly possible that forecasting emotions and visceral reactions are special cases. Take guilt and regret. We tend to anticipate that we will experience both these emotions more intensely and longer than we actually do. Gilbert and Wilson (2000) argue that our “psychological immune system” helps us temper these reactions. We have great talent for rationalizing post hoc in such a way as to soften our failures, enhance our successes, and so on. It might, therefore, seem as if we “miswant” much of the time, when we seem overly attached to getting things right or doing good. This, at least, is what Gilbert and Wilson suggest. However, we may want to focus on the raison d’être of guilt and regret. The main benefit of these two emotions is not that they give us opportunities to feel miserable about ourselves. Rather, their purpose is likely to be forward looking. Anticipated guilt or regret is tremendously useful in cautioning us to choose our actions more carefully lest we should act in suboptimal or immoral ways. Once we have chosen a suboptimal course of action, however, guilt and regret are less adaptive, hence the experience of them is often lessened. This suggests that our anticipated emotions may be somewhat removed from our actual emotions in the same situations in part because anticipating feeling a certain emotion plays an important role in making the right decision. This may also be why pain is remembered 8   This result “is robust and well-documented,” though it is in the small to medium range as effect-sizes go (Guilbault et al. 2004, p. 113).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

198  Heidi L. Maibom to be worse than it actually is. After all, pain is typically associated with harm, and we  are more likely to survive and thrive, generally speaking, if we avoid it. Thus, imagining how we would feel seems less suited to making precise predictions than to guiding action. So far, our anticipated emotions fit with the general idea that forecasting is inextricably linked to decision-making. Unfortunately, forecasting can be influenced by emotions and visceral reactions in ways that appear to bypass deliberation altogether. If we are now hungry, it alters our forecasts of what, and how much, we want to eat in the future. That is evidently not the result of cool and collected decision-making aimed at producing optimal results. I think the best thing here is to bite the bullet and accept that although forecasting does engage our decision-making capacities, it is a fallible process liable to interferences from current strong visceral and emotional states. Even if this is the case, the data still support this interpretation of forecasting better than offline simulation. I am not alone in thinking that the imagination is the child of forecasting for the specific purpose of planning ahead (cf. Williamson, Chapter 4 this volume). Martin Davies and Tony Stone, and Jane Heal, in her later writings, have also argued that the ability to simulate relies on our capacity to make decisions and is therefore an essentially normative enterprise (Heal 1998, Davies and Stone 1998). For Heal the usefulness of simulation is mostly restricted to transitions among thoughts, and requires both the simulator and the person simulated to be rational. Davies and Stone think we deploy a normative theory about correct reasoning to understand others. Should others fail to act rationally, our theory is no good, and we need to supplement it with an empirical theory of how people actually act. I agree that thinking about ourselves and others is much more normatively infused than we are apt to think. However, both accounts tend to focus rather narrowly on either more theoretical reasoning or belief– desire calculations. I think the way we make decisions is much more complex, is related more intimately to how we feel or how we anticipate feeling, to current social and moral norms, and so on. I therefore expect a fuller account of this type of normative simulation to be much more encompassing and messy than Heal, Davies, and Stone suggest. I cannot develop such an account here. Suffice it to say that the evidence presented suggests that accounts of simulation that stress the normativity of such counterfactual reasoning are on the right track.

4. Human Decision-Making The account of forecasting our own reactions that I have presented might be thought problematic since it conflicts with a popular way of thinking about decision-making. Antonio Damasio (1994) has argued that emotions help limit the space of options that a subject considers when deliberating. In the course of experience, representations of events become associated with emotions; they become somatically marked. For

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowing Me, Knowing You  199 instance, representing sticking one’s hand into the fire becomes associated with pain, which causes the person to avoid such an action in the future. These prior associations with positive or negative affect curtail the act of making decisions. They determine which options are considered in the first place and which option, of the ones that are considered, is the best. If the somatic marker system is impaired—e.g. by damage to the orbitofrontal cortex—a person will experience problems inhibiting actions, carrying out everyday or routine tasks, and deciding what to do. Damasio’s picture fits nicely with the offline simulation theory. That is, it seems that what allows us to decide well is  the fact that we have the ability to conjure up emotional reactions to imagined scenarios.9 But what I have said so far appears to be opposed to this. I have argued that we are typically unable to engage emotional and visceral reactions to imagined ­scenarios. Furthermore, I have relied on empirical evidence. Is the evidence hopelessly confused? One of the key experimental tasks that Damasio uses to support his theory is the Iowa gambling task. This is a card game where subjects choose cards from various decks and they win or lose money depending on the cards that come up. Some decks are high-risk decks, having a few high-paying cards, but many losing cards. Others yield smaller monetary rewards, but there are more lower-paying cards and fewer losing cards (low-risk decks). Normal subjects—which is to say subjects who do not have frontal lobe damage or suffer from psychopathy—usually avoid high-risk decks after a couple of rounds. At this point, their skin conductance increases when they attempt to reach for cards from high-risk decks. This is typically interpreted as an aversive response. Subjects are unaware of this effect, Damasio claims. However, it is this aversive response, Damasio argues, that drives the decision to choose cards primarily from low-risk decks. Representations of choosing from the high-risk decks have become somatically marked. Is the Iowa gambling task a good model for decision-making? Probably not, for it is only an example of what we may call “immediate decision-making.” The negative physiological reaction occurred as a result of an action that was initiated, then aborted. Subsequent avoidance of the high-penalty decks is also the result of a decision made in the situation itself. Consequently, such decisions are not good examples of offline simu­lation. The subject is not imagining being in the situation or otherwise representing being in the situation, but is in the situation. Put differently, their representation of the situation is not composed of memories or of cognitive representations that can be tokened at will, and that are relatively concise, naked, and prototypical. Rather, it is a construction out of sensory perception. It contains details rarely represented in more 9   Damasio (1994) thinks there are a direct online loop and an as-if loop involved in decision-making. The latter can produce the same decisions as the former, only without engaging the emotions, or at least the feeling of such emotions. The operation of the as-if loop is made problematic by the false forecasts that people make, so whether the subject directly experiences emotions or not, Damasio’s account is still in apparent conflict with the data that I have been relying on.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

200  Heidi L. Maibom cognitive non-perceptual ways of thinking about that situation. We can also say that being in that situation engages the subject’s implicit memories which, as we have seen, do have the capacity to engage the psychological system in such a way as to produce responses parallel to those the “original” situation would produce. The subject learnt to correlate gain and loss with the various decks only shortly before. Talking about implicit memories here is quite apposite if indeed people are unaware of what they are learning, and cannot explain why they avoid some decks and not others, other than in relatively inarticulate terms (good deck or bad deck). The Iowa gambling task seems to be an instance of implicit learning. By contrast to this type of decision-making, we have what we might call “longerterm decision-making.” Here you are not in the relevant situation, but you are trying to imagine what would be the best option, the most reasonable conclusion, etc., in an imagined or anticipated situation. This type of decision-making requires you to represent a situation that diverges from the one that you are in. Presumably, you construct these representations using general knowledge about such a situation or, as the psychologists might say, semantic memory. Such representations are less rich than the perceptual representations we produce when we are in the situation itself. In other words, longer-term decision-making creates imagined reactions to representations of a different richness than does immediate decision-making. Consequently, we cannot simply assume that what seems to be true of immediate decision-making characterizes longer-term decision-making. In fact, we have many reasons to think that the two are quite distinct. Damasio can be right about the role that somatically marked representations play in immediate decision-making, and I can be right about longer-term decision-making and its role in the empathic imagination. For most of the cases that we have considered are cases of what I have called longer-term decision-making. We are rarely in the situation or, if you like, almost in the situation, that we are trying to forecast our reaction to. This is part of the reason, I contend, that we often fail to get our reactions right. The other part is that when we project ourselves into counterfactual scenarios, we imagine doing, thinking, or feeling what is reason­ able, right, or good. The empathic imagination—or simulation—is circumscribed by our ability to re­create the relevant situation in our imagination in sufficient detail. We have a limited capacity to recreate decision-making in the situation. We generally do not have the imaginative acuity to recreate situations in all the respects that matter to our psychological reactions. Recreating in the right detail a situation in which we are hungry or own a small, insignificant object is apparently either not possible, or too hard for most people, for it to have the downstream effects that we require to correctly forecast our subsequent reactions. (Of course, at the coarse level we do just fine: when we imagine being hungry, we imagine wanting to eat, and so on. But this may largely be the result of semantic, not episodic, memory.) Though to the subject it may seem that they are re­creating the situation in their mind just fine, they actually fail to recreate the internal

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowing Me, Knowing You  201 psychological environment sufficient to trigger the responses—in their imagination— that they would have in the actual situation. When thinking about the role affect plays in decision-making, we need to consider first that episodic memories of affect fade over time—evidence suggests that it is often simply a matter of hours—and recall becomes increasingly characterized by semantic memory of what the relevant emotion is like, under what circumstances it is experienced, and so on (Robinson and Clore 2002, Levine, Lench, and Safer 2009). Implicit memories, on the other hand, may retain their freshness and vividness, but they cannot be called to mind spontaneously. They are triggered by circumstances very like the ones they were originally experienced in. Levine, Lench, and Safer (2009) suggest that implicit memory “elicits an emotional response in the present with all the accompanying physiological arousal, narrowing of attention to goal relevant information, and consideration of a restricted range of actions” (p. 1069). In this way, implicit memories have the capacity to affect a subject in ways that explicit, and particularly semantic, memory cannot. This supports my proposal of two different ways of deciding (recall that success in the Iowa Gambling Task seems to rely on implicit memories). Now we get to the exceptions. First, we are not invariably incapable of producing emotional and visceral reactions to imagined scenarios. People with phobias and those who have experienced serious trauma can have extremely vivid flashbacks and strong reactions imagining certain scenarios. Indeed, I imagine all of us have had experiences, the remembrance of which brings forth some of the pleasure or suffering they caused us. What is important is that these are exceptions. Because they are relatively rare, I don’t think we should worry about them too much. I take it that philosophers have long abandoned the idea that we need to posit necessary and sufficient conditions. Second, people’s imaginative abilities differ from individual to individual. What I have described is meant to apply to the average person or, if you like, most people. Third, I am talking about what I call the empathic imagination—the ability to recreate or re-enact psychological experiences, whether by ourselves or by others, by means of the imagination. The visual imagination is supposedly quite good for such things as figuring out whether objects fit in certain places. I have no argument with that. However, if I am right, visualizing being in a certain situation is no better than simply imagining being in such situations when it comes to forecasting our own reactions. I take it that none of these exceptions create principled difficulties for my proposal. If what I have said so far is right, it might seem as if we rarely act rationally or even continently. Either we figure out what the best thing to do is and then do something different and worse once we get there, or we change our minds in the situation for the better option. In either case, the operation of these two types of decision-making is a match made in hell! Another way to look at things, of course, is to say that the two complement one another. The more naked way of looking at the essence of things, of

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

202  Heidi L. Maibom ignoring the circumstances, the distracting details, or the invasiveness of visceral experiences, can help us make a better decision in the long run and help us resist temptations. For instance, one decides beforehand to go to the gym after work, so as not to have to rely on feeling motivated once one has finished work. On the other hand, the wisdom carried in our bodies and locked in our implicit memories may sometimes override previous decisions for the better because the details, etc., really are of central importance to making the right decision. Take the idea that a woman alone with a man should confront him openly if he is being sexually inappropriate. Quite likely, that is not the most rational course of action, considering that a woman is likely to be inferior to a man in physical strength, and his sexually inappropriate behavior suggests that he has an attitude towards women that does not bode well for open confrontation. Sometimes our idea of the best thing to do, the most rational conclusion, etc., is wrong, and leads to conflict with our online reactions if they are more attuned to rationality. In summary, then, our capacity to make decisions in the situation uses resources that are not accessible to the imagination. When we are in a situation, we have information available to us through sensory perception of an extraordinary richness that is simply not captured in what we will later recall of the situation or in our semantic representation of that type of situation. Being in the situation unlocks the richness of implicit memory, with all the associated visceral and emotional reactions, which are closed to the pale, stylized representations of explicit memory (semantic or episodic). When we recreate situations by means of the imagination, what we have available are  these impoverished representations. Optimists, like Levine (Levine, Lench, and Safer 2009), think that the representations need to be impoverished because faithful recreation of our emotional or visceral reactions, even if less intense, is not adaptive. We are able to do many of the wonderful things we are able to do exactly because we can work with such desanguinated representations. There may be some truth to this. Representations that yield information only about essential or central characteristics of objects or events allow us to systematize and generalize about the world in a way that rich implicit memory representations cannot. We can be ecumenical about representations, accepting them with open arms in whatever form or shape they come to us. But the imagination typically only offers us pale representations compared to sensory ones, and that means that it creates a very different picture of the world for us to react to than if it had access to the richness of perceptual or, perhaps, implicit memory representations.

Acknowledgements I wrote this chapter originally for the Knowledge Through Imagination conference organized by Amy Kind and Peter Kung. Subsequently, different versions of it were presented at University of Copenhagen and Institut Jean Nicod. The wonderful questions and comments I received on all occasions greatly improved the chapter. I also thank one anonymous reviewer. This research was funded by the Canadian Social Sciences and Humanities Research Council.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowing Me, Knowing You  203

References Beese, Angela, and Stephen Morley (1993). Memory for acute pain experience is specifically inaccurate but generally reliable. Pain 53(2): 183–9. Biderman, Albert D. (1960). Social-psychological needs and “involuntary” behavior as illustrated by compliance in interrogation. Sociometry 23(2): 120–47. Breckler, Steven J. (1994). Memory for the experience of donating blood: Just how bad was it? Basic and Applied Social Psychology 15(4): 467–88. Christensen-Szalanski, Jay J. (1984). Discount functions and the measurement of patients’ values: Women’s decisions during childbirth. Medical Decision Making 4(1): 47–58. Damasio, Antonio R. (1994). Descartes’ Error: Emotion, Reason, and the Human Brain. New York: Putnam. Davies, Martin, and Tony Stone (1998). Folk psychology and mental simulation. In Anthony O’Hear (ed.), Contemporary Issues in the Philosophy of Mind. Royal Institute of Philosophy Supplement 43. Cambridge: Cambridge University Press, pp. 53–82. Doris, John M. (2002). Lack of Character: Personality and Moral Behavior. Cambridge: Cambridge University Press. Dunn, Elizabeth W., Noah D. Forrin, and Claire E. Ashton-James (2009). On the excessive rationality of the emotional imagination: A two-systems account of affective forecasts and experiences. In Keith D. Markman, William M. P. Klein, and Julie A. Suhr (eds), Handbook of Imagination and Mental Simulation. New York: Psychology Press, pp. 331–46. Fischhoff, Baruch (1975). Hindsight ≠ foresight: The effect of outcome knowledge on judgment under uncertainty. Journal of Experimental Psychology: Human Perception and Performance 1(3): 288–99. Fodor, Jerry A. (1987). Psychosemantics: The Problem of Meaning in the Philosophy of Mind. Cambridge, MA: MIT Press. Freud, Sigmund (1958). Psychoanalytic notes on an autobiographical account of a case of paranoia. In Standard Edition of the Complete Psychological Works of Sigmund Freud, ed. and trans. James Strachey, vol. 12. London: Hogarth Press, pp. 9–82. [Original work published 1911.] Gardner, Eliot T., and Joyce H. Lowinson (1993). Drug craving and positive/negative hedonic brain substrates activated by addicting drugs. Seminars in Neuroscience 5(5): 359–68. Gilbert, Daniel T., Michael J. Gill, and Timothy D. Wilson (2002). The future is now: Temporal correction in affective forecasting. Organizational Behavior and Human Decision Processes 88(1): 430–44. Gilbert, Daniel T., and Timothy D. Wilson (2000). Miswanting: Some problems in the forecasting of future affective states. In Joseph P. Forgas (ed.), Feeling and Thinking: The Role of Affect in Social Cognition. Cambridge: Cambridge University Press, pp. 178–98. Gilbert, Daniel T., and Timothy D. Wilson (2007). Prospection: Experiencing the future. Science 317(5843): 1351–4. Godfrey-Smith, Peter (2004). On folk psychology and mental representation. In Hugh Clapin, Phillip Staines, and Peter Slezak (eds), Representation in Mind: New Approaches to Mental Representation. Amsterdam: Elsevier Publishers, pp. 147–62. Goldman, Alvin I. (1989). Interpretation psychologized. Mind and Language 4(3): 161–85.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

204  Heidi L. Maibom Goldman, Alvin I. (1995). Empathy, mind, and morals. In Martin Davies and Tony Stone (eds), Mental Simulation: Evaluations and Applications. Oxford: Blackwell, pp. 185–208. Gopnik, Alison, and Henry M. Wellman (1992). Why the child’s Theory of Mind really is a theory. Mind and Language 7(1–2): 145–71. Gordon, Robert M. (1986). Folk psychology as simulation. Mind and Language 1(2): 158–71. Guilbault, Rebecca L., Fred B. Bryant, Jennifer Howard Brockway, and Emil J. Posavac (2004). A meta-analysis of research on hindsight bias. Basic and Applied Social Psychology 26(2–3): 103–17. Heal, Jane (1986). Replication and functionalism. In Jeremy Butterfield (ed.), Language, Mind and Logic. Cambridge: Cambridge University Press, pp. 135–50. Heal, Jane (1998). Understanding other minds from the inside. In Anthony O’Hear (ed.), Contemporary Issues in the Philosophy of Mind. Royal Institute of Philosophy Supplement 43. Cambridge: Cambridge University Press, pp. 83–99. Johansson, Petter, Lars Hall, Sverker Sikström, and Andreas Olsson (2005). Failure to detect mismatches between intention and outcome in a simple decision task. Science 310(5745): 116–19. Kahneman, Daniel, Jack L. Knetsch, and Richard H. Thaler (1991). Anomalies: The endowment effect, loss aversion, and status quo bias. Journal of Economic Perspectives 5(1): 193–206. Kane, Joanne, Leaf Van Boven, and A. Peter McGraw (2012). Prototypical prospection: Future events are more prototypically represented and simulated than past events. European Journal of Social Psychology 42(3): 354–62. Kind, Amy (2001). Putting the image back in imagination. Philosophy and Phenomenological Research 62(1): 85–109. Kunda, Ziva (1999). Social Cognition: Making Sense of People. Cambridge, MA: MIT Press. Latané, Bibb, and John M. Darley (1970). The Unresponsive Bystander: Why Doesn’t He Help? New York: Appleton-Century-Crofts. Levine, Linda J. (1997). Reconstructing memory for emotions. Journal of Experimental Psychology: General 126(2): 165–77. Levine, Linda J., Heather C. Lench, and Martin A. Safer (2009). Functions of remembering and misremembering emotion. Applied Cognitive Psychology 23(8): 1059–75. Loewenstein, George (1996). Out of control: Visceral influences on behavior. Organizational Behavior and Human Decision Processes 65(3): 272–92. Loewenstein, George, and Daniel Adler (1995). A bias in the prediction of tastes. Economic Journal 105(431): 929–37. Maibom, Heidi (2003). The mindreader and the scientist. Mind and Language 18(3): 296–315. Maibom, Heidi Lene (2007). The presence of others. Philosophical Studies 132(2): 161–90. Mazar, Nina, On Amir, and Dan Ariely (2008). The dishonesty of honest people: A theory of self-concept maintenance. Journal of Marketing Research 45(6): 633–44. Mazar, Nina, and Dan Ariely (2006). Dishonesty in everyday life and its policy implications. Journal of Public Policy and Marketing 25(1): 117–26. Mikulincer, Mario, and Netta Horesh (1999). Adult attachment style and the perception of others: The role of projective mechanisms. Journal of Personality and Social Psychology 76(6): 1022–34. Milgram, Stanley (1963). Behavioral study of obedience. Journal of Abnormal Social Psychology 67(4): 371–8.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Knowing Me, Knowing You  205 Milgram, Stanley (1974). Obedience to Authority: An Experimental View. New York: Harper & Row. Morley, Stephen (1993). Vivid memory for “everyday” pains. Pain 55(1): 55–62. Navarrete, C. David, Melissa M. McDonald, Michael L. Mott, and Benjamin Asher (2011). Virtual morality: Emotion and action in a simulated three-dimensional “trolley problem”. Emotion 12(2): 364–70. Newman, Leonard S., Kimberley J. Duff, and Roy F. Baumeister (1997). A new look at defensive projection: Thought suppression, accessibility, and biased person perception. Journal of Personality and Social Psychology 72(5): 980–1001. Nisbett, Richard E., and Lee Ross (1980). Human Inference: Strategies and Shortcomings of Social Judgment. Englewood Cliffs, NJ: Prentice-Hall. Nisbett, Richard E., and Timothy deCamp Wilson (1977). Telling more than we can know: Verbal reports on mental processes. Psychological Review 84(3): 231–59. Osiatyński, Wiktor (1992). Choroba Kontroli [The Disease of Control]. Warsaw: Instytut Psychiatrii i Neurologii. Ravenscroft, Ian (1998). What is it like to be someone else? Simulation and empathy. Ratio 11(2): 170–85. Read, Daniel, and Barbara van Leeuwen (1998). Predicting hunger: The effects of appetite and delay on choice. Organization Behavior and Human Decision Processes 76(2): 189–205. Robinson, Michael G., and Gerald L. Clore (2002). Episodic and semantic knowledge in emotional self-report: Evidence for two judgment processes. Journal of Personality and Social Psychology 83(1): 198–215. Schacter, Daniel L. (1996). Searching for Memory: The Brain, the Mind, and the Past. New York: Basic Books. Schkade, David A., and Daniel Kahneman (1998). Does living in California make people happy? A focusing illusion in judgments of life satisfaction. Psychological Science 9(5): 340–6. Sieff, Elaine M., Robyn M. Dawes, and George Loewenstein (1999). Anticipated versus actual reaction to HIV test results. American Journal of Psychology 112(2): 297–311. Stich, Stephen, and Shaun Nichols (1992). Folk psychology: Simulation or tacit theory? Mind and Language 7(1–2): 35–71. Reprinted in Martin Davies and Tony Stone (eds), Folk Psychology: The Theory of Mind Debate. Oxford: Blackwell, pp. 123–58. Teper, Rimma, Michael Inzlit, and Elizabeth Page-Gould (2011). Are we more moral than we think? Exploring the role of affect in moral behavior and moral forecasting. Psychological Science 22(4): 553–8. Terry, R., and K. Gijsbers (2000). Memory for the quantitative and qualitative aspects of labour pain: A preliminary study. Journal of Reproductive and Infant Psychology 18(2): 143–52. Trivers, Robert (2011). Deceit and Self-Deception: Fooling Yourself the Better to Fool Others. London: Allen Lane. Van Boven, Leaf, David Dunning, and George Loewenstein (2000). Egocentric empathy gaps between owners and buyers: Misperceptions of the endowment effect. Journal of Personality and Social Psychology 79(1): 66–76. Van Boven, Leaf, and George Loewenstein (2003). Social projection of transient drive states. Personality and Social Psychology Bulletin 29(9): 1159–68.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

206  Heidi L. Maibom Van Boven, Leaf, and George Loewenstein (2005). Empathy gaps in emotional perspective taking. In Bertram F. Malle and Sara D. Hodges (eds), Other Minds: How Humans Bridge the Divide Between Self and Others. New York: Guilford Press, pp. 284–97. Van Boven, Leaf, George Loewenstein, Edward Welch, and David Dunning (2012). The illusion of courage in self-predictions: Mispredicting one’s own behavior in embarrassing situations. Journal of Behavioral Decision Making 25(1): 1–12. Wilson, Timothy D. (2002). Strangers to Ourselves: Discovering the Adaptive Unconscious. Cambridge, MA: Harvard University Press. Wilson, Timothy D., and Daniel T. Gilbert (2005). Affective forecasting: Knowing what to want. Current Directions in Psychological Science 14(3): 131–4. Wilson, Timothy D., Thalia Wheatley, Jonathan M. Meyers, Daniel Gilbert, and Danny Axsom (2000). Focalism: A source of durability bias in affective forecasting. Journal of Personality and Social Psychology 78(5): 821–36. Wirtz, Derrick, Justin Kruger, Christie Napa Scollon, and Ed Diener (2003). What to do on spring break? The role of predicted, on-line, and remembered experience in future choice. Psychological Science 14(5): 520–4. Woodzicka, Julie A., and Marianne LaFrance (2001). Real versus imagined gender harassment. Journal of Social Issues 57(1): 15–30.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

9 Imagination Through Knowledge Shannon Spaulding

1. Introduction Imagination, as I shall use the term, is a cognitive activity that involves forming a quasi-sensory mental representation (Walton 1990). Imagining is distinct from believing, desiring, perceiving, remembering, and any combination of these mental states. In other words, imagination is a distinctive cognitive attitude, that is, a content-bearing representational state with a distinctive functional role. This conception of imagination as a distinctive cognitive attitude is not universally accepted (Langland-Hassan 2012), but it is the predominant view in the imagination literature (Schroeder and Matheson 2006). Even with the restrictions stipulated above, imagination is an incredibly diverse category of mental activities. It includes deliberate and spontaneous imagination, creative and recreative imagination, propositional and non-propositional imagination, objectual and active imagination, and conscious and non-conscious imagination, among other kinds (Gendler 2011, Van Leeuwen 2013). There is no single, unified account of imagination, nor is there a generally accepted, exhaustive taxonomy of the varieties of imagination. Moreover, the varieties of imagination listed above overlap unsystematically. Thus far, imagination has resisted comprehensive, systematic characterization. The primary reason is that many fields study imagination, including philosophy of mind, psychology, aesthetics, epistemology, and phenomenology. The features of imagination emphasized by a particular field differ and, in some cases, conflict with the features highlighted by other fields. For example, the capacity of imagination phenomenologists study bears little resemblance to the sort of imaginative activity that psychology and philosophy of mind investigate. The latter fields posit non-conscious imagination, but this idea would be nonsensical in a phenomenological framework. The sort of imagination posited in one field often has little in common with imagination in other fields. The result is that no single mental activity can do the job of imagination in all of these domains.1   Amy Kind (2013) offers a persuasive argument for this idea. She considers the role of imagination in fiction, pretense, mindreading, and modal epistemology. She argues that the features that are essential to 1

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

208  Shannon Spaulding The diversity of kinds of imagination makes it difficult, if not impossible, to answer questions about imagination per se. This collection addresses the question of whether and how we can have knowledge through imagination. This question seems intractable without specifying a particular kind of imagination. In this chapter, I shall focus on both deliberate and spontaneous imagination. Deliberate imagination is the kind of mental activity involved in philosophical and scientific thought experiments, whereas spontaneous imagination is the sort of mental activity involved in daydreams and dreams. Because deliberate imagination is at least prima facie a promising prospect for knowledge through imagination, I shall focus on whether and how we can have knowledge through either deliberate or spontaneous imagination.2 More specifically, I shall discuss whether deliberate and spontaneous imagination can give us new knowledge of contingent facts about the world. Presumably everyone acknowledges that imagination can highlight what we already know. The more interesting question is whether imagination plays a more robust epistemic role, that is, whether it yields new knowledge. Furthermore, imagination may play a role in coming to know necessary truths or truths about what is possible. This is the subject of an important and interesting debate about the source of our knowledge of modal truths (Gendler 2004, Hill 2006). However, my focus here will be on the role of imagination with respect to new knowledge of contingent facts. Hence, there are two questions to answer. Does deliberate imagination give us new justified, true beliefs about contingent facts, and does spontaneous imagination give us new justified, true beliefs about ­contingent facts? I shall argue that although it seems that deliberate imagination is a better candidate for providing knowledge, deliberate and spontaneous imagination are equal with respect to their capacity to generate knowledge. Neither capacity is sufficient to bring about new knowledge of contingent facts about the world.3 An important step in this argument is the claim that imagining is distinct from the epistemic evaluation of the ideas imagined. I shall provide evidence for this conclusion by considering the cognitive capacity of mindreading. Mindreading consists in attributing mental states to another person in order to explain and predict their behavior. It is the process through which we get knowledge of other minds. Consideration of imagination-based accounts of mindreading shows two things: (a) deliberate and spontaneous imagination are insufficient for knowledge of other minds, and (b) both can be supplemented, in a particular way, to get knowledge of other minds. The lessons from mindreading apply more generally to the puzzle of knowledge through imagination. imagination in one domain (e.g. affective responses to fiction) are irrelevant in other domains (e.g. modal epistemology). Moreover, in some cases the essential features of imagination are incompatible with the essential features of imagination in the other domains (e.g. the offline role of imagination in mindreading and the online role of imagination in pretense). See also Walton (1990, p. 19). 2   The concept of knowledge in use here is the ordinary notion of justified true belief. I do not presuppose any particular theory of epistemic justification. 3   See also Peter Langland-Hassan’s discussion of deliberate and spontaneous imagination in this volume (Chapter 2), which considers a related skeptical challenge to knowledge through imagination.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagination Through Knowledge  209 The layout of this chapter is as follows. In the next section, I present a more detailed discussion of deliberate and spontaneous imagination. In Sections 3 and 4, I discuss knowledge of other minds through imagination. In Section 5, I extrapolate from knowledge of other minds to the general puzzle about knowledge through imagination. In Section 6, I consider the lessons for knowledge through imagination and situate my view in relation to other views on knowledge through imagination.

2.  Deliberate and Spontaneous Imagination As stated above, imagination is a distinctive cognitive attitude, that is, a content-­ bearing representational state with a distinctive functional role. This is a fairly typical conception of imagination (Schroeder and Matheson  2006). Imagining is distinct from believing, desiring, perceiving, remembering, and any combination of these mental states. Moreover, belief, desire, perception, and remembering can occur in the absence of imagination, and imagination can occur in the absence of these attitudes. Although these mental states may occur together and influence each other in a particular psychological episode, they are conceptually and psychologically distinct ­attitudes. This will turn out to be an important fact in my argument. The question I am interested in answering is whether imagination itself generates knowledge. Before we can answer that question, we need a more careful explanation of the distinction between deliberate and spontaneous imagination. My characterization of deliberate and spontaneous imaginings is based on Kendall Walton’s distinction (Walton 1990, pp. 13–16). Deliberate imaginings are conscious quasi-sensory mental events that are under our voluntary control. They are under our control in the sense— and to the extent—that we can choose whether to imagine (initiation) and how the imagining goes (elaboration). An imagining is fully deliberate when both the initiation and elaboration are under our control and only partly deliberate when one of these elements is not under our control. A paradigmatic case of deliberate imagination is making up a bedtime story. When your child asks for a bedtime story you choose to imagine, and you choose the plot, the characters, and other details. Deliberate imagination is, or at least it seems to be, useful for finding practical solutions to problems. When we are unsure about how a colleague will react to some news—that they have been voted by the faculty to be the next department head—we deliberately imagine various strategies for breaking the news: perhaps blurting it out as soon as we see them, or softening them up first with a joke about the joys of administrative duties, or leaving an anonymous note in their mailbox. Deliberately imagining the conversation seems to help us figure out which news-breaking strategy will work best. Spontaneous imaginings are quasi-sensory mental representations over which we have relatively little control. An imagining is fully spontaneous when both the initiation and elaboration are not under our control and only partly spontaneous when one of these elements is under our control. Spontaneous imaginings include mental

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

210  Shannon Spaulding events such as daydreams and dreams, in which we simply find ourselves immersed.4 Whereas we consciously, voluntarily control the initiation and development of fully deliberate imaginings, this is not the case for fully spontaneous imaginings, which seem to have a life of their own. When we spontaneously imagine winning the lottery, it is as if we are experiencing a fictional account of what it is like for us to be lottery winners. We are, in a sense, participants rather than creators of our fully spontaneous imaginings. Unlike deliberate imagination, a subject typically does not utilize spontaneous imagination to solve some task or practical problem. We sometimes simply find ourselves in a free-flowing imaginative engagement. The sense that we are mere spectators rather than creators of our spontaneous imaginings is part of what makes them enjoyable, surprising, engrossing, or scary. Spontaneous imaginings may be quite vivid. However, we need not be consciously aware of our spontaneous imaginings. Imaginings, as I am understanding them, simply are quasi-sensory mental representations, and spontaneous quasi-sensory mental representations are representations that occur more or less independently of our volition. In some cases we are aware of our spontaneous quasi-­ sensory mental representations, but we need not be. Imagination is analogous to ­perception in this particular respect. Philosophers and psychologists recognize the existence of conscious and non-conscious perception. Just as we may or may not be consciously aware of our sensory mental representations, we may or may not be consciously aware of our quasi-sensory mental representations. Thus, spontaneous imaginings may be conscious or non-conscious.5 Deliberate imagination and spontaneous imagination are not entirely discrete categories. The paradigm case of deliberate imagination is one where both the initiation and elaboration are fully under our control, so that we can choose whether and how to engage in the imaginative episode. The paradigm example of spontaneous imagining is one where we have no choice over the initiation or elaboration of the imaginative episode. Beyond these paradigm examples, things are more complicated. Imaginings can be more or less deliberate and more or less spontaneous. We can have more or less control over either initiation or elaboration. Moreover, a single imaginative episode may involve both deliberate and spontaneous imagination, and an imaginative episode of one kind can turn into the other kind. These complications suggest that the distinction between deliberate and spontaneous imagination is fuzzy at the borders. Nevertheless, the distinction is straightforward in the paradigmatic cases. These nuances will not affect my argument, and for simplicity I will use paradigmatic cases of deliberate and spontaneous imagination in the rest of the chapter.

4   Walton (1990, pp. 16, 47) argues that dreaming is one form of spontaneous imagination. See also Ichikawa (2009). 5   Unconscious imagining perhaps is unorthodox, but there is a strong precedent for it. For further development of the idea of non-conscious imagining, see Church (2008), Goldman (2006), Nanay (2013), Van Leeuwen (2014), and Walton (1990).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagination Through Knowledge  211

3.  Imagining Other Minds We get knowledge of other minds through mindreading. Mindreading consists in attributing a mental state to a target in order to understand the target’s behavior and anticipate future behavior. The majority of theorists studying mindreading subscribe either to the Theory Theory (TT) or the Simulation Theory (ST) or some hybrid version of the two.6 Theory theorists argue that we understand others by employing a folk psychological theory about how mental states inform behavior. With our folk psychological theories, we infer from a target’s behavior what his or her mental states probably are. And from these inferences, plus the psychological laws in the theory connecting mental states to behavior, we predict the next behavior of the target (Carruthers and Smith 1996, Davies and Stone 1995a). The capacity that underlies the theorizing of TT is supposition, which is distinct from imagination. At this point, a brief aside is necessary to defend the distinction between imagination and supposition, as this distinction is relevant for my argument and the comparison to other views in Section 6. Alvin Goldman (2006) distinguishes between suppositional imagination (S-imagination), which involves merely supposing, positing, or assuming that P is the case, and enactment imagination (E-imagination), which involves mentally enacting what it would be like if P were the case.7 S-imagination has no sensory aspect to it, whereas E-imagination consists in the creation of quasi-sensory mental representations. Applying this distinction to theories of mindreading implies that the TT involves S-imagination, and the ST involves E-imagination. I think using the terminology this way is a mistake. On my view, imagination is distinct from supposition. First, supposing does not generate affective responses like imagination does (Kind 2013). Imaginatively engaging with fiction, daydreaming, dreaming, and deliberate imagination can produce affect. Imagining Desdemona’s fear of Othello produces in us  fear and anxiety. Daydreaming about a romantic getaway with one’s significant other causes one to feel joy. Deliberately imagining the death of a loved one produces considerable negative affect. Moreover, the difference in our affective responses to imagination and supposition is not due to the content of what we imagine. Supposing that Desdemona fears Othello, or that one will go on a romantic getaway, or that one’s loved one has died simply does not generate affect. Perhaps this is because imagination involves elaborating a scenario, filling in some of the details of what it would be like, whereas supposition does not involve such elaboration. In any case, one difference between imagination and supposition is that supposition does not generate affect but imagination often does. 6   There are other accounts of mindreading as well, e.g. Dennett’s Rationality Theory and Maibom and Godfrey-Smith’s Model Theory (Dennett 1987, Godfrey-Smith 2005, Maibom 2009). The TT and ST dominate the literature, though. 7   Currie and Ravenscroft (2002) offer a different way of distinguishing supposition and imagination, according to which supposition involves belief-like imagining but not desire-like imagining, and imagination involves belief-like and desire-like imagining.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

212  Shannon Spaulding A second difference is that we can suppose blatant contradictions—for example, that we have squared the circle—but we cannot imagine blatant contradictions. Some may think that we can imagine blatant contradictions, but this is only because we sometimes mistake imagining ourselves imagining—meta-imagining—for imagining (Sorensen 2006). That is, we imagine ourselves imagining that we have squared the circle and mistakenly conclude that we are imagining a squared circle. Conflating meta-imagining with imagining may lead some to conclude that we can imagine blatant contradictions. However, this is a mistake. We cannot imagine blatant contradictions. Or, if we can, it is difficult and rare (Gendler 2000, Weatherson 2004). However, it is quite easy to suppose blatant contradictions. We often do this in philosophical discussions, for example, in reductio ad absurdum arguments. Finally, conflating supposition and imagination inaccurately minimizes the difference between theorizing and imagining. Theorizing, the sort of cognitive activity we engage in when doing science or mathematics, is based on supposition. If we do not distinguish between supposition and imagination, then we lack a sharp distinction between theorizing and imagination. But theorizing and imagination are different cognitive activities. Theorizing in empirical matters consists in employing something like the hypothetico-deductive model, whereas imagination does not. Imagination essentially involves forming quasi-sensory mental representations, whereas theorizing does not. Moreover, this view that supposition is a kind of imagination implies that the cognitive activity in dreaming or daydreaming is different only in degree, not in kind, from the cognitive activity involved in, say, constructing models in theoretical physics. This is implausible, though. Thus, for many reasons, I regard supposition as distinct from imagination, and hence I do not characterize the TT as imagination-based.8 In contrast to the TT, the ST offers an imagination-based account of mindreading. The ST holds that we understand others via imaginative simulation. That is, we imagine ourselves in the target’s situation, and we imagine what our mental states would be and how we would behave in that situation. On the basis of this imaginative simulation we attribute to the target beliefs and desires, which we use to explain and predict the target’s behavior. Simulation-based mindreading is one of the paradigmatic examples of imagination, and many theorists regard the ST as an intuitively plausible account of how we understand other people. It clearly is relevant to the discussion of knowledge through imagination. The ST provides an explanation of how we can get knowledge of other minds through imagination.9 Imagining what it is like to do, feel, and experience what a target does, feels, and experiences can give us knowledge of the target’s mental states and future behavior. More specifically, we observe the target’s behavior and retrodictively simulate what the target’s mental states could have been to cause the observed behavior. 8   See Gendler (2006) and Weinberg and Meskin (2006) for further arguments that supposition is distinct from imagination. 9   See Heidi Maibom’s contribution to this volume (Chapter 8).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagination Through Knowledge  213 Then we run the imagined mental states through our own cognitive machinery. If it is a successful retrodictive simulation, we take the resulting imagined mental states and attribute them to the target. On the basis of those attributed mental states, we make further predictions about what the target will do (Davies and Stone 1995b). The following example illustrates the simulation heuristic. Suppose I see John making fun of Mary. I wonder why he is doing that, so I imagine myself engaged in his behavior. I imagine that I dislike Mary and want to humiliate her. I imagine that I like Mary and want to get her attention. I imagine that I am indifferent about Mary and simply want to entertain myself. I evaluate the plausibility of these imagined mental states given the observed behavior, and I conclude that I would behave as John is behaving if I liked Mary. I attribute this motivation to John, and from this attribution, I predict his future behavior. I predict that John will continue to pester Mary until she loses interest, at which point he will use a different strategy to try to keep her attention. The simulation routine described above is a manifestation of deliberate imagination. In the mindreading literature, the simulation routine described is characterized as “high-level” simulation. It involves quasi-sensory information, is consciously accessible, voluntary, subject to the agent’s control, and targets mental states of a relatively complex nature, such as propositional attitudes. In addition to high-level simulation, simulation theorists also posit “low-level” simulation. Low-level simulation is a manifestation of spontaneous imagination. Low-level simulation also is quasi-sensory, but is relatively automatic. Whereas for high-level simulation the development of the simulation is under voluntary control, for low-level simulation both the initiation and development of the simulation are not under the subject’s control. Though the low-level simulational process often occurs below the level of consciousness, the product of the simulation is consciously accessible. Lowlevel simulation targets mental states of a less complex nature, for example, basic intentions, sensations, and basic emotions. In the ST literature, it is widely accepted that the mechanism for low-level simulation is the mirror neuron system.10,11 Mirror neurons are the subject of much debate in psychology and philosophy. These neurons, some argue, are the basis of our abilities to interact socially, understand others’ thoughts and emotions, and communicate using complex language. Some have gone so far as to claim that “the discovery of the mirror neuron system will do for psychology what DNA has done for biology” (Oberman and Ramachandran 2009, p. 39). My own view is that mirror neurons are not nearly as 10   Mirror neurons were originally discovered in the brains of macaque monkeys. The existence of mirror neuron systems in monkeys has now been confirmed by a variety of methods (fMRI, transcranial magnetic stimulation, single cell recordings). There is good evidence that there are mirror neuron systems in human brains as well (Gallese, Keysers, and Rizzolatti  2004, Keysers and Gazzola  2009, Rizzolatti and Craighero 2004). 11   Both ST proponents and theorists studying mirror neurons have argued that mirror neurons are strong evidence in favor of ST (Gallese and Goldman  1998, Goldman  2006,  2009, Gordon  2005, Hurley 2005, Iacoboni 2009). Though see Spaulding (2012) for an argument to the contrary.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

214  Shannon Spaulding important as that (Spaulding 2013), but the issues about the relative importance of mirror neurons need not detain us here. For our purpose, all that matters is their role in low-level simulational mindreading. Mirror neurons are multimodal neurons that fire during the execution and observation of particular behaviors. Scientists have discovered several mirror neuron systems in the human brain. The empirical evidence suggests that humans have action, emotion, and sensation mirror neuron systems. The action mirror neuron system is found in the premotor cortex and the posterior parietal cortex, regions involved in sensory guidance of movement and the production of planned movement. Action mirror neurons activate when a subject performs a particular action and when the subject observes a target performing that same action. The same neurons that produce and guide an action, for example, grasping an object, selectively activate when the subject observes a target grasping an object (Rizzolatti and Craighero 2004). There are similar mirror neuron systems for experiencing and observing certain emotions. When I experience disgust and when I observe another person experiencing disgust the same collection of neurons in the insula activates (Calder et al. 2000, Wicker et al. 2003). Corresponding findings hold for the experience and observation of fear (Adolphs et al. 1994), anger (Lawrence et al. 2002), pain (Singer et al. 2004), and touch (Keysers and Perrett 2004). In each of these cases, groups of neurons are endogenously activated when the subject acts, emotes, or feels a certain way, and these same groups of neurons are exogenously activated (at an attenuated level) when the subject observes or even simply imagines another acting, emoting, or feeling in those same ways. Though many neurons fire for a wide variety of stimuli, mirror neurons are unique because only they selectively activate for the execution and observation of the very same behavior. The ST holds that mirror neurons are subpersonal, neural instanti­ ations of the simulation heuristic. Our own cognitive machinery is employed to simulate the target’s mental states, and this simulation generates mental state ascriptions that we use to explain and predict the target’s behavior. Consider the following case. Suppose you have just witnessed something that you find horribly disgusting. (I will let readers generate their own disgusting examples.) When you are disgusted your face naturally contorts in a particular way. Your nose wrinkles and your upper lip and cheeks are raised. When I observe your disgusted facial expression, neurons in my insula selectively activate (at an attenuated level) in the same way as if I were disgusted. I may unknowingly mimic your disgusted facial expression and even come to experience a weak feeling of disgust.12 The suggestion is 12   This emotional contagion can be particularly vivid when observing physical or emotional pain. Some readers will recall the basketball game between the Louisville Cardinals and the Duke Blue Devils in the 2013 NCAA Men’s Division I Basketball Championship Tournament. In this game, Kevin Ware, a Louisville player, fell awkwardly after attempting to block a three-point shot. When he fell, his leg fractured and six inches of his tibia protruded from his leg. Fans who saw the fall reported feeling nauseous and in pain just in virtue of seeing the fall. Watching this horrific injury was so disturbing for viewers that television

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagination Through Knowledge  215 that the very same mechanism is responsible for experiencing and perceiving disgust. I understand your disgust—more generally, your emotions, feelings, and intentions— because my mirror neurons simulate what it is like to experience what you are experiencing. According to the ST, mirror neurons realize low-level simulation. This neural simulation is, for the most part, automatic. Although we are not consciously aware of or in control of the simulational process13 in the way that we are with high-level simulation, the products of the simulation, emotional contagion and the mental state attribution, are consciously accessible. As I noted above, high-level simulation is one form of deliberate imagination, and low-level simulation is one form of spontaneous imagination. Like deliberate and spontaneous imaginings, high-level and low-level simulations exhibit variation within each category. Simulations can be more or less high-level and more or less low-level (de Vignemont 2009). High-level and low-level simulation are best understood as two ends of a continuum.

4.  The Threat of Collapse Simulation theorists argue that low-level and high-level simulation generate know­ ledge of other minds. High-level simulation proceeds through retrodictive simulation, and low-level simulation proceeds through the mirror neuron model described above. As compelling as this account may otherwise be, it faces the following skeptical challenge. It is not clear how imagination alone can give us new knowledge of others’ mental states. Imagination itself does not tell us which of the mental states imagined, if any, are likely to be correct. This is what is known as the “threat of collapse” (Davies and Stone 1995b). It is called the threat of collapse because, upon inspection of imagination-based simulation, it is evident that the ST needs the theoretical knowledge posited by the TT. Thus, it is argued, perhaps a bit hyperbolically, the ST simply collapses into the TT. The threat of collapse looms for both high-level and low-level simulation. The threat of collapse involves three related problems. First, a particular behavior is compatible with indefinitely many mental states. Recall the example of high-level simulation that I discussed earlier. I observe John teasing Mary and retrodictively simulate the mental states that could have caused this behavior. I considered three sets of mental states: John likes Mary and is trying to get her attention, he dislikes her and is trying to humiliate her, or he is indifferent to her and is simply amusing himself. These are not the only explanations compatible with John’s behavior. Perhaps John is trying to ­ rograms would not show the video. They would, however, show the pained and tearful expressions of the p players, coaches, and fans. Simply watching the facial expressions of those who had seen this injury caused real emotional distress for some observers.   I allow for the possibility of unconscious quasi-sensory imagining. See footnote 5.

13

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

216  Shannon Spaulding distract Mary to steal her wallet. John may be trying to hide his homosexuality by ­flirting with a woman. Or maybe John is seeing whether Mary would be a good fit for another friend who is looking for love. We can generate in imagination indefinitely many explanations that are compatible with John’s behavior, and these explanations have very different implications. Second, imaginative simulation provides no way to judge the plausibility of the ­various imagined mental states.14 The simulation does not tell us whether the simulation that involves John’s teasing is more plausible than the simulation that involves a ploy to steal Mary’s wallet. Both simulations are coherent and compatible with what we observe. If we were to try to figure out, with simulation resources only, what our mental states could have been to cause us to behave like John, our retrodictive simulation would have no way to decide between radically different belief-desire combinations that would explain the behavior. Third, the simulation provides no stopping point. Because there are numerous realistic mental states compatible with the observed behavior, and because the simulation itself provides no way to evaluate the plausibility of each of these imagined mental states, the retrodictive simulation, in principle, could go on forever. The simulation itself provides no way to determine when we have landed on a good-enough explanation of the observed behavior and can stop simulating. Thus, imagination-based simulation cannot, all by itself, provide knowledge of other minds. The threat of collapse applies to low-level simulational mindreading, as well. Recall that simulation theorists regard mirror neurons as the mechanism of low-level simulation. Suppose that they are right about this. The problem is that an observed behavior or facial expression is compatible with a number of different basic intentions or emotions. A blush may indicate embarrassment, happiness, anger, or even just a hot flash. The same applies even more clearly for basic intentions. A given behavioral movement may indicate an intention to eat, give, tease, throw, play with, put away, and so on. Pointing out that spontaneous imagination is realistic does not help here. We need more information than the simulation heuristic provides in order to be justified in attributing to a target a particular intention or emotion. As with high-level simulation, the threat of collapse for low-level simulation involves three related problems. First, the observed behavior is compatible with a number of mental states. Second, the low-level simulation itself provides no way to determine the plausibility of the candidate emotions or intentions. Third, there is no 14   One way to understand this problem is in terms of likelihood and probability. The likelihood of a hypothesis is the probability of the observation given the hypothesis. The simulation is relevant for comparing the likelihoods of the various imagined mental states. That is, it is relevant for determining whether the observed behavior would follow from the imagined mental states. However, the simulation provides no information about the probability of the various imagined mental states. The probability of a hypothesis is the probability of the hypothesis given the observation. The simulation procedure itself provides no information about which mental states are more probable given the behavior we observe. There is nothing in the simulation that could tell us that. This problem is compounded by the fact that any behavior is compatible with indefinitely many mental state combinations.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagination Through Knowledge  217 sufficing heuristic or stopping point built into the operation of the low-level simulation. We need other information to discriminate among the intentions that could cause the behavior, or the emotions that could cause the blush. This may for example be information about the target’s recent history, their personality, how certain situations make them feel, or perhaps folk psychological platitudes about how behaviors relate to ­mental states. Such information is not part of either high-level or low-level simulation. This is not accidental. Indeed, the appeal of the ST account is that it does not involve this kind of information. One of the selling points of the ST is that, unlike the TT, it does not require access to large bodies of information about folk psychology. Simulation merely requires an ability to imagine oneself in a target’s position and decide what one would feel, think, and do in that situation. One simply redeploys one’s own cognitive mechanisms for the purpose of mindreading (Goldman 2006). It is in this sense that the ST is an information-poor mindreading process, whereas the TT is an information-rich mindreading process. Although being information-poor is an attractive feature of the ST model of mind­ reading, the problem is that it is too information-poor. It appears that imagining other minds, all by itself, is not sufficient for new knowledge of other minds. Imagination does not tell us which mental states the target is more likely to have. Thus, for both high-level and low-level simulation, the simulation routine is not sufficient for new knowledge of others’ mental states. Most mindreading theorists hold that we can get knowledge of other minds through mental simulation, but not solely through mental simulation. The typical response to the threat of collapse is to admit that the ST must be supplemented with non-simulational resources. Mindreading theorists widely acknowledge that the ST requires some TT methods. The ST needs TT methods to evaluate the plausibility of different simulated mental states and to signal when we have hit upon a good-enough explanation of the behavior, or, failing that, when to give up. These TT resources are not limited to post hoc evaluation of simulated mental states. They can also modulate the selection of possible mental states to run through the simulation. Thus, we need not initiate the simulation with randomly selected imagined mental states.15 To evaluate the possible mental states compatible with a target’s behavior, imagination-based simulation requires theoretical knowledge about folk psychology, general background knowledge, and the cognitive capacity for inference to the best explanation, all of which are elements of the TT. The process of evaluating the possible mental states is not simulational, and it is not based on imagination. Judging the accuracy of imaginings involves a more general theoretical cognitive capacity to form and evaluate suppositions and to make inferences to the best explanation. 15   The TT is not subject to the threat of collapse, but it faces its own underdetermination problem. For any particular observed behavior, the data underdetermine the theorized mental states. The most common solution to the threat of collapse is to combine theoretical and simulational resources to yield an epistemically better process of mindreading. Thanks to Neil Van Leeuwen for discussion of this point.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

218  Shannon Spaulding Once we add theoretical knowledge and general cognitive resources described above to the imagination-based simulation, we have a compelling account of how we could get new knowledge of others’ mental states. It is an open empirical question whether or how often we in fact get knowledge of others’ mental states in this way. Nevertheless, this hybrid account of mindreading is an adequate account that avoids the threat of collapse.

5.  The Puzzle of Knowledge Through Imagination The puzzle of knowledge through imagination is that imagination seems to be epistemically useful in some contexts but also limited to the mere creative generation of ideas. The ST’s threat of collapse is a specific case of the more general puzzle of know­ ledge through imagination. It is a useful case study because it has been thoroughly examined, and the lessons learned are relevant to the more general case. Can we get new knowledge about contingent facts through deliberate or spontaneous imagination? On the face of it, it seems that at least deliberate imagination can give us such knowledge. Imagining carrying a couch through a doorway can give us know­ ledge about moving strategies. Imagining a picture hanging on a particular spot on the wall can give us knowledge about decorating a room. Imagining sending a singing telegram to inform a colleague that they have been elected new department head can give us knowledge about communicating unwelcome news. In these cases, deliberate imagination seems to generate new knowledge about contingent facts. In fact, it seems that deliberate imagination is designed precisely to give us such knowledge. However, spontaneous imagination cannot give us new knowledge of contingent facts. Spontaneous imagination is freer than deliberate imagination—we have no ­control over fully spontaneous imaginings—and this undermines its epistemic value. Spontaneous imagination could incidentally spur an idea, and we could go on to deliberate about the idea, but the spontaneously imagined idea itself does not constitute knowledge. The spontaneously imagined idea must be believed and justified in order to count as knowledge. And though in some cases the idea may be believed simply in virtue of being spontaneously imagined, this is not sufficient for justification. To justify the idea, we have to go through a conceptually and psychologically distinct evaluative process. One way to think of this intuitive asymmetry is that spontaneous imagination is part of the context of discovery, whereas deliberate imagination is part of the context of justification. Deliberate imagination is a tool for solving practical problems, which qualifies it as part of the context of justification, but spontaneous imaginings merely creatively generate ideas, which can lead to new knowledge of contingent facts only indirectly through deliberate imagining, practical reasoning, or some other cognitive intermediary. Hence, on first appearances, there is an asymmetry between deliberate and spontaneous imagining with respect to knowledge. It seems that we can get new

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagination Through Knowledge  219 knowledge directly through deliberate imagining but not directly through spontaneous imagining. I shall argue that the first appearances are mistaken. Deliberate imagination does not lead to new knowledge more directly than spontaneous imagination. Deliberate and spontaneous imaginings are on a par with respect to new knowledge. Much like high-level and low-level simulation, both capacities must be supplemented somehow in order to bring about new knowledge of contingent facts. For the reasons described above, spontaneous imagination does not directly yield new knowledge of contingent facts. It may creatively generate ideas, but it does not directly lead to knowledge. Deliberate imagination has a similarly limited epistemic role. We deliberately imagine many ideas. For example, in deliberately imagining carrying a couch through a doorway, we imagine several scenarios. We imagine pushing the couch straight through the doorway. We imagine turning the couch on its side and pushing it through. We imagine angling the couch diagonally through the doorway and turning it as we push it through. As a matter of fact, some of these strategies may work, and some of them may not. The problem is that imagination itself does not tell us which moving strategy, if any, will work. Deliberate imagination may reveal possible strategies, but these may be mixed in with impossible strategies, and imagination cannot tell us which are the strategies that would work in the actual world. Note the parallel with high-level simulational mindreading: the simulation alone cannot tell us which imagined scenario, if any, is likely to be correct. That evaluation requires additional non-simulational resources. Imagination generates the ideas, but distinct cognitive capacities are responsible for evaluating the plausibility of these ideas. The capacities that evaluate the plausibility of ideas are general cognitive capacities for deductive, inductive, and abductive reasoning, supposition, perception, and long-term and working memory. Above I offered an argument that supposition is distinct from imagination. The same considerations apply here, as well. These cognitive capacities are distinct from imagination. They can be employed in the absence of imagination, and imagination can occur without employing these cognitive capacities. Indeed, they have their own distinct functional roles. Thus, the cognitive capacities that evaluate the imagined ideas are distinct from the capacity that generates the ideas. In distinguishing imagination from other cognitive capacities, I am arguing for a narrow conception of imagination. Claiming that deliberate imaginings often are realistic does not help here, either. Even if all the scenarios imagined are realistic, this in itself does not tell us which imagined scenario, if any, is likely to be correct. To generate knowledge of contingent facts about the world, imagining moving the couch must be able to bring about a new, true, justified belief about how to get the couch through the doorway. Deliberate imagination alone cannot do this. The evaluation of ideas generated by imagination is independent from the capacity for imagination, just as the theoretical resources used to evaluate simulated ideas are conceptually independent from the simulation.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

220  Shannon Spaulding One could object that this argument works only for cases where we imaginatively generate several ideas and must evaluate each of them in order to generate knowledge. What about a case where we imaginatively generate only one idea? In such cases, there does not seem to be an independent evaluation step. Perhaps, one could argue, in these cases imagination leads directly to knowledge.16 I think this is a mistake. It is true that in some cases I imagine only one option. In these cases, I could imagine other options as well, but I do not. The reasons why I imagine only one option have something to do with my goals, motivation, cognitive load, and other psychological factors. Sometimes we imaginatively generate only one option, and there is an assumption of justification. But that does not entail that the one option is justified or that it would be selected if we bothered to imaginatively generate other options and evaluate them. The psychological fact that we sometimes do not evaluate our ideas does not imply that we are not epistemically required to evaluate them for them to count as knowledge. One could maintain that our deliberate imaginings usually are accurate, and thus the beliefs they generate reliably are true and hence justified. There is something to this idea, and I shall come back to it in the next section. However, the blanket statement that our deliberate imaginings reliably are accurate is dubious. Of course, our deliberate imaginings are accurate some of the time. In fact, sometimes we seem to know that our imaginings are accurate. Other times we coincidentally accurately imagine a scenario. Perhaps in considering the various imagined strategies for moving the couch through the doorway, unbeknownst to us, one of the strategies we imagine is in fact accurate. However, in many other cases our deliberate imaginings are not accurate. Sometimes we deliberately imagine fanciful scenarios. Other times, we attempt to imagine accurate scenarios and we can tell that our imagining is inaccurate, despite the fact that we have no problem constructing the mental representation. While imagining moving the couch, I may have no problem imagining various scenarios while knowing that none of these imagined scenarios will work in the actual world. This further suggests that some other cognitive capacity is responsible for evaluating the accuracy of our imagined ideas. Finally, sometimes we try to imagine scenarios accurately, and it seems to us that we got it right, but we are wrong. The response that imagination directly generates knowledge because imaginings usually are accurate fails because, as it turns out, our imaginings often are inaccurate. One could push this objection further by arguing that perception is fallible but nevertheless a source of knowledge, so perhaps liability for error is not so problematic for imagination. However, perception is epistemically different from imagination. First, perception is not free like imagination. We cannot perceive anything we want at will. We can perceive only what is there to perceive. We sometimes misperceive, but the ways in which we misperceive are directly related to the perceptual environment and our perceptual mechanisms. Perceptions cannot freely depart from reality like imagination, which easily and readily departs from reality in dramatic ways.   Thanks to Peter Kung for pressing this objection.

16

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagination Through Knowledge  221 The second difference between perception and imagination is that the mechanisms that test our perceptions (e.g. eye saccades, trans-saccadic information integration, mechanisms of visual attention) are internal to perception. These perceptual mechanisms serve as reality checks, reducing the likelihood of misperceptions and correcting misperceptions that exist. For imagination, the mechanisms that serve as reality checks are independent of imagination. These include general background information, theoretical knowledge pertaining to the particular subject matter, and general cognitive capacities for abductive, inductive, and deductive reasoning, memory, and perception. Similarly, mental simulation does not lead directly to knowledge of other minds; justification of these mental simulations is the job of general theoretical resources that are distinct and independent from simulation. Thus, the analogy between perception and imagination fails. Perception can be a direct source of know­ ledge, but imagination cannot. Neither deliberate nor spontaneous imagination leads directly to knowledge. Both may creatively generate ideas that we may come to believe, but these ideas must be justified in order to count as knowledge. Cognitive capacities distinct from imagination evaluate spontaneously or deliberately imagined ideas. Thus, spontaneous and deliberate imagination are on par epistemically; neither directly generates knowledge.

6.  Lessons for Knowledge Through Imagination Imagination seems to play an epistemic role in thought experiments, deliberating about practical problems, mindreading, and other contexts. However, the puzzle of knowledge through imagination suggests that imagination is not sufficient for new knowledge of contingent facts about the world. Just as there is nothing in mental simulation itself that could evaluate the plausibility of various mental states that would explain a target’s behavior, there is nothing in the capacity of imagination itself that could evaluate the accuracy of the possibilities we imagine. I have argued that the cognitive capacity to imagine scenarios is distinct from the cognitive capacities that underlie our ability to judge the accuracy of our imaginings. All of these considerations suggest a fairly pessimistic evaluation of the epistemic role of imagination. However, the solution to the threat of collapse suggests a solution to the general puzzle of knowledge through imagination. Just as imagination-based mental simulation must be supplemented with general knowledge, folk psychological information, and general cognitive capacities to evaluate hypotheticals and infer the best explanation, imagination in general also must be supplemented in order to generate knowledge. My suggestion is that for deliberate and spontaneous imaginings to yield new knowledge of contingent facts, they must be supplemented in the same way. Specifically, they must be supplemented with general background information, theoretical knowledge pertaining to the particular subject matter, and general cognitive capacities for

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

222  Shannon Spaulding abductive, inductive, and deductive reasoning. Call these capacities and supplementary information knowledge-plus. Knowledge-plus is distinct and independent from imagination. These capacities can be employed in the absence of imagination, and no particular aspect of knowledge-plus is necessary for imagination. Moreover, these cognitive capacities have their own functional roles. Thus, there is good reason to think that imagination is conceptually and psychologically distinct from knowledge-plus. It is plausible that knowledge-plus interacts with imaginings in two ways: it evaluates the imaginings we entertain, thereby allowing us to conclude that what we have imagined is (or is not) an accurate representation of the world; and it modulates our imaginings, thus influencing the sorts of imaginings we entertain in the first place. If this is right, it explains why we sometimes have no problem imagining a scenario but have difficulty knowing whether it is accurate. Our ability to imagine is independent from our ability to judge the accuracy of our imaginings. It also explains why imaginings are not (or at least not always) completely random. For example when we deliberately imagine moving the couch, we do not need to run through all of the logically possible ways to get the couch through the doorway. The scenarios we entertain in ­imagination typically are plausible solutions to the problem because they are modulated by knowledge-plus. A consequence of my argument is that spontaneous imagination and deliberate imagination are equal with respect to knowledge. In terms of the distinction between the context of discovery and the context of justification mentioned in Section 3, it turns out that both spontaneous and deliberate imagination are associated with the context of discovery. Nevertheless, both kinds of imagination can be supplemented, in the same way, with knowledge-plus, to generate new knowledge of contingent facts. One may spontaneously or deliberately imagine a scenario, employ knowledge-plus to evaluate the accuracy of this scenario, and thereby come to know a new contingent fact. Thus, both spontaneous and deliberate imagination can lead indirectly to knowledge. It will be instructive to situate my account in relation to a few well-known views about the epistemic role of imagination. Specifically, I will compare my view to three other views on the spectrum that range from highly skeptical to enthusiastically supportive of knowledge through imagination. This discussion is meant to be an illustrative comparison rather than an argument for or against these accounts. Consider first John Norton’s (2004) account of thought experiments, which in my terminology are a kind of deliberate imagination. Norton argues that imagination-based thought experiments can yield knowledge but only because the contemplation of imaginary scenarios consists in the execution of an argument. The role of imagination in thought experiments is merely decorative. The quasi-sensory mental representations simply are ­picturesque clothing for inductive or deductive arguments. My account of the epistemic role of imagination is quite different from Norton’s account. I do not think that thought experiments always are covert arguments, nor do I think that imagination simply is a decorative addition. My view is that imagination

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagination Through Knowledge  223 has a distinctive and more important epistemic role than this. Imagination generates ideas that may not be available through perception, memory, or other cognitive capacities. We may go on to consider these ideas, and this may lead to new knowledge of  contingent facts. Unlike Norton, I do not regard imagination’s role as merely decorative. Tamar Szabό Gendler (2004) argues that the contemplation of imaginary scenarios evokes quasi-sensory intuitions, the contemplation of which reliably yields true beliefs about contingent features of the world. It is not entirely clear whether quasi-sensory intuitions differ from quasi-sensory mental representations that I posit. A further ambiguity is whether the role of imagination simply is to produce these intuitions or to contemplate them, as well. If quasi-sensory intuitions are the same thing as quasi-sensory mental representations, and if the role of imagination simply is to produce these representations, then Gendler’s view is not different from mine. Using Gendler’s terminology, my view is that imagination produces quasi-sensory intuitions, but imagination does not contemplate these quasi-sensory intuitions. Other cognitive capacities are responsible for the evaluation of quasi-sensory intuitions. Finally, in this volume (Chapter 4), Timothy Williamson argues that the primary function of imagination is to provide a means for knowledge. Both deliberate and spontaneous imagination (voluntary and involuntary in his terminology) can yield new knowledge of contingent facts.17 Williamson offers an evolutionary argument for this claim.18 He argues that because imagination is selective and reality-oriented, it enables us to prepare for possibilities, avoid dangers, solve practical problems, and take advantage of opportunities. This capacity would confer an evolutionary advantage for humans. Thus, it is plausible that imagination was selected for providing a non-perceptual means for knowledge of contingent facts about the world. Williamson’s view of the imagination differs from mine in two salient respects. First, though his notions of voluntary and involuntary imagination are similar to deliberate and spontaneous imagination, his conception of imagination is much broader than mine. In addition to quasi-sensory mental representations, he includes non-sensory mental representations, supposition, forming and evaluating subjunctive conditionals as part of imagination. Second, Williamson assumes that being selective is part of the function of the faculty of imagination. In Section 1, I argued that imagination is distinct from cognitive capacities like ­perception, memory, and reasoning, and in Section 3 I gave three arguments that ­supposition should be treated as distinct from imagination, as well. These cognitive capacities operate independently of imagination, and have their own functional roles. Thus, I concluded that there are good reasons to think they are conceptually and psychologically distinct from imagination. I argued that the selectivity of imagination is   See also Kind (Chapter 6, this volume).   Ichikawa (Chapter 5, this volume) similarly offers an evolutionary argument for our ability to process quotidian modalities reliably. 17 18

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

224  Shannon Spaulding not part of the function of imagination itself. Other sources of knowledge and general cognitive capacities, the collection of which I refer to as knowledge-plus, are responsible for the selectivity of imagination. These cognitive capacities are general and not unique to imagination, thus there are good reasons to think they too are distinct from imagination. On my view, imagination generates ideas, whereas knowledge-plus modulates imaginings and allows us to know when our imaginings are accurate. My project here is to get very clear on what imagination is and determine the epistemological consequences that follow from this. While my disagreement about the conception of imagination and what properly follows from it may not challenge Williamson’s conclusion that imagination evolved to provide a means for knowledge, it may affect the proposed evolutionary story for how imagination came to have this function. Despite the pessimistic conclusion about whether imagination directly yields knowledge, I do think imagination plays an important epistemic role. It can be a non-­ perceptual means for knowledge, but only through knowledge-plus. Distinguishing the role of imagination from the role of knowledge-plus is important especially in this volume, the goal of which is to determine whether and how we can get knowledge through imagination. On my view, what we imagine is modulated by knowledge-plus, and to the extent that our imaginings are epistemically useful it is in virtue of know­ ledge-plus. Thus, rather than knowledge through imagination, a more appropriate slogan for my project is imagination through knowledge.

Acknowledgements The idea for this chapter was conceived at the Knowledge Through Imagination workshop at Claremont McKenna College. I am grateful to Amy Kind and Peter Kung for organizing such an interesting, fruitful workshop. I had particularly useful discussions related to this chapter topic with Peter Langland-Hassan, Shen-yi Liao, Heidi Maibom, Neil Van Leeuwen, and Timothy Williamson. Thanks to the two reviewers for this book whose comments on my chapter were challenging and useful. Finally, thanks to the participants in my graduate seminar on imagin­ ation at Oklahoma State University for helping me more clearly articulate my arguments.

References Adolphs, R., D. Tranel, H. Damasio, and A. Damasio (1994). Impaired recognition of emotion in facial expressions following bilateral damage to the human amygdala. Nature 372(6507): 669–72. Calder, Andrew. J., Jill Keane, Facundo Manes, Nagui Antoun, and Andrew W. Young (2000). Impaired recognition and experience of disgust following brain injury. Nature Neuroscience 3(11): 1077–8. Carruthers, Peter, and Peter K. Smith (eds) (1996). Theories of Theories of Mind. Cambridge: Cambridge University Press.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Imagination Through Knowledge  225 Church, Jennifer (2008). The hidden image: A defense of unconscious imagining and its importance. American Imago 65(3): 379–404. Currie, Gregory, and Ian Ravenscroft (2002). Recreative Minds: Imagination in Philosophy and Psychology. Oxford: Oxford University Press. Davies, Martin, and Tony Stone (eds) (1995a). Folk Psychology: The Theory of Mind Debate. Oxford: Blackwell. Davies, Martin, and Tony Stone (1995b). Mental Simulation: Evaluations and Applications. Oxford: Blackwell. de Vignemont, Frédérique (2009). Drawing the boundary between low-level and high-level mindreading. Philosophical Studies 144(3): 457–66. Dennett, Daniel C. (1987). The Intentional Stance. Cambridge, MA: MIT Press. Gallese, Vittorio, and Alvin Goldman (1998). Mirror neurons and the simulation theory of mind-reading. Trends in Cognitive Sciences 2(12): 493–501. Gallese, Vittorio, Christian Keysers, and Giacomo Rizzolatti (2004). A unifying view of the basis of social cognition. Trends in Cognitive Sciences 8(9): 396–403. Gendler, Tamar Szabό (2000). The puzzle of imaginative resistance. Journal of Philosophy 97(2): 55–81. Gendler, Tamar Szabό (2004). Thought experiments rethought—and reperceived. Philosophy of Science 71(5): 1152–63. Gendler, Tamar Szabό (2006). Imaginative resistance revisited. In Shaun Nichols (ed.), The Architecture of the Imagination: New Essays on Pretence, Possibility, and Fiction. Oxford: Oxford University Press, pp. 147–73. Gendler, Tamar Szabό (2011). Imagination. In Edward N. Zalta (ed.), The Stanford Encyclopedia of Philosophy (Fall 2011 edn). Godfrey-Smith, Peter (2005). Folk psychology as a model. Philosophers’ Imprint 5(6): 1–16. Goldman, Alvin I. (2006). Simulating Minds: The Philosophy, Psychology, and Neuroscience of Mindreading. New York: Oxford University Press. Goldman, Alvin I. (2009). Mirroring, simulating and mindreading. Mind and Language 24(2): 235–52. Gordon, Robert M. (2005). Intentional agents like myself. In Susan Hurley and Nick Chater (eds), Perspectives on Imitation: From Neuroscience to Social Science, vol 2: Imitation, Human Development, and Culture. Cambridge, MA: MIT Press, pp. 95–106. Hill, Christopher S. (2006). Modality, modal epistemology, and the metaphysics of consciousness. In Shaun Nichols (ed.), The Architecture of the Imagination: New Essays on Pretence, Possibility, and Fiction. Oxford: Oxford University Press, pp. 205–35. Hurley, Susan (2005). The shared circuits hypothesis: A unified functional architecture for control, imitation, and simulation. In Susan Hurley and Nick Chater (eds), Perspectives on Imitation: From Neuroscience to Social Science, vol. 1: Mechanisms of Imitation and Imitation in Animals. Cambridge, MA: MIT Press, pp. 177–93. Iacoboni, Marco (2009). Imitation, empathy, and mirror neurons. Annual Review of Psychology 60: 653–70. Ichikawa, Jonathan (2009). Dreaming and imagination. Mind and Language 24(1): 103–21. Keysers, Christian, and Valeria Gazzola (2009). Unifying social cognition. In Jaime A. Pineda (ed.), Mirror Neuron Systems: The Role of Mirroring Processes in Social Cognition. New York: Humana, pp. 1–35.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

226  Shannon Spaulding Keysers, Christian, and David I. Perrett (2004). Demystifying social cognition: A Hebbian perspective. Trends in Cognitive Sciences 8(11): 501–7. Kind, Amy (2013). The heterogeneity of the imagination. Erkenntnis 78(1): 141–59. Langland-Hassan, Peter (2012). Pretense, imagination, and belief: The Single Attitude theory. Philosophical Studies 159(2): 155–79. Lawrence, Andrew David, Andrew J. Calder, Stephen W. McGowan, and Paul M. Grasby (2002). Selective disruption of the recognition of facial expressions of anger. Neuroreport 13(6): 881–4. Maibom, Heidi (2009). In defence of (model) theory theory. Journal of Consciousness Studies 16 (6–8): 360–78. Nanay, Bence (2013). Between Perception and Action. Oxford: Oxford University Press. Norton, John D. (2004). On thought experiments: Is there more to the argument? Philosophy of Science 71(5): 1139–51. Oberman, Lindsay M., and V. S. Ramachandran (2009). Reflections on the mirror neuron system: Their evolutionary functions beyond motor representation. In Jaime A. Pineda (ed.), Mirror Neuron Systems: The Role of Mirroring Processes in Social Cognition. New York: Humana, pp. 39–62. Rizzolatti, Giacomo, and Laila Craighero (2004). The mirror-neuron system. Annual Review of Neuroscience 27: 169–92. Schroeder, Tim, and Carl Matheson (2006). Imagination and emotion. In Shaun Nichols (ed.), The Architecture of the Imagination: New Essays on Pretence, Possibility, and Fiction. Oxford: Oxford University Press, pp. 19–39. Singer, Tania, Ben Seymour, John O’Doherty, Holger Kaube, Raymond J. Dolan, and Chris D. Frith (2004). Empathy for pain involves the affective but not sensory components of pain. Science 303(5661): 1157–62. Sorensen, Roy (2006). Meta-conceivability and thought experiments. In Shaun Nichols (ed.), The Architecture of the Imagination: New Essays on Pretence, Possibility, and Fiction. Oxford: Oxford University Press, pp. 257–72. Spaulding, Shannon (2012). Mirror neurons are not evidence for the Simulation Theory. Synthese 189(3): 515–34. Spaulding, Shannon (2013). Mirror neurons and social cognition. Mind and Language 28(2): 233–57. Van Leeuwen, Neil (2013). The meanings of “imagine” Part I: Constructive imagination. Philosophy Compass 8(3): 220–30. Van Leeuwen, Neil (2014). The meanings of “imagine” Part II: Attitude and action. Philosophy Compass 9(11): 791–802. Walton, Kendall L. (1990). Mimesis as Make-Believe: On the Foundations of the Representational Arts. Cambridge, MA: Harvard University Press. Weatherson, Brian (2004). Morality, fiction, and possibility. Philosophers’ Imprint 4(3): 1–27. Weinberg, Jonathan M., and Aaron Meskin (2006). Puzzling over the imagination: Philosophical problems, architectural solutions. In Shaun Nichols (ed.), The Architecture of the Imagination: New Essays on Pretence, Possibility, and Fiction. Oxford: Oxford University Press, pp. 175–202. Wicker, Bruno, Christian Keysers, Jane Plailly, Jean-Pierre Royet, Vittorio Gallese, and Giacomo Rizzolatti (2003). Both of us disgusted in my insula: The common neural basis of seeing and feeling disgust. Neuron 40(3): 655–64.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

10 Thought Experiments in Ethics Peter Kung

1. Introduction Imagine this scenario: [Wolf] is a great transplant surgeon. Five of his patients need new parts—one needs a heart, the others need, respectively, liver, stomach, spleen, and spinal cord—but all are of the same, relatively rare, blood-type. By chance, [Wolf] learns of a healthy specimen, [Lamb,] with that very blood-type. [Wolf] can take [Lamb’s] parts, killing him, and install them in his patients, saving them. Or [Wolf] can refrain from taking [Lamb’s] parts, letting his patients die. (Thomson 1976, p. 206)

This scenario is a thought experiment. If you are familiar with the philosophical literature in ethics, chances are you have imagined this case, or other famous cases, many times before. Thought experiments are central to ethical theorizing; many articles become famous for the memorable thought experiments they feature. In addition to the transplant case, some of the best-known thought experiments are Thomson’s (1971) kidnapped violinist, Foot’s (1967) trolley problem, Williams’s (1973b) Jim and the Indians, and Rawls’s original position (1999). These thought experiments are justly famous because they have driven, and continue to drive, debate in ethics. The methodology of thought experiments is commonplace not just in ethics, but in the entirety of philosophy, though admittedly there is not universal agreement about how and why it works. I use “thought experiment” in a very inclusive way, to cover any  case, example, scenario, or hypothetical situation intended to provide data for ­theorizing.1 In this chapter, I will focus primarily on thought experiments in ethics that are used as counterexamples—I will call these counterexample thought experiments (or CTEs, for short)—and I will argue that we have to be much more cautious about how we use CTEs in ethics. The reason is that the ethics CTEs that I focus on have a distinctive feature: they feature forced choices with fixed outcomes. For instance, in the 1   Compare Gendler’s definition: “To draw a conclusion on the basis of a thought experiment is to make a judgment about what would happen if the particular state of affairs described in some imaginary scenario were actually to obtain” (1998, p. 398).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

228  Peter Kung transplant case above, Wolf is presented with exactly two choices, to harvest Lamb’s organs or not. Wolf is forced to choose between those two options, and those two options only. Wolf cannot choose another course of action. If you respond to this case by suggesting, say, that Wolf should contact other countries for potential donors with the rare blood type, then you are not playing by the rules of the thought experiment game. The thought experiment is intended to focus your attention on a stark moral choice between two and only two options. Furthermore, the outcome of each choice is mandated in the thought experiment: if Wolf harvests Lamb’s organs, the five patients live, but Lamb dies. If Wolf does not harvest Lamb’s organs, then Lamb lives, but the five patients die. To respond, “Wolf does not harvest Lamb’s organs, but saves the five patients another way” is again not to play by the rules. It is to avoid the dilemma the thought experiment sets out. To set out that dilemma, the outcomes need to be more than merely probable. They need to be fixed. Forced choices with fixed outcomes matter because, according to a generally (though not universally) accepted understanding of the role of CTEs in philosophical methodology, CTEs work by presenting genuinely possible scenarios. I will argue that imagining CTEs gives us no reason to believe that forced choices with fixed outcomes are genuine possibilities. The resulting picture of ethics CTEs is that, to provide data for theorizing, CTEs must describe more realistic scenarios, realistic in the sense that scenarios cannot abstract away from the fact that choices have many possible outcomes. In reality, outcomes are not guaranteed: it is possible that even the overwhelmingly likely alternative will not occur. Good thought experiments will also promise no guarantees. This means that any ethical view that counts outcomes as ethically relevant will have to take seriously moral risk.2 Consideration of moral risk must be built into the foundation of our ethical theorizing, and not just added as an afterthought once we have crafted our ethical theories. Here is the plan for the chapter. Section 2 reviews a generally accepted way to understand the method of counterexample, and explains why thought experiments need to present genuine possibilities. I then turn to imagination in Section 3, and explain how we imagine thought experiments. A key distinction emerges in Sections 3.1 and 3.2 between the pictorial and non-pictorial content of imagination. I use that distinction in Section 3.3 to argue that not every CTE we imagine presents a genuine possibility. I  apply these considerations about imagination and genuine possibilities to ethics CTEs in Section 4, and I conclude that imagining ethics CTEs does not give us reason to think forced choices with fixed outcomes are genuine possibilities. Section 5 briefly reviews whether there are other ways to salvage ethics CTEs with forced choices with fixed outcomes, and offers a tentative negative verdict. I conclude that the prevalence of CTEs featuring forced choices with fixed outcomes is a genuine problem for the methodology of contemporary ethics. 2   In the chapter I focus on consequentialism, but the upshot holds for a broader class of theories. Commonsense morality, for example, counts outcomes as ethically relevant.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Thought Experiments in Ethics  229

2.  The Method of Counterexample Let me begin by sketching the method of counterexample. At the end of this ­section and then again at the end of the chapter, I will address other functions of thought experiments besides counterexamples. I take it the following back and forth is familiar from philosophy articles and philosophical discussions. You have a theory T. T dictates that in circumstances C, proposition P is true. But lo, some smart person comes up with a counterexample: a case in which C, yet P is false. Too bad for your theory. Here is a straightforward explanation of what is happening: if theory T is a run-ofthe-mill philosophical theory, it is necessarily true if true. Smart person’s thought experiment points to a possibility, a metaphysically possible situation, in which your theory T delivers the wrong result. Hence theory T cannot be a necessary truth. This points to one simple way that CTEs can be illuminating. Philosophical theories purport to be necessary truths. Disproving a putative necessary truth takes just a single metaphysically possible case, which the thought experiment provides. The case has got to be metaphysically possible to be inconsistent with the putatively necessarily true theory. In the very best cases, it will be intuitively quite clear that theory T delivers the wrong answer. Much of philosophy involves conjuring up just the right case. It is indisputable that some famous thought experiments are intended to fit this model. Let me offer some examples of philosophers who explicitly stress the metaphysical possibility of the counterexamples they offer. Consider these discussions of radical skepticism: The hypothesis that all of our present experiences are the deceptions of an evil demon is not absurd. It seems to be a genuine metaphysical possibility. So we can’t reject that hypothesis out of hand. If we do know that we’re not being deceived by an evil demon, it’s plausible that that knowledge would have to rest on things we know about our environment on the basis of perception. (Pryor 2000, p. 524) 1. Massive error is possible about the external world. 2. A belief is prima facie justified only if formed in a way where massive error is not possible. 3. Thus, it is a priori that no belief about the external world is prima facie justified. (Graham 2007, p. 29)

Graham explicitly labels (1) a metaphysical thesis. Here is a philosopher of mind examining Descartes’s argument for the real distinction between mind and body: The father of modal thought experiments in modern philosophy of mind is Descartes. Descartes held that it is metaphysically possible that he exists and there be no spatially extended objects. (Taliaferro 1986, p. 95)

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

230  Peter Kung And finally, here is a discussion from the free will literature about Frankfurt cases: If one is to establish by means of a FSE [Frankfurt-style example] that people can exercise their will without there being an available alternative sequence in which they do not so exercise their will then that FSE must express a metaphysical possibility. (Cain 2003, p. 221)

It seems clear that in at least some cases ethics uses the same methodology as in other areas of philosophy. When used as counterexamples, thought experiments in ethics illuminate by showing us metaphysically possible situations in which an ethical theory delivers the wrong answer.3 It seems evident, for example, that many (but perhaps not all) of the famous cases I listed in the first sentence of this chapter— Thomson’s kidnapped violinist, Foot’s trolley problem, Williams’s Jim and the Indians, and Foot’s transplant “donor”—have been taken to present counterexamples to ­ethical claims.4 Two clarifications. First, this is not to say that the sole or even most important purpose of these cases is to serve as counterexamples in the sense of providing metaphys­ ically possible cases that falsify a theory. The literature describes a variety of different functions for thought experiments, only one of which is to serve as a counterexample in that sense.5 The fact that, if my main analysis below is correct, these thought experiments do not provide metaphysically possible cases that falsify theories does not show that they cannot serve some other purpose. At the end of the chapter, I will touch briefly on the prospects of thought experiments to serve other purposes if they are not suit­ able to be counterexamples. But I think providing us with counterexamples is an important function of thought experiments, so we do need to take seriously an argument that thought experiments in ethics frequently have features that render them unsuitable to be counterexamples. Second, this is not exactly a new worry. Philosophers do worry about the metaphysical possibility of cases sometimes, though they do not always put it in those terms. Sometimes philosophers complain that a case is “too outlandish” or “too recherché.” Or they pick at some particular feature of the case that bothers them.6 But absent some type of principled, comprehensive way to sort the “safe” cases from the “too outlandish” cases, this kind of complaint can seem ad hoc, or like special pleading. Philosophy is rife with unrealistic, recherché cases. Why reject this recherché case that happens to be a counterexample to your favored theory? Perhaps you are prepared to jettison philosophical thought experiments entirely. But absent principled reasons to abandon such a central philosophical tool, this looks suspiciously like a heroic attempt to save your theory.

  Henceforth when I use “possible” without qualification, I shall mean metaphysical possibility.   By contrast, Rawls’s original position case never seemed intended to serve as a counterexample. 5   For a helpful summary, see Walsh (2011). 6   See Wilkes (1988), Elster (2011), and Walsh (2011). 3

4

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Thought Experiments in Ethics  231

3.  Modal Epistemology Our starting point is that serving as counterexamples is an important role for ethics thought experiments, and that, to be a good counterexample, a thought experiment must give us a metaphysically possible case. I will argue that we need not check CTEs in ethics case by case for metaphysical possibility. We can instead draw general conclusions about the ability of whole classes of CTEs in ethics to tell us about metaphysical possibility. To make this case, we will need a modal epistemology. Given the theme of this volume, let us start with a popular view in the modal epistemology literature: that we gain evidence for metaphysical possibility via sensory imagination.7 Working with sensory imagination will give us a framework to understand when sensory imagination cannot provide evidence for metaphysical possibility. I will eventually argue that the considerations about sensory imagination and metaphysical possibility apply more broadly, to non-sensory imagination as well as to other putative non-imaginative sources of evidence for possibility. But those arguments will come later. Let us begin with sensory imagination. Sensory imagination is a faculty that combines and rearranges the things discovered in the actual world in new ways. Consider visual imagination; visual imagination allows us to literally picture these new combinations, and that is how it provides us with evidence that the new combinations are possible.8 Now sensory imagination is not limited to what we can depict, in the sense that we can imagine situations that contain more than what is pictured. Peacocke (1985, p. 19) offers the example of visually imagining a cat completely hidden behind a suitcase. We picture the suitcase but not the cat. It is nonetheless true that we have imagined that there is a cat behind the suitcase. Let us call the things that we imagine but that are not pictured the non-pictorial content of the imagining. The assumption that imaginings have non-pictorial content is widespread, if often undefended.9 You picture the suitcase and you imagine—without picturing—that there is a cat behind it. Just to be clear, as I am using this terminology, a  single episode of imagining has both pictorial and non-pictorial content. The non-pictorial content is part of the imagining. Furthermore, this is a theoretician’s distinction. I am not claiming that there are two introspectively distinguishable processes, each of which generates its own kind of content. Nor am I am claiming that there are two distinct psychological processes that create the two different kinds of content, or that imagining is a two-stage process, like writing in street names on a map you’ve drawn. To the imaginer, there is one seamless psychological episode of imagining a cat 7   The most famous example is Hume (1978, I.ii.2). Others include Chalmers (2002), Geirsson (2005), Gendler (2000b), Hart (1988), Hill (1997), and Yablo (1993). 8   I will primarily discuss visual imagination in this chapter, though everything I say about visual imagination should apply mutatis mutandis to imagining in other sensory modalities. 9  Not all authors are explicit, but see Byrne (2007), Kind (2001), Noordhof (2002), and Williams (1973a).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

232  Peter Kung hidden behind a suitcase; the pictorial and non-pictorial content come to mind at once as a package. Let’s examine a little more closely each kind of content. We’ll begin by clarifying pictorial content, mainly by explaining how not to be misled by the term “picture.” Then we will turn to non-pictorial content, before returning briefly to clarify a lingering puzzle about pictorial content.

3.1  Pictorial content When I speak of imagining having “pictorial content,” I mean that sensory imagination involves mental imagery in a familiar way: picturing in the mind’s eye, sounds in the mind’s ear, and so on. As I noted above, I will focus on the visual modality, hence the term “picturing,” but everything I say about picturing applies to other sensory modalities. The “paint” of these mental pictures is the conscious qualitative properties of the imaginative experience. I assume that these conscious qualitative properties— after this subsection, for readability I’ll refer to them as “paint” again—have representational properties.10 Just as, according to many theories of perception, the conscious qualitative properties of perception experience present in a direct and immediate way the world around us, I think it is plausible that the conscious qualitative properties of imaginative experience also present in a direct and immediate way. When you imagine Michelle Obama with short hair, you have a conscious experience with qualitative properties. Many of those qualitative properties are of colors; to use Peacocke’s (1985) terminology, a portion of your experience has black´ properties, and it is in virtue of those black´ properties of your experience that your experience is of something black. This is just a technical way of saying that when you paint the picture of Michelle Obama in your mind’s eye, you use black mental paint for her black hair. The fact that you use black mental paint is why your mental picture is of something black. While familiar, terms like “picture” and “mental image” have one unfortunate drawback: they invite thinking of imagining as examining an internal picture, or an internal mental image. A sense-data theory understands mental pictures or mental images this way. Criticisms of those internal pictures or internal images are familiar, so I will not review them here.11 My use of “picture” and “image” is not intended to refer to internal pictures or images. I use the terms simply as a shorthand, a less clumsy way to refer to the conscious qualitative properties of the imaginative experience.

3.2  Non-pictorial content Let us now turn to non-pictorial content. Imagine Michelle Obama clinking a glass in preparation for a toast at your birthday party. The appearance of a figure who resembles Michelle Obama and the other guests 10   Those who think experience isn’t to be understood in terms of representational content can recast this assumption in their favored terms. 11   See Kind (2001) and McGinn (2004).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Thought Experiments in Ethics  233 are certainly pictured. The particular shade of the birthday cake’s icing, the light and shadow on the table, and the shape of the table the cake sits on seems pictured as well. What is not pictured? One way to test for the non-pictorial content is to ask ourselves whether we could imagine something different while keeping the picture the same.12 Can you, while keeping the mental picture the same, imagine Michelle Obama’s duplicate giving a toast at your birthday party? (It’s too dangerous to let the actual first lady loose in the world, so the government keeps a duplicate on hand for that purpose.) It seems like you can. Similarly, if you imagined your mother at the party, you could imagine instead that your mother’s long-lost identical twin made it to the party when your mother could not. Again, it seems like the pictorial content is the same. Hence the test suggests that identities are non-pictorial content. When you imagine a particular individual (Michelle Obama), or a particular object (my kitchen table, that very object), the image doesn’t, in virtue of its qualitative features, depict particularity. The image does not distinguish between qualitatively identical tokens of the same type. But we can capture those distinctions in imagination. I can imagine Jason Collins rather than his identical twin Jarron. To make that distinction, we need non-pictorial content. The test also reveals that many types are not pictured. You imagine Michelle Obama, and you imagine that she is a human being. The mental picture, however, does not distinguish between a human being and a convincing cyborg duplicate. So again, there has to be some non-pictorial content that settles the fact that you imagine a human being rather than a convincing cyborg.13 The test also reveals that many relationships are not pictured. Perhaps you imagine that you are friends with Michelle Obama, and that is why she has come to your party. But there is no mental picture of your friendship. The same mental picture can be used in imagining that you and Michelle are friends, or that you and Michelle have just met and, natural politician that she is, Michelle excels at forming instant bonds with people. For our purposes in this chapter, another kind of non-pictorial content will be crucial. The varieties of non-pictorial content discussed so far specify individuating details about the things you picture. You picture someone who looks like Michelle Obama; you imagine that it—the someone who looks like Michelle Obama—is Michelle Obama, that she is a human being, and that she is your friend. But many of the things we imagine may have little or nothing to do with the things we picture. You imagine, for example, that Michelle is giving a toast for your birthday in Philadelphia. The fact that it is your birthday, rather than say, your retirement party, or a reception for you to announce that you are running for Congress is the kind of fact I will call   See Wittgenstein (1958), Peacocke (1985), Kind (2001), and Kung (2010).   I have suggested that you imagine Michelle is a human being. Does that mean you imagine that she is a mammal? Or that she exerts some miniscule gravitational attraction on the objects around her? Questions like that raise issues similar to those discussed in truth in fiction, which do not directly bear on the matter here. See Walton (1990, ch. 4) for discussion. 12 13

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

234  Peter Kung background information.14 Our ability to include background information in our sensory imaginings is a huge part of imagination’s power. You can construct in imagination a very elaborate backstory about how Michelle Obama comes to be at your birthday party, while restricting the picturing to just the party itself. You can specify all sorts of details about the event that are not pictured: it’s a Tuesday. The party is in Philadelphia’s Rittenhouse Square area. Michelle Obama is tired from a grueling week of fundraising events.15

3.3  Non-pictorial content and metaphysical possibility I have sketched a view according to which in any episode of sensory imagination, some content is pictorial and some content is non-pictorial. More work is needed to make that distinction precise.16 But the sketch I have already given suffices for us to draw some important modal epistemological conclusions relevant to our assessment of CTEs in ethics. The key observation is that non-pictorial content does not provide evidence for metaphysical possibility. Let me explain. Start with the category of background information I described above: it seems too easy to imagine something by just specifying it in the background information. There are few if any limits on our ability to imagine background information. Imagination is celebrated for this near-limitless power. But it is this same power that makes imagining via background information unsuitable for guiding us towards metaphysical possibility. Limits on imagination have been discussed in the literature responding to the puzzle of imaginative resistance. The puzzle was first posed in terms of imagining countermoral truths. Can you imagine that female infanticide is morally permissible—not just that people in the imagined situation think female infanticide is morally permissible, but that they are right to think so (Gendler 2000a)? The answer seems to be no. Commentators observed that there are other cases that fit the same general pattern: Can you imagine that a figure shaped like the Canadian maple leaf is ovular (Yablo 2002)? Again the answer seems to be no.17 I offer the following explanation of these limitations. When we are absolutely certain that p is false, we find it difficult to imagine that p.18 Everything else we can imagine. 14   There are of course ways to picture things that suggest the background information. You might picture words on the cake saying “Happy Birthday,” or picture what looks like a window behind Michelle through which you can see William Penn at the top of Philadelphia’s City Hall. Two points. First, that kind of picturing is not necessary. Even if you picture a plain cake, and picture no decorations that say “Happy Birthday!”, you can still be imagining that it is your birthday. Second, you can imagine that the suggestive information is misleading. You can imagine that the cake says “Happy Retirement Xiang Yue!” even if you are imagining that it is your birthday and that your name is Xiang Long. 15   In other work (Kung 2010) I give a more technical treatment of non-pictorial content; I call it assigned content. See also Peacocke (1985) and Ichikawa and Jarvis (2012). 16   I try to make the distinction precise in Kung (2010). 17   See Weatherson (2004) for a helpful overview. 18   Gendler (2000a) notoriously argues that we can imagine the falsity of even absolutely certain propositions, like 5+7=12. I disagree, but as we shall see, even if Gendler is right, that makes the situation worse, not better for non-pictorial content as a guide to possibility.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Thought Experiments in Ethics  235 So if you are certain that female infanticide is morally impermissible, then you will find it difficult to imagine that it is permissible. If you are absolutely certain that the Canadian maple leaf shape is not ovular, you will not be able to imagine that it is ovular. For any proposition that we believe with less than absolute certainty, the fallibilist worry that our belief could be mistaken is live. When we have that worry, we can imagine ways for the proposition to be false. For example, you believe that you have a body, but you are not absolutely certain of it. Hence you can imagine when you read the First Meditation that you do not have a body.19 I propose, then, that one can imagine via background information that p if and only if one is not absolutely certain that p is false. If you are absolutely certain that p is false, then you cannot imagine, or will have difficulty imagining, that p, even via background information. But if you are not absolutely certain that p is false, then you can imagine via background information that p. Of course, in most cases where you are not absolutely certain that p is false, you will also not be absolutely certain that p is true, and hence will be able to imagine via background information that p is false. The modal epistemological consequences of this view are easy to trace. Failing to be less than certain is not a modal epistemic credential. The fact that I am not absolutely certain that not-p, and hence can imagine that p via background information, should not by itself count as evidence that p is metaphysically possible. After all, I might have lots of evidence not just that p is false, but that p is a necessary falsehood. But my evidence might not be so good that I am absolutely certain that p is false. In those cases, I predict that we will be able to imagine via background information that p. For example, I have lots of evidence that Fermat’s Last Theorem is a necessary truth. But I am not absolutely certain. So I can imagine that it turns out to be false: I can imagine opening up the New York Times home page and seeing a headline announcing that mathematicians have discovered a flaw in Wiles’s proof and that the “Theorem” is actually false. As this example suggests, I think it is quite easy to imagine the impossible. Here I diverge from Kripkean orthodoxy. In the a posteriori necessity cases that Kripke discusses, the Kripkean view is that we do not imagine what we take ourselves to imagine. We take ourselves to imagine a situation his theory deems impossible (e.g. water ≠ H2O), but in fact we imagine a genuinely possible situation (the clear, colorless liquid in the lakes and streams ≠ H2O), and we mistake it for an impossible situation. I argue elsewhere that this error theory is unmotivated.20 Let’s pull all the pieces together and recap. Our sensory imaginings have two kinds of content, pictorial and non-pictorial content. One category of non-pictorial content is background information. Our ability to specify background information in our 19   My proposal applies to propositions, and not sentences. While I am absolutely certain that I exist, that is because I know how the sentence “I exist” works: whenever it is thought or uttered by me, it is guaranteed to be true. The content expressed by that sentence or that thought, that Peter Kung exists, is not certain. My proposal predicts that I can imagine that I do not exist because the propositional content that Peter Kung exists is not certain. 20   See Kung (in press). See also Ichikawa and Jarvis (2012, sect. 2).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

236  Peter Kung s­ ensory imaginings explains imagination’s power, but it also leads to modal epistemological problems. The only limit on our ability to imagine background information is absolute certainty—we cannot imagine background information that p when we are absolutely certain that not-p. But the absence of that limitation, being less than absolutely certain that not-p, is not a reason to believe that p is metaphysically possible. So imagining background information is not, by itself, evidence for metaphysical possibility. The conclusion I have argued for here is intentionally quite narrow. As a matter of fact I think the modal skeptical conclusions about background information generalize to all non-pictorial content.21 But here, it will suffice to stick with non-pictorial background information, because, as we will see in the next section, the elements of CTEs in ethics that I want to focus on—forced choices with fixed outcomes—are imagined background information. My narrow conclusion is a modal skepticism about one very specific aspect of imagination that is essential to CTEs in ethics. This narrow modal skepticism naturally leads to questions of whether the modal skeptical conclusions generalize. I’ll address those questions at the end of the chapter, after we analyze CTEs in ethics.

4.  Modal Epistemology and Ethics Thought Experiments It is now time to apply the modal epistemological lessons of Section 2 to CTEs in ethics. It will be easier to work out the implications on a concrete case, so let us revisit the transplant case that opened this chapter. Suppose you are a consequentialist. You think (roughly) that in any circumstances, the morally right thing to do is whatever brings about the best overall consequences. A clever objector devises TRANSPLANT. TRANSPLANT Patient Lamb enters the hospital for a routine checkup. It turns out patient Lamb is a perfect match for five patients in the hospital who will die without transplants. Dr. Wolf has two options. Option 1: kill patient Lamb, harvest Lamb’s organs, and save the five by sacrificing one. Option 2: allow the five patients to die, and leave patient Lamb unmolested. Dr. Wolf chooses option 1. He kills Lamb, harvests Lamb’s organs, and saves five people by sacrificing one. Four fewer people die, so Dr. Wolf has brought about the best overall consequences.22

Many people have the fairly strong intuition that Dr Wolf did not do the morally right thing, and that that makes the case a counterexample to consequentialism. According to the model we are considering, that is because the case points to a 21   In Section 5.1, I will argue that the modal skeptical considerations about background information in sensory imagination apply equally well to non-sensory imagination. 22   From Foot (1967) and Thomson (1976).

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Thought Experiments in Ethics  237 g­ enuine metaphysical possibility where the morally right action is not the one that brings about overall consequences, meaning consequentialism cannot be a necessary truth. But does the case point to a genuine metaphysical possibility? Or, more accurately, does imagining the transplant case provide evidence that such a case is metaphysically possible? It might seem like the answer ought to be a straightforward yes, which is why no one to my knowledge has raised the metaphysical possibility of ethics thought experiments as a concern. However, when we apply the modal epistemological conclusions of Section 2, we will reach a different result. When you imagine the TRANSPLANT case, you picture some of the things in the situation: Doctor Wolf in a white lab coat. Maybe a nurse checking Patient Lamb’s blood pressure while he sits uncomfortably in a blue paper robe. Recalling the discussion from Section 2.2, it is clear that some things you imagine non-pictorially. The identities of Wolf and Lamb are imagined non-pictorially. The fact that Wolf is a doctor and Lamb is the patient is non-pictorial content. So far, nothing that you imagine about the TRANSPLANT case distinguishes it from thought experiments in other subfields of philosophy. Now let’s focus on the ­elements I identified in the opening of the chapter, forced choices with fixed outcomes. It looks like a lot of important facts about the imagined scenario are not pictured; they must come from non-pictorial content. Here are four important facts we imagine about TRANSPLANT: i. Background information: that these two choices are the only morally relevant choices. ii. Background information: that the outcome of option 1 is that the five transplant matches live and only Lamb dies. iii. Background information: that the outcome of option 2 is that Lamb lives, but the five transplant matches die. iv. Background information: that Dr Wolf knows (i)–(iii). Why is it important to imagine that TRANSPLANT includes these facts? The whole enterprise of thought experimentation as it is practiced in contemporary ethics consists in abstracting away from the messiness of the real world and distilling an ethical situation down to a choice between two alternatives with fixed outcomes. In TRANSPLANT, facts (ii) and (iii) specify those alternatives. Fact (i) makes it clear that (ii) and (iii) are the only alternatives to consider. And (iv) makes all of this transparent to the subject in the imagined scenario making the choice. Once the case is presented, consequentialists might naturally question whether Wolf really has brought about the best overall consequences. To forestall objections, we also need to imagine the following additional facts are true in TRANSPLANT: v. Background information: that should Dr Wolf pursue option 1, nobody will ever know about it.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

238  Peter Kung vi. Background information: that should they survive, Lamb and each of the five other patients will lead roughly equivalent lives with respect to consequences. vii. Background information: that there are no further relevant consequences downstream to either choice. viii. Background information: that Dr Wolf knows (v)–(vii) for certain. Fact (v) makes it clear that Wolf ’s actions will not deter potential patients from visiting hospitals. Fact (vi) dictates that all six people are equivalent in terms of their future impact on the world; it isn’t like Lamb will cure cancer or one of the five transplant matches is the next Pol Pot. And again, (viii) makes all of this transparent to the subject in the imagined scenario making the choice. None of (i) through (viii) is pictorial content. All of (i) through (viii) are background information and hence non-pictorial content. So by the arguments of Section 2, imagining the transplant scenario does not give us evidence that (i) to (viii) are metaphysically possible. On reflection, this should not strike us as all that radical a suggestion. Take the knowledge claims in (iv) and (viii), which specify that Dr Wolf knows contingent a posteriori facts for certain. Are we really justified in thinking that kind of knowledge is metaphysically possible? The whole skeptical challenge seems premised on the idea that it is not. At best, our epistemic position seems to be that we are not completely certain that certain knowledge of contingent a posteriori facts is impossible; that is why we can imagine it as background information. But as we saw in Section 2, our less-than-ideal epistemic position regarding the italicized proposition does not seem to constitute evidence that the italicized proposition is genuinely possible. Or take claims (ii) and (iii). Is it really metaphysically possible that an action in this world has one and only one outcome? The scenarios we imagine are not possible worlds; they are not precise enough to pick out a single possible world. They are more like possibility types, compatible with many metaphysically possible worlds. There are many metaphysically possible worlds that fit the setup of TRANSPLANT before Wolf makes a choice. In some of those metaphysically possible worlds, when Wolf harvests Lamb’s organs, Wolf thereby saves all five transplant patients. But this is not guaranteed by the setup we imagine (other than (ii) itself). In other metaphysically possible worlds, Wolf ’s deciding to harvest Lamb’s organs leads to a different outcome. What happens if we remove the knowledge claims (iv) and (viii)? Then, more realistically, Wolf faces a risky situation. Wolf must calculate the risks involved in choosing between options 1 and 2, and part of the calculation includes worrying about things not turning out as expected. How might the risks affect our assessment of TRANSPLANT? Let us concede that one person dying who would not otherwise die is pretty bad. And let’s concede that five people dying of natural causes is also very bad. Let us not take a stand on which situation is worse yet. The way the case is supposed to be imagined, one of these bad things will happen. One, and only one. But if we remove (iv) and (viii) from the imagined scenario, then Wolf must consider the possibility that both patient Lamb dies and all the transplant patients die.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Thought Experiments in Ethics  239 How should consequentialists assess the risk of that situation? One way would be to calculate expected value: for each option, take the sum: the probability of this outcome multiplied by the badness of this outcome. Choose the option with the highest expected value (i.e. the lowest expected badness). That is one way to perform the calculation, but it is not obvious to me that a consequentialist has to say this. Patient Lamb and the five transplant patients all dying is a catastrophic outcome, arguably much worse than either of the other two situations. Another way to assess risk is to weigh losses much more heavily than gains; it is well established that we do this quite naturally anyway.23 A risk-averse consequentialist might weigh outcomes so the better choice is the one least likely to lead to the catastrophic outcome. Since killing one to save five opens up the possibility of killing all six, a risk-averse consequentialist will counsel against doing it. Discussions of risk are familiar from the literature on Rawls’s Theory of Justice and the appropriate risk policy behind the veil of ignorance; it is a vexed and open question how open to risk one “ought” to be.24 Risk has also been discussed in the medical ethics literature, of course, and there is some discussion of risk in the mainstream ethics literature as well.25 However, if I am right, risk has to be discussed in foundational theorizing. We cannot first use counterexamples featuring forced choices with fixed outcomes to eliminate theories (like consequentialism) and then, with the foundational theory issues more settled, turn to risk. Our ethical theories have to factor in risk right from the start. The foregoing analysis has focused on whether the agent in the imagined scenario, Wolf, makes the correct moral decision. But the points I make above do not depend on assessing Wolf ’s decision-making. Suppose again that we remove (iv) and (viii) from the scenario. Imagine Wolf just acts rashly, without much evidence that (ii) and (v) are true. Imagine further that Wolf has chosen option 1, killing patient Lamb and harvesting all Lamb’s organs. And imagine that things pan out and Wolf manages to save all five transplant patients. We can agree that because Wolf acted rashly, Wolf did not do the subject­ ively right thing. But we can still ask whether Wolf did the objectively right thing. The answer is the same as in our original case. In order for theorists to evaluate Wolf ’s action, we need to know certain counterfactuals. We need to know what would have happened had Wolf acted differently. Had Wolf not killed Lamb, would the five transplant patients have died? As the case is set up, we are to imagine that counterfactual, had Wolf not harvested Lamb’s organs, the five transplant patients would have died. The truth of that counterfactual in the imagined scenario isn’t pictured. It is non-pictorial background information content. So by the argument of Section 2, we have no reason to think that that counterfactual is metaphysically possible. This means that we theorists   e.g. see Horowitz (1998).   See Freeman (2008), esp. sect. 6.1 for a summary. 25   See for example Zimmerman (2008). In Zimmerman’s view an agent ought to do the “prospectively best” option (p. 56), which is cashed out as maximizing expected value given evidence. But Zimmerman does note that “some prospects are so momentously bad or good in terms of their evident value that no matter how low the probability of attainment, they overwhelm all other considerations” (p. 52). 23 24

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

240  Peter Kung must consider additional possibilities, like the possibility that had Wolf not killed Lamb, the five patients would have survived anyway; that is, we need to consider the possibility that Wolf killed Lamb unnecessarily. This seems like a fairly catastrophic possibility. As above, it seems an open question how consequentialism ought to handle this kind of risk.

5. Scope Let me now take up some issues about the scope of my claims that remain open. Given space considerations, the remarks in this section are tentative and have to remain somewhat schematic.

5.1  Other sources of modal evidence To reach the modal skeptical conclusion about the metaphysical possibility of forced choices with fixed outcomes that I defend above, I took sensory imagination as my starting point. I argued that certain ways of using the sensory imagination do not provide evidence for metaphysical possibility. A natural question is whether there are other ways to justify the metaphysical possibility of these situations involving forced choices with fixed outcomes. Some philosophers claim that non-sensory imagination provides evidence for metaphysical possibility. Frequently non-sensory imagination is characterized simply as imagination that is not grounded in imagery.26 A consequence of the account I sketched in Section 2 is that non-sensory imagination is imagining background information with no picturing. Note that non-sensory imagining is still distinct from supposition; non-sensory imagining is subject to the limitations on imagination, like imaginative resistance, that I described in Section 2.3, while supposition is not. Of course, if non-sensory imagining is just imagining background information, then all the skeptical considerations from Section 2 still apply. Our ability to imagine background information is almost completely unconstrained. As I said above, unless we are absolutely certain that not-p, we can imagine p to be true as part of the background information. But, to reiterate, failing to be absolutely certain is not a modal epistemic credential. So on this account, non-sensory imagining (aka imagining background information) provides no modal evidence. What about another putative source of modal evidence that philosophers frequently mention, conceivability? On many accounts, conceiving is a type of imagining; those accounts would be subject to the modal skeptical challenges I sketched in Section 2. If conceiving is supposed to be distinct from imagining, then we need an account of conceiving, and in particular how it is distinct from non-sensory imagining.27 While I do not have the space to survey all accounts of conceivability in the literature here, let me   See Chalmers (2002), Walton (1990), and Yablo (1993).   See Magdalena Balcerak Jackson’s helpful contribution to this volume (Chapter 1).

26

27

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Thought Experiments in Ethics  241 point to one common deficit. Many accounts of conceivability fail to offer practical advice on how to distinguish good cases of conceiving from bad cases, meaning they are not equipped to provide useful epistemological guidance on the forced choice/ fixed outcome issue. This is a very familiar problem; Descartes claims that clear and distinct perception is infallible, but he has little to say about how to distinguish true clear and distinct perception from false perceptions that merely appear clear and distinct. For a more recent example, Menzies’s (1998) response-dependent account of conceiving, which proposes a biconditional linking metaphysical possibility and what an ideal conceiver can conceive, offers no practical recipe for determining when we are conceiving like an ideal conceiver and when we are making a mistake.28 I conclude, then, that ethicists ought to take seriously the question of whether we know that forced choices with fixed outcomes are possible. The challenge raised here needs to be addressed, and it is not at all clear that there is a simple way to address it.

5.2  Modal skepticism? A second natural question is whether the skeptical considerations about imagination’s background content in Section 3.3 lead to a pervasive modal skepticism.29 I think the answer is no, but I concede that responding effectively to modal skepticism requires separate paper-length treatment. But let me give some indication why I think we need not be modal skeptics. When we literally picture a situation, we display or stage for ourselves a way that it could be or happen. Right now I am looking at the graphic from Wikipedia’s “Missing shade of blue” page.30 I close my eyes and visually imagine the missing blue rectangle. The fact that I can actually picture a rectangle colored in the missing shade of blue suggests that there could be—i.e. that it is metaphysically possible for there to be—a rectangle of just that shade. By composing a novel tune in my mind’s ear, I give myself evidence that such a sequence of sounds could really exist. These kinds of cases suggest that picturing does at least some of the time provide defeasible evidence for possibility. This is not a conclusive argument—nobody to my knowledge has supplied a conclusive argument—but it does suggest that pictorial content might be a good place to ground a modal epistemology. And it does suggest that we need not submit to a thoroughgoing modal skepticism. I think that pictorial content can serve as the foundation for a fairly powerful modal epistemology, one that allows us to justify belief in the possibility of, say, Michelle Obama coming to your birthday party. But as I already noted, that is a project for another occasion.31 28   Bealer (2002), Chalmers (2002), and Ichikawa and Jarvis (2012) offer (very different) accounts where idealization plays a key role. 29   Other arguments for modal skepticism can be found in Byrne (2007), Fiocco (2007), Tidman (1994), and van Inwagen (1998). 30  . 31   See Kung (2010). In brief, the idea is that for many of the elements imagined via non-pictorial content, like the identity of Michelle Obama, we have independent reason to think that those contents are metaphysically possible. In Michelle Obama’s case, we know that Michelle Obama is metaphysically possible

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

242  Peter Kung

5.3  Other uses of thought experiments The problem I raise here applies specifically to CTEs in ethics. I do not claim that the sole purpose of thought experiments is to provide counterexamples; I claim only that it is an important purpose. Walsh (2011, sect. 3) identifies four roles for thought experiments, the first of which is to serve as counterexamples. The others are (2) to serve as intuition pumps, (3) to serve as clarificatory devices, or (4) to allow us to reframe or reimagine a debate. While technically I leave open the possibility that thought experiments in ethics featuring forced choices with fixed outcomes can serve other purposes, I do think it is worth questioning how well thought experiments can serve other purposes if they cannot be counterexamples. Let me close with some tentative remarks about the role of counterexamples. The points here need further elaboration to be more than merely suggestive. I hope to provide that further elaboration in other work. Recall our rough first pass at the method of thought experiments. Proponent ­proposes an ethical theory. Opponent replies with a thought experiment—a counterexample—eliciting intuitions that conflict with what Proponent’s theory predicts. Proponent’s ethical theory is necessarily true if true. Opponent’s thought experiment points to a possibility, a metaphysically possible situation, in which Proponent’s ethical theory delivers the wrong result. Hence Proponent’s ethical theory cannot be a necessary truth. It seems clear (to me anyway) that Proponent should not be particularly worried by Opponent’s thought experiment if there is reason to think that it’s impossible. Why worry that your theory cannot account for a situation that (you are justifiably confident) cannot ever possibly arise? Should Proponent worry if Opponent’s counterexamples lack modal evidence? That is, suppose we have no reason to think the situation presented by the thought experiment is possible. But we have no reason to think the thought experiment is impossible either. We simply lack evidence. Is lack of evidence for the impossibility of a counter­ example enough? This is a question about burden of proof. I’m inclined to answer no, it’s not enough. The reason: to be a counterexample to a necessary truth, the situation must be (metaphysically) possible. It’s unclear that a thought experiment that falls short of disproving a necessary truth does any work. Opponent has not yet made the case that Proponent’s theory has a problem. Proponent’s theory remains untarnished. Hence I’m inclined to believe that we must have evidence that Opponent’s thought experiment presents a metaphysically possible situation. because she is actual. Because we know that fact, we can use her in a scenario that we picture. Contrast the case of Michelle Obama with Sherlock Holmes. Imagine Sherlock Holmes and Michelle Obama talking. With Michelle Obama, we have independent reason to think that she could exist. Not so with Sherlock Holmes. The mere fact that we can imagine Sherlock Holmes does nothing to answer Kripke’s (2011) worries about the metaphysical possibility of fictional characters.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Thought Experiments in Ethics  243 Sometimes people claim that it’s enough that scenarios are conceptually possible.32 Conceptual possibility, even without evidence for metaphysical possibility, is enough to raise a problem for Proponent’s theory. A thought experiment presenting a conceptually possible situation in which the right thing to do is not what brings about the best overall consequences suffices, even absent evidence that the situation is metaphysically possible, to show that there’s a problem—a conceptual problem—with Proponent’s theory. This puzzles me, at least in this context. Here is a gloss on conceptual possibility. Suppose we’ve got a concept of X and we’re curious what being X entails. Our concept of X might leave it open whether X is F. That can certainly happen when the proposition is a posteriori. However when is agreed on all sides to be an a priori matter, I’m less sure whether we can make any sense of the claim that our concept of X leaves it open whether X is F while, at the same time, the conceptual possibility of X being F is nonetheless a significant result as alleged in the previous paragraph. There are two situations to consider: first, that we have definitively concluded that X being F entails no contradiction, and second, that we are just so far unable to see any contradiction in X being F. In the first situation, we would be in a position to conclude that X being F is metaphysically possible after all, making the appeal to conceptual possibility in the previous paragraph gratuitous. So I take it the objection of the previous paragraph considers us to be in the second situation. This makes sense; often we resort to thought experiments when we are not in a position to draw out all the entailments of our concepts. I have pointed out that our ability to imagine the forced choice/fixed outcome conceptual possibility is driven by our less-than-ideal epistemic situation, raising the concern that we simply have not reflected carefully enough on whether this situation is in fact coherent. Consider the issue briefly now: How can there be fixed outcomes? Imagine the setup of TRANSPLANT any way you like such that outcomes are supposedly fixed: for example, an all-powerful deity guarantees that if Wolf kills Lamb, Lamb’s organs will save the five transplant patients; Wolf goes ahead and kills Lamb.33 Despite that setup, and despite the presence of the putative guarantee, we can imagine the scenario unfolding in a number of ways: Wolf harvests Lamb’s organs and saves the five transplant patients; Wolf harvests Lamb’s organs, the deity changes its mind and decides not to intervene, and four of the transplant patients die on the operating table; and so on. Fixed outcomes are best understood as a feature of a set of worlds, not a feature of any particular world. Fixed outcomes say: take any world where the setup facts S are true; in that world, the outcome facts O will be true as well. The problem is that it seems conceptually possible for S to be paired with outcomes other than O. It is not at all clear to me that in the kinds of cases that arise in the ethics   See Leon and Tognazzini (2010), who are addressing Frankfurt-style cases in the free will literature.   Remember we are explicitly setting aside metaphysical possibility, and concerning ourselves with conceptual possibility. So we need not worry about whether all-powerful deities are metaphysically possible. 32 33

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

244  Peter Kung l­ iterature forced choices with fixed outcomes are coherent conceptual possibilities. At the very least, I hope the remarks here prompt further reflection on the matter. These remarks about conceptual possibility make me question whether a thought experiment can play Walsh’s second role, which is to serve as an intuition pump, if it does not also play the first role. (Here I take no stand on Walsh’s third and fourth categories, thought experiments as clarificatory devices and thought experiments as tools to reframe debate.) As Walsh defines them, intuition pumps use single thought experiments to establish a very general conclusion. His example is using the trolley case to establish that numbers do not count morally. But if the thought experiment cannot serve as a counterexample, and hence cannot show there is a problem in even a particular case, how can it establish a general case? If the trolley problem fails to establish that numbers do not matter in the trolley case, how can it establish that numbers do not matter at all? I conclude that the metaphysical possibility of forced choices with fixed outcomes is a serious and live problem for the methodology of contemporary ethics.

Acknowledgements Thank you to Amy Kind, Neil Mehta, Masahiro Yamada, Yuval Avnur; audiences at Cal State San  Bernardino, University of Nevada–Las Vegas, and the National University of Singapore; students in my 2012 Knowledge Through Imagination seminar, particularly Brendan ­ Schneiderman; and attendees at the 2012 Knowledge Through Imagination conference at Claremont McKenna and Pomona Colleges.

References Bealer, George (2002). Modal epistemology and the rationalist renaissance. In Tamar Szabό Gendler and John Hawthorne (eds), Conceivability and Possibility. Oxford: Clarendon Press, pp. 71–125. Byrne, Alex (2007). Possibility and imagination. Philosophical Perspectives 21(1): 125–44. Cain, James (2003). Frankfurt style examples. Southwest Philosophy Review 19(1): 221–9. Chalmers, David J. (2002). Does conceivability entail possibility? In Tamar Szabό Gendler and John Hawthorne (eds), Conceivability and Possibility. Oxford: Clarendon Press, pp. 145–200. Elster, Jakob (2011). How outlandish can imaginary cases be? Journal of Applied Philosophy 28(3): 241–58. Fiocco, M. Oreste (2007). Conceivability, imagination and modal knowledge. Philosophy and Phenomenological Research 74(2): 364–80. Foot, Philippa (1967). The problem of abortion and the doctrine of double effect. Oxford Review 5: 5–15. Freeman, Samuel (2012). Original position. In Edward N. Zalta (ed.), The Stanford Encyclopedia of Philosophy (Spring 2012 edn). Geirsson, Heimir (2005). Conceivability and defeasible modal justification. Philosophical Studies 122(3): 279–304.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Thought Experiments in Ethics  245 Gendler, Tamar Szabό (1998). Galileo and the indispensability of scientific thought experiment. British Journal of Philosophy of Science 49(3): 397–424. Gendler, Tamar Szabό (2000a). The puzzle of imaginative resistance. Journal of Philosophy 97(2): 55–81. Gendler, Tamar Szabό (2000b). Thought Experiment: On the Powers and Limits of Imaginary Cases. London: Routledge. Graham, Peter J. (2007). The theoretical diagnosis of skepticism. Synthese 158(1): 19–39. Hart, W. D. (1988). The Engines of the Soul. Cambridge: Cambridge University Press. Hill, Christopher S. (1997). Imaginability, conceivability, possibility and the mind-body problem. Philosophical Studies 87(1): 61–85. Horowitz, Tamara (1998). Philosophical intuitions and psychological theory. Ethics 108(2): 367–85. Hume, David (1978). A Treatise of Human Nature (ed. L. A. Selby-Bigge and P. H. Nidditch, 2nd edn). Oxford: Clarendon Press. (Original work published 1739–40.) Ichikawa, Jonathan, and Benjamin Jarvis (2012). Rational imagination and modal knowledge. Noûs 46(1): 127–58. Kind, Amy (2001). Putting the image back in imagination. Philosophy and Phenomenological Research 62(1): 85–109. Kripke, Saul A. (2011). Vacuous names and fictional entities. In Philosophical Troubles: Collected Papers, Volume I. New York: Oxford University Press, pp. 52–74. Kung, Peter (2010). Imagining as a guide to possibility. Philosophy and Phenomenological Research 81(3): 620–63. Kung, Peter (in press). You really do imagine it: Against error theories of imagination. Noûs. doi: 10.1111/nous.12060 Leon, Felipe, and Neal A. Tognazzini (2010). Why Frankfurt-examples don’t need to succeed to succeed. Philosophy and Phenomenological Research 80(3): 551–65. McGinn, Colin (2004). Mindsight: Image, Dream, Meaning. Cambridge, MA: Harvard University Press. Menzies, Peter (1998). Possibility and conceivability: A response-dependent account of their connections. European Review of Philosophy 3: 255–77. Noordhof, Paul (2002). Imagining objects and imagining experiences. Mind and Language 17(4): 426–55. Peacocke, Christopher (1985). Imagination, experience, and possibility: A Berkeleian view defended. In John Foster and Howard Robinson (eds), Essays on Berkeley: A Tercentennial Celebration. Oxford: Clarendon Press, pp. 19–35. Pryor, James (2000). The skeptic and the dogmatist. Noûs 34(4): 517–49. Rawls, John (1999). A Theory of Justice (rev. edn). Cambridge, MA: Harvard University Press. Taliaferro, Charles (1986). A modal argument for dualism. Southern Journal of Philosophy 24(1): 95–108. Thomson, Judith Jarvis (1971). A defense of abortion. Philosophy and Public Affairs 1(1): 47–66. Thomson, Judith Jarvis (1976). Killing, letting die, and the trolley problem. The Monist 59(2): 204–17. Tidman, Paul (1994). Conceivability as a test for possibility. American Philosophical Quarterly 31(4): 297–309.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

246  Peter Kung van Inwagen, Peter (1998). Modal epistemology. Philosophical Studies 92(1): 67–84. Walsh, Adrian (2011). A moderate defence of the use of thought experiments in applied ethics. Ethical Theory and Moral Practice 14(4): 467–81. Walton, Kendall L. (1990). Mimesis as Make-Believe: On the Foundations of the Representational Arts. Cambridge, MA: Harvard University Press. Weatherson, Brian (2004). Morality, fiction, and possibility. Philosophers Imprint 4(3): 1–27. Wilkes, Kathleen V. (1988). Real People: Personal Identity without Thought Experiments. Oxford: Oxford University Press. Williams, Bernard (1973a). Imagination and the self. In Problems of the Self. Cambridge: Cambridge University Press, pp. 26–45. Williams, Bernard (1973b). A critique of utilitarianism. In J. J. C. Smart and Bernard Williams, Utilitarianism: For and Against. Cambridge: Cambridge University Press, pp. 77–150. Wittgenstein, Ludwig (1958). The Blue and Brown Books. Oxford: Blackwell. Yablo, Stephen (1993). Is conceivability a guide to possibility? Philosophy and Phenomenological Research 53(1): 1–42. Yablo, Stephen (2002). Coulda, woulda, shoulda. In Tamar Szabό Gendler and John Hawthorne (eds), Conceivability and Possibility. Oxford: Clarendon Press, pp. 441–92. Zimmerman, Michael J. (2008). Living with Uncertainty: The Moral Significance of Ignorance. Cambridge: Cambridge University Press.

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

Index abduction  28, 113, 119, 122, 219, 221–2 acceptance  24, 43, 51–4 agency  26, 85–7, 90, 93, 95, 107 alternate perspectives  165–7, 169, 176 a posteriori necessities  17–18, 135, 235, 238, 243 arbitrary names  120 attention  1, 15, 56, 89, 90, 103, 115, 172, 175, 178, 193, 201, 213, 215, 221 attitude imagining, see imagination, propositional automatic nervous system  95 Bealer, George  16, 124, 127, 130, 133–4, 241 belief belief and acceptance  51–4 belief and certainty  234–6 belief in possibility  16–19, 22, 24, 51, 54, 124–5, 141, 147–8, 149 n8, 241; see also metaphysical possibility belief ’s influence on imagination  32, 79, 93–4, 98, 104, 160 belief-formation  126, 220 conditional belief  117–18 conflicting beliefs  177–80 imagination as belief-like  3, 15–16, 20, 65, 211 n7 imagination justifying contingent beliefs  146–7, 156, 208, 219–20, 223 imagination as simulated belief  20, 24, 49, 53–4, 212, 216 imagining as mirroring belief ’s inferential structure  25–6, 68–71, 73–5, 81 imaginings are beliefs  25, 68–9 perceptual belief  163–4 perspective of the subject of rational belief 56–8 predicting another’s beliefs  185–6, 198, 212 unjustified beliefs, see skepticism Bush administration  97–8 Campbell, John W.  29–30, 149–52, 154–5, 158 Chalmers, David  4, 16, 19, 42, 54–7, 61, 124, 148, 231, 240–1 change constraint, see constraints, change cognitive penetration  168, 172 collapse  33, 56, 215–18, 221 conceiving  7–10, 16, 18, 23–4, 41–5, 52 n30, 54–9, 124–5, 147 n5, 154, 240–1 conditionals  25, 28–9, 43, 63, 68, 81, 116–19, 122, 135–41, 223, 241

consequentialism  34, 52, 68, 91–3, 98, 118–19, 228, 236–7, 239–40 constraints  on belief  56–7 change constraint  29, 151–3, 155–7 on imagination  2–3, 13, 18–19, 21–7, 29, 32, 61, 63, 67–75, 77–82, 94 n18, 146, 151–7 reality constraint  27, 29, 94, 114, 116, 151–3, 155–7, 223; see also reality congruence constructive imagination, see imagination, constructive context of justification  44, 115, 218, 222 contextualism  32, 79, 139, 196–7 counterexamples  20, 34, 148, 227–31, 236, 239, 242, 244 counterfactuals  28–9, 32, 75, 103, 124, 135–41, 174, 185, 188, 191–2, 198, 200, 239 Currie, Gregory  14–15, 20, 23, 43, 49, 53, 61, 67–9, 211 Damasio, Antonio  97–9, 106, 198–200 Davies, Martin  185, 195, 198, 211, 213 daydream  1, 5, 14, 32, 44, 79, 156, 208, 210–12 decision-making  1, 19–20, 22, 32, 34, 42, 49, 75–6, 97, 99, 117, 132, 145–6, 185–6, 188–9, 195–202, 239 deduction  104, 119–20, 122, 178, 212, 219, 221–2 deliberate imagination, see imagination, deliberate Descartes, René  4, 6–8, 10–11, 16, 21, 175, 178–9, 229, 241 discovery  44, 115, 122, 218, 222 Einstein, Albert  147 Ekman, Paul  89, 95–6 empathy  26, 31, 85, 99–102, 179, 185–6, 191, 194, 200–1 epistemology of perception  164 equivocation  2–5, 21, 23 ethics  33–4, 102, 187, 227–8, 230–1, 234, 236–7, 239, 241–4 evolution  95, 106, 129–30, 172, 175 and the purpose of imagination  27–9, 33, 113–14, 123, 223–4 and quotidian modals  132, 135–6, 141–2 expectation  31, 63, 76, 81, 119, 173

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

248  index fantasy  1, 5, 11, 14–15, 21, 79, 94, 151, 158 fiction  1, 13–15, 18, 20, 23, 72, 85–8, 91, 96, 97  n23, 103–6, 111, 123, 149, 151, 187, 192 n6, 207–8 n1, 210–11, 233 n13, 241–2 n31 fixed outcomes  33–4, 227–8, 236–7, 239–44 Foot, Philippa  227, 230, 236 forced choices  33–4, 227–8, 236–7, 239, 241–2, 244 forecast 185–201 forward model  63, 70–1, 73–4, 77 freedom of imagination, see imagination, freedom of free play  13, 22 Gendler, Tamar  4, 16, 23, 43, 47–8, 90, 97–9, 103, 130, 147–8, 207–8, 212, 223, 227, 231, 234 Goldman, Alvin  4, 49, 61, 185–6, 195, 210–11, 213, 217 Grandin, Temple  29, 154–7 Hill, Christopher  18, 124, 136–7, 208, 231 Hume, David  4, 6, 8–13, 16–18, 21–2, 43, 93, 104, 146, 231 hypothetical reasoning  27, 43–4, 58, 117, 123, 221, 227 I-C-E-C  26, 85–6, 90–106 iconic memory  171, 175 illusion, see imagination, illusion imagery  4, 6, 18, 22–3, 26, 34, 46, 64, 66, 77–9, 81, 86 n2, 87–8, 90–5, 98–9, 102–3, 106, 117, 121, 147, 160, 192, 194, 232, 240; see also imagination, imagistic; imagination, sensory imagination attitude imagination, see imagination, propositional constructive imagination  86–7, 94 n18, 97–8, 107 deliberate imagination  32, 175, 207–13, 215, 218–23; see also imagination, voluntary freedom of imagination  18, 25, 62, 66, 68, 76, 82, 146 imagination box  68 n9, 73–4 imaginative illusion  125, 157–8 imaginative resistance  23–4, 47–8, 79, 86, 88, 103, 234, 240 imagining from the inside  4–5, 20, 48 imagining from the outside  4–5, 20 imagistic imagination  4–6, 8, 10, 18, 26, 34, 64, 70, 78, 85–8, 90–6, 98–102, 104, 106, 147; see also imagery; imagination, sensory instructive use of imagination  1–5, 8–23 involuntary imagination  22, 27, 115–17, 120, 123, 168, 223 non-human imagination  113, 233

non-pictorial imagination  4–5, 34, 228, 231–9, 241 n31 productive imagination  11–13, 22, 171, 175 propositional imagination  25, 32, 52, 64–73, 78–81, 86–7, 93–5, 98, 207, 213, 235 recollective imagining  11, 91–2, 171, 174, 177 n34 recreative imagination  23–4, 49–51, 54, 58, 207 reproductive imagination  11, 13, 22, 171, 175 sensory imagination  16, 25, 32, 49, 64–7, 69–71, 73–4, 77, 80–1, 92, 207, 209–13, 215 n13, 222, 231–2, 234–6, 240; see also imagery; imagination, imagistic spontaneous imagination  32–3, 207–10, 213, 215–16, 218–19, 221–3 transcendent use of imagination  1–5, 6–12, 14–18, 20–1, 25, 27 voluntary imagination  19, 21, 25–7, 32, 43–5, 50, 52, 58, 99, 115–18, 123, 145, 168, 209–10, 213, 223; see also imagination, deliberate immediacy  48, 69, 161, 163–4, 167, 170, 179 n37, 180, 199–200, 232 implicit memory  194, 201–2 induction  119–20, 122, 178, 219, 221–2 infusion effect  30–1, 167, 171, 173, 175–6, 193, 198 intention  14, 25, 61–7, 70–1, 74–81, 91 n12, 115, 161, 173 n27, 175, 186, 215–17 introspective evidence  76, 167–9, 170, 231 intuition  242, 244 Kantian intuition  11–13 linguistic intuition  134 modal intuition  16 n20, 55, 122, 124–7, 242, 244 sensory intuition  178, 223 social intuitionist model of moral appraisal 100 involuntary imagination, see imagination, involuntary Jackson, Frank  17, 20 justification for belief in possibility  16, 24, 51, 54, 58, 66, 124–6, 147; see also belief, in possibility; knowledge, of possibility; modal epistemology context of discovery vs. context of justification  44, 115, 122, 218, 222 for contingent beliefs, see belief, imagination justifying contingent beliefs perceptual justification  163–4, 170 skepticism about imagination’s ability to supply  36, 44–5, 57, 174, 219–21, 238; see also skepticism about knowledge through imagination

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

index  249 Kant, Immanuel  4, 6, 11–13, 22, 30, 171, 173 knowledge aesthetic knowledge  13, 22, 80 background knowledge’s influence on imagination  32–3, 62, 73, 94, 98, 200, 215–18, 221–4; see also belief ’s influence on imagination conditional knowledge  116–17, 119, 137, 239 moral knowledge  26 of necessity  28, 54, 66, 124–5, 127–8, 130–1, 133, 136–8, 140 of other people  20, 31–2, 160–1, 174–6, 178–9, 185, 208–9, 211–12, 215–18, 221; see also mindreading perceptual knowledge  22, 161 n2, 176 of possibility  10–11, 17, 28, 34, 54, 124–5, 127–30, 135, 137, 141; see also belief in possibility and modal epistemology from rational insight  6–7 self-knowledge 185–6 of sensorimotor regularities and contingencies  70, 81, 165 see also fiction; possibility; quotidian modals; skepticism about knowledge through imagination Kripke, Saul  17–18, 22, 125, 130, 134, 141, 235, 242 LeDoux, Joseph  88 Lewis, David  127–8, 136–7 make-believe  1, 13–15, 81, 107 mathematics  41, 86, 122, 149, 179, 187, 212 mental imagery, see imagery metaphysical modality  28–9, 122, 124–6, 129–37, 141–2 mindreading  4, 13, 17, 19–20, 22, 31–3, 61, 207–8, 211–14, 216–19, 221; see also knowledge, of other people modal epistemology  10, 13, 16–18, 22, 124–32, 148, 207, 231, 234–7, 241; see also belief in possibility; knowledge, of necessity; knowledge, of possibility modality  29, 90, 124–6, 128–37, 141–2, 165–7, 223, 231–2 moral risk  34, 228, 238–40 motor control  69 necessity, see a posteriori necessities and knowledge of necessity Nichols, Shaun  61, 67–9, 73, 81, 142 n28, 185 Noë, Alva  12, 164–6, 168 non-human imagination, see imagination, non-human non-pictorial imagination, see imagination, non-pictorial Norton, John  222–3

objectivity  12–13, 128–9, 142, 161, 164–7, 170, 239 offline processes  22, 27–8, 31, 50, 118–20, 185–6, 193–6, 198–9 online processes  27–8, 118–19, 202, 208 paradox of fiction  86, 103, 105 P-C-E-C  26, 90, 93, 95, 99, 102, 106 perspective-taking  23–4, 45, 47, 50, 54, 56 phenomenology  of belief  56 of imagination  46, 48, 56, 207 of imaginative resistance  103, 104 n29 of perception  30, 161–7, 170–4 play  1, 13–14, 22, 103, 115, 123, 216; see also make-believe possibility conceptual possibility  80, 133, 243–4 epistemic possibility  130–1, 135 logical possibility  80, 133–4 metaphysical possibility  10–11, 16–19, 22, 24, 28–9, 34, 50 n27, 51, 54, 57–8, 63, 66, 86, 124–5, 127–31, 133–5, 136, 138, 140–1, 147–9, 167, 174, 228–31, 234, 236–44 physical possibility  131, 133–4, 140 rational possibility  135 realistic vs. unrealistic possibilities  27, 33, 97–8, 107, 114–15, 216, 219–22, 224 prediction  of future sensory input  25, 69–71, 75 via imagination  1, 4, 19, 31, 121, 160 n1, 174–5, 185–98, 208, 211–14 predictive coding  169, 172 pretense  13, 61, 64, 67, 72–3, 75, 80–1, 207–8 productive imagination, see imagination, productive propositional imagination, see imagination, propositional quotidian modals  28, 124, 129–30, 132–3, 135–8, 140–2, 223 Ramsey, Frank  118 Ravenscroft, Ian  20, 23, 43, 49, 53, 61, 67–9, 186, 211 reality congruence  21, 26, 86, 93–6, 99, 102, 104–5; see also constraint, reality reality constraint, see constraint, reality recollective imagining, see imagination, recollective recreative imagination, see imagination, recreative reductio reasoning  24, 41, 43, 53, 146, 212 Reichenbach, Hans  115 reproductive imagination, see imagination, reproductive

OUP CORRECTED PROOF – FINAL, 01/11/2016, SPi

250  index Sartre  45, 145 science fiction  29, 149, 151 Script Elaborator  73–6, 81 Searle, John  20, 62, 66 sensory imagination, see imagination, sensory simulation  emotional response to  98 in problem solving  155 in Simulation Theory of mindreading  19–20, 22, 24, 31–3, 47 n14, 49–50, 160–1, 166, 174, 178, 185–6, 191–6, 197–200, 211–19, 221; see also mindreading of subject of rational belief  24, 56–8; see also belief, perspective of the subject of rational belief supposing as simulation of belief  49, 51, 53; see also supposing skepticism about knowledge through imagination  4 n7, 8, 11, 17, 20 n27, 27–8, 31–4, 44–5, 119, 128, 174, 208 n3, 215, 222, 236, 240–1 social emotions  26, 100–1 Spinoza, Baruch  8 spontaneous imagination, see imagination, spontaneous Stalnaker, Robert  24, 48, 52–3 Stich, Stephen  61, 67, 69, 73, 81, 175, 185 Stone, Tony  221 supposing  4, 10, 17, 23–4, 43–59, 98 n25, 120–1, 123, 147, 211–12, 217, 219, 223, 240

tableau method  120–1 Tesla, Nikola  29, 154–7 testimony 118 theorizing  29, 34, 141, 211–12, 217, 227–8, 239 thought experiments  16–17, 20, 28, 32–4, 42, 54–5, 122, 208, 221–2, 227–31, 237, 242–4 transcendent use of imagination, see imagination, transcendent use transplant case  33, 227–8, 236–7 updating  28, 118 visceral responses  32, 90, 188, 193–4, 197–9, 201–2 voluntary imagination, see imagination, voluntary Walton, Kendall  3–4, 14–16, 48, 61, 81, 147, 207–10, 233, 240 Williamson, Timothy  27–9, 33, 50, 87, 93, 124, 135–8, 140–1, 198, 223–4 Wittgenstein  18, 45, 162, 168, 175–6, 233 Yablo, Stephen  4, 16, 19, 50, 61, 124, 163, 231, 234, 240 zombies  16, 41–2, 44, 54–8, 72