Minds, Ethics, and Conditionals: Themes from the Philosophy of Frank Jackson

Minds, Ethics, and Conditionals

This page intentionally left blank

Minds, Ethics, and Conditionals Themes from the Philosophy of Frank Jackson

Ian Ravenscroft

C L A RE N D O N P RE S S · OX F O RD

1

Great Clarendon Street, Oxford ox2 6dp Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide in Oxford New York Auckland Cape Town Dar es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi New Delhi Shanghai Taipei Toronto With offices in Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan Poland Portugal Singapore South Korea Switzerland Thailand Turkey Ukraine Vietnam Oxford is a registered trademark of Oxford University Press in the UK and in certain other countries Published in the United States by Oxford University Press Inc., New York  Ian Ravenscroft 2008 The moral rights of the authors have been asserted Database right Oxford University Press (maker) First published 2008 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this book in any other binding or cover and you must impose the same condition on any acquirer British Library Cataloguing in Publication Data Data available Library of Congress Cataloging in Publication Data Data available Typeset by Laserwords Private Limited, Chennai, India Printed in Great Britain on acid-free paper by Biddles Ltd., King’s Lynn, Norfolk ISBN 978–0–19–926798–9 10 9 8 7 6 5 4 3 2 1

Acknowledgements Thanks are due to Michael Smith for initially suggesting that I edit a collection of papers on Frank Jackson, and for his subsequent advice and encouragement. Thanks are also due to Peter Momtchiloff at Oxford University Press for his support, to Flinders University for giving me a teaching release grant whilst the collection was finalized, and to Tamara Zutlevics for a number of useful discussions. I would also like to thank all the contributors for entrusting their papers to me and giving me their patience. Finally, I would like to thank Frank Jackson for agreeing to be the ‘‘target’’, and for teaching me much philosophy.

This page intentionally left blank

Contents Introduction: Themes and Criticisms Ian Ravenscroft

Part 1: Metaphysics and Conceptual Analysis

1

21

1. Analysis, Description, and the A Priori? Simon Blackburn

23

2. Physicalism, Conceptual Analysis, and Acts of Faith Jennifer Hornsby

43

3. Serious Metaphysics: Frank Jackson’s Defense of Conceptual Analysis William G. Lycan

61

4. Jackson’s Classical Model of Meaning Laura Schroeter and John Bigelow

85

5. The Semantic Foundations of Metaphysics Huw Price

111

6. The Folk Theory of Colours and the Causes of Colour Experience Peter Menzies

Part 2: The Knowledge Argument 7. Consciousness and the Frustrations of Physicalism Philip Pettit 8. Jackson’s Change of Mind: Representationalism, a Priorism and the Knowledge Argument Robert Van Gulick

Part 3: Ethics 9. Analytical Moral Functionalism Meets Moral Twin Earth Terry Horgan and Mark Timmons

141

161 163

189

219 221

viii

contents

10. Consequentialism and the Nearest and Dearest Objection Michael Smith

237

11. The ‘Actual’ in Actualism Julia Driver

267

Part 4: Conditionals and the Purposes of Arguing

281

12. Conditionals, Truth, and Assertion Dorothy Edgington

283

13. Conditionals: A Debate with Jackson Graham Priest

311

14. Two Purposes of Arguing and Two Epistemic Projects Martin Davies

337

Part 5: Replying to My Critics

385

15. Replies to My Critics Frank Jackson

387

Index

475

Introduction: Themes and Criticisms IAN RAVENSCROFT

Frank Jackson’s philosophical interests are wide, ranging across philosophical logic, the philosophy of language, metaphysics, the philosophy of mind and ethics. Much but not all of his work develops and applies an approach to metaphysical issues built on the twin foundations of supervenience and conceptual analysis. The range and depth of Jackson’s thought presents a challenge to anyone who sets out to edit a collection such as this. To some extent I have had to sacrifice width in order to do justice to Jackson’s depth. Thus in the pages that follow you will not find discussion of Jackson’s (1977) development of the representative theory of perception, but you will find a number of papers exploring Jackson’s reliance on conceptual analysis. The papers in Part 1 of this collection examine from various angles both Jackson’s general approach to metaphysics and the application of that approach to the philosophy of mind and the philosophy of colour. Part 2 consists of papers discussing Jackson’s work on phenomenal consciousness. I have separated the papers on consciousness from the papers on more general issues in the philosophy of mind because his papers in this area have been so influential (especially Jackson 1982 and 1986). The papers in Part 3 discuss Jackson’s work in ethics—both his normative ethics and his metaethics. Finally, Part 4 contains papers related to Jackson’s work on conditionals. I have included Martin Davies’s paper on two sorts of epistemic tasks in Part 4 because it takes as its starting point a distinction Jackson originally made in his 1987 monograph on conditionals. This collection closes with Part 5, replies by Jackson to his critics. This section deserves especial attention because it contains what are, in my opinion, some of the clearest expressions of Jackson’s ideas. In this paper I sketch some key aspects of Jackson’s thought. My discussion is in four sections, corresponding to the four parts of the collection. The

2

ian ravenscroft

sketch provides me with a framework for briefly introducing some of the ideas developed in this collection. I hasten to add that all the papers are far more richly detailed—and more nuanced—than my brief comments might suggest.

1. Metaphysics In his From Metaphysics to Ethics (1998, especially chapters 1–3), Jackson develops what we might describe as his meta-metaphysics.¹ His aim is to elucidate and defend an approach to what he calls ‘‘serious metaphysics’’. The serious metaphysician regards a proper subset of the (actual) properties as in some sense fundamental, and then seeks to explain—or elucidate—or understand—the remaining properties in terms of the fundamental subset. Physicalism provides a good example of serious metaphysics since the physicalist takes as fundamental a set of properties posited by physical science, and aims to explain (or elucidate or understand) all the remaining properties in terms of the fundamental physical properties. It’s worth noting that, as Jackson uses the expression, ‘‘serious metaphysics’’ is a term of art: he does not dispute that some philosophers have taken seriously a kind of radical pluralism which denies that any one set of properties is metaphysically basic.² He uses the term to pick out a position which he regards as being, upon reflection, obviously true. He notes, for example, that the property of density can be understood in terms of the distinct properties of mass and volume, and so there is an obvious sense in which the properties of mass and volume are more basic than, and able to explain, the property of density. Jackson points out that any serious serious metaphysician must face up to what he calls the ‘‘location problem’’; indeed, they must face up to a series of roughly analogous location problems. Following Jackson I will take physicalism as my example of serious metaphysics. The physicalist faces the problem of locating the psychological properties amongst the physical properties. With which physical property, or complexes of physical properties, is the property believes that snow is white to be identified? Similarly, the physicalist faces the problem of locating the semantic properties and the ethical properties amongst the physical properties. One option here, of course, is to turn eliminativist ¹ Jackson 1998, chs. 1–3, is an extension and development of earlier works, especially Jackson 1992 and 1994. ² He quotes a passage from Searle 1992 which would appear to endorse something close to what I have called radical pluralism.

introduction: themes and criticisms

3

about any putative set of non-fundamental properties. Jackson is, however, robustly realist about the psychological, the semantic and the ethical. I said somewhat vaguely that the serious metaphysician wants to explain or elucidate, or understand one set of properties in terms of another. Exactly what relation does Jackson think exists between the fundamental properties and the target properties—between, for example, the physical properties and the psychological properties? According to Jackson, if physicalism is true then the way things are physically entails the way things are psychologically (1994, 1998: chapter 1). He arrives at this conclusion by first articulating physicalism’s central claim as a supervenience thesis. According to Jackson, if physicalism is true at the actual world then: Any world which is a minimal physical duplicate of our world is a duplicate simpliciter of our world. (1998: 12) Physicalism about the psychological then drops out as a special case: Any world which is a minimal physical duplicate of our world is a psychological duplicate of our world. Notice, first, the qualifier ‘‘minimal’’. The physicalist should not accept that any world which duplicates the actual physical properties is a psychological duplicate of this world, for presumably there are worlds which duplicate the actual physical properties but which include, in addition, a range of epiphenomenal mental properties. Consequently, Jackson defines physicalism in terms of minimal physical duplication, where a minimal physical duplicate is one to which no non-physical properties have been added. Second, notice that on this account physicalism is at best a contingent truth about our world, not a necessary truth about all worlds. And this sounds right: the physicalist need not deny that there are ‘‘spooky’’ worlds in the more remote parts of logical space; they need insist only that ours is not a spooky world. (For a dissenting opinion see Simon Blackburn’s contribution to this volume (p. 26). Blackburn suggests that any substrata both sufficiently stable and sufficiently rich to support cognition would probably deserve the label ‘‘physical’’.) With this account of physicalism in hand, we can return to Jackson’s claim that, if physicalism is true at our world then the way things are physically entails the way things are psychologically. (See Jackson 1998: 25–6.) Consider two sentences, φ and ψ. φ is a complete description of the physical nature of the actual world and includes a ‘‘stop clause’’, analogous to the idea of minimal duplication, which indicates that no further properties are instantiated. According to Jackson, φ is true at the actual world and at all the minimal physical duplicates of the actual world. The other sentence, ψ, is a true sentence describing

4

ian ravenscroft

some psychological aspect of our world. Now, if physicalism (as defined by Jackson) is true, then ψ is also true at all the minimal physical duplicates of our world. Consequently, any world at which φ is true is also a world at which ψ is true. That is, φ entails (in the sense of necessarily determining) ψ. We have seen that, for serious metaphysicians, metaphysical problems will often take the form of location problems, for example, locating the psychological properties amongst the physical properties. As Jackson points out, to solve any location problem we need to be able to pick out what it is we want to locate; that is, we need to get clear on the subject of the problem (1998: 28–31). According to Jackson, the best way to do this is the ‘‘method of cases’’—we ask ourselves whether, in a range of actual and counterfactual cases, some object or property instance is a member of the target kind (Jackson 1998: 31–7). For example, we might ask ourselves whether Swampman is a believer; whether the watery stuff on Twin-Earth is water; or whether Putnam’s superstoic is in pain. Jackson calls the process of picking out the subject of a location problem via the method of cases ‘‘conceptual analysis’’. One of the things we can learn by conceptual analysis, according to Jackson, is that (a) Water is the stuff that plays the watery role, where the expression ‘‘the watery role’’ abbreviates a long folksy description which include ‘‘flows from taps, fills lakes, falls from the sky, ... ’’³ In addition, science has discovered that (b) H2 O is the stuff that plays the watery role, and that (c) 60% of the Earth is covered by H2 O. It follows that (d) 60% of the Earth is covered by water. (See Jackson 2003. I have labelled the propositions differently to Jackson.) Jackson accepts, of course, that (b) and (c) are empirical claims. However, he insists that (a) is a priori and that the entailment of (d) by the conjunction of (a), (b) and (c) is a priori. Consequently we have an ‘‘a priori passage’’ from two empirical truths, (b) and (c), neither of which mention water, to a truth about water. ³ Better: ‘‘the watery role’’ abbreviates a long disjunction of conjunctions of most of the folk platitudes concerning water.

introduction: themes and criticisms

5

Closely related to Jackson’s views on conceptual analysis is his advocacy of a distinction, originally due to Davies and Humberstone 1980, between A-intensions and C-intensions. Consider a term like ‘‘water’’. The A-intension of ‘‘water’’ is determined by the reference-fixing description associated with the term. Let’s use the phrase ‘‘watery’’ to describe any substance which satisfies the reference-fixing description associated with ‘‘water’’. Then the A-intension for water is that set of worlds which contain a watery substance. The C-intension of ‘‘water’’ is given by the natural kind which, at the actual world, is the watery stuff. That is, the C-intension of ‘‘water’’ is the set of worlds which contain H2 O. According to Jackson, the method of cases reveals the reference-fixing description of a term like ‘‘water’’ and thus reveals its A-intension. Conceptual analysis is one of the most controversial aspects of Jackson’s philosophy. Simon Blackburn (Chapter 1) raises a number of concerns about both conceptual analysis and the kind of a priori passage which Jackson endorses. I will mention only one of Blackburn’s worries here. Blackburn doubts that ‘‘N = the F’’ is a priori for all cases in which ‘‘F’’ is a description which serves as the reference-fixer of ‘‘N’’ (p. 32). Imagine we know that some unique person invented the zipper, but that is all we know about the zipper’s inventor. And imagine that we decide to use the name ‘‘Julius’’ to refer to that person—whoever they happen to be—who invented the zipper. Do we therefore know a priori that Julius invented the zipper? Blackburn’s view is that we do not, on the grounds that it is not ‘‘a priori knowable of anyone that they invented anything. You have to know the way the world is to know things like that.’’ (p. 32) Similarly, whilst we can use the name ‘‘water’’ to refer to whatever it is that (uniquely) plays the water role, we don’t know a priori that water plays the water role. To find out what plays the water role you have to go and look. Jennifer Hornsby (Chapter 2) accepts—at least for her present purposes—Jackson’s claim that any world which is a minimal physical duplicate of our world is a psychological duplicate of our world. She denies, though, Jackson’s further claim that ‘‘everyday psychological explanations are understandable in terms of, or are reducible to, physical ones’’ (p. 44; Hornsby is here quoting Jackson 2000: 18). According to Hornsby, a person has different properties from the assemblage of physical particles of which they are constituted. It follows by Leibniz’s law that a person is a distinct object from the assemblage of physical particles of which they are constituted. (Her argument here echoes the more familiar example of the lump of clay and the statue. Several authors have argued that the statue has different properties from the lump of clay from which it is fashioned and is therefore a distinct object. See

6

ian ravenscroft

Hornsby’s chapter for references.) Since people are not physical objects and do not have exclusively physical properties, people-level explanations will not reduce to physical level explanations.⁴ Hornsby thinks that Jackson has simply assumed that people are identical to the assemblage of physical particles of which they are constituted. But I am not sure that he does. After all, Jackson endorses arguments of the following sort: 1. Mental state M = the occupant of causal role R. 2. The occupant of causal role R = neural state N. Therefore, 3. Mental state M = neural state N. (See for example Braddon-Mitchell and Jackson 1996: 92–3. Jackson may want to restrict (2) in certain ways, but we can ignore that for present purposes.) Now the argument from (1) and (2) to (3) looks like an argument to the effect that a ‘‘person-level’’ property (having M) can be identified with a physical-level property (having N). So it seems to me that Jackson has not merely assumed the identity of persons and the assemblage of physical particles of which they are constituted; rather, he has indicated a general approach to establishing that claim. How do we obtain premiss (1)? Here we return to the issue of conceptual analysis, for according to Jackson we establish premiss (1) by the method of cases. Hornsby, though, objects to conceptual analysis for two reasons. First, she draws attention to the striking lack of successfully completed analyses. Second, she doubts that there is some principled way of distinguishing the everyday psychological properties from all the other properties. If there is no such way, then the task of analysing the everyday psychological properties in terms of strictly non-psychological properties looks hopeless. William G. Lycan (Chapter 3) also has a number of concerns about conceptual analysis, of which I will mention just one. The outcome of a successful attempt at Jacksonian conceptual analysis will be a description which serves to fix the reference of the term in question. Lycan accepts that people often do deploy reference-fixing descriptions, but thinks that such reference-fixers are make-shift and fallible (p. 72–3). Consequently, we cannot expect to obtain a unique reference-fixing description for a term like ‘‘water’’. This ties into Lycan’s views on A-intensions. According to Lycan, English ⁴ It is important to stress that Hornsby is not endorsing any kind of dualism. While she accepts, for present purposes, Jackson’s physicalism, she thinks that physicalism leaves room for properties which are neither included in, nor reducible to, the properties over which physics quantifies.

introduction: themes and criticisms

7

words don’t have A-intensions because there is no stable and distinctive set of reference-fixers held by all users of a word. Reference-fixing descriptions vary from person to person, and may vary for a single person over time. Lycan defends this view by pointing out that all the usual candidates for fixing the reference of ‘‘water’’ are expendable. We could help someone unacquainted with the term grasp the intension of ‘‘water’’ even in the absence of lakes or taps or rain or ... (p. 72–3). Schroeter and Bigelow (Chapter 4) begin by pointing out Jackson’s commitment to what they call the classical model of meaning. Let us take the kind term ‘‘water’’ as our example. On the classical model, competent speakers of English possess a set of tacit assumptions about the reference conditions of ‘‘water’’—assumptions which allow them to determine the reference of water across a range of real and hypothetical situations. Co-reference is possible because the vast majority of competent English speakers share pretty much the same set of reference-fixing assumptions about water. Notice that, if the classical model is correct, reflection on real and hypothetical situations should, in principle, yield an a priori analysis of ‘‘water’’. Schroeter and Bigelow, however, have doubts about the classical model. One of their principle concerns is Quinean: they doubt that a sharp distinction can be made between those assumptions about water which fix the reference of ‘‘water’’ and those that do not. In place of the classical model they urge, firstly, that co-reference emerges not from the common possession of a fairly determinate set of assumptions, but from a commonly shared intention to co-refer; and secondly that the intention to co-refer typically leads to the acquisition of overlapping sets of assumptions by the speakers involved. Crucially, the explanatory direction has been reversed. On the classical model a set of shared assumptions underpins co-reference; on Schroeter and Bigelow’s ‘‘jazz’’ model, recognition of co-reference tends to lead to the possession of overlapping sets of assumptions. According to Schroeter and Bigelow, the jazz model will not support the kind of a priori analysis Jackson requires. In some ways, the objections raised by Huw Price (Chapter 5) to Jackson’s metaphysical project are the most radical of those presented in this volume. Jackson assumes that the semantic properties are substantial; that is, he rejects deflationism about the semantic properties. Jackson’s view is accepted—at least for present purposes—by most of Jackson’s critics. Whilst they take issue with Jackson on semantics questions, they leave unchallenged his anti-deflationism. In contrast, Price endorses global expressivism—the view that the uses of language can be adequately understood in non-representational terms. Global expressivism is ontologically conservative in the following way. If we accept that some term T stands for or represents something in the world—say X—then

8

ian ravenscroft

our using T typically commits us to the existence of X.⁵ It is then open to us to worry about how X fits into the physicalist picture. But if we eschew the idea that T stands for or represents X, we are no longer committed to accepting X into our ontology. In this way, global expressivism is conservative: it is committed to the existence of talk about X, but not to the existence of X itself. And if we are not committed to the existence of X, then we are no longer committed to resolving the location problem for X. Price accepts, of course, that there can be real location problems (p. 121). However, they do not emerge from semantic considerations, and their resolution does not depend on semantic claims. One location problem to which Jackson gives special attention concerns the colours. Jackson accepts the following as an axiom of the folk theory of colour: Causal axiom. Red is the property of an object which is the normal cause of its looking red. Science tells us that the normal cause of an object’s looking red is a complex of physical properties. It follows that red is a complex of physical properties. However, Jackson also accepts the following further axiom of the folk theory of colour: Representation axiom. Red is the property that looking red represents objects as having. These two axioms do not sit comfortably together, for my colour experiences do not represent objects as having complex physical properties like reflectance profiles. Jackson is aware of this difficulty, and proposes that the folk theory of colour in effect renders ‘‘red’’ ambiguous. Sometimes we use ‘‘red’’ in accordance with the causal axiom to pick out a complex physical property; sometimes we use it in accordance with the representation axiom to pick out the second-order property of having a property which occupies the role associated by the folk with red. That role includes being more like pink than green, being more like orange than blue, and so on. Peter Menzies (Chapter 6) offers three reasons for rejecting Jackson’s resolution of the tension between the causal and representational axioms (pp. 147–8). First, he doubts that it is true to the phenomenology of colour experience. Second, he interprets the representation axiom as requiring that colours are simple intrinsic properties of objects rather than second-order properties as Jackson suggests. And finally he takes it that, according to the folk theory, red is a single property which is both represented by our red ⁵ Typically but not always: we can accept that some terms fail to refer.

introduction: themes and criticisms

9

experiences and the cause of those experiences. Menzies does not, however, want to reject one or other of the folk axioms. Rather, he aims to defuse the tension between them. Why not say that colours are intrinsic properties of objects as required by the representation axiom, but are realized by complex physical properties as required by the causal axiom? Menzies calls this position ‘‘primitivism’’. Primitivism is commonly rejected on the grounds that it entails causal overdetermination. Consider a causal claim involving colour: the red traffic light caused the driver to stop. Which property of the traffic light is doing the causing: the property of being red or the complex realizer property? Surely it can’t be both, because that would be a case of overdetermination. In reply, Menzies sketches a model-relative theory of causation. On this view, no overdetermination is involved in claiming both that the traffic light’s being red caused the driver to stop and that the complex physical realizer property caused the driver to stop because these causal claims are being made relative to different models. In this way, Menzies defends primitivism as the correct approach to the folk theory of colour.

2. The knowledge argument In recent years Jackson has argued against the knowledge argument (see for example Jackson 2003). His response relies on (a version of) representationalism about qualia—the doctrine that the phenomenal aspects of our experiences are determined by their representational properties: no phenomenal difference without representational difference. Representationalism is, of course, controversial. Some authors have advanced examples of phenomenal experiences which, it is alleged, are not representational. Thus Ned Block has claimed that the feeling of orgasm represents nothing (Block 1994: 216). Others have argued that there can be phenomenal difference without representational difference (for discussion see Tye 2000 and Byrne 2001). Rather than pursue this debate, I will turn to the way Jackson deploys representationalism about qualia in his reply to the knowledge argument. According to representationalism, the phenomenal properties of Mary’s post-release experience of a red tomato are determined by its representational properties. Now whilst the details remain controversial, it is widely agreed that the representational properties of mental states are determined by the physical properties of the organism (or by the physical properties of the organism and its environment including, perhaps, the selective history of the species).

10

ian ravenscroft

Consequently Mary could, in principle, have anticipated the phenomenology of seeing a red tomato. Jackson still thinks, however, that some change occurs when Mary leaves the black and white room. Whilst pre-release Mary will have detailed propositional knowledge of the nature of the representational vehicle formed when normal people gaze at a ripe tomato, she will not herself possess such a vehicle until she is released. And her lack of such a vehicle will be revealed by her lack of certain skills—skills associated with recognizing, recalling and imagining red objects. In other words, Jackson arrives at a version of the Nemirow–Lewis response to the knowledge argument.⁶ Philip Pettit (Chapter 7) largely agrees with Jackson’s physicalism, including Jackson’s representationalist rebuff of the knowledge argument. He is struck, though, by the unsatisfying nature of the representationalist reply. Even if we give our intellectual assent to the premisses of the representationalist argument, the conclusion—that the phenomenal is determined by the physical—does not carry much intuitive conviction. According to Pettit, this lack of conviction is due to our complete lack of experience of similar sorts of cases. We simply don’t have a large background of relevant examples to draw on, examples in which grasping the physical facts about a human agent leads us to an appreciation of their phenomenal experiences. We are, in this area, cognitively limited. It is worth noting, though, that Pettit’s position is quite different to that of ‘‘mysterians’’ like Colin McGinn (1993). McGinn denies that humans are intellectually capable of formulating the true physicalist theory of phenomenal consciousness. In contrast, Pettit asserts that we already possess at least the outline of such a theory, but that the conclusion will not ‘‘pop-out’’ (his term) from the premisses in the way the conclusion of a simple modus ponens ‘‘pops-out’’ from the premisses. Robert Van Gulick (Chapter 8) does not share Pettit’s sanguine attitude to Jackson’s reply to the Knowledge Argument. Van Gulick’s objections turn not so much on his representationalism as on his a priorism. The a priorist physicalist claims that the truth of physicalism entails that all truths (including truths about phenomenal experience) can be derived from the microphysical facts. Consequently, the seeming impossibility of Mary knowing what red is like prior to her release challenges a priori physicalism. In contrast, non-a priori physicalists like Van Gulick deny physicalism entails that all truths can be derived from the microphysical level, and are consequently nonplussed by Mary’s ignorance. According to Van Gulick, the a priori physicalist has confused metaphysical issues with epistemic ones: whilst the microphysical necessitates the phenomenal, facts about the former cannot always be derived ⁶ See Nemirow 1980 and 1990, and Lewis 1990.

introduction: themes and criticisms

11

from the totality of facts about the latter. Van Gulick locates the a priorist’s error in a mistaken conception of scientific theories. He rejects the idea that theories are fundamentally representational, preferring a much more pragmatist approach. To borrow a memorable phrase from Richard Rorty, the task is not so much to copy the world as to cope with it.⁷ There is no guarantee that one useful coping strategy can be derived a priori from another.

3. Ethics Jackson has defended positions in both metaethics and normative ethics. Not surprisingly, his metaethical position is realist and thus there arises a version of the location problem introduced above (Jackson 1998: chapters 5 and 6). If we reject the idea that the moral properties are either sui generis or supernatural, we face the following question: with which descriptive properties or complexes of such properties are the moral properties to be identified? In the philosophy of mind analytic functionalists treat our everyday platitudes about the mind as a term-introducing theory (Lewis 1970, 1972). The set of platitudes (or their systematization) is called, of course, ‘‘folk psychology’’. Similarly, Jackson proposes that we treat our everyday moral platitudes as a term-introducing theory (Jackson 1992, 1998: chapter 6). More accurately, Jackson proposes that we should treat the set of moral platitudes at which we would arrive after the process of Rawlsian reflection is complete as a termintroducing theory. He calls that set of platitudes (or their systematization) ‘‘mature folk psychology’’ (Jackson 1998: chapter 5). He then applies the method of ramsification familiar from analytic functionalism, and obtains definitions of the moral terms in purely descriptive language. With such definitions in hand we can, in principle, identify the moral properties with (complexes of) descriptive properties. Terry Horgan and Mark Timmons (Chapter 9) offer a Twin Earth style objection to Jackson’s semantic theory of the moral vocabulary. Say that on Earth the mature folk morality on which the process of Rawlsian reflection would converge is a consequentialist theory, C. Moral Twin Earth is, as we would expect, strikingly similar to Earth except that, due to subtle differences in Twin Earth psychology, Twin Earthers would converge on a deontological theory, D. Now for the test of semantic intuitions: when Twin Earthers and Earthers meet and disagree about moral questions, does their disagreement arise ⁷ Cited in Blackburn 2003. As Blackburn rightly points out, an accurate representation of the world can be very useful in coping with it.

12

ian ravenscroft

because they mean different things by words like ‘‘good’’ and ‘‘right’’ and are consequently talking past each other, or because they hold different substantive views about the good and the right? Horgan’s and Timmon’s intuitions favour the latter alternative. Jackson, though, would appear to be committed to the view that, because C and D are different theories, Twin Earthers and Earthers are deploying different—though orthographically identical—moral vocabularies. Generalizing on this result, Horgan and Timmons go on to articulate their own anti-descriptivist position. What of normative ethics? Jackson defends a decision-theoretic version of consequentialism (Jackson 1991). On this view, we should maximize i Pr(Oi /Aj ) × V(Oi ), where Pr is the agent’s subjective probability function, V is consequentialism’s value function, and Oi are the possible outcomes of action Aj . In other words, ‘‘whereas decision theory enjoins the maximization of expected utility, consequentialism enjoins the maximization of expected moral utility’’ (Jackson 1991: 464). One of the advantages Jackson claims for decision-theoretic consequentialism is that it yields a solution to the nearest and dearest objection. Numerous authors have worried that consequentialism places too little emphasis on the people and projects closest to our hearts. Say that I can choose between purchasing my son a $10 present or donating $10 to famine relief. Plausibly, my $10 is more likely to maximize happiness or preference satisfaction or ... if donated to famine relief than if used to buy (yet another) present for my son. According to consequentialism I am therefore morally required to donate to famine relief rather than purchase the present. And yet it has seemed to many that relationships exemplified by that between parents and children lie at the very heart of a worthwhile life (see for example Williams 1973). If this is right, consequentialism instructs us to perform acts which will render our lives not worth living. In response, Jackson draws attention to the fact that the agent’s subjective probability function plays a significant role in decisiontheoretic consequentialism. Whilst the consequentialist value function assigns a higher value to my giving to famine relief than it does to my giving a present to my child, giving a present to my child may have a higher expected moral utility. This will be the case if I am much more confident about the positive effects my gift will have on my son than I am about the positive effects my gift will have on the famine victims. Jackson contends that my close proximity to my son and my great distance from the victims of famine pretty much guarantees that I will be very much better informed about my

introduction: themes and criticisms

13

son than about the famine victims. In addition, given the nature of human motivational biases, I can be more confident that I will carry through on my son-orientated project than on any alternative stranger–orientated project. Thus I have, after all, good consequentialist reasons to favour my nearest and dearest. Michael Smith (Chapter 10) suggests that the nearest and dearest objection is best understood as attempting to show that consequentialism is, in Parfit (1984: 27) terms, ‘‘indirectly collectively self-defeating’’. The concern is that if significant numbers of us act to maximize neutral value we will end up living lives which are not worth living—not worth living because we will have given up precisely those projects that make our lives worth living. Consequentialism is thus indirectly collectively self-defeating: adhering to the injunction to act so as to maximize neutral value is likely to render us all worse off. In reply, Smith follows Parfit in noting that consequentialism enjoins us to apply its principle of evaluation to both acts and the kinds of lives we might live. With respect to acts, it tells us to act so as to maximize neutral value; with respect to the lives we might live, it tells us to live so as to maximize neutral value. These injunctions are logically independent: what acts I should perform is one thing; what life I should lead quite another. Of course, in leading the life which maximizes neutral value I will perform various acts, but they may not be the acts which consequentialism, applied to acts, would enjoin me to perform. In short, the premiss of the nearest and dearest argument—act so as to maximize neutral value—does not entail the conclusion—live a life not worth living. In addition to defending decision-theoretic consequentialism, Jackson defends actualism. According to actualism, when an agent deliberates about whether to A she should consider what would actually occur if she were to A. In contrast, according to possibilism an agent deliberating about whether to A should consider at least some of the possible outcomes of her action. Jackson and Pargetter (1986) have offered the example of Procrastinator who is asked to write a book review. The best outcome would be for Procrastinator to agree to write the review and do so on time; the worst outcome would be for him to agree to write the review but fail to do so on time. The value of declining to write the review lies somewhere in between. The actualist says that Procrastinator should decline to write the review because Procrastinator won’t actually write the review on time; the possibilist says that Procrastinator should agree to write the review because it is possible that Procrastinator will overcome his usual sloth and write the review on time. So Jackson is a decision-theoretic actualist: an agent should maximize expected

14

ian ravenscroft

moral utility, considering only what will actually occur subsequent to her action. In addition to the distinction between actualism and possibilism, Peter Railton (1984) has distinguished between objective and subjective utilitarianism. According to the objectivist, the right action is the one which would, as a matter of fact, bring about the best outcome; according to the subjectivist, the right action is the one which the agent believes will bring about the right action. Jackson is opposed to objective consequentialism. He offers the case of Jill, a doctor whose patient John has a minor skin rash. Jill has two drugs at her disposal: X and Y. X has a 90 per cent chance of curing John and a 10 per cent chance of killing him. Y has a 50 per cent chance of curing John and no adverse side effects. (Jackson 1991: 467.) According to Jackson, Jill should prescribe drug Y—and surely he is right about this. However, notice that it is drug X which has the highest probability of getting the best outcome. Consequently the objectivist seems to be committed to the entirely counterintuitive result of advising Jill to prescribe X. According to Jackson, the right action is that endorsed by a particular version of subjective consequentialism: actualist decision theoretic consequentialism. However, Julia Driver (Chapter 11) argues for a distinction between right action and what an agent ought to do. When it comes to determining the right action, Driver endorses objectivism: the right action is the one which would bring about the best result. She notes, though, that there is a separate, epistemic, question: given that we do not always know in advance which action will bring about the best result, the issue arises of what we ought to do. In other words, we have to give thought to decision procedures. At this point, Driver joins with Jackson, and endorses decision-theoretic consequentialism as the best account of what an agent ought to do.

4. Conditionals & the purposes of arguing In his 1987 monograph, Conditionals, Jackson offers truth and assertibility conditions for both indicative and subjunctive conditionals. For the indicative conditional, the truth conditions are those of the corresponding material conditional, and the conditional is assertible if and only if its consequent is probable with respect to its antecedent.⁸ For the subjunctive conditional, the truth conditions are given by the familiar Stalnaker–Lewis semantics (Stalnaker ⁸ Jackson 1987 places further restrictions on his account of assertibility, but they need not concern us here.

introduction: themes and criticisms

15

1968; Lewis 1973), and the conditional is assertible if and only if its probability is high. On Jackson’s view, the assertibility of a conditional makes a contribution to its meaning: ‘‘There is more to meaning than truth-conditions’’ (Jackson 1987: 36). Distinguishing between truth and assertibility conditions gives Jackson a strategy for handling conditionals that seem to resist interpretation as material conditionals (in the case of indicatives) or Stalnaker–Lewis conditionals (in the case of subjunctives). The problematic conditionals do in fact have the truth values given by Jackson; however, their meaning is not captured by their truth conditions alone because assertibility plays a role in determining meaning. Once we take assertibility into account, the difficult cases are readily resolved. Like Jackson, Dorothy Edgington (Chapter 12) thinks that the conditional probability of the consequent given the antecedent is crucial to understanding indicative conditionals. However, unlike Jackson, she denies that the truth value of an indicative conditional is given by the corresponding material conditional. She notes, to begin with, the difficulties which embedded conditionals (i.e. conditionals which have conditional antecedents and/or consequents) present to Jackson. Jackson is aware that embedded conditionals present problems to his approach, and argues that ‘‘If A then (if B then C)’’ is regarded by English users as interchangeable with ‘‘If (A and B) then C’’. Since the latter contains no embedded conditionals, the problem is avoided. (See the appendix to Jackson 1987.) However, Edgington is concerned that the proposed solution fails to generalize in satisfactory ways to connectives like ‘‘but’’ and ‘‘even’’. In addition, she is concerned that Jackson’s approach to conditionals is restricted to conditionals used to make assertions, and does not readily generalize to conditionals used in other kinds of speech acts. Edgington’s response to the difficulties with taking the material conditional as giving the truth conditions for the indicative conditional is not to seek an alternative, more satisfactory account of the truth conditions of indicative conditionals; rather, she endorses the ‘‘no truth conditions’’ position originally proposed by Ernest Adams (see Edgington’s chapter for references). In addition to worries about the truth conditions of indicative conditionals, Edgington has concerns about the distinction between indicative and subjunctive conditionals. As she points out, future-orientated indicatives can be expressed quite naturally in the subjunctive mood: ‘‘If you touch that wire you will get a shock’’ and ‘‘If you were to touch that wire you would get a shock’’ are interchangeable in most contexts. In addition, if the antecedent turns out to be false, the indicative is naturally expressed using a counterfactual: ‘‘If you had touched that wire you would have got a shock’’. So it seems that the same indicative thought can naturally be expressed using the subjunctive or

16

ian ravenscroft

the counterfactual. If this is right, the indicative/subjunctive distinction does not mark an important semantic boundary. Graham Priest (Chapter 13) has a number of concerns about Jackson’s approach to conditionals. To begin with, he doubts that the distinction between indicative and subjunctive conditionals withstands careful scrutiny. Counterfactual conditionals—that is, conditionals with counterfactual antecedents—are often taken to be a subset of the subjunctive conditionals. However, Priest offers examples of subjunctive conditionals which are not counterfactual and of indicative conditionals which are counterfactual (pp. 313–4). In addition, Priest has doubts about Jackson’s account of assertibility. According to Priest, there are highly assertible conditionals which have improbable consequents with respect to their antecedents (p. 319). He offers the following case. Say that (i) all children under 10 live with their parents; (ii) all the inhabitants of Autumn’s Retreat are aged pensioners; and (iii) Fred’s parents live at Autumn’s Retreat. Then someone might reason as follows: If Fred is under 10 he lives with his parents. But in that case he lives at Autumn’s Retreat, in which case he is an aged pensioner. Thus, if Fred is under 10 he is an aged pensioner. It follows that Fred is not under 10. The italicized conditional is highly assertible in the context, and yet Pr(Fred is an aged pensioner/Fred is under 10) is very low. Priest is also concerned about the Stalnaker–Lewis semantics. According to Stalnaker–Lewis, any counterfactual conditional with an impossible antecedent is false. However, many people share the intuition that some counterfactuals with impossible antecedents are true. Priest offers the following example (p. 318): (1) If you had shown Fermat’s last theorem to be false in 1990, you would have become famous. Since Fermat’s last theorem was proved true in the mid-1990s, the antecedent is impossible and so, on the Stalnaker–Lewis semantics the conditional is false. However, intuitively (1) is highly probable. Finally, Priest doubts that the indicative conditional can be successfully interpreted as the material conditional. He offers the following example. Say that you don’t know where John is, but you do know that if he is in Paris then he is in France, and if he is in Beijing then he is in China. Interpreting the indicative conditional as the material conditional we have (with the obvious symbolism): (2) (P ⊃ F) & (B ⊃ C)

introduction: themes and criticisms

17

From (2) it follows that: (3) (B ⊃ F) & (P ⊃ C). However, it is clearly not the case that if John is in Beijing then he is in France, and if John is in Paris then he is in China. In Conditionals (1987: chapter 6), Jackson distinguishes two purposes of arguing: teasing out and convincing. Consider an argument A from premisses P1 ... Pn to conclusion Q. When A is articulated for the convincing purpose, the speaker intends the hearer to accept Q on the basis of P1 ... Pn . Typically, the speaker also offers reasons in support of the premisses, and the degree of support the reasons give to the premisses depends on background assumptions. The speaker’s background assumptions may differ from the hearer’s background assumptions, and so the degree of support the reasons provide to the premisses may differ from speaker to hearer. When A is articulated for the teasing out purpose, the aim is not to convince the hearer that Q. The hearer may already accept Q, or resolutely deny it. Rather, the purpose is to remind the hearer of the relevance of P1 ... Pn to Q. When the hearer already accepts Q, they may simply note the inference from P1 ... Pn to Q; when they reject Q they may carefully check that they reject at least one of P1 ... Pn . Martin Davies (Chapter 14) transposes Jackson’s convincing/teasing out distinction from the dialectical domain to the epistemological one. Analogous to the convincing purpose of argument is the epistemic task of settling the question; and analogous to the teasing out purpose of argument is the epistemic task of deciding what to believe. The latter involves noting that some of one’s beliefs (P1 ... Pn ) entail Q. Q is then added to one’s corpus of beliefs or, if one already has good grounds to deny Q, one’s grounds for accepting P1 ... Pn are reassessed. In contrast, settling the question involves deciding what to believe about Q. Davies explores a variety of complexities that arise in the case of settling the question. In Chapter 6 of Conditionals, Jackson gives an account of begging the question as a shortcoming of arguments proposed for the convincing purpose. Davies argues that the traditional conception of begging the question as proposed by, for example, Copi (1961) is a shortcoming of arguments proposed for the teasing out purpose. He then identifies a pair of epistemic shortcomings associated with settling the question and deciding what to believe. Long and sophisticated, Davies’s paper is more than a set of important distinctions: he deploys his subtle conceptual apparatus to investigate recent work on warrant and transmission, and Moore’s (1939) ‘‘hands’’ argument against external world scepticism. Flinders University

18 ian ravenscroft

References Blackburn, S. 2003. ‘‘Richard Rorty’’. Prospect Magazine 85. Block, N. 1994. ‘‘Consciousness’’. In S. Guttenplan (ed.), A Companion to the Philosophy of Mind (Oxford: Blackwell). Braddon-Mitchell, D. & Jackson, F. 1996. Philosophy of Mind and Cognition (Oxford: Blackwell). Byrne, A. 2001. ‘‘Intensionalism defended’’. Philosophical Review 110: 199–240. Copi, I. (1961). Introduction to Logic (New York: MacMillan). Davies, M. & Humberstone, L. 1980. ‘‘Two notions of necessity’’. Philosophical Studies 38: 1–30. Jackson, F. 1977. Perception: A Representational Theory (Cambridge: Cambridge University Press). Reprinted, Gregg Revivals, 1994. 1982. ‘‘Epiphenomenal qualia’’. Philosophical Quarterly 32: 127–36. 1986. ‘‘What Mary didn’t know’’. Journal of Philosophy 83: 291–5. 1987. Conditionals (Oxford: Blackwell). 1991. ‘‘Decision-theoretic consequentialism and the nearest and dearest objection’’. Ethics 101: 461–82. 1992. ‘‘Critical notice of Susan Hurley, Natural Reasons’’. Australasian Journal of Philosophy 70: 475–88. 1994. ‘‘Armchair metaphysics’’. In M. Michael and J. O’Leary-Hawthorne (eds.), Philosophy in Mind (Dordrecht: Kluwer). 1998. From Metaphysics to Ethics (Oxford: Oxford University Press). 2000. ‘‘On the metaphysics of human agency’’. (Discussion with Jennifer Hornsby and Lynne Rudder Baker.) Philosophical Explorations 3: 188–201. 2003. ‘‘Mind and illusion’’. In A. O’Hear (ed.), Minds and Persons, Royal Institute of Philosophy Supplement 53 (Cambridge: Cambridge University Press). & Pargetter, R. 1986. ‘‘Oughts, options, and actualism’’. Philosophical Review 95: 233–55. Lewis, D. 1970. ‘‘How to define theoretical terms’’. Journal of Philosophy 67: 427–46. 1972. ‘‘Psychophysical and theoretical identifications’’. Australasian Journal of Philosophy 50: 249–58. 1973. Counterfactuals (Oxford: Blackwell). 1990. ‘‘What experience teaches’’. In W. Lycan (ed.), Mind and Cognition (Oxford: Blackwell). McGinn, C. 1993. The Problem of Consciousness (Oxford: Blackwell). Moore, G. 1939. ‘‘Proof of an external world’’. Proceedings of the British Academy 25: 273–300.

introduction: themes and criticisms

19

Nemirow, L. 1980. ‘‘Review of T. Nagel, Mortal Questions’’. Philosophical Review 89: 475–6. 1990. Physicalism and the cognitive role of acquaintance. In W. Lycan (ed.), Mind and Cognition (Oxford: Blackwell). Parfit, D. 1984. Reasons and Persons (Oxford: Oxford University Press). Railton, P. 1984. ‘‘Alienation, consequentialism, and the demands of morality’’. Philosophy and Public Affairs 13: 134–71. Searle, J. 1992. The Rediscovery of Mind (Cambridge, MA: MIT Press). Stalnaker, R. 1968. ‘‘A theory of conditionals’’. In N. Rescher (ed), Studies in Logical Theory (Oxford: Blackwell). Tye, M. 2000. Consciousness, Color, and Content (Cambridge, MA: MIT Press). Williams, B. 1973. ‘‘A critique of utilitarianism’’. In J. Smart and B. Williams, Utilitarianism: For and Against (Cambridge: Cambridge University Press).

This page intentionally left blank

PA RT I

Metaphysics and Conceptual Analysis

This page intentionally left blank

1 Analysis, Description, and the A Priori? SIMON BLACKBURN

I This paper is an expanded version of a review of Frank Jackson’s From Metaphysics to Ethics that I wrote for the Australasian Journal of Philosophy. I owe thanks to the editor for permission to make use of overlapping material, and apologies to Frank Jackson, and readers of that journal, for revisiting old pastures. I still find the problems that I tried to highlight sufficiently interesting to bear additional scrutiny, and I hope some other philosophers do so as well. It concerns, after all, one of the deepest-rooted of all philosophical perennials: the way to locate difficult things like minds and values in the natural world. The subject of my problem is ‘analytic descriptivism’, or the ‘a priori passage principle’ that Jackson stands by, generating ‘the view that for each true statement concerning our world, there is a statement in physical terms that a priori entails that statement’ (Jackson 2003). But before discussing that principle, I want to stand back a little. I think it is illuminating, if perhaps a little depressing, to compare the current state of the mind–body problem with the state it was in at the end of the seventeenth century. When Locke pondered the relationship between the ‘motions’ in the body and brain on the one hand, and the sensations of heat or pain on the other, he could only shrug and say that it was ‘God’s good pleasure’ to annex such sensations to such motions. Leibniz protested vigorously that it was ‘not God’s way to act in such an unruly and unreasoned fashion’. He sought instead a way in which one thing expresses another ‘through some orderly relationship between them’, and gives as an example the way in which an ellipse has some relationship to a circle of which it is a projection on a plane. The relationship is not only orderly, but necessary, and it can be rationally grasped. It is a priori.

24

simon blackburn

Today we have machinery unknown to Locke and largely unknown to Leibniz. We have roles and realizers, the contingent a priori, a posteriori necessities, possible worlds, supervenience, Ramsey sentences, counterparts, de se properties, higher-order properties, and two-dimensional modal logic. Or at least, some of us like to operate this machinery, and Jackson is enthusiastically among them. Yet, curiously, when the machinery grinds to a halt, the options seem to be left surprisingly close to where our great predecessors left them. Either we have God’s good pleasure, which in the philosophy of mind amounts to mysterianism, or we must find a priori or ‘analytic’ relations between the difficult stuff we want to understand—such as the mind or the normative—and the familiar stuff we suppose we do understand—here, the physical. Jackson endorses this choice, and his deployment of the analytic and a priori suggests that he is as good a rationalist as Leibniz, although we shall eventually see some reason to qualify this. Furthermore, Jackson holds that the shape he gives the debates is found in many areas of metaphysics. Where there is something that threatens to transcend the physical or the natural, the way to demystify it is to ‘locate’ it in the natural order; this location means using conceptual analysis en route to showing how facts about it are deducible a priori from facts about the natural order. The key to establishing these deductions is conceptual analysis, and when it is provided, along with other materials we come to in due course, we have the a priori deducibility of the one kind of fact from the other. Many writers would agree with Jackson that a fundamental task of metaphysics is what he calls the location problem: showing how to locate the mystifying area in the natural world. Jackson is using the same metaphor as C. D. Broad who of course called his great book on the philosophy of mind, Mind and its Place in Nature. We have to find a place for mind in nature, or for colours in physical reality, or for the normative in the natural. I shall say that in such cases we have facts or properties or stuff at an unhappy level, a level that gives us discomfort, for one reason or another. The facts or properties of this unhappy kind seem not to fit well with what we take ourselves to know about our world and what it contains. To alleviate our discomfort, we need to place them in a world we like, which means relating them, and relating them happily, to facts or properties at a happy level. The happy level is likely to be that of physical facts, or properties, or stuff. So sometimes I shall simply refer to it as that. To say this is no more than to write the prospectus for naturalism, and to go further and provide goods answering the prospectus, we need to be sure what the levels are, what the sources of unhappiness may be, and what a happy relating account would have to look like as well.

analysis, description, and the a priori

25

Jackson argues that the way to approach the location problem is first, by defending supervenience, showing that there can be no variation at the unhappy level without variation at the happy (physical) level. He carefully defines this notion in terms of ‘minimal physical duplicates’ of our world (Jackson: 12). He goes on to argue that the supervenience thesis so-defined gives us entailment, although he is also clear that entailment is not a priori or ‘conceptual’ in character. It is more like ‘fixing’ or perhaps better ‘metaphysically fixing’. To explain this a little further, to philosophers of my generation, entailment was thought of as a logical relation. But the relation between logic and metaphysics was left fairly indeterminate. When that became a subject of intense scrutiny, with the work of Kripke and Lewis, it was realized that there were two notions that might qualify for the title. There was the old logical relationship, thought of in terms of a priori or rational deducibility. And there arrived a new notion corresponding to metaphysical necessity. These only kept together if metaphysical necessities corresponded completely with a priori deducibilities. But with the widely-accepted accounts of the necessity of identity, and identities flanked by terms for substances and properties and kinds, that correspondence vanished. Generally speaking, whether it was true in all possible worlds that something held became hostage to scientific discovery, of which the ‘water = H2 O’ identity serves as a paradigm. Jackson’s entailment at this point is a metaphysical one. That is, it gives us that the happy (the physical) determines the truth about, or fixes, the unhappy (e.g. the mental), in whatever way it does in this world, in all possible worlds. I am not myself sure that this implication actually holds; it looks to me as though it only holds given a somewhat flexible conception of the physical. Thus consider the physical and the mental. Jackson, like Lewis, is hospitable to the possible existence of stuff other than the physical and that might carry mentality—ectoplasm, entelechies and ‘such-like rubbish’ as Lewis called them. This is why supervenience is originally introduced only in terms of minimal physical duplication (Jackson 1998: 12). The idea is that if ours is a physical world, so that mentality is carried by the physical, then any world where we exactly duplicate the physical and stop right there is a mental duplicate of our world. Stopping right there, presumably, means not adding to it a dollop of entelechy or other non-physical way of carrying mentality. But entailment is a thesis about every world at which some complete physical story ϕ is true. Jackson effects the transition by saying ‘Let ϕ be the story as told in purely physical terms, which is true of the actual world and all the minimal physical duplicates of the actual world, and false elsewhere’ (Jackson 1998: 25). He claims that every world at which ϕ is true is one at which the

26

simon blackburn

true psychological nature of our world is duplicated. But given that ectoplasm and entelechies are in the modal cupboard, on the face of it this will not be true. In outer modal space there may be possible worlds in which the physics is exactly as in this world, but because of the ectoplasm, the psychological is not. Clearly this requires that the absence of entelechies is not itself part of ϕ: it is supposedly a truth invisible to physics, just as entelechies are. If that idea were dropped, then there could be no two physically identical worlds with entelechies sprinkled on one and none sprinkled on the other. ‘No entelechies’ would be a truth of physics. I have some sympathy with this. I suspect that there is an important sense in which anything capable of ‘carrying’ the mental must be physical, even if it is not confined to particles and energies and forces that determine our actual physics. Anything carrying the mental would need to endure at places, and for times, and it would need to have causal roles, varying with its history, in the same way that the physical itself carries traces of what has happened to it. It seems to be coming close to being physical, to me. But in that case the whole Lewisian caveat about entelechies could be dropped. In all worlds the mental would supervene upon the physical (just as in all worlds software supervenes upon hardware, that being whatever endures, has stable causal properties and so on), even if in some worlds the physical encompasses different stuff than in other worlds. But while entelechies and such like rubbish are thought of as non-physical, and so the possibility of non-physical bearers of mentality remains, it undermines the entailment thesis. This is especially so if we are even moderately hospitable to the holism of the mental. For that would prevent us even from sympathy with a bipartite story, whereby if ϕ is the complete physical story about our world, and determines a mentality Ø, then if we add to ϕ some mentality-carrying entelechies, we do not change Ø but merely add to it. For given moderate holism, addition will mean change. You cannot just add the belief, say, that New York lies in Australia to your present mindset, without its impact ricocheting around a bit. Or, if its impact does not do this, then your mind has changed by becoming muddled, which again would be a change, let us hope, from its antecedent state Ø.

II I do not think the struggles with minimal physical duplicates and entailment, in this metaphysical sense, are critical: there seems to be room to shuffle the

analysis, description, and the a priori

27

pieces in a number of different ways. As I have indicated, I am myself inclined to a supervenience claim that will give us metaphysical fixing, albeit with a generous conception of the physical. Jackson, as I understand him, thinks that a ‘stop clause’ that entails ‘no entelechies’ is itself part of the physical description of the world. But however this is solved, it is the next step, to conceptual analysis, that is more surprising, and that gives Jackson’s work its distinctive point. Clearly most modern metaphysicians would get as far as the physical ‘determining’ the mental, in the sense of necessary truth-fixing, and then baulk at supposing that this has anything to do with good old-fashioned analytic reductionism. They would take the ‘water = H2 O’ model to have shown that we can have identities, and metaphysical entailments, without analytic reductions. After all, there is no serious location problem for water, given physical reality. God has only to fix the distribution of H2 O to fix the distribution of water. Yet we need no a priori analytical approach to ‘water’ to understand this. On the face of it, we needed only chemistry. Perhaps this last remark is too strong. Chemistry itself needed a subject matter. It needed to know when it was faced with water (although as it grew up, it could also become legislative, determining whether stuff was water in terms of whether it is H2 O). In the beginning, in order to be confident that water is H2 O there must be a going practice of identifying water. We have to know what we are talking about. Now, we can certainly ruminate on the underpinnings of that practice. It might depend upon stereotypical examples and experiences, causal-historical connections, vague bunches of platitudes, and so on. But not, surely, by definition in any old-fashioned, pre-Quinean sense? How does Jackson come to think otherwise? First of all, it is not entirely clear how far he does think otherwise. He is, for instance, quite content to pay due homage to Quine, at least to the extent of not caring very much about conceptual identity versus paraphrase (Jackson 1998: 44). And perhaps he need not care too much about the status of what are often called the ‘platitudes’ that make up the ‘folk theory’ of water, or the mental, or the ethical, although in fact he does evince such care, arguing in general for their a priori status. I believe that Jackson’s principal argument for the omnipresence of conceptual analysis in metaphysics is quite short. It is, first, that we need to identify ‘our ordinary conception’ of the unhappy facts. The next step is that this ordinary conception is only revealed by partitioning possible cases into ones in which the targeted facts obtain and ones in which they do not; and ‘that in turn is to do conceptual analysis’ (Jackson 1998: 42). Jackson further argues that

28

simon blackburn

theories, such as causal-historical theories or theories based on stereotyping, that might appear to contravene this model, in fact do not, since they too appeal in various ways to the role of possible cases, and are therefore, in his sense, dependent on conceptual analysis. In fact it is clear that Jackson will not readily recognize any alternative: ‘... the crucial point, here, and generally, is that our classifications of things into categories ... is not done at random and is not a miracle’ (Jackson 1998: 64). Of course, the opposite of a miracle is something natural, and some might suppose that the lesson of Wittgenstein, or of Quine, or for that matter of little babies and chicken-sexers, is that classification can be natural without being a matter of the application of hidden rules or analyses. In fact, one might go further and say that a method of possible cases cannot possibly remove the taint of ‘miracle’ from a classification, since it will only reveal a reaction to the possible cases in terms of other words, which will then attract the miracle charge, leading to a regress. Thus, for instance, children grow into a world where they can tell what is a brick (out of the candidates in the contexts they live in) and what is a plank. If this seems miraculous, it will not help to defuse the miracle by listening to stumbling attempts at ‘analysis’. A brick, one might say, is a hard oblong stable smallish component of walls and buildings. Not a very good shot admittedly, since some decorative bricks are trapezoid for example, and stones are not bricks. Nevertheless, over-simple as it is, as far as removing miracles go it is useless, since the miracle now simply transposes to ‘oblong’ and the rest. And this would get worse if we started to add more qualifications, trying to remove the charge of over-simplicity. This is the standard objection to Fodor’s route to an innate ‘language of thought’, whereby learning any term presupposes a breakdown into more primitive terms, until we arrive at those we were born with, at which point we just cite God’s good pleasure at having given us an innate linguistic repertoire. For those unconvinced of this innateness, the terms of any analysis just raise the same questions, and if Jackson is generally right, they will require ‘analysis’ in just the same way, forever. There is a further worry suggested by the ‘water is H2 O’ paradigm. Did analytic chemists actually have to identify our ‘ordinary conception’ of water, in order to begin their work? They had to identify the stuff, certainly. But even if we concede that they had to use some conception of the stuff in order to have something to work on, as it were, it is a further step to say that they had to identify that conception. They certainly did not have to worry about whether XYZ on twin earth would count as water, any more than analytic chemists working out the constitution of jade had to worry about

analysis, description, and the a priori

29

what would be said if it came in various different chemical kinds, as in fact it does. Our chemists could be more or less totally relaxed about what to say under various possibilities. It was philosophers, much later on, who got exercised about that. I have been quite careful to describe water as stuff. The talismanic equation ‘water is H2 O’ is often taken as an equation of properties, but that strikes me as very likely a category mistake. Stuff like water has properties, but it does not follow that it is a property. It may be harmless to say that the identity is short for ‘the property of being water is the property of being H2 O’. But it may not—after all, a more natural rendering of what that sentence seems to be getting at would be ‘the property of being liquid and good for bathing and so on is the property of being composed of H2 O’, which strikes me as just untrue. And when we think of the extension, for instance to putative identities like ‘the property of being good is the property of being such as to satisfy the maximum number of desires, impartially considered’ I think caution is in order. For in recent philosophy the chemical model is the Trojan horse introducing the idea of property identities that escape Moore’s open question objection, since, it is said, that applies to concepts rather than properties. But it may not even do that: Moore, or versions of Moore’s argument, might well lead one to doubt whether those ‘two properties’ are one. I am not entirely prepared to be terrorized by the modern consensus into saying that Moore shows us that there are two concepts (in the realm of sense) but not that there are two properties (in the realm of reference). Similar remarks apply if we use the chemical model to persuade ourselves that we know what is meant by identifying mental and physical properties. However, I merely raise these worries, and nothing I go on to say depends upon sympathy with them. Jackson’s claim on behalf of analysis need not be quite as radical (or retrograde) as it sounds. Consider the above quotation from p. 42 of From Metaphysics to Ethics. It is only on a very thin conception of analysis that teasing out attitudes to possibilities is enough to count as doing analysis. It certainly does not entail finding classical necessary and sufficient conditions. Perhaps it is true that I do not know the ordinary folk conception of ‘red’ until I know the folk reaction to at least some range of actual and possible cases: perhaps orange, purple, red in the dark, red in queer lighting, surface red, solid red, after-image red. But finding where the boundaries lie is not the same as finding that the folk are applying a concept built from component concepts in the way that classical analysis demanded. And in fact it seems that rather little description even of where the boundaries lie may be sufficient to solve the location problem. Consider water again. Do

30

simon blackburn

we need to know much about how the folk will react to a whole budget of possibilities, in order to be confident that the location problem is solved by the identity of water and H2 O? I do not think so. Heroine who thinks that XYZ is water on twin earth is no worse off in this respect than Hero who thinks it is not. So long as neither of them is hospitable to the idea that we might have H2 O without it being water, neither of them is countenancing the dreaded superaddition—God’s good pleasure—that stops the unhappy from being in satisfactory metaphysical linkage with the happy. The same is true if they differ over mundane indeterminacies: cucumbers are not water, although they contain a higher proportion of H2 O than the Dead Sea, which is. Should that bother us as metaphysicians? Of course not. The folk drew their lines according to superficial similarities and their own interests, and while the Dead Sea looks more like a stretch of pure water than a cucumber does, they may have been swayed more by the thought that you can more easily immerse yourself in the Dead Sea than in a cucumber. Nietzsche said that no worthwhile concepts can be defined. Instead they have histories: they have grown in particular historical circumstances in response to particular historical pressures and human interests. They evolve into the often messy complexes that they now are, just as living organisms have evolved into the even more messy complexes that they now are. They are a mixture of adaptations and kludges. Consider for instance, the idea of ‘liberalism’ as a political kind. If Nietzsche is right this is better understood by a history or genealogy of the leading components (various kinds of freedoms, various attitudes to equality, various ideals of government, various attitudes to property) than by legislating a set of necessary and sufficient conditions for being a liberal. So one might be uninterested in ‘conceptual analysis’, contenting oneself with an understanding of the notion that serves the purposes of communication well enough, in the political circumstances we actually meet. We know enough to be able to say with confidence that Sweden is a liberal society, whereas Saudi Arabia is not, for example, and understanding the tradition enables us to do that, just as when we grow into our physical environment we have learned to tell bricks from planks, without ever thinking of a ‘conceptual analysis’ of either. Nevertheless, we might remain absolutely certain of a ‘metaphysical’ fixing: whether a society is liberal supervenes upon an underlying set of facts about its legal and governmental organization, system of sanctions, and so on. A society cannot change from being liberal to not being so without something in that region changing. Similarly, you cannot turn a brick into a non-brick without changing its physical nature. So, on the face of it, the ‘placing’ problem can be solved without very much conceptual analysis at all.

analysis, description, and the a priori

31

III Jackson turns from his defence of conceptual analysis to defence of a more remarkable thesis yet, namely that physicalism is committed to ‘conceptual entailments’ or a priori deducibility from the physical to the psychological (Jackson 1998: 68). It is important to notice that this is a further claim, not implied by his defence of conceptual analysis. For, to stick with the example, even if we set more store by investigating the partition of possible cases made with the unhappy vocabulary than I have suggested we should, we might still think that it took Lavoisier rather than any rationalist to show how, having fixed the physical facts, God had nothing left to do to make the world watery. Again, how does Jackson argue otherwise? Once more, there is a sense in which he doesn’t. Jackson knows that the bridge proposition: ‘H2 O is the watery stuff of our acquaintance’ is contingent and empirical (Jackson 1998: 82). His idea is a more subtle one. In terms of the exposition in his (2003), Jackson wants to relate (d) (water serving as an example of an ‘unhappy’ truth) to (a), a happy physical truth: (a) 60 per cent of the earth is covered by H2 O. (d) 60 per cent of the earth is covered by water. This passage, Jackson admits, is not a priori. But now we bring in the idea of ‘the watery role’. This sums up the role of water in our lives: its empirical manifestations. It includes everything—paradigms, necessary and sufficient conditions—that in normal life identifies water, and indeed did so well before Lavoisier. It is a role that can be identified without identifying, physically or chemically, the stuff that plays the role. And in principle different stuff could play the watery role. Now we have two further claims to consider: (b) Water is the stuff that plays the watery role. (c) H2 O is the stuff that plays the watery role. Jackson regards (b) as a priori. (c) he knows is empirical. But, (a) & (b) & (c) plausibly a priori entail (d). And since (b) is a priori we have two H2 O truths, (a) and (c), neither of which mention water, but that together a priori entail (d). We do have a priori passage from (a) and (c), the happy truths, to (d), the unhappy truth. The principle behind this, which I shall not contest, is that if A & B a priori entail C, and B is itself a priori, then A a priori entails C. Block and Stalnaker worried about the uniqueness claim (water is the stuff), pointing out that it is hardly a priori that water comes in just one chemical

32

simon blackburn

kind. Jade, samples of which appear very similar and play all the folk roles equally, does not. Jackson believes he can answer this query, and although I have some sympathy with it, I shall put it to one side. For, however that falls out, there is still a sense of something fishy going on. One point of attack would be the a prioricity of (b). This is argued by relying on the Kripkean account, especially illustrated by the case of the metre rod, of the logical behaviour of reference fixing descriptions. The general pattern is that we have sentences such as ‘N = the F’, where the description fixes the reference of the name ‘N’. Is it true, as Jackson claims (2003: 87) that ‘N = the F’ is a priori when ‘F’ is a description that specifies the reference fixer for ‘N’? I believe this claim is false. Consider Evans’s descriptive names. Suppose we none of us know who invented the zip, but let us suppose it is common knowledge that some unique person did. Now I introduce ‘Julius’ on the back of the description: ‘Let us call the man who uniquely invented the zip, Julius’. What is now a priori? We might say (if we are fastidious we might even baulk at this, but I shall let it pass) that it is a priori that the sentence ‘Julius invented the zip’ is true. But is it a priori known or knowable that Julius invented the zip? I do not think so, and I do not think it follows from the claim about the sentence being a priori true. I don’t think it is a priori knowable of anyone that they invented anything. You have to know the way of the world to know things like that. And since that is true, you cannot know a priori of Julius that he invented the zip either. If later on we discover empirically who invented the zip, we cannot say, ‘Hey, we knew all along it was you’. That’s just what we did not know before the detective work.¹ Agreeing now with Kripke that proper names are scopeless, since we do not know a priori of Julius (or anyone) that he invented the zip, we do not know a priori that Julius invented the zip either. It is no good saying that we can bridge this with another dose of a priori truth. Someone might try arguing that it is a priori that the sentence is true, and it is a priori that the sentence says that Julius invented the zip, so it must be a priori that Julius invented the zip. But to see that this is wrong, compare the standard account of indexicals. Someone saying ‘I am here now’ knows that the character of the sentence is such that it is bound to express a truth. But she does not know which truth (which proposition) it expresses, unless she knows where she is. The character is the standing meaning of the sentence, that anybody who can speak English understands. The content is only given in a context, including here the place the person is standing. ¹ I first gave this argument in Blackburn 1984: 334.

analysis, description, and the a priori

33

It is enough for our purposes to suppose that the content is entirely given by the place, so that the same content (proposition, thought, object of knowledge) is expressed by any two people standing at the same place and saying ‘I am here now’, regardless of whether one of them is lost in a fog, another is quite deluded about where he is, and only a third remains geographically competent.² What this means is that the content is given by a context, but the identity of the context does not impress itself on you unbidden, just through you being in a particular context. You only know the content if you know the context—which in this case means knowing where you are. So although each of these persons could say ‘I am here now’ and each be perfectly certain that they have voiced a truth, we ourselves should hold that only one of them knew that he was here (and we designate the place, say with an ostension) then. The deluded and the lost did not know this. That might be why, for instance, they remained insouciant although they were on the edge of a precipice. It might be why one of them, after bravely enunciating ‘I am here’, took a fatal step forward. He would not have done that if he had known where he was—had he known he was here (and we show where he was). So we block the idea of a sneaky way of obtaining empirical knowledge a priori, as if from the meaning of the sentence alone we could work out where we are, without compass or GPS. It may be that the spatial metaphor of ‘externalism’ in the philosophy of knowledge and the philosophy of mind equally has disguised this from people. It allows them to think of the mind extruding itself over the context (bringing the place within its ambit, as it were) without requiring any abilities on the part of the subject. So much the worse for it, and for the spatial metaphor which feeds it.³ How does this relate to the existence of conceptual entailments? Consider Evans’s example again. Most writers suppose that ‘de re’ knowledge is a different matter from ‘de dicto’ knowledge. Russell himself, Strawson, Kripke, Evans and many others have insisted on the distinction. But Jackson could use the same argument to ‘bridge the gap’. De re knowledge is a priori deducible from de dicto knowledge! Suppose, that having been shown how unobvious and tricky zips are, John asserts: (a) The chap that thought this up must have been brilliant. ² In other words, nothing in my position depends on either accepting or denying the idea that fine-grained Fregean senses can somehow be married to theories that primarily locate content by reference. ³ I cast further aspersions on it in Blackburn 2006.

34

simon blackburn

He can by Jackson’s reasoning then deduce a priori that (b) Julius was brilliant. For he has at his disposal the allegedly a priori (c) Julius is the chap who invented the zip. And by the same principle that if A & B a priori entail C, and B is a priori, then A by itself a priori entails C. Obviously, the oddity lies in supposing that (b) can both be a priori and can fill the demanding role of turning our position from that where (a) represents the sum of our knowledge of zip invention, to one where we know who did it. This is wrong: you cannot know that Julius invented the zip without knowing, of the person Julius, that he invented the zip. Once more, as Kripke says, names are scopeless. To know, of the person Julius, that he invented the zip requires much more than is given just by knowing that someone (uniquely) invented the zip. And what more it requires is not a priori, and not gained by choosing a descriptive name in a mock baptism without a person to be baptized. I suspect that all this is a little less obvious than it should be (apart from the fact that nothing about use and mention ever seems obvious) because Kripke chose the standard-setting example of the metre rod. ‘Let a metre be the length of that rod’, the committee says, and lo and behold, it seems as though they know that the rod is one metre long, and since it took no investigation or measurement, it sounds as though they must know it a priori. I suspect this case is plausible because it combines several different features which disguise its real nature. First, it is a case of prescription, rather than description. The subjects are saying what a metre is to be. Second, in our imaginations we place the committee in a scenario in which the rod is before them, and by some standards we know the length of things that are before us well enough. And thirdly, we imagine it ‘given’ that they have the authority to set the standard—but you could not know you had any such authority a priori. Let us consider a case where these features are different. Suppose the committee believes itself to be the committee determining the metre, when in fact it is not. Committees often make mistakes about the scope of their authority. They might be no more dispositive than I would be if I stood on the platform where the dignitary will later be, and announced that I name this ship ‘Seaspray’. I do not know a priori, or in any other way, that ‘Seaspray’ is the name of this ship, for since I have no authority to name it, it probably is not.

analysis, description, and the a priori

35

So suppose this committee has mistaken its brief. Their role is just advisory, the letters patent are a forgery, their decision may be overturned by Parliament, or whatever. Then the metre will very likely not be the length of that rod, and if it will not be that in the future, then it is not that now either. Their determination of the metre misfires, as Austin would have said. Second, imagine the case where the rod is hidden from them. Suppose now that the committee is indeed dispositive. And the King has a famous but hidden sword Excalibur, and to flatter him they say ‘let’s make the metre the length of Excalibur’ although they do not know whether that is (say) about the length of a grown man, or about the length of his forearm. Do they now know that the metre is the length of Excalibur, or (for it is equivalent) that the length of Excalibur is a metre? No: they still don’t know what the length of Excalibur is, by any standards. So they still don’t know what length the term ‘metre’ is going to designate, and in particular they don’t know that it is going to designate a metre. It might be nearer two metres, for all they know. Since they do not know what length the term ‘metre’ will designate they do not know of anything, including Excalibur, that it is a metre long—although given that they are as authoritative as they take themselves to be (a contingent and a posteriori matter) they may perhaps know that whatever length Excalibur turns out to be, will be called a ‘metre’. But at this point you cannot disquote, and give them any knowledge at all about metres, and which things they measure. The a priori status of Jackson’s bridge principles is therefore highly unreliable. But I now want to turn to a different and perhaps yet more troubling problem, illustrated by a metaphysically more heavyweight example.

IV Suppose I have got myself unhappy about McTaggart’s A-series, or in other words about tense. I can see no way in which physics makes room for a ‘now’, and hence for a future or a past. Like many thinkers, I believe that the universe which physics reveals is a ‘block’ universe, in which there is before and after, but no past and future.⁴ So I am happy about ordinary B-series determinations of time, including the tagging of times by dates. But it would be God’s good pleasure, an extra-physical determination, that would have been needed to make it so that 2004 is the present, now. Just think, it might have been that now ⁴ The best case for this is made by Price 1996.

36

simon blackburn

was 1500 bc, I ponder, and become very upset about the metaphysical extra, the surd, needed to make that not so.⁵ Jackson’s proposal then has this shape. We can use conceptual analysis to ‘locate’ tenses, even in the block universe, because tensed ascriptions are a priori deducible from propositions ascribing dates. If you know that the date of the battle of Waterloo is 1815, you can deduce a priori that it happened before now! Corresponding to the water/H2 O argument above we have: (a) The battle of Waterloo happened in 1815. (d) The battle of Waterloo happened before now. That sounds amazing—how does it work? True, the account continues, there is no a priori deducibility yet. Your mere English understanding of the sentence ‘it happened before now’ does not guarantee that you can perform this deduction. In Kaplan/Strawson terms mere understanding of its character is sufficient for this ‘linguistic’ understanding. In this sense, Shakespeare understood the sentence just as you do, yet if he performed the deduction he would get from a true premiss to a false conclusion. But wait. You only know which proposition ‘it happened before now’ expresses if you know both character and context for the indexical ‘now’. Knowing context here means knowing the date, and presto, if you know that you can indeed deduce, a priori, that the battle happened before now. It is as if the sentence ‘It happened before now’, said now, expresses the proposition ‘it happened in 2004’ and we can deduce a priori that any event happening in 1815 happened some 189 years earlier. That is, (a), together with something like: (b) Now is the time that plays the role of being the present, and (c) 2004 is the time that plays the role of being the present, give us premisses from which to deduce (d). And, Jackson may continue, (b) is a priori, and (c) is simply empirical, established with a calendar or clock. Why would a B-theorist feel short-changed by this? He could express his dissatisfaction a number of ways. He might say that the metaphysical problem he had was how there could be anything, within the block universe, to play ‘the role of being the present’. He does not see how times, one after another, succeed to and then lose that role. He thinks that times ought to be static, as it were, with no role attaching to one date that does not attach to another, for that is how physics sees it. Thinking like this, it is (b) that conceals the whole mystery. Or, he might focus on (c). It is not that (c) is false, or cannot be ⁵ This is not intended to represent the mindset of actual B-theorists. Nor am I endorsing it! But it is a recognizable mindset, and I am only using it to make the point.

analysis, description, and the a priori

37

established by using a clock. But he does not see how (c) is possible, how there can exist the kind of truth of which (c) is an example. Either way, if we could not locate tense in the block world before, we cannot do it after this argument is added. All these examples—the tense example, the location example (‘I am here’) and the Julius example—show that a prioricity is considerably more elusive than Jackson realizes. It is no part of my claim that names of stuff behave exactly like indexicals. Remember that if Shakespeare said ‘I am here’, when he was standing on a particular spot, such as the edge of the Ben Nevis plateau, we can only say that he knew he was there if he actually knew where he was. Otherwise he merely knew he spoke a truth, but not knowing where he was, did not know which truth it was. But do we really want to say, for instance, that when Shakespeare said that we write men’s evil deeds in brass but their virtues in water, he didn’t really know which truth he was expressing? This would interpret him as knowing only the character of what he said, but not its content, since he did not know he was in an H2 O world as opposed to an XYZ world. I take it that we could think that. After all, there must be at least a sense in which it is far from a priori that water is the stuff that plays the watery role. An eighteenth-century chemist coming across H2 O that he had just synthesized from burning hydrogen might well not recognize it, and need convincing that it ever occurred outside the laboratory. He needs convincing that this stuff—water—plays the watery role round here. The convincing will be an empirical process. And unless they have hydrogen and oxygen and have done the synthesis, nobody on twin earth knows or is in a position to learn that water plays the watery role anywhere at all. They couldn’t work that out a priori. The stuff that splashes round them is not water, but XYZ, and all they know because of that is that XYZ, not water, plays the watery role. A priori knowledge should be shareable between us and them (it is, as it were, the gift of reason which we share alike), in which case the only a priori knowledge they or we have is that whatever plays the watery role plays the watery role. However, I freely concede that we usually allow Shakespeare to have meant, and know he meant, water by ‘water’. I believe this should diminish our enthusiasm for assimilating stuff terms to rigid names or indexicals. But that, once more, is not a fish I want to fry here. How does all this affect either psychology or ethics? It is bad, surely, if Jackson’s strategy needs it to be true that we do not know which proposition someone is asserting when they tell us they are in pain, until we have an actual physical ‘realizing’ state for pain. It would follow that we do not know what

38

simon blackburn

someone believes or fears if she believes or fears that she will be in pain, until science does its work. It is bad that we do not know which proposition the moralist would have us believe until we know her standards, or perhaps her mature community standards. We do not know what someone would have us think, if they tell us that rape is bad, until we know whether their mature folk morality is utilitarian or something else. These are surprising claims, and each tramples on very robust intuitions. Jackson may try to let Shakespeare off the hook by saying that ‘meanings ain’t in the head’, or in other words being sufficiently ‘externalist’ about the concepts that enter into our descriptions of what people know. So it might be suggested that in natural kind cases merely being in a certain context (a world in which the watery stuff around us is H2 O) even when we do not know that context, suffices to know propositions expressed in terms of ‘water’—the propositions, that is, whose identity is only given by the chemistry. That is not so in epistemic and doxastic contexts in general, or you could swap ‘water’ and ‘H2 O’ at will, leading to the surprising historical revision that simply by knowing that we write mens’ virtues in water, Shakespeare beat Lavoisier to it.

V In the tense case, we discovered that the imagined solution to the problem of placing tense in a block universe required a fact of us. It required us to be in a context in which, when we talk of ‘now’, we refer to one date rather than another. But a way of putting the A-theorist’s puzzle is that he doesn’t see how there could be such a context. It is precisely that kind of fact that eludes his metaphysics. Just announcing it seems to him to be announcing the very mystery he wanted to dispel. And the same problem awaits any attempt to apply the machinery to the philosophy of mind. Thus suppose we protect the a priori deducibility of the mental from the physical in the same way. In a context in which pain is C-fibres firing, the character + context account gives us that we can know that someone is in pain only if we know that his C-fibres are firing, and if we know that his C-fibres are firing we can deduce, a priori, that he is in pain, since that is what the proposition amounts to, in this context. But the mysterian, for instance, will be entirely unimpressed, since for him the metaphysical miracle, God’s good pleasure, lay behind the very idea that you can have a context in which pain is C-fibres firing. The place of the contingent and empirical discoveries in Jackson’s story distinguishes Jackson from a Leibnizian rationalist, for whom deducibility

analysis, description, and the a priori

39

would be a priori in the much stronger sense, namely one in which our understandings would not require empirical identities in order to operate. As we put it before, the Leibnizian a priori would be shared with those on twin earth. But what I have suggested is that the additional ingredient is the very one that makes the ‘placing’ problem so difficult. If it was hard to place tense in the static block world of physics, then this is not a solution, for it conceals the suspect fact in its very set-up. And I fear it is not a solution in the mental case either.

VI I shall pass over the treatment of colours, to concentrate upon that of normative terms, to which the same machinery is applied. Again, it is important not to misunderstand what is achieved by a priori deducibility. In the case of ethics it might seem like a defence of rationalism against sentimentalism: Wollaston against Hutcheson, or Kant against Hume. But Jackson intends nothing of the sort. The ‘descriptive properties’ which will give us the reduction are located as the standards that would be applied in ‘mature folk morality’. There is no proof that there is only one such morality, on which all folk must converge. And more importantly there is no reason to deny that it is their sentiments rather than their reason that lead the folk to their morality, or indeed to any morality at all. Hutcheson and Hume can sleep easy. So suppose we set out to perform an analytic deduction of ethics from the descriptive: (a) Genghis wishes harm to his neighbours. (d) Genghis is bad. We need an a priori principle: (b) Badness is what plays the role of (deserving hate, exciting disgust ...), and a bridge: (c) wishing harm to one’s neighbours plays the role of (deserving hate, exciting disgust ...) And now we have two premisses (a) and (c), which with a priori (b) in the background, together deductively entail the moral (d). In the light of what I have said, it should be clear that each of (b) and (c) needs a lot of attention. If (c) is a purely descriptive claim about the folk,

40

simon blackburn

including our folk, then we cannot be at all sure of (b)—not without endorsing the folk in question. It will be an open question whether what these people dislike indeed deserves their dislike. We can close off the open question by reading (c) as having a concealed ‘rightly’ in it, so one is asserting that wishing harm to one’s neighbours rightly deserves hate and disgust. But then the moral nature of the premiss is obvious. This is not the place to rehearse all that an expressivist, or for that matter a more robust rationalist, might say in response to Jackson’s moral functionalism. I will simply record that from my point of view, like Cornell realism it concentrates too heavily on reference and not enough upon sense. In other words, it sees the problem of normativity as one that is in essence solved if we can identify normative properties with descriptive properties, so that ‘any ethical way of carving up how things are is equivalent to some descriptive way of carving up how things are’ (Jackson 1998: 139). This bypasses the question of what is distinctive about those people who are operating the ‘ethical way of carving up how things are’: it ignores their special take on the situation, or in other words forgets to ask how valuing is different from describing. Of course, I share with Jackson the basic metaphysics. Ours is a physical world, which has given rise to animals, and those animals that we are wield norms and insist on conduct and lament the actions of their fellows. And in some deep sense, this is all there is. But it does not follow that all mental activity, including these ethical activities, is a matter of describing how things stand, and there is a serious problem about the relation between the results of such description, and the more active, prescriptive, states of mind that at least on the face of it have the opposite direction of fit to the world. I do not think that the a priori is adapted to crossing that gap. It takes the passions to do it. But the passions will not do it by becoming just more of the things we describe. This is the moral of Moore’s argument. We can describe the likes and dislikes of people forever, but unless we share them, we will not make the passage to sharing the valuations that they fund. I have tried to sketch what seems to me the essence of Jackson’s approach to metaphysics and ethics, and to indicate some of the places where it is surprising and, inevitably, controversial. I have not been able to highlight the incidental details, of which there are many, and which give his writings their extraordinary depth and quality. If I would not have everybody convinced by what they read, I would still have them read it, and they will learn a great deal, as I have. Cambridge University

analysis, description, and the a priori

41

References Blackburn, S. 1984. Spreading the Word (Oxford: Oxford University Press). 2006. ‘George Berkeley and Julius Caesar play leapfrog’. In C. Macdonald & G. Macdonald (eds.), McDowell and his Critics (Malden, MA: Blackwell). Jackson, F. 1998. From Metaphysics to Ethics (Oxford: Oxford University Press). 2003. ‘From H2 O to water’. In H. Lillehammer & Rodriguez-Pereyra G. (eds.), Real Metaphysics (London: Routledge). Price, H. 1996. Time’s Arrow and Archimedes’ Point (New York: Oxford University Press).

This page intentionally left blank

2 Physicalism, Conceptual Analysis, and Acts of Faith JENNIFER HORNSBY

Not very long after the publication of From Metaphysics to Ethics¹, I spoke about the role of conceptual analysis in the vindication of physicalism at the ANU’s Philosophy Thursday Seminar Series. The Thursday seminar being on a Monday, and Frank Jackson at that time being Director of the Institute, he had to go out of his way to be present. This he generously did. It was predictable not only that Frank would disagree with almost everything I said but also that he would respond in the gracious and urbane way that is Frank’s alone. Frank helped me to see that I had allowed the argument between us to turn on matters of relative detail. The invitation to contribute to a volume in his honour gives me an opportunity to step back. Frank Jackson and I each take the other to hold a position in philosophy of mind that it is extremely difficult to sustain. Here I shall confine myself to trying to say something about how that can be.² My objective is more to demonstrate the sanity of Jackson’s opponents and the fragility of his own position than to hold out for the truth of any particular doctrine. Chiefly I want to bring to the surface an assumption in ontology, which I see as a crucial part of the grounding of Jackson’s particular version of physicalism. Once it is appreciated that this assumption is contestable, Jackson’s opponents may be seen in a different light from the one in which they appear in his ¹ Jackson (1998). Otherwise unattributed page references here will be to this. ² Jackson’s recent work has provoked a vast amount of debate about species of necessity. But I want to avoid engaging in that particular debate here. Another issue set to one side here is how physicalism might be affected by considerations about consciousness—specifically about ‘qualia’ which have been argued to be ‘epiphenomenal’. This is an issue that Jackson did not address in (1998), assuming there (as I am more than happy to assume here) that such considerations make no special problem for physicalism.

44

jennifer hornsby

writings. More generally, a connection will appear between the vast literature on physicalism as a topic in philosophy of mind and the equally vast literature on material constitution as a topic in metaphysics. I should start by setting up the difference of opinion in broad outline. I can use Jackson’s own words to characterize both what he and opponents may agree about and what they certainly disagree about. What (at least for present purposes) everyone may agree about is (P).³ (P) Any world which is a minimal physical duplicate of our world is a psychological duplicate of our world. (p.14) The thing that we anti-reductionists dissent from is that Everyday psychological explanations are understandable in terms of, or are reducible to, physical ones. (2000, p.188) The assumption which I want to suggest separates physicalists of Jackson’s stripe from many of those who reject this reductionist thesis is, as I said, an assumption in ontology. We shall see that it has opponents in the final section below (Section 4). I start by drawing attention to what Jackson takes (P) to import, which, thanks to his making the assumption, goes well beyond what I think (P) actually implies. Then I look at the claims about conceptual analysis which Jackson uses to secure the commitments he takes (P) to incur (Section 2). Jackson says that those of us who aren’t willing to undertake these commitments have to engage in ‘an act of faith’. But I shall suggest that physicalists who reject his reductionism can make a virtue of faithlessness as much as anyone else (Section 3).

1. Physicalism One could quibble about (P). But my main quarrel is with what Jackson hopes to get out of it, so that I want to bypass quibbles and to work as if (P) could do duty for a doctrine which is likely to appeal to anyone who doesn’t espouse any substantial dualism. Call this doctrine, which (P) encapsulates, physicalism. In order that it should have a definite content, we need to be explicit about what ‘physical’ and ‘psychological’ mean in (P). By ‘psychological’ is meant what is variously called everyday or commonsense or folk psychological. ‘Physical’, ³ The significance of ‘minimal’ in (P) need not detain us now; but see n. 25 and §4 below. And see n. 26 for the explanation of my ‘(for present purposes)’ qualification.

physicalism, conceptual analysis, and acts of faith

45

meanwhile, as Jackson says, must have a somewhat specialized sense; and I shall speak of physical science to signal this.⁴ Consider someone a (Anna, as it might be), and suppose that some everyday psychological property ψ is correctly attributed to her (it might be that Anna believes that John is now in Norway). According to (P), a’s not being ψ would involve some difference in the physical-scientific facts. So where  is the hugely complicated conjunction saying in the terms of physical science how things in the domain of physical science are at some time, we have:  → ψa⁵ (P) is a thesis of global supervenience inasmuch as it admits that objects other than x may play a part in fixing what psychological properties x possesses. In the case at hand, there are two points which explain why one might want to endorse only supervenience theses that are global in this sense.⁶ First, well-known anti-individualist arguments demonstrate that how things in our environment are contributes to determining the contents of our psychological states: facts about people’s surroundings may be relevant to whether anyone is ψ.⁷ Secondly, it may be doubted that the things to which everyday psychological predicates apply—people, as I shall say⁸—are the same as the objects which physical science describes from moment to moment. These two points may be thought of as relating respectively to the predications of everyday psychology—to its ideology—and to the things of which it makes its predications—to its ontology. A global supervenience thesis can accommodate both points inasmuch as it allows that facts about other particulars than those mentioned in  can be determined by . ⁴ Jackson gives three accounts of what might be meant by ‘physical’ in (P) which all advert to science. He says that it doesn’t matter which we pick (pp. 6–8). ⁵ The ‘→’ of this conditional corresponds to what Jackson calls ‘entailment’: he says that ‘any putative psychological fact [such as our fact that ψa] has its place in the physicalists’ world view only if it is entailed by ’. Some people might not want to use ‘entail’ at this point, and I hope that we will soon be able to understand their reservations. I shan’t make an issue of the word, however, because, as I see it, their reservations turn on more than a merely terminological matter. ⁶ See Haugeland (1982) and (1984) and Petrie (1987) on what makes global supervenience attractive. ⁷ At n. 11 to ch.1 (pp. 8, 10), Jackson notes that there are ‘issues about the role of a subject’s environment in settling intentional content and psychological nature’—issues which he says don’t have to be settled if our focus is ‘physicalism in the wider sense that concerns us’. ⁸ One does not need to choose a single term for the things of which everyday psychological predications are made unless one thinks (as Jackson does—see further below) that a network of specific concepts constitutes everyday psychology. Obliged to choose a single term for the sake of argument, I use ‘person’ with the plural ‘people’. But I acknowledge that things that we would not call people might possess some (or perhaps in a view like Jackson’s very many) everyday psychological properties: see further n. 19 below.

46

jennifer hornsby

Jackson’s formulations suggest that he resists the second point and takes it for granted that physical science and everyday psychology share an ontology. He often speaks of ‘how things are in one vocabulary being made true by how things are as told in some other vocabulary’. But a question then must be ‘Which things’? Call the things in physical science’s domain φ-things (so that  is an account of the φ-things) and the things in everyday psychology’s domain -things (so that the fact about a—that she is ψ —is a fact about one of the -things). Given (P), variation in how any -thing is requires variation in how one or more φ-things are: if a weren’t ψ, for instance, then some physical thing would be different from how it actually is. But this does not mean that there is any vocabulary in which it could be told how things in general are—how the φ-things are and how the -things are. It is compatible with (P) that our world, in which there are fundamental particles of physical science and people (and other things besides, no doubt), should not be characterized exhaustively in vocabulary suited to saying everything there is to say about the φ-things—about the particulars favoured by physical science. Thus physicalists may take issue with Jackson when he writes: A complete account of what our world is like ... can in principle be told in terms of a relatively small set of favoured particulars, properties, and relations—the physical ones. (p. 6)

The response to this now, on the part of some anti-reductionists, will be that a complete account of what our world is like must tell one all the facts about people, as well as the facts about the φ-things. And the facts about people cannot be told in the favoured vocabulary of physical science even if variation in those facts ensures variation at the level of physical science. Presumably Jackson will counter by saying that people can be identified in a vocabulary suitable for describing physical science’s favoured particulars, so that the facts about people just are some of the facts about fundamental physical particles. But the present point is that if Jackson makes this claim in ontology, then he is going beyond the thesis (P). It is compatible with (P) that a specification in the language of physical science of a duplicate of our world should say nothing about any person. This is not to deny that physical scientists’ vocabulary has the resources to speak about things the size of people. (One might be inclined to think of physical scientists as speaking of very tiny things. But the language of physical science can include mereological words like ‘part’ and ‘sum’ without losing its pretensions to be scientific, so that its terms can denote arbitrarily large things.) What a physicalist, who assents only to (P), can perfectly well deny is that a person is the same as any sum of things that

physicalism, conceptual analysis, and acts of faith

47

physical science is concerned with. Indeed an anti-reductionist physicalist is likely to hold that one could never identify a person by identifying such a sum. She may think that people do not belong in the domain of physical science.⁹ Well, Jackson would surely say that anyone who resists his claim in ontology is not a ‘serious metaphysician’. For Jackson tells us that the idea ‘that a full inventory of the instantiated physical properties and relations would be a full inventory simpliciter’ is part of ‘physicalists’ distinctive doctrine’ (p. 9). And he tells us this after he has outlined his project of serious metaphysics but before he invokes (P) as the way to capture physicalism. Still, given that someone can be committed to (P) without endorsing ‘the distinctive doctrine’, the question has to be why Jackson should think that someone is less than serious if they settle for (P) but reject his claim in ontology. Jackson says that ‘metaphysics is about what there is and what it is like’ (p. 4). And it must be that he thinks that no serious physicalist can countenance among ‘what there is’ anything except the φ-things. Certainly it seems right to think that a physicalist will subscribe to an ontological doctrine (about what there is), as well as to an ideological one (about what the things there are like). Physicalists of any persuasion will rule it out that people are psychologically, as they are in virtue of their mental life, being sustained by immaterial stuff. Those who take Jackson’s side might claim that in order for this really to be ruled out, what there is in our world, the particulars there, must be restricted to the φ-things.¹⁰ But I would disagree. We can find an answer to their claim by looking now at Jackson’s own distinction between two senses in which properties may fail to be physical. On the one hand, there are the (putative) properties whose instantiation is inconsistent with physicalism—anti-physical properties, one might say: the property of being made of ectoplasm would be an example. On the other hand, there are properties which, although not themselves physical, need not be thought of as anti-physical, at least not by physicalists. Jackson calls these latter properties ‘onlookers’ in the debate about physicalism (p. 16). Everyday psychological properties are examples of onlooker properties. Now a parallel distinction may be made in respect of particulars, between two senses in which they may fail to be physical. On the one hand, there are those (putative) particulars whose existence is inconsistent with physicalism—anti-physical ⁹ The anti-reductionist need be no more precise than this about what she denies: it is for the reductionist to say exactly how an ontology of people is supposed to be captured in the language of the φ-things. See further n. 28 below. ¹⁰ I speak of those who take Jackson’s side, rather than of Jackson, at this point because Jackson himself wants to rule this out in a different way: see n. 25 below.

48

jennifer hornsby

things, one might say: anything made of ectoplasm would be an example. On the other hand, there are those particulars that physicalists have no need to think of as anti-physical. And just as physicalists accept the instantiation of what Jackson calls onlooker properties, so they may accept the existence of things that are not φ-things. A, for instance, might not be identifiable in vocabulary suitable for describing physical science’s favoured particulars without being an anti-physical thing. It is easy now to see that physicalists can take the existence of particulars that are not among the φ-things in their stride. Even if they deny that everything can be picked out using the language of physical science, they can still assert that everything is physical according to one quite ordinary conception of this. This ordinary conception requires no more of something than that it not be (or be composed of) any anti-physical thing (or things). Any person is made up exclusively of physical matter, so that many of their parts, and certainly all of their tiniest parts, instantiate only physical science’s favoured properties. This fact about the material constitution of any person suffices for people to be physical in an intuitive sense. And this in turn is at least enough to ensure that people need not be thought to be anti-physical just because they are refused membership in the domain of physical science. Physicalism, then, is no more contravened when people are mentioned in saying how things are than when onlooker properties are applied in saying how things are. The facts of everyday psychology can supervene on facts of physical science even if none of the facts of physical science is a fact about any person. For present purposes, the assumption that Jackson brings to his understanding of (P) is the principal bone of contention.¹¹ Uncovering this assumption helps to explain why Jackson should so often talk of ‘the psychological nature of our world’ as ‘entailed by the physical account of our world’. And it highlights one very fundamental point of dispute. Anti-reductionists, who say that everyday psychological explanations cannot be reduced to physical scientists’ explanations, may insist upon a distinction between things in whose nature it is to have everyday psychological properties and things whose nature is the concern of physical scientists. That distinction can be part of a general distinction between material things and their matter. (A familiar example, on which I say more below: a statue is not the same as the clay from which it is ¹¹ Not that it is the only assumption of Jackson’s that physicalists may want to resist. Antireductionists find supervenience theses appealing because (we say) holding them is consistent with holding that an understanding of supervening terms could not be conveyed in the vocabulary of the level supervened upon. So anti-reductionists’ quarrel with Jackson cannot be confined to the particular assumption that he brings to a claim of global supervenience (even if it suits me so to confine it for the purposes of the present paper).

physicalism, conceptual analysis, and acts of faith

49

made.¹²) Henceforth I shall take anti-reductionist physicalists to deny and reductionist physicalists to accept, that everything (temporal¹³) belongs in the domain of physical science.¹⁴

2. Conceptual Analysis Jackson assumes that all of (P)’s adherents think that , which is cast exclusively in the language of physical science, entails every other truth. Thus in Jackson’s book, physicalism brings in its train many ‘entailment theses between matters described in the vocabulary of physical science and matters described in other vocabularies’ (p. 28). The job of securing these entailment theses is the job of conceptual analysis—of using one’s intuitions to determine what we would say about possible cases. When it comes to everyday psychology, Jackson treats its concepts as comprising a network, the identity of each being secured by its place within the network. ‘The network itself’, Jackson says, ‘is the theory known as folk psychology, a theory we have a partly tacit and partly explicit grasp of’ (p. 130). His idea is that, by using conceptual analysis, one could tease out the patterns that implicitly guide us when we apply psychological concepts, and thus reach a theory containing a host of truths ‘that could only come false by virtue of meaning change’ (p. 45). This theory in its turn could be used to provide an implicit definition of all our psychological terms, the definition being given in non-psychological vocabulary. Here we have the doctrine that Jackson has defended for many years, to which he gave the name ‘analytic functionalism’.¹⁵ ¹² Familiar because it was an example used early on to illustrate the non-identity of a thing with that which materially constitutes it: see e.g. Wiggins (1967) and Perry (1970). ¹³ I qualify with ‘temporal’ in order to set aside issues about numbers and the like. There may be disagreements about how best to set them aside. But since Jackson sets them aside himself (p. 8), it is safe to assume that believing in numbers etc. is not enough to make one deny the assumption which I want to treat as the bone of contention. ¹⁴ I acknowledge that there are anti-reductionist physicalists who do not fit within this scheme of things. Davidson’s anomalous monism, which is the best-known version of anti-reductionist physicalism, addresses an ontology of mental events, and it may incorporate a local supervenience thesis to which mental/physical identities—sc. between events—are presupposed. (I comment on this in n. 34 below.) Note that my anti-reductionist denies that people are identifiable in vocabulary suitable for describing physical science’s favoured particulars, whatever these are. And reductionist physicalists may want to claim that events (or states: see n. 22) are what any person is a sum or fusion of. My focus here, however, is on a reductionist physicalism which takes material things (not in the category of events or states) to be what a person is composed from, since this indeed proves to be Jackson’s view: see §4 below. ¹⁵ See e.g. Jackson and Pettit (1990).

50

jennifer hornsby

Some people do not share Jackson’s optimism about the possible achievements of conceptual analysis in this area. They have concentrated on objections of two sorts. I shall rehearse these now, wanting then to connect them with the larger project of physicalism and with the present bone of contention. The first objection has us notice the paucity of candidates for analytic truths that employ everyday psychological language. It is all very well for Jackson to tell us that ‘those good enough at theory construction could extract and articulate the patterns that guide us in classifying the various possible cases’. The trouble is that those who have attempted to state some relevant analytic truths have very soon found themselves very stuck.¹⁶ The second objection is that even such analytic truths as might be accepted lend no immediate support to entailment theses of the kind Jackson believes in. To understand this objection, it helps to remind oneself what a rich and subtle collection of concepts figure in the psychological understanding of people. To mention just a few candidates (out of hundreds, presumably), we have: babble, baptize, believe, belligerent, bemoan, bisexual. If we consider the sorts of claims containing concepts such as these, which we might be willing to accept on the basis of engaging in conceptual analysis, then the non-psychological terms within them will certainly not be those of physical science. For the properties that might be connected with everyday psychological ones as a part of some a priori philosophical endeavour are surely those in terms of which people think of one another and of the other things around them. These are (roughly) macrophysical properties, rather than the physical scientists’ properties which interest Jackson.¹⁷ Well, presumably Jackson envisages a two-stage process: everyday psychological concepts are to be seen as caught up with the everyday macrophysical ones, and the everyday macrophysical ones in turn are to be subjected to analysis to connect them with the concepts of physical science. This means that we need to believe in very many more analytic entailment theses than those that were supposed to be uncovered in the first instance. Again we may feel simply stuck. If questions are raised about which categories—macrophysical? psychological? scientific physical?—concepts belong in, then one starts to wonder how the line is supposed to be drawn between the everyday psychological ones ¹⁶ See e.g. Schiffer (1987) ch. 2, §2.4. ¹⁷ Functionalists say that when mental states are defined in the context of a network of their interrelations and their relations to inputs and outputs, inputs and outputs have to be ‘describable in non mental terms’. ( This is said to be ‘crucial to [the] enterprise’ in Braddon-Mitchell and Jackson (1996) p. 53.) The present point is that when analytic functionalism is a component of a physicalism characterized by reference to physical science, inputs and outputs have to be describable, or at least come to be shown to be describable, in the terms of physical science.

physicalism, conceptual analysis, and acts of faith

51

and the rest. (When the doctrine of analytic functionalism first had its adherents, the question of where this line comes may not have seemed pressing. The project then was to deal with a specific group of what can appear to be particularly problematic everyday psychological concepts, sometimes said to figure in ‘the belief-desire theory’. But when analytic functionalism takes its place as part of a larger metaphysical enterprise of establishing an overall physicalist doctrine, the question what does and what does not count as comprising everyday psychology can seem more pressing.¹⁸) If you consider our short list of six (babble, baptize, etc.), then you may think that some of them are borderline, or at least that it isn’t clear whether they are everyday psychological or not. But however that may be (however exactly the line is supposed to be drawn), the analytic functionalist relies upon there being a distinctive group of interdependent concepts, interdependencies between which are to be uncovered by armchair reflection. What makes the group distinctive and holds the theory together is surely the concepts having application to things of a certain kind (to people, as we have been saying¹⁹). Thus even if it is not clear exactly which concepts have the status of the everyday psychological, it can be clear that the obvious candidates for such status are some of those we actually apply to people. (People sometimes babble; people usually believe many things; people may be belligerent, and so on.) When Jackson speaks of a ‘sentence which tells us something about the psychological nature of our world’, he makes it sound as if psychological terms might latch onto nothing in particular (just to ‘the world’). And when he and other analytic functionalists gesture towards truths of the sort that they think belong in folk psychology, they are apt to say such things as ‘it is platitudinous that sharp pain typically causes wincing’; and here it seems as if nothing need exist of which ‘feels pain’ and ‘winces’ may be true. Presumably, though, ‘a sentence about psychological nature’ says something about the nature of some individual person (or, as it might be, of several, or many, or all, individual people). And presumably the reason why any putative platitude of folk psychology recommends itself is that it appears ¹⁸ Though it is not a question Jackson himself attempts to answer in (1998): his own examples of everyday psychological concepts there are ‘believe’ and ‘desire’ (p. 63). ¹⁹ One natural way to distinguish folk psychology (for those who accept that it is a theory) is to think of it as a theory whose application is to the very things that use it. If this was right, then we should use ‘member of the folk’ to denote things of the kind in question. Some will want to allow things other than people into the domain of folk psychology. Others may want to tell a more complicated story in which some everyday psychological concepts have a wider range of application than others (see n. 8 above). This would make the ontological question that I insist upon raising more tangled, but it would not make it go away.

52

jennifer hornsby

to record some generalization about folk psychological subjects: any folk psychological subject x, is such that x’s feeling sharp pain causes x to wince, or whatever. Simplifying drastically, and ignoring relativization to times, we can say that the form that the theory that analytic functionalists call folk psychology needs to assume is ‘T[z, ψ1, ψ2, .., ψn ]’, where ‘z’ is a variable whose range is folk psychological subjects.²⁰ So even when the theory is ‘ramsified’²¹ in order to ‘define’ its psychological terms ψ1, ψ2, .., ψn , people will not be defined out of existence. This may become more obvious when one thinks of particular everyday psychological terms. Imagine that one characterizes someone as belligerent in order to explain a pugnacious reaction on her part. One then takes this person’s reaction to be explained by her possessing a particular property. If the property is thought of along functionalists’ lines, it will be treated as ‘a state that plays a causal role’. But an ontology of people is still not dispensed with. For the state is then ‘a state in her’—or so the functionalists say.²² (And presumably a state of belligerence in her directly causes her pugnacious reaction rather than anyone else’s.) Even if folk psychology could be spelled out in a theory, then, and even if that theory could be used in some definition of its theoretical terms, no ontological elimination would be effected.

3. Acts of Faith A debate about what conceptual analysis affords can reach a standoff. Jackson’s opponents think the paucity of known candidates for analytic truths that employ everyday psychological terms argues for an actual absence of many such truths. Jackson, on the other hand, thinks that we can know our concepts to be amenable to analysis without actually being able to provide analyses: abortive analytical endeavours only expose philosophers’ inability explicitly to articulate what they tacitly know. It was in order that we should be able to get ²⁰ See Loar (1981), ch. 1 on this. ²¹ I follow others in using ‘ramsification’ as the name of the method of David Lewis (1970) which draws on work of Ramsey. See pp. 140–1 for Jackson’s employment of it in respect of ‘mature folk morality’, on which I comment in n. 29 below. ²² I don’t have to quarrel with this talk of states here, since my anti-reductionist takes no definite stand on what things they are whose existence we may be committed to in making everyday psychological predications (except inasmuch as I think that we are committed to the entities of which such predications are made): see n. 14 above. So my objection here is not to treating what Lewis (1970) calls T-terms as names. My claim is that in so far as they are names, they name states of things of some sort.

physicalism, conceptual analysis, and acts of faith

53

past this standoff that I located Jackson’s assumption in ontology as one source of his belief that conceptual analysis is guaranteed in principle to yield certain results. When this assumption is allowed to be a real bone of contention, the whole character of the dispute between Jackson and anti-reductionists is affected. The first thing to notice is that Jackson’s defence of conceptual analysis is in one way indirect.²³ He says that IF his physicalism (i.e. reductionistphysicalism) is true, then the prospects for conceptual analysis must be bright. As he puts it: if serious metaphysics ‘is committed to an account of how things are in one vocabulary being made true by how things are as told in some other vocabulary, then it had better have to hand an account of how accounts in the two vocabularies are interconnected’ (p. 29). But anyone who thinks that the vocabularies of everyday psychology and physical science are not interconnected would wish to contrapose. No conceptual truths link everyday psychology with science, we think. So physicalism of Jackson’s sort is false. Jackson’s conviction that conceptual analysis can deliver entailment theses comes to look like an act of faith at this point. Of course Jackson sees things differently. From his point of view, it is anti-reductionists who make an act of faith. Physicalists who are not eliminativists about intentional states have to say something about how the physical story about our world makes true the intentional story about it. Otherwise their realism about intentional states will be more an act of faith than anything else. For they will have nothing to say to one who insists that their view that a complete account of the nature of our world can be given in purely physical terms without recourse to intentional vocabulary is precisely the view that there are no intentional states. (pp. 29–30)

I believe that Jackson’s picture of the dialectical situation results from foisting his ontological assumption onto all physicalists. So I think that we can turn the tables on Jackson now. (1) Eliminativism is actually a very serious threat for his own position. (2) Anti-reductionist physicalists do not take their realism on faith. (3) Nor do they take physicalism on faith. I shall argue these things in turn, before concluding with a brief look at why Jackson’s ontological assumption may be contested even when questions about physicalism are not at issue (Section 4). ²³ Jackson has other routes to defending conceptual analysis. (a) He has things to say to opponents of the possibility of conceptual analysis generally (not only to opponents of the idea that physicalism needs conceptual analysis for its defence): see his ch. 4. (b) He believes in an account of necessity which would ensure that there are plenty of a priori claims such as conceptual analysis is supposed to yield. (On his account of necessity, conceptual analysis comes to embrace the elucidation of A-intensions—a subject I don’t touch on here: see n. 2 above.)

54

jennifer hornsby

(1) Suppose that some concept resists containment within Jackson’s analytic physicalist project—that it does not occur in enough conceptual truths to provide entailment theses of the needed sort. The concept in that case fails to live up to the standards of those which, in Jackson’s view, belong in a ‘description of reality’. By his lights, it should be dispensed with. If concepts in some area suffer this fate, then Jackson is set on a path to eliminativism about the area. When it comes to everyday psychology, the eliminativism takes a particularly serious form. Jackson says ‘our classifications of things into categories ... is not done at random and is not a miracle’ (p. 64). In so far as a concept is not amenable to his analytic treatment, then, Jackson will be forced to conclude that we give a perfect appearance of going in for classification using that concept even while we actually proceed at random or miraculously. This is worrying. The worry could seem to be a relatively local one: if conceptual analysis failed us only here or there, then it would only be in particular areas that we unknowingly proceeded at random. But it is much worse than this. For to possess any concept is to be in an everyday psychological state. So a black mark against any concept is a black mark in turn for everyday psychology. But everyday psychological concepts were supposed to comprise a network, so that a threat in one place would, at least potentially, be a threat to the whole. Jackson’s entailment theses now might be thought of as potentially constituting a huge edifice, based on endorsing realism about everyday psychology, but ready to topple if, at any point, it cannot be supported by deliverances of the method Jackson favours. Of course Jackson is convinced that conceptual analysis is capable of turning up the goods; but his conviction, as we saw, is bound to strike his opponents as an act of faith. (2) We saw that it is not the view of every physicalist that ‘a complete account of the nature of our world can be given in purely physical terms’. (This is the view only of those of (P)’s proponents who allow themselves to be boxed into a corner by Jackson when he treats it as a foregone conclusion that everything is a φ-thing.) Certainly purely physical terms can be used to give a complete account of the φ-things—of the particulars recognized by physical scientists. But a complete account of our world must record the facts about people. And it is only to be expected that a complete account of the φ-things will not contain all the facts about people if, as the anti-reductionist physicalist says, people are not themselves φ-things. In one sense, however, Jackson is right that anti-reductionists ‘have nothing to say to justify calling themselves realists rather than eliminativists’. For antireductionist physicalists do not accept that when it comes to claims about

physicalism, conceptual analysis, and acts of faith

55

the reality of anything, the burden of proof is on serious metaphysicians. Thinking as they do that there are people, they take it for granted that people think things, and want things, and so on—just as anyone who ever attributes an intentional state of mind to herself or to another takes this for granted. And whatever might be said for or against such realism, Jackson himself is obliged to assume it when he engages in the philosophical project of proving it. For Jackson’s entailment theses, which are supposed to reveal our concepts as in good physicalist standing, are meant to record things that we all (perhaps only tacitly) know. Unless it is assumed, then, that we are psychological subjects, who, among other things, know things, Jackson’s project cannot get started. If the anti-reductionist physicalist really did require an act of faith, then that act would be required of the reductionist physicalist too. (3) Whether or not realism about everyday psychology is assured, it may be thought an objection to the anti-reductionist that she takes her physicalism on faith. How else is it to be ruled out that the world contains what I earlier called anti-physical things? Well, the easy answer is that, a hundred and fifty years on from Darwin, it is a reasonable view that nothing more than natural processes in a world of regular material stuff has ensured the presence here of beings with everyday psychological properties. The alternative can nowadays seem simply unmotivated. Of course arguments for Cartesian dualism need to be answered.²⁴ But I suggest that in the twenty-first century, rather little needs to be said in favour of a rather unexciting sort of physicalism. It is a question why Jackson himself should be confident that when the conceptual analysts and the physical scientists had done their work, the actual world would be revealed as containing only material stuff. This question arises for Jackson because he endorses only a contingently true physicalism, thinking that there are possible worlds which are ‘physically exactly like ours but which contain as an addition a lot of mental life sustained in non-material stuff’ (p. 12).²⁵ If he were to tell us why he is willing to affirm that actually persons are made only of material stuff, then perhaps he too would need to appeal to the rather obvious considerations of the sort contained in what I call the easy answer. Suppose it can indeed be made to seem rather obvious that actually, despite the absence of anything anti-physical, we exist and have our everyday psychological properties. Then it should not need very much work to make ²⁴ And also arguments for epiphenomenalism: see n. 2 above. ²⁵ This explains why Jackson formulates (P) as a doctrine only about minimal physical duplicates: see n. 3 above, and see further text below.

56

jennifer hornsby

it credible that physical conditions suffice in some sense for there to be beings with psychological properties. If people and all the macrophysical things to which they relate are made up entirely of tiny physical things, then any variation in facts about people may be expected to go hand in hand with variation at the level of physical science. An unexciting physicalism can then be dressed up as physicalism.²⁶

4. Material Things The assumption isolated in Section 1 as the bone of contention was that everything (temporal) belongs in the domain of physical science. The assumption hardly comes to the surface in Jackson. The effect of Jackson’s formulating (P) as a doctrine about minimal physical duplicates of our world is to dissociate ontological questions—about whether there are anti-physical things—from ideological questions—about whether anything has anti-physical properties. When Jackson claims that the psychological facts are fixed by minimal physical duplicates of our world, he intends to rule it out that there is actually any of the additional non-material stuff that he believes there to be in some possible worlds. Thus in so far as any ontological assumption is in place in Jackson, it is simply that (actually) the only stuff is material stuff. The only point in (1998) where ontological questions are salient is when Jackson illustrates what he calls conceptual analysis in its (in his view unacceptable) immodest role: an argument against four-dimensionalism is used as the example there (pp. 42–3).²⁷ It is a mark of how well concealed Jackson usually keeps his ontological assumption that the discussion of folk psychology contains no hint that he takes folk psychological subjects to be φ-things. But we see an explicit sign of the assumption in his (2000), where Jackson speaks of human beings as among other ‘parcels of matter’ (p. 190). ²⁶ Or anyway something like physicalism. But quibbles about (P), which I set aside at the beginning, might enter at this point. I should acknowledge that one may well not endorse (P) if one is opposed to Jackson’s reductionist thesis. Still, the point of introducing the anti-reductionist physicalist (whose endorsement of (P) is a matter of my stipulation) has been to uncover a position that Jackson keeps hidden, rather than to arrive at a definitive view of exactly what those who reject the reductionist thesis but who don’t consider themselves to be anti-physicalists ought to say. ²⁷ It is true that Jackson locates his disagreement with those whom he calls Cornell realists (in ethics) as a disagreement over ‘ontological descriptivism’ (p. 146); but no ontological question of the kind I am pressing here arises in his arguments against the position (pp. 147–50). I guess that Cornell realists want to say about ethics what my anti-reductionist physicalists say about everyday psychology (and to make the point about supervenience in n. 11 above). Certainly the present bone of contention is not special to any particular subject matter. And unsurprisingly I think that Jackson rides roughshod over ontological questions in his treatment of ‘mature folk morality’ at pp. 140–1.

physicalism, conceptual analysis, and acts of faith

57

Many philosophers have rejected Jackson’s assumption. Their reason is the fact that everyday macroscopic material things appear to differ in their properties from parcels of matter or coincident φ-things,²⁸ so that, by Leibniz’s law, such material things cannot be the same as the φ-things.²⁹ In the familiar case of the statue made of clay, the statue may have come into existence at t, and be well made, Romanesque and admired ..., even while the coincident piece of clay existed before t, and is not well-made or Romanesque or admired. (I take over Kit Fine’s examples here.³⁰ Fine’s paper is particularly useful in a dispute with Jackson, because although Fine’s considerations are mostly conceptual/linguistic, they do not give intuitions about possibilities a large place, thus give to conceptual analysis what Jackson would call a ‘modest role’, and thus ought to be acceptable to Jackson.) Such differences in properties as these can seem to reveal the point of having statues and other works of art in our ontology. Someone who wanted to know why we talk about these things would need to grasp what kind of properties they may have—that they are created, that they may be defective, well or badly made, valuable, ugly, Romanesque, admired ... It is no help to grasp that they may have properties that are attributable to mere pieces of matter. Now Jackson’s reductionist thesis concerns explanations of everyday psychology. (It says, to recall, that everyday psychological explanations are understandable in terms of, or are reducible to, physical ones.) We start to see why such a thesis may be rejected when we appreciate that the concepts ²⁸ Some reductionists might care about whether they call something made of material stuff a parcel or a piece of matter, or a collection or a sum of parts: ‘collection/sum of parts’ implies, as ‘parcel or piece of matter’ perhaps does not, an articulated something. But here I can speak of coincident φ-things, intending to include all of the candidates for reductionist identities. Jackson’s parcels of matter surely must be articulated: unless they are supposed to be composed of things in the domain of physical science, it is impossible to understand the claim that how they are is ‘made true by how things are as told in physical vocabulary’. Once the domain is acknowledged to contain articulated things, there may not be very much point in the reductionist’s caring which word (’parcel’ or ‘sum’ or whatever) is used. The obvious way to ground Jackson’s assumption in ontology is to endorse what Rea (1995) calls the thesis of mereological extensionality: (x)(y)(x is composed of m1, ..mn, and y is composed of m1, ..mn → x = y); and if this were true, then there would only be, as it were, one candidate for identity with any material thing. ²⁹ Admittedly it is controversial whether this appearance can be sustained, and some think that they have arguments in favour of the disputed identities. In n. 12, I mentioned Wiggins and Perry; more recent articles on the anti-reductionists’ side include Doepke (1982), Johnston (1992), Baker (1997), and Fine (2003); see also Wiggins (2001). All these authors use Leibniz’s Law to argue for non-identities: the examples are many and various. ³⁰ p. 206 (2003). Fine shows that the case against the reductionists’ identities is best made without fixating, as so many discussions have, on temporal and modal properties. And he demonstrates the intolerable consequences that accrue from attempts to disallow the conclusion of the arguments from Leibniz’s Law.

58

jennifer hornsby

used in such explanations may be denied to have application to φ-things. No doubt the point of having people in our everyday ontology is rather different from the point of having statues. But just as being in a position to talk about statues goes hand in hand with using concepts applicable to them, so recognizing that there are people goes hand in hand with being able to give explanations using the vocabulary of everyday psychology. And if people are not identical with physical scientists’ items, which now seems very plausible (given that physical scientists’ concepts are generally of no actual help in understanding people), then explanations of facts about people using everyday psychological concepts will not be understandable in the terms of physical science.³¹ When Jackson outlines his account of theory reduction, he makes use of a classic example: temperature in gases is mean kinetic energy.³² Well, it is only to be expected that the behaviour of a (quantity of) gas can be explained in terms of the behaviour of its parts. Discoveries about gases included the decisive discovery that (quantities of) gases simply are collections of molecules. There seems to be no similar expectation of theoretical reduction in respect of subject matters which deal with things that are not collections or sums or parcels of their parts. So I think that when Jackson’s assumption in ontology is in the open, we shall question his overall attitude to the possibilities of reduction. Someone who rejects the ontological assumption will say that it is simply impossible to give ‘the interesting account’ which Jackson tells us that physicalists ‘must give’ of ‘how and why the physical account of our world makes true the psychological account of our world’ (p. 30). Thus physicalists can be absolved from the hard work demanded of Jackson’s serious metaphysicians. (Not that they need have anything against honest toil: we saw that anti-reductionist physicalists at least cannot be accused of taking their philosophical doctrines on faith.) ³¹ See Lynne Rudder Baker (1997) and (2000) for (a) an extended defence of the claim that individuals may be constituted by fundamental particles without explanations in terms of fundamental particles being explanations of the behaviour of the individuals, and (b) an account of how things which aren’t φ-things (i.e. not in the domain of particulars favoured by physical science) can perfectly well have some of the properties of φ-things. I note that Baker’s own view about persons—that they are not the same as the human bodies that constitute them—might or might not be endorsed by anti-reductionist physicalists. The analogy with statues may help to show that it need not be endorsed. And the analogy may also help to show that in so far as I fail to share Jackson’s ‘conception of humans, [which] regards them as fully a part of nature’ (2000, p. 190), I must, in his view, be implicated in a conception of statues which regards them as not fully part of nature. Jackson thinks that my position ‘belongs with the dualists’’. But I should surely be reckoned a pluralist by his lights, thinking as I do that things of many kinds are not φ-things. ³² pp. 57–60. Jackson’s purpose in using the example is to show that conceptual analysis is involved in scientific reduction—a claim that need not be at issue here.

physicalism, conceptual analysis, and acts of faith 59 I may have managed to make it look very hard to defend physicalism of Jackson’s sort and very easy to defend physicalism of a different sort. But if I have made anything look hard or easy, then that is because I have trodden a path so much of my own choosing. I have relied on Jackson’s bold and lucid statement of his version of physicalism in order to pit it against another. I have relied upon setting things up so as to avoid assessing certain claims which Jackson uses in support of his own position (see n. 23 above). And I have only gestured towards arguments for the claim on which I am suggesting a defence of physicalism of an anti-reductionist sort might rest—arguments for rejecting Jackson’s assumption in ontology. Perhaps, however, there is a general lesson here. Physicalists need not subscribe to the idea that ‘the world’ or ‘reality’ is something of which accounts are given.³³ Accounts are accounts of things; and there can be more to the things of which some accounts are given than the material stuff which constitutes them.³⁴ Birkbeck College

References Baker, Lynne Rudder 1997. ‘Why constitution is not identity’, Journal of Philosophy 94: 599–621. 2000. Persons and Bodies: A Constitution View (Cambridge: Cambridge University Press). Braddon-Mitchell, David and Jackson, Frank 1996. Philosophy of Mind and Cognition (Oxford: Blackwell). Doepke, Fred 1982. ‘Spatially coinciding objects’, Ratio 24: 45–60.

³³ Using ‘the world’ in Jackson’s way, one readily assumes the position of what Fine (2003) calls the fanatical mono-referentialist, who thinks that every singular term refers to the ‘One’ (p. 202). Jackson is not alone in speaking in this sort of way: many philosophers speak as if ‘reality’ were something that we describe. It is a question whether mental events (and/or states) depend for their existence upon the existence of the things to which everyday psychological concepts have application. If the answer is Yes, then the ontological assumption at issue here might also be put in question when assessing versions of physicalism very different from Jackson’s (such as anomalous monism, see n. 14 above). ³⁴ For helpful comments on the paper given at the ANU, which I mentioned at the outset, I thank Michael Ridge and Scott Sturgeon. And for helpful comments on the present paper, I thank Scott Sturgeon (again) and David Wiggins. For very useful discussion of a draft, I thank all those who attended a seminar at the Centre for Advanced Study in the Norwegian Academy of Science and Letters, especially Dagfinn Føllesdal, Olav Gjelsvik, Carsten Hansen and Øystein Linnebo. And I am most grateful to the Centre for support.

60

jennifer hornsby

Fine, Kit 2003. ‘The non-identity of a material thing and its matter’, Mind 112: 195–234. Haugeland, John 1982. ‘Weak supervenience’. American Philosophical Quarterly 19: 93–103. (Reprinted as ch. 5 of his Having Thought: Essays in the Metaphysics of Mind (Cambridge, MA: Harvard University Press, 1998).) 1984. ‘Ontological supervenience’. Southern Journal of Philosophy 22: 1–12. (Reprinted as ch. 6 of his Having Thought.) Jackson, Frank 1998. From Metaphysics to Ethics: A Defence of Conceptual Analysis (Oxford: Oxford University Press). 2000. ‘On the metaphysics of human agency’. (Discussion, with Jennifer Hornsby and Lynne Rudder Baker). Philosophical Explorations 3: 188–201. Jackson, Frank and Pettit, Philip 1990. ‘In defence of folk psychology’. Philosophical Studies 59: 31–54. Johnston, Mark 1992. ‘Constitution is not identity’. Mind 101: 89–105. Lewis, David 1970. ‘How to define theoretical terms’. Journal of Philosophy 67: 427–46. Loar, Brian 1981: Mind and Meaning (Cambridge: Cambridge University Press). Menzies, Peter 1988. ‘Against causal reductionism’. Mind 97: 551–74. Petrie, Bradford 1987. ‘Global supervenience and reduction’. Philosophy and Phenomenological Research 48: 119–30. Perry, John 1970. ‘The same F’. Philosophical Review 79: 181–200. Rea, Michael C. 1995. ‘The problem of material constitution’. Philosophical Review 104: 525–52. Schiffer, Stephen 1987. Remnants of Meaning (Cambridge, MA: MIT Press). Wiggins, David 1968. ‘On being in the same place at the same time’. Philosophical Review 77: 90–5. 2001. Sameness and Substance Renewed, 2nd edn. (Cambridge: Cambridge University Press).

3 Serious Metaphysics: Frank Jackson’s Defense of Conceptual Analysis WILLIAM G. LYCAN

According to Stich and Weinberg (2001: 637), Frank Jackson’s From Metaphysics to Ethics (1998a) ‘‘is, by a long shot, the most sophisticated defense of the use of conceptual analysis in philosophy that has ever been offered.’’ I agree. But the book is also very difficult. In this paper I shall work my way through its three main chapters, trying to clarify its basic notions and its argument, and taking issue where I see fit.¹

1. Serious metaphysics and the ‘‘location problem’’ As Jackson uses the phrase, ‘‘serious metaphysics’’ is the attempt to give a comprehensive account of some subject-matter—the mind, the semantic, or, most ambitiously, everything—in terms of a limited number of more or less basic notions. ... Serious metaphysics ... seeks comprehension in terms of a more or less limited number of ingredients, or anyway a smaller list [of kinds of things] than we started with. (pp. 4–5)

The particular smaller list of privileged ingredients that interests Jackson is that provided by science, broadly speaking. His central example (p. 6) is that of explicating the psychological in terms of ‘‘the kinds of properties and relations needed to give a complete account of things like ... tables, ¹ Unless otherwise indicated, all page references are to Jackson 1998a.

62

william g. lycan

chairs, mountains, and the like’’, meaning, I assume, chemical, physical and microphysical properties and relations. For any familiar everyday property of anything (artifactual, mental, semantic, social, economic ...), the ‘‘location problem’’ vis-`a-vis that property is to say how and why the property does or does not ‘‘get ... a place in the scientific account of our world’’ (p. 3). One option is eliminative: in some cases we may find that the property in question does not get a place in the scientific account, and we conclude that the property is unexemplified. Or the property may be ‘‘implicit in’’ the scientific account, in that its ascription is semantically entailed by physical fact (Jackson’s example is that of the taller-than relation being implicit in the sentence ‘‘Jones is six foot and Smith is five foot ten’’ (p. 3)). Or, we will wish to add, the property may bear some other, more complex reductive relation to science. ‘‘[T]here are inevitably a host of putative features of our world which we must either eliminate or locate’’ (p. 5). Jackson turns to defend his first main contention, which he calls ‘‘the entry by entailment thesis’’: that ‘‘the one and only way of having a place in an account told in some set of preferred terms is by being entailed by that account’’ (p. 5). On its face, that thesis appears radical, because it suggests that a semantic relation holds between the set of preferred terms and the relevant macroscopic term, as if a psychological ascription were to be semantically entailed by a set of sentences about chemical compounds, in the same analytical way that ‘‘Jones is six foot and Smith is five foot ten’’ entails ‘‘Jones is taller than Smith.’’ But on p. 25 Jackson explicitly disavows this intention. By ‘‘entailment’’ he does not mean entailment in the usual tight semantic sense, or even more broadly ‘‘conceptual’’ or ‘‘a priori’’ entailment. Rather, somewhat neologistically, he means ‘‘simply the necessary truth-preserving notion—call it ‘necessary determination’ or ‘fixing’ if you prefer.’’ Thus, the entry by entailment thesis is only the linguistic counterpart of the claim that the relevant macroscopic properties must supervene on the preferred properties, which is not terribly controversial. Now, ‘‘[w]hy should a commitment to entailment [i.e., merely necessitation] theses between matters described in some preferred vocabulary and matters described in various other vocabularies require serious metaphysicians to do conceptual analysis?’’ (p. 28). That is the question exactly, if by ‘‘conceptual analysis’’ one means anything like traditional Oxford-style analysis, the search for analytically necessary and sufficient conditions, a matter of linguistic meaning.² It is the question even if one means more loosely ² Jackson offers the following comment on his use of the word ‘concept’: I should enter an explanation ... concern[ing] my word ‘concept’. Our subject is really the elucidation of the possible situations covered by the words we use to ask our questions [about possible cases] ... I use

jackson’s defense of conceptual analysis

63

a priori enterprise, since there is no obvious a priori connection between psychology and chemical composition, or for that matter between water and H2 O.

2. Conceptual analysis ‘‘The short answer is that conceptual analysis is the very business of addressing when and whether a story told in one vocabulary is made true by one told in some allegedly more fundamental vocabulary’’ (ibid.). That short answer is at first puzzling, because empirical science addresses the same sort of issue when it achieves a theoretical reduction, and empirical science is not conceptual analysis in any recognized sense of the term (note that Jackson said ‘‘is the very business,’’ not ‘‘is in the very business’’). But he immediately proceeds to narrow the focus: citing Chisholm, Ayer and Gettier on knowledge and justified true belief, Jackson says that their common effort ‘‘counted as a piece of conceptual analysis because it was intended to survive the method of possible cases’’ (p. 28), the test of intuitive judgments about hypothetical examples. That sounds right, because it returns us to the idea of seeking analytically necessary and sufficient conditions (in this case for knowing). But traditional analysis is not all that Jackson means to include under the term.³ He seems to hold that whenever a thesis is defended by appeal to intuitive judgments about hypothetical cases, conceptual analysis in his sense is being performed. Immediately following his mention of the Gettier problem, and as a second example, he mentions Putnam’s and Kripke’s refutation of standard description theories of reference; although such theories were intended as giving necessary and sufficient conditions for reference, I do not think anyone claimed that the conditions were supposed to be analytic. Still further afield, Jackson cites the example of Putnamian intuitions about Twin the word ‘concept’ partly in deference to the traditional terminology which talks of conceptual analysis, and partly to emphasize that though our subject is the elucidation of the various situations covered by bits of language according to one or another language user, or by the folk in general, it is divorced from considerations local to any particular language. (p. 33, italics original) On our conception, ... we are simply concerned with making explicit what is, and what is not, covered by some term in our language. (p. 51) ³ Which is just as well, because—as is insufficiently appreciated—traditional analysis has been an utter failure. At least, despite vigorous pursuit by brilliant practitioners, it has had no notable successes. The Gettier problem remains unsolved. The most satisfactory analyses I know of any philosophically interesting concept are the more complicated versions of Grice’s analysis of speakermeaning (for a good engaged survey of the latter, see Avramides (1989)), and none of those is unproblematic.

64

william g. lycan

Earth, as (allegedly) showing that none of water’s observable properties are essential to that substance, but that its chemical composition is. We might also mention Twin Earth’s role in convincing us that propositional attitudes are ‘‘wide’’ in that their contents do not supervene on the physical states of their subjects’ heads alone. These deliverances of ‘‘intuition’’ do not take the form of traditional Socratic or Oxford-style analyses, because they do not issue in what seem to be analytic truths—though we must grant Jackson’s emphatic contention that they come from the armchair and not from the streets. Thus, I would point out, there are intuitions and intuitions, though Jackson seems to take no account of the differences. Here is a pertinent taxonomy. First, there are logical intuitions, such as that ‘‘Sally is athletic and Fred admires her’’ entails ‘‘Sally is athletic.’’ Second, there are more broadly linguisticsemantical intuitions, such as that ‘‘Winifred recklessly swam the moat in 45 seconds’’ entails ‘‘Winifred swam’’ and ‘‘Winifred did something reckless.’’⁴ Third, there are (arguably) a priori modal intuitions, such as: that what is known must be believed, that if a speaker means something in making an utterance, then s/he must have some intentions, that two distinct purely physical objects cannot occupy exactly the same space at the same time, that water could have lacked all the stereotypical features associated with water, and that Jackson could not have been David Lewis or a ham sandwich. Fourth, there are the empirically tutored modal intuitions called to our attention by Kripke and Putnam: that nothing can be water without being H2 O, that gold must have the atomic number 79, that this table could not have been made of ice instead of the wood it is made of, and so on.⁵ I shall return to these distinctions below. Jackson says that one’s intuitive judgments about cases manifest one’s ‘‘theory of’’ the relevant subject-matter. And ‘‘[t]o the extent that our intuitions coincide with those of the folk, they reveal the folk theory’’ (p. 32). (Jackson laudably emphasizes that the method of possible cases is fallible in each of several ways, and that the extraction of ‘‘theory’’ from it requires careful handling.) So, in Jackson’s sense, our theory of knowing is that knowing is justified true belief plus some further element of ‘‘non-flukey success.’’ But there is some unclarity in his notion of ‘‘folk theory,’’ that emerges in a criticism he proceeds to make of my own Putnamian view of beliefs as a natural kind. ⁴ It is a disputed question whether intuitions of the first kind are merely a special case of intuitions of the second. ⁵ Of course, there are many other sorts of ‘‘intuitions’’: syntactic, epistemological, moral, aesthetic. Lycan (1988) argues that all these have exactly equal epistemological status as evidence for the respectively relevant theories.

jackson’s defense of conceptual analysis

65

3. ‘‘Folk theory’’ In his section, ‘‘Folk Theory and the Causal-Historical Theory of Reference: A False Opposition’’ (pp. 37–41), Jackson seems to contest my claim (Lycan (1988), 31–32) that ‘‘belief’’ is a Putnamian natural-kind term and that we can succeed in referring to beliefs even if the vast majority of our commonsense beliefs about beliefs are mistaken. (I had said that ‘‘I am entirely willing to give up fairly large chunks of our commonsensical or platitudinous theory of belief ... and decide that we were just wrong about a lot of things, without drawing the inference that we are no longer talking about belief. ... [T]he ordinary word ‘belief’ (qua theoretical term of folk psychology) points dimly towards a natural kind that we have not fully grasped and that only mature psychology will reveal. I expect that ‘belief’ will turn out to refer to some kind of information-bearing inner state of a sentient being ... but the kind of state it refers to may have only a few of the properties attributed to beliefs by common sense.’’) Apparently against that, Jackson says, ‘‘I of course hold against Lycan that if we give up too many of the properties common sense associates with belief as represented by the folk theory of belief, we do indeed change the subject, and are no longer talking about belief’’ (p. 38). ‘‘This will sound like the Lewis–Ramsey–Carnap theory of the reference of theoretical terms. ... And, accordingly, my defense of conceptual analysis will sound committed to a controversial theory of reference’’ (p. 37). Indeed it does sound like the Lewis–Ramsey–Carnap view. If ‘‘folk theory’’ means what it did to David Lewis, namely, the set of commonsensical but mostly contingent platitudes about the subject-matter in question, then we do obviously disagree about changing the subject, and I would want to know why Jackson supports Lewis–Ramsey–Carnap against Putnam. But it becomes clear that that is not what he means by ‘‘folk theory.’’ And instead, Jackson opines that I, WGL, ‘‘misconstrue ... the relevance to folk theory of what we learnt from Putnam (and Kripke)’’ (p. 38). Rather, he says, Putnam’s anti-Lewisian theory of natural-kind terms, the Twin-Earth argument that supported it, and Kripke’s Causal–Historical (hereafter C–H) account of the referring of kind terms themselves ‘‘reflected our folk theory’’ (p. 39). ‘‘We agreed with Putnam ... [and o]ur agreement was endogenous’’ (p. 38).⁶ So by ⁶ In a footnote (12), Jackson adds that ‘‘[s]ome of us agreed with him less whole-heartedly than others.’’ He says he thinks that prior to the Putnamian revolution it was at best indeterminate whether ‘‘water’’ meant H2 O or just watery stuff in general. Off the record (we are in a footnote here, not main text), I think I agree. I am not even sure that in ordinary as opposed to philosophers’ English, ‘‘water’’ does not mean something like watery stuff rather than a natural kind—except that

66

william g. lycan

‘‘folk theory’’ he does not mean a Lewisian set of platitudes.⁷ Yet, as noted above, he does not mean a traditional Socratic set of analytically necessary and sufficient conditions either. Rather, by ‘‘folk theory’’ I think he means just whatever set of beliefs or cognitive dispositions involving a term underlies such intuitive judgments containing the term in question. [Y]our intuitions reveal your theory. To the extent that our intuitions coincide, they reveal our shared theory. To the extent that our intuitions coincide with those of the folk, they reveal the folk theory. (p. 32)

‘‘Our agreement [with Putnam about ‘water’] ... reflected our folk theory of water. Putnam’s theory is built precisely on folk intuitions’’ (pp. 38–9). Our folk theory of water is that water is ‘‘whatever actually is both watery [i.e., satisfies the Lewis–Ramsey–Carnap platitudes regarding water] and is what we are, or certain of our linguistic forebears were, acquainted with [which acquaintance eventuated in our use of the term]’’⁸ (p. 39). Although this is not a traditional Socratic set of analytically necessary and sufficient conditions, it is (has been revealed to be) a conceptually necessary and sufficient condition.⁹ for reasons to be given in Section 5 below, I believe ‘‘watery’’ is ill-defined. Certainly it is hard to get novices to buy Putnam’s Twin-Earth argument; at best, usually, they end up being persuaded that there is a sense in which XYZ is not water. ( Thus, though ‘‘we were not under external instruction from some linguistic dictator to agree with [Putnam]’’ (p. 38), we were under some heavy socioprofessional pressure.) But like Jackson, I will publicly and officially go along with the Putnamian orthodoxy. ⁷ What would Jackson say about the parallel issue concerning proper names? I myself find it obvious that virtually all our beliefs about person X expressed using a proper name of X could be false, yet X would still be the person those beliefs were about. In a lecture, David Kaplan once gave the real-world example of ‘‘Robin Hood’’; according to some researchers circa 1970, Robin Hood was a real person, whom they had identified, but he had none of the properties stereotypically associated with him (except for being a male human alive in, probably, the fourteenth century), including being called ‘‘Robin Hood’’. Would Jackson disagree? It seems not. In note 16 (p. 40) he repudiates the traditional sort of description theory of names, though he thinks it likely that in any counterexample case such as that of ‘‘Robin Hood,’’ a speaker still knows a nontraditional description uniquely true of the referent—perhaps, a´ la Plantinga (1978) and Ackerman (1979), a description such as ‘‘whoever stands at the origin of the appropriately shaped causal-historical chain leading up to our present uses of ‘Robin Hood’.’’ In any clear case of successful referring, the speaker ‘‘could give a procedure for identifying the thing referred to if all went well. But then they know something that individuates the thing in question: it is that which would be identified via its being thus and so if this and that happened.’’ See also Jackson (1998b). ⁸ This is not Putnam’s view exactly. Putnam notably did not require that water actually is watery; our water stereotype may (not just might) be false. ⁹ On this interpretation, Jackson’s antecedent, ‘‘if we give up too many of the properties common sense associates with belief ’’ (my italics), was an inappropriate and misleading understatement, since his Putnamian analysis features only two properties (being actually watery, and being what we or our linguistic forebears were acquainted with), and to give up either one of these would on Jackson’s view to be instantly to change the subject.

jackson’s defense of conceptual analysis

67

And that is why, as Jackson’s section subtitle has it, there is only ‘‘A False Opposition’’ between ‘‘folk theory’’ and C–H. Now, I do not think Jackson could seriously say that the property of being ‘‘whatever actually is both watery and what we are, or certain of our linguistic forebears were, acquainted with’’ is associated by common sense with water, or that the Putnamian analysis is itself a platitude. What common sense associates with water is just the set of Lewis–Ramsey–Carnap platitudes. This means that Jackson is dissociating ‘‘folk theory’’ from common sense, at least as I would use the term.—Fine; it does not matter exactly what we call ‘‘common sense.’’ Does Jackson, then, really contest my Putnamian theory of belief? Not seriously, as it turns out. On p. 40 he says, ‘‘As it happens [my italics], I do not find very appealing Lycan’s view that the term ‘belief’ is a term for an informational natural kind whose identity will be revealed by psychological investigation. ...’’ Perhaps despite his being a Putnamian about less controversially natural-kind concepts and his regarding the Putnamian view as the relevant ‘‘folk theory’’, he would just rather be a Lewis–Ramsey–Carnapian about the concept of belief in particular.

4. The defense: defining/changing the subject With the foregoing notions in place and tentatively understood in the ways I have outlined, how does the defense of ‘‘conceptual analysis’’ go? Jackson argues, ‘‘[S]urely it is possible to change the subject, and how else could one do it other than by abandoning what is most central to defining one’s subject?’’ (p. 38). He recapitulates it in a section called ‘‘The Case for Conceptual Analysis in a Sentence (or Two)’’: Serious metaphysics requires us to address when matters described in one vocabulary are made true by matters described in another. But [1] how could we possibly address this question in the absence of a consideration of when it is right to describe matters in the terms of the various vocabularies? And [2] to do that is to reflect on which possible cases fall under which descriptions. And [3] that in turn is to do conceptual analysis. [4] Only that way do we define our subject—or rather, only that way do we define our subject as the subject we folk suppose is up for discussion. (pp. 41–2; reference numbers interpolated)

[1] should be uncontroversial. So, I think, should [2], though Jackson is right to feel that the point deserves extra emphasis. Philosophy cannot proceed without intuitions about possible cases, and we sometimes forget

68

william g. lycan

that.¹⁰ Notice that if to appeal to modal intuitions about hypothetical cases were eo ipso to do conceptual analysis, then ‘‘conceptual analysis’’ does not need any defense, save against the most callous Quinean naturalist who claims simply to read philosophy off science. But ‘‘conceptual analysis’’ in this broad sense is not the Oxford-style meaning analysis that used to be designated by the term and that is widely thought to have been trashed by Quine. Now, for the reason given in Section 2 above, I balk at Jackson’s [3]. Some intuitions are ‘‘conceptual,’’ if you like,¹¹ but others are merely modal. I would identify the former with the linguistic-semantical, as in ‘‘Winifred recklessly swam the moat in 45 seconds’’ ’s entailing ‘‘Winifred swam,’’ and ‘‘sister’’ ’s entailing ‘‘female’’ and ‘‘sibling.’’ Intuitions of necessitation that do not seem to line up with the meanings of words are modal only—for example, that two distinct purely physical objects cannot occupy exactly the same space at the same time, that nothing can be water without being H2 O, that the biological supervenes on the physical.¹² (There are disputable, indeed disputed, borderline cases. Jackson’s Gettier example is one such. In the 1960s, ¹⁰ All too often, some of our more empirically minded colleagues have appealed to scientific results as if to suggest that those results by themselves have settled the relevant philosophical issues: e.g., ‘‘Science has shown that physical objects are not really colored.’’ ¹¹ At this point I must confess that I myself am a Quinean skeptic about truth by virtue of meaning and about the a priori more generally (Lycan (1994), chs. 11 and 12). But in assessing Jackson’s program we should set that well aside; the critique of his defense should not depend on all-out Quinean skepticism. (Also, Jackson gives a nice argument against such skepticism (pp. 53–4), a version of which I partly accept (1994), 273–80). For the rest of this paper I shall assume that there are lots of conceptual truths in roughly the sense of truths by virtue of meaning. (But as a fallback I would still defend the view of Putnam (1966), according to which although there are analytic truths of several sorts, based on logic, on stipulations, or on simple and nontechnical ‘‘one-criterion’’ concepts such as ‘‘brother’’ or ‘‘bachelor,’’ scientific terms do not generate analytic truths.) ¹² The point is somewhat obscured by Jackson’s tendency to treat modal intuitions as linguistic intuitions. (He likes to speak of them as intuitions about ‘‘how various ... merely possible cases ... are described ...’’ (p. 31) or about whether a word, such as ‘‘knowledge’’, correctly applies to this or that hypothetical case; and notice the full linguisticizing of serious metaphysics in the passage quoted above.) This practice makes some of the modal intuitions seem more semantic and/or conceptual than I believe they are. Serious metaphysics, as originally defined in the quotations with which we began, is not a metalinguistic activity. But allusions to language started taking over very early, as witness Jackson’s odd appeal to ‘‘entailment’’ in his nonstandard sense, a linguistic counterpart of supervenience, instead of just to supervenience. And note hybrids such as on p. 5, where Jackson speaks of a feature being ‘‘entailed’’. Of course, the operative relations between linguistic expressions and the world are a main concern of Jackson’s in the book, and he defends a very sophisticated theory of those relations, but that is not the metaphysics part. It is philosophy of language. Metaphysics is not about language (except of course on those few occasions when it happens to be). At one point (p. 30), Jackson offers a little argument for his linguisticizing: ‘‘Although metaphysics is about what the world is like, the questions we ask when we do metaphysics are framed in a language, and thus we need to attend to what the users of the language mean by the words they employ to ask their questions.’’ Since that conditional’s antecedent is a tautology, nothing of interest follows from

jackson’s defense of conceptual analysis

69

most Gettier aficionados thought that they were analyzing the concept of knowing and delineating the meaning of ‘‘know.’’ Others then repudiated that interpretation and insisted, a bit contemptuously, that they were saying nothing about concepts or the meanings of words, but merely investigating the metaphysical nature of knowing by appeal to modal intuitions.) Of course, there is an anti-metaphysical theory of modality, the Logical Positivists’, according to which all necessity is linguistic-semantical; the Positivists would have rejected the foregoing distinction on the spot. But since Jackson grants the existence of a posteriori necessities, he is not just regressing to Positivism and is no conventionalist about necessity (though as we shall see, he denies the distinction between metaphysically possible and merely conceptually possible worlds). He does use the term ‘‘a priori’’ very broadly, to encompass all except the ‘‘empirically tutored’’ modal intuitions. (‘‘What we can know independently of knowing what the actual world is like can properly be called a priori’’ (p. 51).) I shall assume, then, that ‘‘conceptual’’ is conceptually narrower than ‘‘a priori,’’ and if Jackson wants to move from this-or-that kind of truth’s being ‘‘a priori’’ to its being ‘‘conceptual,’’ argument will be required.¹³ That brings us back to [4]. Defining the subject should not be an issue. Kripke gave us the theory of reference-fixing by descriptions. Scientists (and others) fix reference to phenomena or underlying mechanisms or substances using linguistic and mental descriptions—a loose, casual, shifting set, varying in response to empirical discoveries but in no systematic way. And each reference-fixer is fallible; it need not correctly apply to the referent, but only be thought to be correct for some reason grounded in the referent. But now, Jackson’s changing-the-subject defense, which may be taken as an argument for [3]. The idea seems to be that if we are to theorize and solve the location problem about Xs, we must have an initial characterization of Xs that takes the form of a conceptual analysis; if we try to abandon a condition for being an X that is in fact conceptually necessary, we simply change the subject and stray into irrelevancy. What exactly does Jackson mean by ‘‘change the subject’’? I must assume he means, to be using the same word but no longer it—in particular, not its consequent. (Which of course is not to deny that working metaphysicians sometimes need to ask what other metaphysicians mean by the words they use.) ¹³ In Jackson (1994), he gave an argument that grants this: If a particular supervenience is merely a posteriori, ‘‘how the purely physical makes psychological statements true is rendered an impenetrable mystery’’ (p. 3). This truth-making is remarkable and needs explaining. The obvious explanation would be that it was a logical or conceptual necessity; if it is not one of those, why is it necessary at all? Modal supervenience intuitions are to be nontrivially explained by the semantic/conceptual. That argument is not reprised in From Metaphysics to Ethics. But a further reason why ‘‘conceptual’’ must be narrower than ‘‘a priori’’ will be noted in Section 6 below (note 22).

70

william g. lycan

be talking about the same phenomenon one was before; or, a more familiar situation, to be using the same word as another person but using it to refer to a different phenomenon, i.e., to be talking past each other. Well, here is how else the latter could happen, according to the Kripke– Putnam C–H theory: the respective causal–historical chains might be, unbeknownst to the speakers, grounded in different phenomena. I say something about ‘‘Rod Stewart,’’ meaning the philosopher at Austin College in Texas; you take me to mean the now aging rock star, and for a little while we talk at cross purposes. ‘‘Meaning’’ there, says the C–H theory, is just a matter of the causal grounding aforementioned. And the same for natural-kind terms. This sort of story is a little harder to envisage for the intrapersonal case; the speaker’s later uses of the word would have not to be primarily grounded in her/his earlier uses. But it seems perfectly possible, and also, intrapersonal ‘‘changing the subject’’ is itself harder to envisage in the first place.¹⁴ According to my interpretation of Jackson’s ‘‘False Opposition’’ idea (and cf. again Jackson (1998b)), he will simply co-opt the C–H theory for the terms in question, counting it as the verdict of ‘‘folk theory.’’ But why should he think that what effects the subject-changing is one’s abandoning the Putnamian ‘‘folk’’ condition associated with the term, rather than just the difference in causal grounding itself, when it is the latter that seems to be doing all the work? Whatever the outcome(s) of those cavils, we shall see that Jackson’s defense gets more ambitious, in each of three ways.

5. A-intensions What, then, about the Kripke–Putnam a posteriori necessities? Jackson maintains that they are grounded in a priori necessities, which he further believes are conceptual; that is the first way in which his defense of ‘‘conceptual analysis’’ reaches farther. He appeals to the now familiar distinction between two sorts ¹⁴ There is a genuine issue about changing the subject, even within the C–H camp. A number of people have argued that the C–H theory implies that phlogiston really exists: Phlogiston is oxygen; the original theorists just had lots of false beliefs about it, such as that in combustion it is given off rather than taken on, etc. Well, yes. I confess that I have always wondered why people have not taken that line; I secretly tend to believe it myself. But many would regard it as a reductio. The best discussion I know of these issues is in section 11.2 of Stich (1996). Stich argues convincingly that in historical cases, the ‘‘reduce or eliminate’’ choice has been largely verbal and governed by contingent sociological and/or political factors. If true, that counts against any conservative descriptivist view of the matter, though it does not directly support my libertine view. ( Jackson himself is somewhat sympathetic to Stich’s thesis. See, e.g., pp. 44–5.)

jackson’s defense of conceptual analysis

71

of intension that a sentence or term may have (Davies and Humberstone (1980), going back both to Evans (1979) and to two-dimensional modal logic). What we ordinarily think of as the intension of a general term, for example, is the function from possible worlds to the term’s instances at those worlds; the intension of ‘‘water’’ sucks up a world and spits out all the water, i.e., the H2 O, that inhabits that world. Jackson calls this the term’s ‘‘C-intension,’’ because it is what we get when we ask ourselves ‘‘what the term applies to under various counterfactual hypotheses’’ (p. 48). But in addition, Jackson maintains, the term has an ‘‘A-intension,’’ which is what we get when ‘‘we are considering, for each world w, what the term applies to in w, given or under the supposition that w is the actual world, our world’’ (ibid.).¹⁵ The latter characterization is both obscure and vexed.¹⁶ But I believe the idea is this: To find ‘‘water’s’’ A-extension at a world w, find out what ‘‘water’’ would refer to if it were being used by a denizen of w, meaning what it would mean there in that person’s mouth, as opposed to what it means in our speech, if various other things in the other world are held fixed relative to our actual world. But what things? Here is my guess: Let us stipulate and hold fixed that at the relevant Twin World, ‘‘water’’ is a natural-kind term and works semantically much as it does here. That is, it refers to an underlying chemical compound that is the scientific essence of the familiar liquid in question, and its reference is fixed by superficial descriptions such as ‘‘the stuff that fills the lakes’’ and ‘‘the stuff that comes out of the taps.’’ Everything else about English is held fixed also. Under those suppositions, what does ‘‘water’’ mean at Twin World? Everything is the same except that the underlying substance that ‘‘water’’ refers to there is XYZ. So at Twin World, ‘‘water’’ means XYZ, not H2 O, in the same way that in our language ‘‘water’’ does mean H2 O. Now we can generate Jackson’s A-intension for ‘‘water.’’ At any relevant world, i.e., any world at which the sign-design ‘‘water’’ is used as we use ¹⁵ Jackson’s ‘‘A-intension’’ is what some two-dimensional modal logicians have called the ‘‘diagonal’’ (see Stalnaker (1978)) and what Chalmers (1996) calls the ‘‘primary’’ intension. ¹⁶ Think of a world as a maximally consistent set of propositions. What does it mean to suppose that a world ‘‘is actual’’? Obviously, to suppose that the collectively false set of propositions were true. I shall try supposing that an XYZ world is actual. So I am to suppose that, among other things, it is true that the lakes are filled with XYZ and XYZ comes out of the taps. What does ‘‘water’’ refer to at that world? To H2 O, if anything, just as always, and there is not any H2 O there, so ‘‘water’’ is empty at that world, contra Jackson’s intentions. ( That must be a bad argument, but I am not perfectly sure why. Very probably I erred in using the word ‘‘water’’ with its actual meaning in real-world English, when the Ur-idea of two-dimensional modal logic was to evaluate a term at a world w, not according to what it means in our language, but according to what the term would mean at w. Yet ‘‘what the term would mean at w’’ is hardly well-defined.) For more serious difficulties with the notion of ‘‘considering a world as actual,’’ see Stalnaker (2001). For a more recent explication, see Jackson (2004).

72

william g. lycan

it except for there being a different underlying substance, it will mean, not H2 O, but the underlying substance at that world. It will refer to whatever ‘‘plays the watery role’’ there (p. 50), which role is given by standard Kripkean this-worldly reference-fixers for the term (p. 49). So (this is Jackson’s bold inference): even for us in the actual world, ‘‘water’’ has a kind of flaccid intension or meaning along with its normal, rigid referential meaning. Thus, ‘‘water’’ ’s flaccid meaning for us is, stuff that plays the watery role. And that intension is a world-indifferent kind of meaning. N.b., there must be some such kind of meaning, because ‘‘water’’ had a meaning in English long before Watt and Lavoisier, much less Dalton, and we do not suppose that the word now means something different. Also, Jackson is now able to explain the existence of our a posteriori necessities: They arise when a term’s C-intension differs from its A-intension. We know a priori, Jackson maintains, that water is whatever actually plays the water role; empirical input is needed to discover that H2 O plays it. But, Jackson insists, we could not have made the empirical discovery without relying on the a priori A-intension, a deliverance of conceptual analysis. Even scientists must at least tacitly rely on conceptual analysis. Here I must register another, fairly fundamental dissent. I do not believe that English words have A-intensions. To generate an A-intension, one needs a transworld ‘‘role,’’ as in ‘‘plays the watery role.’’ Such roles are supposed to be constituted by reference-fixing descriptions, that are the same across the relevant worlds. Which presupposes that an English word has a distinctive and stable set of reference-fixers. And that presupposition I deny. Reference-fixers are rarely enshrined in the public language;¹⁷ they are private to individual speakers at particular times. The presupposition is not silly. Dictionary entries contain common stereotypical information about water, and philosophers have had no trouble coming up with sets of stereotypical reference-fixers such as those mentioned above. There is a loose body of information about water that we twenty-first-century Americans share with our Australian friends. That seems to constitute a ‘‘watery role.’’ But there being such a body of information is a highly contingent fact. One could still have the word ‘‘water’’ explained to one even if there were neither lakes nor taps, by reference to some other mode of acquaintance with water. In fact, even the collective body of all the reference-fixers that have ever been mentioned in the Putnam literature is expendable. Under unusual circumstances, one could have the word ‘‘water’’ explained to one even if ¹⁷ I have added the qualification ‘‘rarely’’ because of examples like ‘‘Jack the Ripper,’’ in which a name is publicly associated with a single reference-fixer.

jackson’s defense of conceptual analysis

73

there were neither lakes nor taps nor rain nor drinking nor colorless liquid nor ..., so long as there were (real or imaginary) water around affecting us in some way—just by making a trickling sound, or looking dark blue, or feeling hot. And ‘‘water’’ would still mean just what it does now, in real-world English, even if its reference were fixed, in some context, by descriptions entirely different from the usual ones. If that point needs defending: Two English speakers who happen to have in their heads different reference-fixers for ‘‘water’’ do not make the word ambiguous in English. No more would two English-speaking communities which generally had different stereotypes for ‘‘water’’, a seaside fishing village in rainy country and a desert camp that has all water delivered by camel train. I see no obvious reason why the same would not hold for two planets whose inhabitants bore even more disparate everyday relations to water. (There is the fact that all humans need to drink water, but that is a highly atypical feature of this example, and does not carry over to other natural-kind terms.)¹⁸ Perhaps surprisingly, it is not a linguistic fact that ‘‘water’’ satisfies the descriptions that we find in dictionaries. Dictionaries contain lots of nonlinguistic information. The difference between what information turns up in dictionaries and what information does not is only the difference between information that is, though entirely contingent, widely known among a particular dictionary’s intended readership, and information that is not so widely known.¹⁹ No distinctive, stable set of reference-fixers, no stereotypical ‘‘role’’. No stereotypical role, no Jacksonian A-intension.²⁰ Incidentally, even if I am wrong and there are A-intensions of Jackson’s sort, their specifically linguistic ¹⁸ It is an interesting question whether people or communities that use different reference-fixers for ‘‘water’’ have different speaker-meanings behind utterances containing the word. I think they would not, but in any case speaker-meaning is not our present concern. ¹⁹ What, then, is the pre-chemistry meaning of ‘‘water’’? That is a tough question, I admit, and it should receive much more attention from philosophers of language. The leading candidate would be something indexical, something like ‘‘whatever real stuff-kind shares the nature of this and that there’’. But there are two daunting objections to that proposal. First, there is no linguistic evidence known to me that natural-kind terms contain any indexical element. Second, the demonstratives that figure in the foregoing description might, on any given occasion, be demonstrating something that is thought by the speaker to be water but is not. ( The lack of a convincing nondescriptivist theory of the pre-chemistry meanings of natural-kind terms is another reason for doubting Putnamian orthodoxy; cf. note 6 above.) ²⁰ I should emphasize that my objection is to A-intensions considered as a type of specifically linguistic meaning, as, e.g., Kaplanian characters are a type of linguistic meaning. Jackson and Chalmers (2001) back off the idea that A-intensions are public linguistic meanings or types of meaning: [E]ven if these conditions of application are not part of the semantics of ‘‘water’’ in English, this does not entail that a subject’s application of the term to epistemic possibilities is not justified a priori. ... [I]t may be the case that the relevant [reference-fixers for] ... a term may vary as between users of a term

74

william g. lycan

elusiveness distinguishes them from Kaplanian characters, to which it seems Jackson would like to assimilate them (pp. 72–3). Everyone agrees that indexicals have Kaplanian characters and that those characters are part of what one must know if one claims to know English. And the characters are easily, if not exceptionlessly, articulated: ‘‘ ‘I’ refers to the speaker’’, ‘‘ ‘Now’ refers to the time interval containing the utterance,’’ etc. The putative A-intensions are not accessible in this straightforward way. There is a further problem about A-intensions. It seems perfectly possible that a term’s reference should get fixed by descriptions that are one and all ‘‘near-misses’’ in the sense of Donnellan (1966); recall his example of indicating a party guest by calling him ‘‘the man drinking the martini’’ when in fact the guest’s martini glass contains plain water with an olive in it. Suppose there is a kind term T that refers to an underlying substance S, but none of T’s reference-fixers is strictly correct. Nothing plays the ‘‘T-ish role,’’ though S does things that approximate that role. On Jackson’s view we are supposed to know a priori that S is whatever actually plays the T-ish role. It seems to follow that the term T fails to refer, which is contrary to hypothesis.

6. Entry by conceptual entailment On p. 68 Jackson turns to a crucial distinction. As we saw in Section 1 above, his ‘‘entry by entailment’’ thesis amounted just to the claim that the properties being ‘‘located’’ must supervene on the preferred properties. It is a further question whether this supervenience is semantic or conceptual. ‘‘Is physicalism committed to an a priori deducibility thesis in addition to an entailment one in the [weak nonsemantic] sense we have been giving to entailment?’’ And, boldly, Jackson’s answer is affirmative. He is concerned to defend, as he did in Jackson (1994), what I shall call the Strong entry by entailment thesis, that the one and only way of having a place in an account told in some set of preferred terms is by being semantically or conceptually entailed by, and ‘‘deducible’’ from (pp. 68, 83), that account. This Strong thesis is startling and highly controversial; it is the second way in which Jackson’s defense gets more ambitious.²¹ (so that the corresponding conditions of application are not built in to the term’s semantics in English), but that each user’s knowledge of the [relevant] conditionals is justified a priori all the same. The point is correct: A mental A-intension could be constructed from an individual speaker’s mental state at a time. But that takes us beyond the scope of this paper. (I would make a different Quinean objection to such an analogue.) ²¹ The Strong thesis figures crucially in Chalmers’ (1996) argument for mind–body dualism, which argument is rebutted in Lycan (2003). Jackson (1994) suggested the same argument, but drew only

jackson’s defense of conceptual analysis

75

On the Kripke–Putnam view, the type-identities that figure in scientific reductions are a posteriori, and so they do not afford semantic entailments going upward from micro- propositions to macro- ones. That is one reason why we should expect the Strong thesis to fail. There are other good reasons. A second is that the vocabulary of macro- theories bears no semantic relation to that of micro- ones; what possible semantic relation could there be between ‘‘lung,’’ or ‘‘loan,’’ and ‘‘lepton’’? A third is that the taxonomies of the special sciences—biology, psychology, economics, etc.—all cross-classify with each other and with chemistry; there are no ‘‘bridge principles’’ of the sort assumed by the positivists. A fourth is the prevalence of idealization at each level of nature: Rarely is there a smooth fit between a reduced concept and the constellation of lower-level reducing entities; even if truths about the lower level did a priori entail something about the upper level, what they would entail would not be that the reduced concept applies, but only some more complex fact still expressed in lower-level terms (to which the reduced concept somehow approximates by idealization). So, to say the least, the Strong thesis requires argument.²² Mobilizing his apparatus of A-intensions, Jackson begins by offering us a model for deriving sentences containing ‘‘water’’ from sentences containing ‘‘H2 O.’’ We will be able to move a priori from ... sentences about the distribution of H2 O combined with the right context-giving statements, to the distribution of water. ... [F]or consider: (2) H2 O covers most of the Earth; (2a) H2 O is the watery stuff of our acquaintance; (3) Therefore, water covers most of the Earth. Although the passage from (2) to (3) is a posteriori, the passage from (2) together with (2a) to (3) is a priori in view of the a priori status of ‘Water is the watery stuff of our acquaintance’. (p. 82)

This is very compressed. It is crucial to see that the bridging premise, ‘‘Water is the watery stuff of our acquaintance,’’ is supposed to be known a priori in virtue of one’s grasp of the A-intension of ‘‘water.’’ This idea goes back the weaker but still alarming conclusion that either materialism is false or an ‘‘analytical’’ form of materialism, such as Analytical Behaviorism or so-called ‘‘Analytical Functionalism,’’ is true. Of course the Weak entry by entailment thesis would not support any such arguments. ²² This is a second reason why Jackson must be using ‘‘conceptual’’ more narrowly than ‘‘a priori’’: Otherwise, the Strong deducibility would be too easily achieved, e.g. as in, ‘‘We all agree that biology modally supervenes on physics. That modal intuition—like every other intuition except empirically tutored ones—is a priori. Thus, we have just a priori deduced biology from physics.’’

76 william g. lycan to Evans (1979): Suppose someone stipulates, ‘‘Let us use ‘Julius’ to refer to whoever [actually] invented the zip.’’ Then, ‘‘(If anyone uniquely invented the zip) Julius invented the zip’’ is known a priori. Likewise, ‘‘(If there is a unique occupant of the ‘K’ role) the actual occupant of the ‘K’ role is K.’’ Jackson holds it to be true in virtue of (A-)meaning that water is whatever actually plays the watery role. A first question is, how is (2a) itself supposed to be conceptually derivable from microphysics? But let us put that aside, because the point of the example is only to show how Jackson would deal with the obvious hard case of naturalkind terms. There are more serious objections. One is that (again) English does not specify any ‘‘watery role,’’ and there are no A-intensions to begin with. Another objection echoes a complaint made against Evans by Donnellan (1979) and Blackburn (1984: 333–5), who question whether Evans’s sample truths are really known a priori. They argue convincingly that the latter claim rests on a use-mention fallacy. To wit, referential a prioritude does not survive disquotation. Take a simple example first: Although in virtue of your grasp of Kaplanian characters you know a priori that ‘‘I am here now’’ as uttered by me is true, you do not know a priori that I am here now; for the latter knowledge you need perception and memory. Nor, for the same reason, do I know a priori that I am here now. I do not know a priori that if anyone uniquely invented the zip, Julius did, even though I do know a priori that the sentence ‘‘If anyone uniquely invented the zip, Julius did’’ must be a true sentence. (Of course, the corresponding T-sentences are not themselves known a priori.) And similarly, even if we know a priori that the sentence ‘‘The actual occupant of the ‘watery role’ = water’’ must be a true sentence, we do not know a priori that the actual occupant of the ‘‘watery role’’ = water (i.e., H2 O). That had to be discovered empirically, and we know it through testimony. Thus, I deny that one can move a priori, much less conceptually, from (2) and (2a) to (3). Finally, let us remember that even if Jackson did deal satisfactorily with the hard case of natural-kind terms, that does not by itself show that physicalism requires deducibility across the board. Our reasons for expecting the Strong thesis to fail are still, collectively, good reasons. A second argument for the Strong thesis is suggested by a passage on p. 83: [T]he contextual information, the relevant information about how things actually are, by virtue of telling us in principle the propositions expressed by the various sentences ... enables us to move a priori from the H2 O way things are to the water way they are. But if physicalism is true, all the information needed to yield the

jackson’s defense of conceptual analysis

77

propositions being expressed about what the actual world is like in various physical sentences can be given in physical terms, for the actual context is givable in physical terms according to physicalism. Therefore, physicalism is committed to the in principle a priori deducibility of the psychological from the physical.

I think this is another assimilation of A-intensions to Kaplanian characters. To understand a sentence containing an indexical is to know, for any context, how to tell what proposition the sentence expresses in that context. Accordingly, contingent contextual information can license semantically valid inferences that are not valid in virtue of their logical forms: Sue, speaking to me from behind, says ‘‘Your jacket is on fire’’; I infer ‘‘My jacket is on fire,’’ and Frank infers ‘‘Bill’s jacket is on fire,’’ and we all know we are expressing the same proposition. Similarly, Jackson thinks, ‘‘context’’ rules that ‘‘water’’ designates H2 O. But now the question is, how is it that if physicalism is true, ‘‘the actual context is givable in physical terms’’? If physicalism is true, all relevant features of the actual context supervene on (and are reducible to) microphysics. But without begging the question, Jackson cannot assume that those features figure in ‘‘information’’ in the semantic or conceptual sense, so that such information can be used as a premise in a derivation. And short of that, I do not see how the physicality of the contextual features might help. A third argument for the Strong thesis is briefly presented in the section, ‘‘A Simple Argument to Finish With’’ (pp. 83–4): The physical story about amoebae and their interactions with their environments is the whole story about amoebae. ... Now, according to physicalism, we differ from amoebae essentially only in complexity of ingredients and their arrangement. It is hard to see how that kind of difference could generate important facts about us that in principle defy our powers of deduction. Think of the charts in biology classrooms showing the evolutionary progression from single-celled organisms on the far left to the higher apes and humans on the far right: where in that progression can the physicalist plausibly claim that failure of a priori deducibility of important facts about these organisms and creatures emerges?

I shall content myself with one objection and then one very quick answer to Jackson’s rhetorical question. The objection is that although (of course) I too am a physicalist about amoebae, I do not grant in the first place that sentences about amoebae are conceptually derivable from microphysics. That is just another instance of the issue that currently divides Jackson and me. Even if we waive that, my answer to the question, ‘‘[W]here in that progression ... ?’’ would be: At the point in the progression where natural selection cuts in and starts dramatically cross-classifying biological structures and functions against chemical kinds. (Granted, that ‘‘point’’ is pretty vague.)

78

william g. lycan

7. The ‘‘metaphysical’’/‘‘conceptual’’ distinction It is clear throughout the book that Jackson really does not like ‘‘the now famous distinction between metaphysical and conceptual necessity’’ (p. 68), though he has to grant at least a superficial version of it for the case of the a posteriori necessities. He is concerned to emphasize that the distinction is superficial at best and is standardly mischaracterized by us Kripkeans. Consider the standard picture of logical space, featuring ever-larger concentric circles. We can start with the usual three grades of possibility, nomic, metaphysical, and conceptual; the nomically possible worlds are a proper subset of the metaphysically possible, which in turn are a proper subset of the conceptually possible. (I myself would distinguish logical from conceptual possibility, on the grounds that very little is ruled out by formal logic alone, and I would add plenty of logically impossible worlds outside that pale.) Of course, the usual three grades are only a tiny subset of all the grades or types of possibility there are. Biological possibility, legal possibility, moral possibility, ... And notice that no ordinary English sentence expresses an unrestricted alethic modality. We do not hear mention of logical necessity, logical possibility or entailment outside a philosophy department. Rather, all everyday modalities, expressed by English modal auxiliaries, are restricted, relative to contextually determined sets of background assumptions, and few of even those streetlevel restriction classes themselves correspond to recognizable philosophical categories.²³ The usual way of generating and distinguishing all the various types and grades of possibility is in terms of consistency or compatibility. Nomic possibility is consistency with the laws of nature that govern our world, legal possibility is consistency with the relevant civic or criminal laws, epistemic possibility is consistency with what is known; and so on. On this model, conceptual possibility outruns metaphysical possibility because (so far as has been shown) a proposition may be compatible with all conceptual truths yet incompatible with some fact or law of metaphysics. But Jackson believes that something is very wrong with that familiar picture. (This is the third way in which Jackson aggravates his defense of conceptual analysis.) He is concerned to deny that there is a difference between ‘‘metaphysical’’ and ‘‘conceptual’’ possibility/necessity (pp. 68ff.). There is a distraction that we need to get out of the way: Jackson denies that there are two senses of the terms ‘‘necessary’’ and ‘‘possible,’’ and he ²³ For some (if I do say so myself ) absolutely splendid examples of real restriction classes, see ch. 8 of Lycan (1994).

jackson’s defense of conceptual analysis 79 continues to refer to his target view as ‘‘the two senses view.’’ But no one (I trust) thinks that the alethic modal terms are ambiguous, as between different senses. Rather, as I put it earlier, there are types and grades of necessity and possibility, designated by adverbs such as ‘‘conceptually’’ and ‘‘legally.’’ Jackson is surely aware of that, and his arguments do not turn on the difference between different senses and different types. What he wants is to deny that there is a difference between metaphysical and merely conceptual possibility/necessity. He offers two reasons for that denial. The first is ‘‘Occamist’’ (pp. 70–4). As its name implies, it is an appeal to parsimony. ‘‘The phenomena of the necessary a posteriori, and of essential properties, can be explained in terms of one unitary notion of a set of possible worlds’’ (p. 70). The main phenomenon that Jackson is concerned to explain is disconcertingly specific: ‘‘... how a sentence can be necessarily true and understood by someone, and yet the fact of its necessity be obscure to that person’’ (p. 71). He gives a predictable and (if we again let pass the issue of the stable set of reference-fixers) not implausible two-dimension-style explanation. He concludes, And the important point for us is that this story about the necessary a posteriori does not require acknowledging two sorts of necessity. The story was all in terms of the one set of possible worlds.

The latter remark is perfectly true, but unresponsive. First, that the ‘‘metaphysical’’/‘‘conceptual’’ distinction is not required to explain the specific phenomenon of how a sentence can be both necessary and a posteriori does not show that it is not required for any explanatory purpose. Second and more importantly, the distinction is not (or not primarily) an explanatory posit in the first place. As before, conceptual possibility outruns metaphysical possibility because, just as a proposition may be (semantically) entailed by the laws of nature without being true in all metaphysically possible worlds, a proposition may be entailed by some fact or law of metaphysics without being a conceptual truth: ‘‘All water is H2 O,’’ ‘‘Nothing is both red all over and green all over,’’ and ‘‘Mark Twain is Samuel Clemens,’’ for example. If we think of possible worlds as sets of propositions, the merely conceptually possible worlds contain propositions that no metaphysically possible world contains. Those are (broadly) logical facts, not hypotheses invoked to explain anything in particular. All the distinctions between grades of necessity and possibility—nomic vs. legal vs. biological vs. moral etc. etc.—remain in place just as before; why should metaphysical vs. conceptual be any different?²⁴ ²⁴ Perhaps it is tendentious to think of possible worlds as sets of propositions. Suppose David Lewis was right, and other worlds are not abstract entities, but gigantic concrete entities like our own universe

80

william g. lycan Jackson’s second argument is brief:

The key point is that the right way to describe a counterfactual world sometimes depends in part on how the actual world is, and not solely on how the counterfactual world is in itself. The point is not one about the space of possible worlds in some newly recognized sense of ‘possible’, but instead one about the role of the actual world in determining the correct way to describe certain counterfactual possible worlds. (pp. 77–8).

I have my usual reservations about that two-dimensional way of putting it, but let us again suspend those. My problem with this second argument is that I do not see how it gets from premise to conclusion. Suppose Jackson’s ‘‘key point,’’ expressed in the first sentence of the foregoing quotation, is correct. How is it supposed to follow that there is no difference between metaphysical and conceptual possibility? Perhaps this is another appeal to parsimony: we can construe the arguments of Kripke and Putnam in the two-dimensional way, so there is no need to posit a special realm of conceptual-but-not-metaphysical possibilities. But if that is the right interpretation, I would (obviously) make the same reply as I did to the first argument. The realm is not a special realm, and certainly not ‘‘some newly recognized’’ one; it is an already recognized sector of logical space like any other sector. But if the argument is not another appeal to parsimony, then I do not know what it is. A better argument against the ‘‘metaphysical’’/‘‘conceptual’’ distinction is suggested by Jackson’s remarks on the ‘‘methodological objection’’ (p. 80). He reminds us that the alethic modalities are not primarily features of sentences. He says that what we should be talking about is possibilities themselves and how many kinds of them there are. Presumably they are propositions or states of affairs. Now, Kripkean–Putnamian believers in a posteriori necessities hold their belief primarily because of a posteriori identities and the Marcus–Kripke but merely disconnected from it. Then there might be a debate over exactly which types of concrete world exist. For example, Lewis himself refused to countenance impossible worlds; Lewis (1986) gave an argument against them. In this concretist context, Jackson’s appeal to parsimony would make more sense, because gigantic concrete entities are more plausibly understood as posits than are sets of propositions. Just three points in response to that. First, on pp. 10–11 Jackson disavows ontological concern about worlds; he explicitly refuses to take sides as between concretists and nonconcretists; nor do I know of any evidence that he is a concretist. Second, concretists are very few, and there is a good reason for that: concretism is, to say the least, highly problematic. (See ch. 4 of Lycan (1994). Of course, there is no unproblematic ontological interpretation of worlds.) Third, even if concretism is true, there is no sound motivation for selectivity about which kinds of worlds to countenance. In particular, Lewis’s argument against impossible worlds fails (Lycan (1994): 39–40). And, to date, I have seen no argument for the existence or ‘‘existence’’ of possible worlds that is not equally an argument for that of impossible ones.

jackson’s defense of conceptual analysis

81

point that genuine identities (identities whose terms are rigid designators) are necessary. But someone might argue that any two true identity sentences whose terms are rigid designators of the same individual express the same proposition. ‘‘Mark Twain = Samuel Clemens’’ expresses just the same singular proposition as does ‘‘Samuel Clemens = Samuel Clemens,’’ namely, the proposition that that person is that person. ‘‘Water = H2 O’’ expresses just the same proposition as do ‘‘Water = water’’ and ‘‘H2 O = H2 O,’’ namely, that that stuff is that stuff. Moreover and more generally, outside intensional contexts coreferring rigid designators may be substituted in sentences salva propositione, so ‘‘Some water is not H2 O’’ expresses the same proposition as ‘‘Some water is not water’’ and ‘‘Some H2 O is not H2 O.’’ So if we are individuating worlds according to sets of propositions, there is (at least as yet) no proposition that holds in a conceptually possible world but not in any metaphysically possible world. In particular, the proposition that some water is not H2 O does not hold in a conceptually possible world, because it is one and the same as the proposition that some H2 O is not H2 O. So too for the proposition that Twain is taller than Clemens, which is just the proposition that Clemens is taller than Clemens. The foregoing argument may seem to break down when it comes to naturalkind terms. For no one thinks that ‘‘water’’ and ‘‘H2 O’’ are synonymous. It is a given of the Kripke–Putnam literature that ‘‘water’’ and ‘‘H2 O’’ differ in meaning. Since ‘‘water’’ and ‘‘H2 O’’ do differ in meaning, ‘‘Some water is not H2 O’’ expresses a different proposition from that expressed by ‘‘Some water is not water,’’ and ditto for ‘‘Some H2 O is not H2 O.’’ So, continues the objection, there is after all a proposition, expressed by ‘‘Some water is not H2 O,’’ that holds in a conceptually possible world but in no metaphysically possible one, and the argument fails. But Jackson has a reply available. He may remind us that ‘‘water’’ and ‘‘H2 O’’ do have the same C-intension. Their difference in meaning is a difference in A-intension. The A-intension of ‘‘Some water is not H2 O’’ is a metaphysical possibility, not a metaphysical impossibility. So there is no single intension or proposition that is metaphysically impossible but conceptually possible. Given the two-dimensional framework and the assumption that there are A-intensions, the foregoing seems (otherwise) to be a sound argument. Yet even given those assumptions it could not settle the issue, because as I have said, I believe there are unrelated types of sentence that do express propositions that are conceptually possible though metaphysically impossible: ‘‘There are distinct physical objects that occupy the same region of space at the same

82 william g. lycan time’’; ‘‘Some abstract entities have causal powers.’’ But Jackson may disagree about that.²⁵ This study has been unsympathetic. It has also focused exclusively on the first three chapters of From Metaphysics to Ethics, which would hardly be fair if this were a book review. As few readers of the present volume will need to be told, Jackson’s other three chapters, on color and on moral properties, are full of good first-order argumentation. (They also defend several views with which I emphatically agree.) The book well deserves its place among Jackson’s many, many fine and provocative contributions to philosophy.²⁶ University of North Carolina

References Ackerman, D. F. 1979. ‘‘Proper names, propositional attitudes and nondescriptive connotations’’, Philosophical Studies 35: 55–69. Avramides, A. 1989. Meaning and Mind (Cambridge, MA: MIT Press). Blackburn, S. 1984. Spreading the Word (Oxford: Oxford University Press). Chalmers, D. 1996. The Conscious Mind. (Oxford: Oxford University Press). Davies, M., and Humberstone, L. 1980. ‘‘Two notions of necessity,’’ Philosophical Studies 38, 1–30. Donnellan, K. 1966. ‘‘Reference and definite descriptions’’, Philosophical Review 75, 281–304. 1979. ‘‘The contingent a priori and rigid designators’’. In P. A. French, T. E. Uehling and H. K. Wettstein (eds.), Contemporary Perspectives in the Philosophy of Language (Minneapolis: University of Minnesota Press). Evans, G. 1979. ‘‘Reference and contingency’’, Monist 62, 161–89. Jackson, F. 1994. ‘‘Armchair metaphysics’’. In J. O’Leary-Hawthorne and M. Michael (eds.), Philosophy in Mind (Dordrecht: Kluwer Academic Publishing). 1998a. From Metaphysics to Ethics: A Defence of Conceptual Analysis (Oxford: Oxford University Press). 1998b. ‘‘Reference and description revisited’’. In J. Tomberlin (ed.), Philosophical Perspectives, vol. 12: Language, Mind and Ontology (Atascadero: Ridgeview Publishing), pp. 201–18. 2004. ‘‘Why we need A-intensions’’, Philosophical Studies 118: 257–77. ²⁵ Further arguments against the ‘‘famous distinction’’ are given by Stalnaker (2003). ²⁶ For extensive discussion of From Metaphysics to Ethics I am grateful to Dorit Bar-On, Mark Bauer, Katya Hosking, Janine Jones, John Roberts, and especially David Braddon-Mitchell. And extra thanks to Frank Jackson for lengthy (and no doubt to him tedious) correspondence about early versions of this paper.

jackson’s defense of conceptual analysis

83

Jackson, F., and Chalmers D. 2001. ‘‘Conceptual analysis and reductive explanation’’, Philosophical Review 110, 315–60. Lewis, D. 1986. On the Plurality of Worlds (Oxford: Basil Blackwell). Lycan, W. G. 1988. Judgement and Justification (Cambridge: Cambridge University Press). 1994. Modality and Meaning (Dordrecht: Kluwer Academic Publishing). 2003. ‘‘Vs. a new a priorist argument for dualism’’. In E. Sosa and E. Villanueva (eds.), Philosophical Issues, vol. 13 (Oxford: Blackwell). Plantinga, A. 1978. ‘‘The Boethian compromise’’, American Philosophical Quarterly 15, 129–38. Putnam, H. 1966. ‘‘The analytic and the synthetic’’. In H. Feigl and G. Maxwell (eds.), Minnesota Studies in Philosophy of Science, vol. iii (Minneapolis: University of Minnesota Press). Stalnaker, R. 1978. ‘‘Assertion’’. In P. Cole (ed.), Syntax and Semantics, Vol. 9: Pragmatics (New York: Academic Press). 2001. ‘‘On considering a possible world as actual’’, Proceedings of the Aristotelian Society, suppl. vol. 65, 141–56. 2003. ‘‘Conceptual truth and metaphysical necessity’’. In his Ways a World Might Be (Oxford: Oxford University Press). Stich, S. P. 1996. Deconstructing the Mind (Oxford: Oxford University Press). Stich, S. P., and Weinberg, J. 2001. ‘‘Jackson’s empirical assumptions’’, Philosophy and Phenomenological Research 62: 637–43.

This page intentionally left blank

4 Jackson’s Classical Model of Meaning LAURA SCHROETER AND JOHN BIGELOW

1. Conceptual analysis and implicit knowledge In his book From Metaphysics to Ethics, Jackson’s defence of what he calls ‘‘conceptual analysis’’ proceeds in two stages. In the first stage, he argues for the existence of common knowledge shared by all speakers of a public language. He then brings this postulated common knowledge to bear on an array of philosophical problems, ranging from puzzles about secondary qualities like colour to meta-ethical concerns about moral appraisals of things as right or bad or immoral. The method of a priori conceptual analysis, he thinks, can yield important metaphysical conclusions by unearthing the commitments implicit in our shared linguistic norms. It is plausible to suppose that speakers of a language have implicit knowledge of the way in which the reference of pronouns like ‘I’ and ‘you’ shift, depending on empirical facts concerning who has uttered them, who else was nearby, and so on. Jackson argues that we have similar implicit knowledge of the way in which the reference of a noun like ‘water’ or an adjective like ‘red’ will depend on empirical facts about the world around us. This sort of knowledge can be conveniently laid out in 2-D matrices, in a manner that has been explained at length by Kaplan, Stalnaker, and others.¹ Such matrixes encode amazingly rich and complex patterns of common knowledge. This common knowledge is important, Jackson believes, because it underwrites the philosophical method of conceptual analysis. Suppose Jackson were right to think that we have a rich stock of common knowledge concerning the ways in which word-world relationships ¹ See Kaplan 1989 and Stalnaker 1978 for seminal work on the 2-D semantic framework.

86

laura schroeter and john bigelow

depend on complicated matters of empirical fact. Suppose he were right to think that common knowledge of this kind is an essential precondition for our speaking a common language. Suppose he were right to think that this common knowledge is in some important sense available to speakers of the language on careful reflection. Then it would be reasonable to conclude with Jackson that we can establish significant results in philosophy that are in some sense ‘‘analytic’’ (that is, their truth is determined solely by facts about the meanings of words) and that we could establish these results by a method that is in some sense ‘‘a priori’’, a method that is conducted almost entirely by careful reflection rather than observation and experiment. Thus, at the opening of chapter 4 of From Metaphysics to Ethics, we find the framework that has been articulated and illustrated in earlier chapters is brought to bear on philosophical problems concerning what are called secondary qualities, like colour. As Jackson says, we already know—near enough—roughly what are the relevant sorts of properties of material objects, light, eyes and brains; and what we know already is enough to present us with what he calls the location problem: Which of those properties in the world are the ones that have colour names as well as scientific descriptions? Jackson thinks his conceptual analysis supports an unequivocal verdict: My answer is the ‘Australian’ view that colours are physical properties of objects: certain physical properties of objects have colour names as well as their physical property names. ( Jackson 1998a: 87)

This is a substantial philosophical result. Jackson claims to establish this result by drawing on semantic theories for colour words modelled on the semantic theories he has sketched for words like ‘water’, which in turn are modelled on the 2-D semantic theories for words pronouns like ‘I’ and ‘you’. The upshot is that he claims to establish the ‘‘Australian’’ theory of colours as an analytic truth, which can be established a priori. ‘‘Them’s fightin’ words!’’ This is one of many substantial results that Jackson’s semantic theory promises to deliver. You might contest the details of some of Jackson’s applications of his semantic theories. Yet even if he were to have made significant errors in application, we could still see the potential that the method would have if we were to grant his initial semantic framework. There would surely be some significant applications of the broad kind that he illustrates with examples like that of the ‘‘Australian’’ theory of colours, even if the right

jackson’s classical model of meaning

87

applications were to differ a little from the sketches that Jackson actually offers. The vulnerable point in Jackson’s project is his opening move. Jackson maintains that language use rests on a deep bed of common knowledge. The coordination required for a natural language does not, he assumes, come about by magic, and he is surely right about that. There must be something that explains it. Jackson maintains that this ‘‘something’’ is speakers’ implicit knowledge of what it takes to be the reference of a word. This implicit knowledge plays two important causal roles in Jackson’s account. The first role is to help causally explain how members of a linguistic community achieve coordination by means of public language symbols. The second role is to causally explain individual speakers’ verdicts about how to apply a word to particular hypothetical cases—thus, grounding the armchair methodology of conceptual analysis. But is shared implicit knowledge of the putative subject matter really the only way to explain these phenomena? Jackson often writes as if his account of public language meanings in terms of descriptivist conventions were just plain common sense. How else are we to explain how different speakers manage to communicate using a public language? And how else can we explain how individuals arrive at confident judgments about the reference of their words in hypothetical scenarios? Our aim in this paper is to show just how controversial the psychological assumptions behind Jackson’s semantic theory really are. First, we explain how Jackson’s theory goes well beyond the commonsense platitudes he cites in its defence. Second, we sketch an alternative explanation of those platitudes, the improvisation model of meaning, which seems psychologically more realistic. Our conclusion is that the psychological picture presupposed by Jackson’s semantic theory stands in need of a more substantial defence than he has so far offered.

2. The role of the 2-D framework Before we examine the psychological underpinnings of Jackson’s account, it’s important to get clear about the role played by his two-dimensional semantic framework. The core claim in Jackson’s account is that speakers implicitly know what it takes to be the reference of their own words. But this claim seems to directly contradict the standard lessons drawn from externalist thought experiments offered by Kripke and Putnam. The moral standardly drawn from Putnam’s Twin Earth case and Kripke’s G¨odel/Schmidt case is that our words

88

laura schroeter and john bigelow

‘water’ and ‘G¨odel’ can have a determinate reference even if we have no accurate knowledge of what it takes to be the reference.² Jackson rejects this moral. He argues that the real lesson to be learned from the externalist thought experiments is that, although we don’t know the essential nature of what’s picked out by names and natural-kind terms, we do know an indirect reference-fixing criterion. In the case of ‘water’, for instance, competent speakers don’t necessarily know that water is H2 O, but they do know (roughly) that water is the actual stuff that fills lakes and oceans, runs from taps, and is good for drinking.³ Since we all agree what would count as water if the actual world turned out to be like Twin Earth, he argues, we must have some shared criterion for what it takes to be the reference of ‘water’. In a slogan: we know how the reference depends on empirical facts about one’s actual environment. So if Jackson is right, our grasp of names and natural-kind terms resembles our grasp of indexicals: in all three cases, competent speakers implicitly know a criterion for identifying the reference on the basis of contingent empirical facts about their environment. A number of externalists have challenged Jackson’s story at this point. The problem they see is that as soon as we actually specify a candidate referencefixing criterion, we can think up counterexamples—situations in which the proposed reference-fixing criterion would yield the intuitively wrong verdict about the reference of ‘water’ or ‘G¨odel’. The worry, then, is that the ease with which counterexamples emerge every time a proposal is actually articulated suggests that there just isn’t any uniform reference-fixing criterion that determines our verdicts for every way the actual world could turn out. (See, in particular, Block and Stalnaker 1999; Byrne and Pryor 2004; and Schroeter (forthcoming).) Jackson is not too worried about this sort of objection. In Jackson’s view, all this sort of objection establishes is that a particular proposed definition does not perfectly capture the criterion we actually rely on. But the fact that we haven’t yet gotten the definition right so far doesn’t show that there is no right answer to be had. Indeed, Jackson thinks that the very fact that we can generate convincing counterexamples to the proposed definitions shows that we must share some implicit reference-fixing criterion. (See Jackson 1994: 33; 1998a: 37–41.) Jackson can even admit that there is no precise, non-homophonic definition of a word like ‘water’ or ‘red’ in English. It’s very hard, for instance, to ² For the classical statement of the thought experiments see Kripke 1972 and Putnam 1975. ³ See Jackson 1994: 37–40 and Jackson 1998a: 46–52. For an explanation of this distinction between grasping the essential nature of the reference and grasping a reference-fixing criterion for recognizing its essential nature, see Schroeter (forthcoming).

jackson’s classical model of meaning

89

precisely characterize what it takes to fall into the extension of ‘red’ without using the term ‘red’, for instance in some such phrase as ‘looks red’. Even so, this doesn’t establish that we have no shared implicit reference-fixing criterion. The impossibility of formulating a precise definition might simply be due to the limitations of our current natural language. Alternatively, the lack of a precise definition might be traceable to facts about our psychology: perhaps our ability to identify water involves a psychologically basic type of pattern recognition that cannot be decomposed in an illuminating way. The point is that the success or failure of Jackson’s descriptivism does not turn on there being short, handy descriptions that sum up the putative reference-fixing criteria we associate with ‘water’. 2-D matrices provide a convenient way of making explicit the sort of implicit knowledge Jackson thinks we associate with words like ‘water’. Along the vertical axis of the matrix, we can array all the possible empirical contexts in which the speaker’s practice with a word might be embedded. And along the horizontal axis, we can array all the possible worlds considered as a circumstance of evaluation. The 2-D matrix can then be used to record the speaker’s ideal, fully informed verdicts about the reference of her word relative to different empirical contexts of use. Such a 2-D matrix would display the speaker’s ultimate commitments about how the reference of her word varies depending on her empirical environment. Jackson has emphasized, however, that this 2-D framework is not essential to his account of meaning. (See, e.g., Jackson 2003: section 2. iv.) Jackson starts with the conviction that we must have implicit knowledge of reference-fixing criteria for words like ‘water’, and he takes the 2-D framework as a convenient way of elucidating the structure of that knowledge and of tacking down its precise content. But nothing important to his account would be lost if we cast aside the 2-D framework and characterized Jackson’s account of meaning directly in terms of the speaker’s implicit criteria for specifying the reference of her words.

3. Lockean descriptivism As we’ve seen, Jackson insists that English speakers must share an implicit reference-fixing criterion for ‘water’, even if we can never actually formulate a failsafe definition. Why is he so confident? The answer lies in his attraction to a broadly Lockean approach to meaning. The core Lockean idea is that linguistic communication is explained by tacit social conventions that require speakers

90

laura schroeter and john bigelow

to associate certain reference conditions (and ultimately truth conditions) with words.⁴ This constitutes Jackson’s basic working hypothesis about the nature of linguistic meaning.⁵ Almost everyone agrees that the connection between a word and its reference is established by conventions.⁶ The controversial bit concerns the precise content of these conventions. This is where Lockean descriptivism provides a simple and powerful explanatory hypothesis. The Lockean account starts with the bold claim that all competent speakers share a core set of assumptions about what it takes to be the reference of ‘water’—an implicit ‘‘folk theory’’ of water—and that this fact is mutually obvious to everyone. If this is right, then the standard ‘‘folk theory’’ can ground a tacit linguistic convention governing how speakers are to understand the reference of ‘water’. Given speakers’ interest in coordinating their linguistic usage with those of others in their community, the fact that most of them associate certain assumptions with ‘water’ gives everyone a strong motive for accepting those assumptions themselves. Since these facts are mutually obvious to everyone, the standard set of assumptions will tend to become conventionally entrenched within the community: everyone will be motivated to accept the standard ‘‘folk theory’’ for ‘water’, but their motivation will be conditional on others’ acceptance of the same theory. Lockean descriptivism is a bottom-up approach to linguistic communication. It starts with an account of individual speakers’ substantive understanding of what it is a word represents, and then claims that this understanding also provides the content of tacit linguistic conventions. Thus the very same theoretical posit—implicit knowledge of a ‘‘folk theory’’—simultaneously helps explain both how individual speakers use words to categorize things and how different speakers manage to coordinate their understanding of words. This is a very natural and attractive account of what different speakers of a common language must share if they are to communicate. To appreciate the explanatory role of the Lockean account, it’s important to be clear about the phenomena to be explained. Jackson is right to emphasize that speakers of a common language share a great many attitudes and dispositions ⁴ This talk of the reference conditions of a word is intended as a neutral way of specifying the reference, either by specifying its essential nature or by giving an indirect reference-fixing condition. ⁵ Jackson advertises himself as a Lockean in a number of places. See, for instance, Jackson 1998a: 114; 1998b: 201–2. It’s worth noting that this Lockean conception of language is not shared by Jackson’s sometime co-author David Chalmers. Although Chalmers advocates a similar type of a priori conceptual analysis, he does not think it captures anything like the conventional linguistic meaning of words. See Chalmers 2002: section 8, for Chalmers’ explicit rejection of this idea. ⁶ Quine once queried this platitude ‘‘Truth by Convention’’ but Lewis has convincingly defended it in Convention (Lewis 1969), and now it can be assumed without fear of significant dispute.

jackson’s classical model of meaning

91

regarding the references of their words. Indeed, everyday coordination would be impossible if we did not. When Monima tells Dirk the sink is leaking water, he immediately understands something about the mess that’s being created. And Monima knows that Dirk knows this: that’s why she told him in the first place. And he knows that she knows that he’ll know this: otherwise he wouldn’t trust her utterance to be a reliable guide to the facts. And so on. Thus, some sort of common knowledge of the reference of words seems a precondition for ordinary communication. We’ll focus on three commonsense observations about language that need to be accounted for. Like Jackson, we’re inclined to regard these as Moorean facts about language. First, individual speakers’ substantive understanding of a word puts them in a position to know about its reference, if provided with sufficient empirical information about their actual environment. For instance, if we describe for Dirk different ways the actual world might turn out to be like in sufficient detail, he will be able to come to correct verdicts about what would count as water in those scenarios. It’s controversial, of course, what exactly the epistemological status of these judgments is. (Are they a priori? Are they analytic?) But it’s not controversial that, given enough empirical information, speakers can come to correct verdicts about many cases. This simple psychological fact is what we’re focusing on here. Second, two interlocutors who share the same language will normally have substantially overlapping sets of assumptions about what it is their words represent—and they are generally aware of the rough extent of this overlap. It’s common knowledge between Dirk and Monima, for instance, that water is clear, odourless drinkable liquid, that it fills rivers and oceans, that it’s useful for washing clothes and making tea, and so on. Notice that the extent of this overlap will vary depending on the interlocutors involved. Since Dirk and Monima are both academics of a certain persuasion, it’s also common knowledge between them that the chemical composition of water is H2 O and that water figures in one of Hilary Putnam’s most famous thought experiments; but these assumptions are not common knowledge between Monima and the barrista at the local caf´e. In general, the list of assumptions that interlocutors treat as common knowledge about water will vary depending on how well they know each other and on what the conversation has previously included. What’s undeniable, however, is that competent speakers are normally in a position to know a great deal about their interlocutors’ conception of the objects, kinds, and properties their words refer to. Third, for there to be successful linguistic communication, members of a linguistic community must know that others in the community use words to

92

laura schroeter and john bigelow

refer to the very same object, kind or property. In short, competent speakers must know that their words co-refer. It’s common knowledge between Dirk and Monima, for instance, that they each use the word ‘water’ to refer to the very same stuff. In fact, the phenomenon is a bit more specific than simple knowledge of co-reference. Intuitively, there is a sharp contrast between the kind of knowledge of co-reference involved when two words are taken to share a meaning (e.g. Dirk’s use of ‘water’ and Monima’s use of ‘water’) and when the words merely share their reference (e.g. Dirk’s use of ‘water’ and Monima’s use of ‘H2 O’). Sharing a meaning makes words co-referential de jure rather than merely de facto. An account of linguistic meaning must be able to explain how this knowledge of de jure co-reference is possible. Lockean descriptivism offers a powerful unified explanation of these three points. The first point concerns speakers’ ability to know the reference of ‘water’ in different empirical scenarios. On the Lockean account, a speaker’s ‘‘folk theory’’ determines the reference conditions for all empirical circumstances. So Dirk can know the reference of the term ‘water’ insofar as his judgments are ‘‘guided by’’ the assumptions that constitute his ‘‘folk theory’’ of water.⁷ The Lockean explanation of the second point is equally straightforward. Interlocutors like Monima and Dirk know they share a common set of assumptions about the reference of ‘water’, according to the Lockean, precisely because there is a conventionally entrenched ‘‘folk theory’’ associated with the term. Thus, it will be mutually obvious to everyone that most speakers associate certain core assumptions with the term ‘water’. The Lockean explanation of the third point builds on the previous two. The reason Monima can know that competent English speakers’ use of the word ‘water’ is de jure co-referential with her own is that she knows they associate the same ‘‘folk theory’’ with the term and she implicitly understands that the ‘‘folk theory’’ associated with a term determines its reference in any empirical circumstance. So Monima can know that other competent English speakers’ use of ‘water’ must be co-referential with her own use, no matter what the actual world is like. To sum up: Lockean descriptivism claims that all competent speakers associate such-and-such reference conditions with the word ‘water’. The idea that an individual speaker implicitly treats these reference conditions as decisive explains her judgments about hypothetical scenarios. The idea that these reference conditions are associated with a term through a tacit linguistic convention explains how different speakers can successfully communicate via ⁷ The image of being guided by one’s ‘‘folk theory’’ is from Jackson’s explanation of the role of conceptual analysis in metaphysics, Jackson 1998a: 31.

jackson’s classical model of meaning

93

language. If this Lockean approach to language is right, then Jackson needn’t be too worried if he cannot produce a precise and informative specification of reference conditions for ‘water’ that perfectly captures our best judgments about possible cases—for his theoretical apparatus assures him that speakers must implicitly rely on such reference conditions. A 2-D semantic matrix can then be pressed into service as a way of making this implicit knowledge explicit. Our aim in this section has been to spell out the Lockean approach and show how it affords a substantial and attractive explanation of uncontroversial features about language. In the next two sections, we’ll consider whether the Lockean explanation is the best explanation of these uncontroversial matters. We’ll first consider the Lockean explanation of individual speakers’ referential abilities, and then turn to the Lockean explanation of communication.

4. Judgements about cases: the Meno fallacy Speakers can know what their own words would refer to in hypothetical situations. For instance, you can know that ‘water’, as you currently understand the term, would refer to XYZ if the actual world were like Putnam’s Twin Earth. It may seem that Lockean descriptivism provides the only viable explanation of this Moorean fact. After all, it’s not as if your verdict is simply a matter of brute stipulation: you could get the facts about the reference of your term wrong. So how do you recognize the correct answer when you do find it? Clearly you must rely on your current implicit understanding of water. Just as bounty hunters rely on a picture on a government handbill in order to catch their man, there must be some implicit understanding you rely on in order to recognize what counts as the reference of ‘water’ in different empirical scenarios.⁸ Otherwise, you’d simply be groping in the dark. So it seems that, in order to explain your ability to make explicit knowledgeable judgments about the reference of ‘water’ in hypothetical scenarios, we must conclude that you have implicit knowledge of reference conditions for the term. This line of reasoning is tempting. It is, however, fallacious. Consider the following analogous argument offered by Plato in the Meno. Plato starts by drawing our attention to an indisputable fact about ordinary subjects’ ability to answer mathematical questions. Merely by asking a series of simple questions, ⁸ The image is Jackson’s; see his 1998a: 31.

94

laura schroeter and john bigelow

Socrates leads an uneducated slave boy to recognize that the Pythagorean theorem is true. It’s clear that the boy doesn’t just accept the theorem on Socrates’ authority—he genuinely understands that it must be correct. How are we to explain this ability to recognize a complex mathematical fact? Clearly, the boy’s answers were constrained by his prior understanding of mathematics. Plato concludes that the boy must have implicitly known the correct answer all along: Socrates’ questioning was just the catalyst that transformed this implicit knowledge into explicit knowledge (through the process of ‘‘recollection’’).⁹ Surely Meno’s slave didn’t need to know anything as complicated and topic-specific as the Pythagorean theorem in order to recognize its truth on the basis of Socrates’ leading questions. All he needed was an implicit understanding of much simpler geometrical facts (such as a rough grasp of what it takes to be a triangle or a right angle) together with some very general reasoning and imaginative capacities. The boy’s implicit understanding of the simple geometrical facts constrained his answers to Socrates’ questions about more complex geometrical facts. Only by chaining these intermediate conclusions together was he able to see (for the very first time!) that the Pythagorean theorem must be true. Given the availability of this alternative account, the Platonic suggestion that the boy already implicitly knew the theorem all along is unsupported by the evidence. In general, the mere fact that a person can answer a particular kind of question doesn’t prove that he is guided by implicit knowledge of a dedicated, topic-specific criterion for answering it. It may be more psychologically realistic to say that the subject relies on general, topic-neutral reasoning capacities to arrive at a genuinely new piece of knowledge. Similarly, when the Lockean descriptivist says you have implicit knowledge of the reference conditions for ‘water’, he is making a substantive and potentially controversial psychological claim. Jackson’s image of the bounty hunters provides graphic illustration of the kind of view he has in mind. Bounty hunters rely exclusively on the relatively restricted set of images and descriptions printed on the handbill in order to recognize the wanted man. In the same way, Jackson thinks you rely exclusively on a relatively restricted set of assumptions that ⁹ Although Socrates’ audience in the dialogue seems to find the hypothesis of recollection quite convincing (‘‘I feel, somehow, that I like what you are saying’’), readers need to be wary of what is known as Platonic irony. Socrates immediately goes on to say: ‘‘And I, Meno, like what I am saying. Some things I have said of which I am not altogether confident. But that we shall be better and braver and less helpless if we think that we ought to enquire ...’’. Of course, hardly anyone nowadays agrees with the hypothesis that a mathematician recollects truths that he has learned ‘‘when he was not a man’’. (Cf pp. 85–6 in Jowett’s translation of the Meno.)

jackson’s classical model of meaning

95

comprise your ‘‘folk theory’’ of water in order to recognize what it takes to be water in any given empirical situation. This limited set of assumptions will then determine your verdicts about the reference of your word ‘water’, since you take the word to refer to water. Thus Jackson assumes that it is only the ‘‘obvious and central’’ assumptions about water that guide your judgments about hypothetical cases (Jackson 1998a: 31). Why must the assumptions be limited in this way? Why not allow that your total state of mind is relevant to determining the reference of your term ‘water’? The answer is that Jackson’s Lockean account of communication requires that most speakers associate the very same ‘‘folk theory’’ with the term ‘water’. But no two speakers share all their assumptions about a given topic. So the Lockean story requires that there be a certain limited set of assumptions about water—the obvious and central ones—that all speakers share and that guide individual speakers’ judgments about hypothetical cases. These assumptions serve as a kind of topic-specific template for recognizing what it takes to be water. So just as different bounty hunters all rely on the same handbill, members of the same linguistic community all rely on the same ‘‘folk theory’’ of water. But is this really the best psychological explanation of how individual speakers figure out what the word ‘water’ refers to? Does a privileged subset of assumptions guide your verdicts about what counts as water no matter what the actual world is like? Not obviously so. It may be that the best psychological explanation of your verdicts doesn’t afford a privileged role to any determinate and fixed subset of your assumptions. Here’s one consideration that might seem to support the Lockean explanation. When considering what would count as water if the actual world were like Putnam’s Twin Earth, you may have an immediate and straightforward conviction about the correct answer: XYZ. This immediacy seems an indication that you’re relying on an implicit criterion for identifying the reference of ‘water’. So it can seem that you must have some fixed template that guides your verdicts about possible cases. As one of us has argued elsewhere, however, this impression is due entirely to a poverty of examples (Schroeter 2003). Twin Earth is specifically designed to vindicate all of your current beliefs about water except for one. It’s no surprise, then, that verdicts about the reference of ‘water’ in this sort of context are fairly straightforward. Things are different, however, when you start to consider contexts that diverge more sharply from your current assumptions about your actual environment. What should you say about the reference of ‘water’ if it turns out that you’ve been interacting with a number of different chemical substances in different contexts? Could ‘water’ turn out to

96

laura schroeter and john bigelow

be ambiguous? Or could it refer to a disjunctive kind? What if there hadn’t been any coherent chemical kind here at all? Could water have turned out to be an ecological kind? Or perhaps a culinary kind?¹⁰ How can you adjudicate among these many possible different interpretations? Clearly you don’t have anything like an automatic template that churns out answers to these sorts of questions. Instead, you need to engage in detailed hermeneutical deliberation about which interpretation makes best sense of our total state of understanding in each particular empirical situation. Hermeneutical reasoning about such cases does not seem to conform to the Lockean thesis that only a circumscribed set of ‘‘obvious and central’’ assumptions are ever relevant to determining the correct verdicts about what counts as water. Different assumptions about water will be crucial to determining your verdict about what counts as water, depending on what you assume about your actual context of use. In some contexts, your conviction that water is a fundamental explanatory kind will guide your verdicts about what water is; in other contexts this assumption will be irrelevant. In some contexts, your conviction that water is potable may be crucial; in other contexts it will count as false. And so on. Why suppose that the assumptions that strike you as obvious and central given your current opinions about your environment will continue to strike you as obvious and central when you accept a very different hypothesis about your environment? From the perspective of the deliberator, virtually any assumption could potentially affect the plausibility of a particular interpretation of your word in some context of use or other. There just doesn’t seem to be a clear demarcation between assumptions that are potentially relevant to determining the nature of water in some hypothetical context or other and assumptions that are always irrelevant to this project. Given the strong intuitive plausibility of epistemological holism, moreover, it seems unreasonable to expect such a thing.¹¹ The psychological picture that naturally emerges from our experience as deliberators is simpler than the Lockean account. It should be granted by all parties that all-purpose reasoning skills are crucial to your ability to recognize the facts about what your words refer to in novel empirical contexts. According to the Lockean descriptivist, your deliberation about what counts as water is ¹⁰ It is chastening to recall that of the four elements postulated by the ancient physical theory—fire, air, water and earth—water was the only one that turned out to be a chemical natural kind. ¹¹ It’s a testimony to the fundamental appeal of epistemological holism, that Bertrand Russell found the doctrine plausible even when he was a logical atomist, and was semantically and metaphysically diametrically opposed to holism (see for instance Russell 1959, p. 140). More recently, epistemological holism has become solidly entrenched in the philosophical landscape through the influence of Quine (e.g. Quine 1953).

jackson’s classical model of meaning

97

always guided, in addition, by a particular circumscribed set of assumptions no matter what you suppose about your actual empirical environment. This fixed set of topic-specific assumptions, together with your hypothesis about the actual world, provides the input for your deliberation about what counts as water. Our suggestion is that this Lockean constraint on inputs is unrealistic and should simply be abandoned. In deliberating about what counts as water you try to make best sense of your total prior epistemic perspective with respect to water, in the light of your new hypotheses about the nature of your environment. We might call this the holistic model of your ability to recognize correct answers about the reference of your words. This holistic model provides a more economical psychological explanation of the phenomena, and one that better meshes with the highly plausible case for epistemological holism. Now, we don’t pretend to have demonstrated that the Lockean model is false. Perhaps further empirical inquiry will reveal that you do implicitly rely on a circumscribed set of assumptions about water that figure as fixed constraints on deliberation about its essential nature. This is an empirical possibility. Further inquiry might also reveal that Meno’s slave boy really did know the Pythagorean theorem all along. Our point is just that this hypothesis is not required in order to explain your judgments about possible cases. And sound psychological methodology counsels against positing extra psychological states unless they are required in order to explain the data.

5. Linguistic communication: Classical or jazz? A proponent of Lockean descriptivism may not be too worried about this point. Perhaps the primary attraction of the view is its ability to explain linguistic communication, rather than its ability to explain individuals’ judgments about reference in controversial thought experiments. Indeed, Jackson himself has stressed that it is the voluntary and conventional aspects of communication that are ‘‘crucial to the plausibility of the description theory of reference’’ (Jackson 1998b: 204). In this section, we’ll consider whether Lockean descriptivism does afford the best explanation of tacit linguistic conventions. The Lockean explanation of communication can seem like plain common sense. Everyone agrees that there are tacit linguistic conventions that associate words with reference conditions. And everyone agrees that interlocutors who share the same language have some common ways of understanding their words. The Lockean explanation of these facts is straightforward: there are

98

laura schroeter and john bigelow

tacit linguistic conventions that associate words with ways of understanding them—i.e. implicit assumptions about what conditions an object or property or event would have to meet in order for it to be the reference of a given expression. Jackson seems to find this style of explanation inevitable. However, the Lockean account of communication faces serious difficulties. It’s helpful to break the Lockean account down into a series of discrete claims. 1. There is a relatively limited set of assumptions about a putative feature of the world—a speaker’s ‘‘folk theory’’—that guides an individual speaker’s judgments about the reference of her word ‘water’, regardless of what she assumes about her actual empirical environment. 2. Most members of a linguistic community associate the very same ‘‘folk theory’’ of a putative feature of their environment with the term ‘water’. 3. This convergence in ‘‘folk theories’’ is mutually obvious to all members of the community. 4. The fact that most people converge in this way gives everyone a justifying and motivating reason to associate the standard ‘‘folk theory’’ with the term ‘water’. The conclusion of this line of reasoning is that the assumptions standardly associated with the term ‘water’ will become conventionally entrenched within the linguistic community: everyone will be motivated to accept those assumptions about the reference of ‘water’, insofar as others accept them. The upshot is that English speakers can know they’re using the word ‘water’ to refer to the very same stuff because they know that their words are associated with the very same guiding assumptions about the nature of the putative reference. We agree that the conclusion is well supported by the initial assumptions. The trouble is that there seem to be good reasons to doubt each of these assumptions. We’ve already seen in the previous section why one might doubt the first step of the Lockean explanation of communication. The Lockean posits a hard and fast distinction between assumptions that can affect a speaker’s verdicts about the nature of water and assumptions that cannot. From the point of view of psychological explanation, however, this seems to be an unnecessary epicycle in the mechanisms guiding a speaker’s judgments. And from the point of view of normative epistemology, the account seems to conflict with the holistic nature of epistemological reasoning and about the extent of our fallibility about familiar things like water. Now consider the second step of the Lockean explanation. Assuming that each individual speaker does rely on a fixed set of guiding assumptions in

jackson’s classical model of meaning

99

deliberation about what it takes to be water, how plausible is it that most speakers will converge on precisely the same set of guiding assumptions? The uncontroversial Moorean fact is that for there to be successful communication on a particular occasion, the interlocutors must share some background assumptions about what it is a word like ‘water’ refers to. But this is a much weaker claim than the Lockean claim that there is determinate set of contextindependent guiding assumptions about what it takes to be a particular kind of stuff that most speakers associate with the term ‘water’. In particular, the Moorean fact is compatible with the idea that each speaker’s ‘‘folk theory’’ overlaps to some extent with that of every other individual speaker, but there is no determinate list of assumptions that everyone is required to accept in order to qualify as competent.¹² One way to test the Lockean suggestion is by consulting your intuitions about what’s required for competence with the term ‘water’. If the Lockean argument is correct, then all competent speakers must associate the very same ‘‘folk theory’’ of water with ‘water’. But on the face of it, there doesn’t seem to be any non-negotiable set of assumptions about water that competent English speakers are required to share. Take any widespread assumption about water—say, that it’s potable, or that it fills the oceans—we can imagine someone who failed to share that assumption and yet who would strike us as linguistically competent with the term ‘water’. Perhaps the speaker was raised in very unusual circumstances—in the Sahara, or on Mars. Or perhaps she has a weakness for eccentric empirical hypotheses about the non-potability of water. It is easy to imagine that we might in such cases have good reason to think that we were talking about the very same stuff, and that we’d be inclined to see such a person as ignorant or misguided rather than linguistically incompetent. Prima facie, then, the Lockean thesis that competent speakers must accept a determinate ‘‘folk theory’’ looks dubious. This brings us to the third step of the Lockean explanation: given that most speakers associate the very same set of assumptions with ‘water’, their convergence on this set of assumptions will be mutually obvious to all speakers. Now, clearly we often do know roughly what another person is assuming about the stuff they call ‘water’. The point we’d like to stress here is a relative one. Normally, facts about the reference of a speaker’s word are a lot more obvious than facts about the speaker’s conception of the reference. For instance, if a ¹² There’s an important scope distinction here. The uncontroversial claim has the form: (∀x)(∀y)(∃ζ )(x & y are successful interlocutors and they share a set of assumptions z) whereas the controversial Lockean claim reverses the scope of the quantifiers: (∃z)(∀x)(∀y)(x & y are successful interlocutors and they share a set of assumptions z).

100

laura schroeter and john bigelow

foreign speaker points at a rabbit and says ‘Gavagai!’ in the right sort of context, it’ll strike you as obvious that her word refers to rabbits. Pace Quine, we think you can even know this fact. However, you may have very little insight into the speaker’s ‘‘folk theory’’ of rabbits. Does she think they’re good to eat? Are they sacred totems? Is there any detailed rabbit lore that everyone hereabouts is privy to? You may have no idea. Similar observations apply to speakers getting initiated into their native languages. When a toddler’s mum points to a furry creature and says ‘Look Trev—a rabbit!’ he’ll immediately acquire a rough idea of what kind of thing her word ‘rabbit’ refers to. But it’s much less likely that he’ll have any reliable insight into her implicit assumptions about rabbits (e.g. ‘they breed quickly, they’re a biological species, they’re related to hares, they’re a pest in Australia, they’re Easter totems, they are often eaten as stew ...’). On the Lockean account, it’s the mutual obviousness of speakers’ conception of the reference that is supposed to explain how it can be mutually obvious that they use words co-referentially. Prima facie, however, this seems to get things precisely backwards. Finally, let’s consider the fourth step of the Lockean explanation. Given that it’s mutually obvious that most people accept a particular ‘‘folk theory’’ about the reference of ‘water’, everyone has a good reason to do likewise. Once again, we think there are grounds for scepticism. For the Lockean, a ‘‘folk theory’’ is a set of assumptions about a putative feature of the world that will guide a speaker’s judgments about reference, regardless of what she assumes about her actual empirical environment. Accepting a standard ‘‘folk theory’’ for water, therefore, does not merely require you to accept that other speakers assume such-and-such about water. In addition, you must accept the standard assumptions yourself, using them to guide your judgments about what counts as water no matter what the actual world turns out to be like. But do you really have reason to accept such beliefs simply because other people accept them? On the face of it, it seems you could have sound epistemological reasons for rejecting standard assumptions about water. And it seems that if you did reject those assumptions, you would not thereby cease to be a competent English speaker. When Galileo insisted that the earth moves, he wasn’t just changing the subject! Each important claim in the Lockean account of communication seems to face difficulties in accommodating the commonsense point of view on language. We hope these observations have served to raise some doubts as to the adequacy of the Lockean approach. But we don’t pretend these purely negative considerations are decisive reasons to reject the Lockean model. In the absence of any plausible alternative explanation of the data about communication, it’s reasonable to hang on to the Lockean account. A leaking

jackson’s classical model of meaning

101

raft, after all, is better than no raft at all. And perhaps it can be patched up. No one should abandon the Lockean model unless there is something more seaworthy in the neighbourhood. As it happens, we think there is an attractive alternative. We’d like to propose a new model of communication. This model, we believe, provides a natural and unified explanation of the points we’ve raised as difficulties for Lockean descriptivism. The best way to appreciate the contrast between the two models is to focus on a central explanatory strategy of the Lockean account of communication. In section 3, we highlighted two commonsense observations about communication that any acceptable theory should be able to explain. 1. Substantive knowledge about the reference: two interlocutors normally have substantially overlapping sets of assumptions about what it is their words represent and they are aware of this overlap. 2. Knowledge of de jure co-reference: two interlocutors normally know that their respective uses of a given term are co-referential and they take this co-reference to be de jure rather than merely de facto. It would be surprising if these two kinds of linguistic knowledge were entirely independent. Lockean descriptivism provides a unified account, according to which interlocutors’ knowledge that their use of a word is de jure co-referential is based on their knowledge that they associate the same set of assumptions with the word. Two interlocutors, say Monima and Dirk, may know that their respective uses of the term ‘water’ are de jure co-referential precisely because they know each of them associates the very same ‘‘folk theory’’ with the term. Since a person’s ‘‘folk theory’’ constrains what counts as the reference of the term no matter what the actual world is like, the fact that Monima and Dirk share the very same ‘‘folk theory’’, means that there are no empirical circumstances in which the reference of Monima’s word ‘water’ could diverge from the reference of Dirk’s word—their words necessarily refer to the same stuff. It is the appreciation of these facts that explains why the two interlocutors take their words to be de jure co-referential. Thus, according to Lockean descriptivism, the first kind of knowledge is more basic: it both justifies and causally sustains the second kind of knowledge. Our starting point is to reverse the direction of explanation posited by the Lockean account: it’s your knowledge that other speakers’ words are de jure co-referential with your own that is more basic than knowledge of others’ substantive assumptions about the reference. When you enter into conversation with someone new, you can know that the other person’s use of ‘water’ is co-referential with your own before you have any beliefs about the details of

102

laura schroeter and john bigelow

their associated assumptions. The knowledge that your interlocutor uses ‘water’ co-referentially, together with your appreciation of the standardizing effects of communication and your assumptions about your interlocutor’s idiosyncratic epistemic history, helps justify and causally sustain your knowledge of his substantive assumptions about water.¹³ But how, you may wonder, could one possibly know that another person’s word was de jure co-referential with one’s own without any information about her assumptions as to the nature of the reference? Wouldn’t this just be a blind leap of faith? Surely knowledge that another person is referring to the same object, kind, or property as oneself must be based on some evidence! And surely the relevant evidence must ultimately concern the speaker’s substantive understanding of the reference of her word. We agree that knowledge of co-reference must be based on evidence. But we don’t agree that you need evidence that the other speaker endorses some determinate list of assumptions about the reference of her term ‘water’. It’s enough if you have evidence that the other person intends to use her words co-referentially with others in her linguistic community. An analogy might help explain the difference here. Whereas the members of a classical string quartet master a common score that serves to coordinate their performance, members of a jazz quartet are simply committed to building on whatever musical themes the other members try out. Each member trusts that the others will try to take his individual contributions on board and incorporate them into a coherent musical structure. Thus, the musicians’ mutually obvious cooperative intentions keep the musical performance on track, developing interesting themes rather than degenerating into a welter of random notes. Because of these cooperative intentions, the musicians do not need to agree on any substantive plan in order to achieve a coordinated performance. We suggest that a similar structure of coordinated intentions can explain speakers’ ability to coordinate on a joint representational practice with public language expressions like ‘water’. We might call our account the jazz model of meaning—but perhaps it would be helpful to give it the more sober and informative title, the cooperative intention model of de jure co-reference. The key idea is that speakers’ cooperative intentions to use a given word co-referentially ¹³ This is not to deny that you can have other, non-linguistic evidence of another person’s conception of water. If you see someone lounging by the pool in an international hotel, with a book and a glass of water by her side, you can be justified in forming some ideas about what her conception of water is like without knowing that she even speaks English. Our point is that you don’t need any precise understanding of the set of assumptions she takes to be crucial to determining what counts as water in order to know that her use of the word ‘water’ is co-referential with your own. In fact, you’re unlikely to have access to anything so detailed, as our ‘Gavagai’ and ‘rabbit’ examples illustrate.

jackson’s classical model of meaning

103

help constitute the fact that their uses of the word are co-referential. When speakers’ words co-refer simply in virtue of such intentions, their co-reference is de jure—otherwise, it’s de facto. On this account, knowledge of de jure co-reference will be based on evidence that another speaker intends her use of words to be co-referential with that of others in her linguistic community. Such evidence is quite easy to come by. For instance, the mere fact that the speaker is addressing you with familiar words and grammatical structures is strong evidence that she intends to invoke common referential practices English speakers have built around those expressions. Speakers wear their co-referential intentions on their sleeves. The obviousness of such cooperative referential intentions makes them well suited to figuring in tacit linguistic conventions. Everyone has an interest in making sure their use of a word is co-referential with that of others in the linguistic community, provided that others are ready to do likewise. And it’s obvious that almost everyone does in fact have the relevant co-referential intentions. So, according to the standard Lewisian account of tacit conventions, there is a tacit linguistic convention to make sure one’s own use of the word ‘water’ refers to the same thing as others’. This sort of bare co-referential intention, we’re suggesting, exhausts the content of tacit linguistic conventions. If you’re sympathetic to the Lockean approach, this linguistic convention may strike you as inadequate. Unlike mere wishes, intentions require that the subject be able to envisage some realistic means of bringing about their fulfilment. And this is precisely what the Lockean account provides—realistic means for guaranteeing that different speakers co-refer. According to the Lockean descriptivist, speakers have a tacit convention to treat such-and-such assumptions as determining the reference of ‘water’, come what may. It’s tempting to suppose that something along these lines is the only realistic means to guaranteeing co-reference: speakers must settle in advance on a determinate plan for how each word is to be understood. However, jazz provides a viable alternative model of how speakers might achieve coordination. We don’t need to settle on any determinate plan for exactly how a particular word is to be understood, as long as we’re willing to try to make the best of everyone’s contributions. Like jazz musicians, we can just improvise together. To see how this might work, let’s consider once again how individual speakers make judgments about what counts as the reference of their own word, ‘water’. In the previous section, we suggested that individual speakers must rely on general hermeneutical reasoning to figure out what ‘water’ would refer to in surprising empirical contexts of use. What are the defining characteristics of water if it turns out that there is no single, underlying

104

laura schroeter and john bigelow

explanatory kind that unites most of your paradigm examples? In responding to variations on this sort of case, it doesn’t seem that you can restrict your attention to some determinate subset of assumptions about water. Instead, you rely on general hermeneutical skills to find an interpretation that makes best sense of your entire prior understanding of water. Now, what happens when we include other speakers in the picture? How are you to make sure that your use of ‘water’ is co-referential with theirs no matter what the actual world turns out to be like? One way, of course, is the Lockean’s: everyone can agree in advance on a determinate plan as to which features are relevant to deciding what counts as water. But there’s another, easier way: you can simply take the assumptions other speakers associate with ‘water’ into account when revising your own beliefs about what counts as water. Of course, you may not know precisely what those assumptions are—you might have to wait for them to tell you. But the key idea is that you take the correctness of a given interpretation of ‘water’ to be answerable to everyone’s assumptions about the reference, not just to your own assumptions. If other speakers are prepared to do likewise, then general hermeneutical skills will tend to mandate the same interpretation of ‘water’ for everyone, no matter what the actual world turns out to be like. The effect, then, of such cooperative hermeneutical intentions is to extend the basic unit of semantic interpretation from an individual’s practice with a word to a whole community’s practice with the word.¹⁴ On this approach to meaning, one must carefully distinguish two different explanatory roles: on the one hand, there are the facts that explain de jure coreference, on the other hand there are the facts that explain why individuals’ words refer to certain things. Our suggestion is that speakers’ (undefeated) co-referential intentions play the first role, whereas holistic hermeneutical reasoning starting from the entire relevant set of assumptions plays the second role. In contrast, the Lockean descriptivist posits a single thing—the ‘‘folk theory’’ a speaker associates with a term—to play both roles. There is plenty of work still to be done in fleshing out and motivating the cooperative intention approach to communication. Still, we think enough has been said to show that it represents a promising alternative to Lockean ¹⁴ We don’t mean to suggest that having the requisite cooperative intention is a failsafe guarantee that your word is co-referential with others’. If the understanding an individual speaker associates with a word is extremely deviant, her co-referential intentions can be defeated. When Mrs Malaprop declares that she finds the pope ‘‘exceedingly meretricious’’, she intends her word ‘meretricious’ to be co-referential with ours. However, she also assumes being meretricious is something like being meritorious—indeed, she hasn’t an inkling of what the rest of us understand by ‘meretricious’. Given such a radical divergence between the individual’s understanding and that of the rest of the group, the individual speaker’s own interests win out in determining the idiolect meaning of her word.

jackson’s classical model of meaning

105

descriptivism. The model does as well as the Lockean model in explaining the Moorean facts about communication, while making much more modest claims about what is mutually obvious to all competent speakers. We think the weight of psychological and epistemological reasons militate favour of our proposal rather than the Lockean descriptivist approach. However, making that case in full is a task for another occasion. Our goal in this section has simply been to cast doubt on Jackson’s view that Lockean descriptivism constitutes the best—perhaps the only—strategy for explaining communication. We’ve argued that Lockean descriptivism is a substantive explanatory hypothesis that faces a number of important difficulties. In addition, we’ve argued that the Lockean approach is not the only viable contender. The moral we draw here will be a modest one: Lockean descriptivism is much less commonsensical than Jackson makes it out to be. Before jumping confidently onto Jackson’s Lockean descriptivist bandwagon, you should insist on seeing a more careful articulation and defence of the view than Jackson has so far put forward.

6. Descriptivism redux? In closing, we’d like to briefly consider a line of response one might be tempted to make on Jackson’s behalf. It might seem that we’ve simply lumbered Jackson with an overly restrictive version of Lockean descriptivism. Perhaps on a more charitable understanding of the Lockean approach, there is no genuine conflict with our preferred explanation of communication. We’ve assumed that Jackson’s Lockean descriptivism requires that speakers have an implicit convention to the effect that such-and-such topic-specific assumptions about water are to guide verdicts about the reference of the term ‘water’ come what may. But perhaps this is too restrictive. Perhaps what’s really crucial to Lockean descriptivism is just two ideas: 1. Speakers implicitly understand what their own linguistic conventions require of them. 2. This implicit understanding of linguistic conventions puts a speaker in a position to identify the reference of her word on the basis of ideal empirical information and ideal rational reflection. If this is all that Jackson is claiming, then the cooperative intention model of communication does not represent a genuine alternative to his Lockean descriptivism. The first of the two claims just follows from the nature of tacit

106

laura schroeter and john bigelow

linguistic conventions: there simply couldn’t be a tacit convention in play unless participants had some implicit understanding of what the convention required of them. The second claim is similarly uncontroversial: it simply amounts to the idea that a speaker can, in principle, learn what her own convention-governed words refer to. Together, these two claims imply that competent speakers know something that puts them in a position to identify the reference of their words. However, the two claims leave it open exactly how speaker’s implicit knowledge is structured. How does the cooperative intention model fit this deflated characterization of Lockean descriptivism? According to the cooperative model, speakers have an implicit convention that requires that verdicts about the reference of ‘water’ must be guided by hermeneutical reasoning based on all the assumptions that anyone in the community associates with the term ‘water’ together with the empirical facts about their actual environment. Speakers implicitly know that this is what is required of them and this implicit understanding puts them in a position to identify the reference of their term on the basis of inquiry into their social and physical environment. The crucial point about this proposal is that it doesn’t require the speaker to rely on any particular topic-specific assumptions about water that serve as necessary guides or templates for identifying the reference of ‘water’. All that is required of a speaker is that she take everyone’s topic-specific assumptions about the putative reference of ‘water’ into account when coming to a verdict about what water is. Would it be more charitable to Jackson to count our cooperative intention model as just a particular version of Lockean descriptivism? We think not. There are two important respects in which this proposal fails to live up to Jackson’s theoretical needs. First, we find the purely formal and meta-linguistic character of the tacit conventions governing ‘water’ extremely hard to square with Jackson’s insistence that speakers grasp an implicit ‘‘folk theory’’ of water. There is no theory of water that the competent speaker is required to accept on this account. Instead, there’s a commitment to treat any theory (of whatever!) that’s associated with the word ‘water’ by a member of the community as an input into hermeneutical reflection. This strikes us as an important departure from the spirit of Jackson’s Lockean descriptivism. Jackson has repeatedly stressed that in order to achieve linguistic coordination, two speakers must share particular object-level assumptions about the stuff they’re talking about—and it is these topic-specific assumptions that he takes to figure in tacit linguistic conventions. Second, and more importantly, the cooperative intention model cannot underwrite the kind of a priori metaphysics that Jackson has powerfully advocated. In his book, Jackson argues that a priori metaphysics is underwritten

jackson’s classical model of meaning

107

by the fact that certain ‘‘obvious and central’’ assumptions about the precise nature of, say, water or redness guide our deliberation about the nature of these things no matter what we assume about our actual environment. But why should these assumptions count as a priori? After all, any specific claim—like ‘water is a natural kind’ or ‘the redness of objects helps causally explain our perceiving them as red’—seems empirically defeasible. Jackson’s answer seems to hinge on a claim about what’s required in order to use a word with a particular meaning: if you don’t accept the relevant assumptions about water or redness as ‘‘obvious and central’’, you simply do not associate the same meaning with the words ‘water’ and ‘red’ as the rest of us. This claim about what it takes to use a word with a particular meaning grounds a kind of analytic truth: it’s true in virtue of the standard conventional meaning of ‘water’ that water is the stuff that satisfies such-and-such assumptions about its nature, or comes suitably close to satisfying them. Since this ‘‘folk theory’’ of water is true (or close enough) in virtue of the meaning of the word ‘water’, it’s reasonable to conclude that it enjoys a kind of a priori status.¹⁵ This defence of a priori metaphysics hinges on the idea that there is a single thing—the ‘‘folk theory’’ a speaker associates with a word—that plays two distinct roles: it provides reference conditions for the speaker’s word and it provides a standard for judging whether the speaker uses the word with the same meaning as others. The cooperative improvisation model, however, postulates two distinct things to play these two different roles: sameness of meaning is settled by (undefeated) co-referential intentions, whereas the actual reference is determined by the relevant set of substantive assumptions associated with a word. Since no particular assumption is required to use a word with the same meaning, there are no quasi-analytic truths of the sort that seem to be required for Jackson’s defence of a priori metaphysics. Our conclusion is that Jackson’s philosophical program presupposes a classical, not a jazz model of meaning. A priori metaphysics and a priori conceptual analysis of the sort Jackson advocates require speakers to treat certain substantive assumptions about the putative reference of their words as true (or close enough to true) come what may. And it requires that we see acceptance of ¹⁵ In the joint paper ‘‘Conceptual Analysis and Reductive Explanation’’, Chalmers and Jackson (2001) advance a somewhat different account of the a priori status of conceptual analysis. They suggest that the conditional claim ‘if the actual world is like Twin Earth, then water is XYZ’ counts as a priori simply because all the relevant empirical information has been included in the antecedent of the conditional. On this approach, the a priori status of metaphysics would be secured by the simple expedient of making all metaphysical claims conditional ones. The same strategy could be used to secure the a priori status of physics or any other empirical discipline. Jackson’s defense of conceptual analysis in his book, however, seems to focus on a stronger, non-conditional kind of a priori knowledge grounded in one’s mastery of meanings. This is the line of thought we address in the text.

108 laura schroeter and john bigelow these substantive assumptions as settling whether a speaker is competent with the meaning of a word. This picture of language affords a powerful, unified explanation of many linguistic phenomena. However, it is not the only way of explaining these phenomena. Nor, we suspect, is it the best way—for the Lockean descriptivist model clashes with important psychological and epistemological commitments. These commitments provide the fundamental grounds for dissatisfaction with Jackson’s defence of a priori metaphysics. University of Melbourne Monash University

References Block, N. and Stalnaker, R. 1999. ‘Conceptual analysis, dualism, and the explanatory gap’, Philosophical Review 108: 1–46. Byrne, A. and Pryor, J. 2004. ‘Bad intensions’. In M. Garc´ıa-Carpintero and J. Maci`a (eds.), The Two-Dimensionalist Framework: Foundations and Applications (Oxford: Oxford University Press). Chalmers, D. 2002. ‘On sense and intension’. In J. Tomberlin (ed.), Philosophical Perspectives 16: Language and Mind (Oxford: Blackwell). Chalmers, D. and Jackson, F. 2001. ‘Conceptual analysis and reductive explanation’. Philosophical Review 110: 315–61. Jackson, F. 1994. ‘Armchair metaphysics’. In M. Michaelis and J. O’Leary-Hawthorne (eds.), Philosophy in Mind, Philosophical Studies Series, vol. 60 (Boston: Kluwer). 1998a. From Metaphysics to Ethics: A Defense of Conceptual Analysis (Oxford: Oxford University Press). 1998b. ‘Reference and description revisited’, Philosophical Perspectives 12: 201–18. 2003. ‘From Water to H2 O: The relevance to a priori passage’. In H. Lillehammer and G. Rodriguez-Pereyra (eds.), Real Metaphysics (New York: Routledge). Kaplan, D. 1989. ‘Demonstratives’. Reprinted J. Almog, J. Perry, and H. Wettstein (eds.), Themes From Kaplan (Oxford: Oxford University Press). Kripke, S. 1972. Naming and Necessity (Cambridge, MA: Harvard University Press). Lewis, D. 1969. Convention (Cambridge, MA: Harvard University Press). Plato 1952. ‘Meno’. In B. Jowett (trans.), The Dialogues of Plato (Chicago: University of Chicago Press). Putnam, H. 1975. ‘The meaning of ‘‘meaning’’ ’. Minnesota Studies in the Philosophy of Science 7: 131–93. Quine, W. 1936. ‘Truth by convention’. In O. H. Lee (ed.), Philosophical Essays for A.N. Whitehead (New York: Longman’s, New York). 1953. ‘Two dogmas of empiricism’. Reprinted in From a Logical Point of View (Cambridge, MA: Harvard University Press).

jackson’s classical model of meaning

109

Russell, B. 1959. The Problems of Philosophy (London: Oxford University Press). Schroeter, L. 2003. ‘Gruesome diagonals’, Philosophers’ Imprint 3: http://www. philosophersimprint.org/003.003/. (forthcoming). ‘The limits of conceptual analysis’. Pacific Philosophical Quarterly. Stalnaker, R., ‘Assertion’. In P. Cole (ed.), Syntax and Semantics vol. ix (New York: Academic Press).

This page intentionally left blank

5 The Semantic Foundations of Metaphysics HUW PRICE

1. From metaphysics to ethics, via semantics? In the first chapter of From Metaphysics to Ethics, Frank Jackson begins, as he puts it, ‘by explaining how serious metaphysics by its very nature raises the location problem.’ (1998: 1) He gives us two examples of location problems. The first concerns semantic properties, such as truth and reference: Some physical structures are true. For example, if I were to utter a token of the type ‘Grass is green’, the structure I would thereby bring into existence would be true ... How are the semantic properties of the sentence related to the non-semantic properties of the sentence? Where, if anywhere, are the semantic properties of truth, content and reference to be found in the non-semantic, physical or naturalistic account of the sentence? (1998: 2)

Jackson notes two possible answers to questions of this kind. The first denies that there are any such semantic properties: We might respond with a sceptical or eliminativist position on truth, meaning and reference. Sentences ... are a species of physical object, and we know that science can in principle tell us the whole story about physical objects. And though we are not, and may never be, in a position actually to give that whole story, we know enough as of now to be able to say, first, that it will look something like ... a story about masses, shapes, causal chains, behavioural dispositions of language users, evolutionary history, and the like ... and, secondly, that in any case it will not contain terms for truth, reference and meaning. But if the complete account does not contain truth, reference, and meaning, then so much the worse for truth, reference, and meaning, runs the sceptical response. (1998: 2)

112

huw price

Jackson contrasts this sceptical response with the response he favours, which rests on ‘distinguishing what appears explicitly in an account from what appears implicitly in it.’ (1998: 2) This is the idea that he goes on to develop in the book in considerable detail. The case of the semantic properties deserves its prominence, I think, though for reasons Jackson himself does not mention at this point. Certain assumptions about language, naturally cast in semantic terms, are crucial to Jackson’s conception of the task of philosophy in the cases he calls location problems. These assumptions underpin the most natural path to the view that there is a distinctively metaphysical problem for philosophy to address, as opposed to a problem of a broadly anthropological kind, about human linguistic behaviour—in this case, about the use of the semantic terms, ‘true’, ‘refers’, and the like. Jackson is aware of the need for these assumptions, of course. However, I think he underestimates the work needed to justify them, and hence the extent of the threat that they pose to the foundations of his program. By making the assumptions explicit, and by arguing that we need to take seriously the possibility that they might fail,¹ I hope to show that the foundations are in need of reinforcement; indeed, that there is a serious issue about the advisability of the enterprise, at least in Jackson’s ambitious form. There’s an alternative conception of what philosophy should be up to in this area—a conception, as I said, that regards the main task as more like anthropology than metaphysics. And the issue as to which is the right conception is not to be settled by philosophy, but by the science of human linguistic behaviour, broadly construed. So there is an important sense in which the anthropological viewpoint should come first. Jackson’s approach is not peculiarly at fault here. On the contrary, the linguistic assumptions in question are implicit in a great deal of contemporary metaphysics. The attraction of Jackson’s program, for my purposes, is that in virtue of its clarity and would-be comprehensiveness, its semantic foundations are comparatively easy to bring into view. But most of the concerns thus exposed apply much more widely, I think.² ¹ A lot will turn on a distinction between two kinds of failure. Roughly, one involves the endorsement of the negation of the relevant semantic assumptions, the other—the more interesting, in my view—the abandonment for theoretical purposes of the semantic notions in terms of which the assumptions are expressed. ² If Jackson’s approach has a distinctive vulnerability, it turns on the fact that it is intended to be comprehensive. More on this later.

the semantic foundations of metaphysics

113

2. Deflationism and the location problem for semantic properties Let’s begin with a familiar approach to semantic notions such as reference and truth, namely, deflationism or minimalism. A number of ideas go under these headings. For present purposes, I’ll take the central claim to be that truth and reference are not ‘substantial’ properties—not properties needed in the theoretical ontology of a mature science of human linguistic behaviour, or of a kind whose nature is properly investigated by science or metaphysics. On the contrary, say deflationists, the task of philosophy is to explain the distinctive role of the semantic terms—‘true’, ‘refers’, and the like—in linguistic practice. And the function of these terms is not to refer to substantial properties, about whose nature there might be a serious philosophical question. Rather, for example, it might be that they play a distinctive grammatical role, providing a logical device for sundry purposes (e.g., for expressing infinite conjunctions, as in ‘Everything Fred says is true’). Deflationist views are cousins of what Jackson calls the ‘sceptical or eliminativist position’ about semantic properties.³ Like that position, one of their claimed attractions is that they make short work of metaphysical puzzles about reference and truth. For a deflationist, these puzzles rest on a philosophical confusion about the linguistic role of the terms concerned. In Jackson’s terms, the claim is that if deflationists are right about the linguistic job of the semantic predicates, then there is no location problem for semantic properties. Deflationism is controversial, of course. I have defended a version of it elsewhere (Price 1988, 2003), but won’t be trying to do so here—or at least not directly. Rather, I want to use the fact that deflationism is a well-recognized position in the relevant landscape, to call attention to some presuppositions of the program that Jackson advocates in From Metaphysics to Ethics and elsewhere. By showing that semantic deflationism would undercut an otherwise attractive foundation for metaphysics as Jackson understands it, I want to show that in its present form, the program presupposes that deflationism is false. This will establish that the program relies on certain empirical assumptions about language, in a way which deserves more prominence than it receives in ³ Nothing much hangs on how closely deflationism is related to the views Jackson has in mind, but it is worth noting this difference. Deflationism is not typically an eliminativist position—not a view that agrees, as Jackson puts it, that ‘if the complete [scientific] account does not contain truth, reference, and meaning, then so much the worse for truth, reference, and meaning.’ On the contrary, say deflationists, it is a mistake to look for truth and reference in the scientific account, and hence no black mark against them if they are not to be found there.

114

huw price

Jackson’s presentation, in my view. I’ll argue that these assumptions are harder to justify than Jackson takes them to be. Jackson’s own stance with respect to deflationism also deserves scrutiny, it seems to me. On the one hand, as we’ll see, he often seems robustly realist about semantic notions, and the representationalist view of language with which they go hand in hand.⁴ On the other hand, in the passage from which I quoted at the beginning, he appears to say that it is clear that a scientific account of linguistic tokens—utterances and written sentences, for example—‘will not contain terms for truth, reference and meaning’. On the face of it, this sounds like a version of the deflationist claim that semantic properties have no essential role in a mature science of human language. If Jackson’s view is interpreted this way, then my argument has particular bite. For as things stand, I maintain, the implicit assumption that semantic properties do have an important theoretical role is playing a crucial role in motivating Jackson’s view of the task of metaphysics, in the semantic case itself as well as in others. If I am right about the role of this assumption, there seem to be two ways in which we might try to reconcile it with the view here endorsed by Jackson, that a scientific account of the occurrence of linguistic tokens will not employ semantic terminology. The first would rely on a direct appeal to folk intuitions and practice—to folk use of the semantic vocabulary, in effect. However, leaving aside concerns we should surely feel about allowing the folk to lead us into theoretical territory where science itself declines to tread, we’ll see that such an appeal is only helpful if we can exclude a deflationary interpretation of the role of the folk vocabulary, which is the very point at issue. The second option (Jackson’s own choice, I think) would be to argue that the semantic facts on which the assumption depends are implicit in the scientific story—the representational view is implicit in the scientific view. This option thus relies on Jackson’s proposed solution to the location problem for semantic properties. As I’ll argue, however, both the problem and the solution depend on the very assumption here at issue, namely, a semantically-characterized representational view of folk use of semantic vocabulary. So there’s no comfort here for someone looking for a scientifically respectable reason to believe that assumption. More at the end of the chapter about this apparent circularity problem, and related issues. For the moment, I’m interested simply in exploring the ⁴ Later in the paper, I want to argue that various commonplace observations about language that Jackson appears to take to support this view do not bear the weight, being thoroughly compatible with deflationist views.

the semantic foundations of metaphysics

115

consequences of deflationism for the task and scope of metaphysics as Jackson understands it—in particular, for the existence of location problems. I’ll need the following two observations about the commitments of semantic deflationism. 2.1 Properties thick and thin First, deflationists needn’t say that there are no semantic properties. On the contrary, to the extent that folk usage is committed to claiming that there are such properties, deflationists can concur—can speak with the folk, in effect. But the deflationist insists that once we understand the functions of the corresponding terms in language, we can see that there isn’t any metaphysical problem about the nature or ‘location’ of these properties. As Paul Horwich says: [I]t is not part of the minimalist conception to maintain that truth is not a property. On the contrary, ‘is true’ is a perfectly good English predicate—and (leaving aside nominalistic concerns about the very notion of ‘property’) one might well take this to be a conclusive criterion for standing for a property of some sort. What the minimalist wishes to emphasize, however, is that truth is not a complex or naturalistic property but a property of some other kind ... The point behind this jargon is that different kinds of property correspond to different roles that predicates play in our language, and that unless these differences are appreciated, we will be tempted to raise questions regarding one sort that can legitimately arise only in connection with another sort. (1998: 37–8)

A caution is needed here about Horwich’s use of ‘standing for’. This sounds like a semantic notion, but if so, a deflationist can’t expect it to bear theoretical weight. The main point of deflationism is that semantic notions do not bear theoretical weight. More on this potential inconsistency in a moment. In the present context, one way to avoid it is simply to note—as the rest of the passage says, in effect—that there is a loose and popular use of the notion of a property, according to which, so to speak, where go predicates, there go properties. In this sense, there is certainly a property of truth. But this use of ‘property’ needs to be distinguished from its use in serious science and metaphysics, precisely because these disciplines need to be sensitive to the fact that predicates may play different linguistic roles, in a way which isn’t evident on the surface. Truth is thus a property in the loose, popular or ‘thin’ sense, says the deflationist, but not in the scientific, metaphysical or ‘thick’ sense. 2.2 Two ways of rejecting semantic theory The second point relates to the kind of inconsistency just noted. Paul Boghossian (1990) has argued that irrealism about semantic properties is incoherent,

116

huw price

because the irrealist wants to claims that terms such as ‘true’ and ‘refers’ do not refer to anything, or that sentences ascribing such semantic properties lack truth values (or, perhaps, are uniformly false). The essence of Boghossian’s objection is that the italicized claims involve, centrally and ineliminably, the very notions with respect to which the semantic irrealist professes irrealism. Surely this involves some sort of vicious circularity? In the present context, our interest is in a specific form of irrealism about semantic notions, namely, deflationism, characterized as the view that these notions play no substantial or ineliminable role in linguistic theory. Against such a view, Boghossian’s charge takes a more specific form. Isn’t the deflationist employing the notions concerned ‘in her theoretical voice’, in characterizing her own deflationist view? This is precisely the inconsistency we noted a moment ago, concerning the semantic notion of ‘standing for’. But the objection is easily side-stepped. We simply need to distinguish between (i) denying (in one’s theoretical voice) that ascriptions of semantic properties have semantic properties; and (ii) saying nothing (in one’s theoretical voice) about whether ascriptions of semantic properties have semantic properties—i.e., simply employing different theoretical vocabulary, in saying what one wants to say about such ascriptions. A deflationist cannot consistently do (i), but can consistently do (ii). Let’s call (i) active rejection and (ii) passive rejection of the theoretical claim that ascriptions of semantic properties have semantic properties. (Like passive aggression, then, passive rejection involves strategic silence.) Here’s an analogy. Unlike Creationists, Darwinians don’t think that the species were created by God. Does this mean that Darwinians must use the term ‘God’ in their theoretical voice, as it were, in order to deny that the species were created by God? Not at all. Darwinians simply offer an account of the origin of the species in which the term ‘God’ does not appear. So rejecting the view that God created the species does not require accepting the following claim: God did not create the species. The alternative—the correct alternative, surely—is passive rejection: simply avoiding theological vocabulary in scientific contexts. Similarly, rejecting the view that ascriptions of semantic predicates are referential—rejecting it as a theoretical view —does not require that we endorse a negative claim, in which the semantic terms are employed in our theoretical voice. Passive rejection provides an alternative, and arguably the only consistent alternative. From now on, then, I assume that semantic deflationism takes this passive form. It doesn’t say that ascriptions of semantic properties lack semantic properties.

the semantic foundations of metaphysics

117

Deflationism thus provides a familiar and apparently coherent view, from the standpoint of which the location problem for the semantic properties seems not to arise. It might be suggested that this impression is mistaken—that the particular form of deflationism just outlined does not escape the location problem. I’ll come back to this, but suppose for the moment that deflationism does avoid the location problem, for the semantic properties. Then, as already noted, it provides a relative of the sceptical or eliminativist approach, that Jackson himself mentions as avoiding this particular location problem. Not particularly new news, in other words. But there is newer news nearby, I think. I want to argue that semantic deflationism not only provides an example of how the location problem may be defused for the semantic properties themselves, but also a strategy for defusing location problems about other topics. More accurately, semantic deflationism blocks a particular route to location problems—a route which otherwise carries a lot of traffic. Blocking this route doesn’t necessarily imply that there are no interesting location problems in metaphysics. There may be other ways of getting to the same place, at least in some cases. But it does suggest that such problems may be rarer than Jackson thinks, and in need of foundations he does not provide. But does semantic deflationism even promise old news, given the constraints of the two observations made above? It is easy enough to see how the location problem for semantic properties is avoided by someone who says that there are no semantic properties, or that ascriptions of semantic predicates do not represent things as being a certain way. In these cases, of course, there can be no distinctive puzzle raised by semantic language about the nature of semantic properties or ‘ways of being’—or about the relation of such things to the properties or ‘ways of being’ talked about elsewhere. For there simply are no such things, full stop. But if a deflationist agrees that there are semantic properties, at least in the loose and popular sense, and does not actively deny that ascriptions of semantic predicates are representational in nature, doesn’t the location problem still bite? I’ll meet this challenge by showing that the new news confirms the old news. As I said, I’m going to argue that there is one very general route to location problems which is blocked by the kind of semantic deflationism just canvassed. Having explained why such a semantic deflationism blocks an otherwise appealing route to location problems in general, I’ll then be able to point out that the argument is self-applicable. This kind of semantic deflationism blocks the same route to the location problem about the semantic properties themselves. The new news thus grounds the old news.

118

huw price

3. Noncognitivism, deflationism and global expressivism The move to side-step metaphysics in the way exemplified by semantic deflationism has a long pedigree. Its most famous incarnation is noncognitivism in ethics. Accordingly, when Jackson turns to ethics later in his book, he notes that he is assuming cognitivism. As he puts it: It is only under the assumption of cognitivism that ethics presents a location problem. If the non-cognitivists are right and ethical sentences do not represent things as being a certain way, there is no question of how to locate the way they represent things as being in relation to accounts told in other terms—descriptive, physical, social or whatever—represent things as being, though there will still, of course, be a need to give an account of the meaning of ethical sentences and of what we are doing when we make ethical judgements (where, of course, to make an ethical judgement better not be literally to take things to be some way or other). (1998: 117)

Jackson here relies on an explicitly semantic characterization of the difference between cognitivism and noncognitivism about ethical claims. Cognitivists take ethical claims to ‘represent things as being a certain way’, and noncognitivists disagree. So understood, cognitivists and noncognitivists are both relying on a non-deflationary view of the semantic properties, for they both use semantic or representationalist notions to characterize the function of ethical talk. What, then, is the impact of adopting deflationism? Exploiting the distinction introduced earlier between active and passive rejection, I want to argue that deflationism about the semantic properties entails a view sufficiently close to noncognitivism in ethics to defuse metaphysics in precisely the way that noncognitivism itself does—by ensuring that it is not the case that ‘ethics presents a location problem’ (or, more exactly, by defeating an influential reason for thinking that there is a location problem). If I am right, the point has wide ramifications. Nothing hangs on the fact that the case in question is that of ethics. Quite generally, then, semantic deflationism blocks a certain route to location problems. As I’ve stressed, this does not mean that there could not be other routes to the same metaphysical issues. But there seem no candidates of the required generality on offer in Jackson’s program, and none in the offing. The upshot is that on Jackson’s conception, metaphysics seems to rest on ‘thick’ semantic foundations: in particular, on a non-deflationary view of semantic properties. In arguing that deflationism favours a view close to noncognitivism, I’m swimming against the tide. Many writers have suggested that deflationism is an enemy of noncognitivism. If truth is minimal, the thought goes, then

the semantic foundations of metaphysics

119

it is easy for sentences to be truth-evaluable, and implausible to claim that moral claims are not truth-evaluable. Versions of the argument may be found in McDowell (1981), Boghossian (1990), Wright (1992), and Humberstone (1991). The argument is endorsed by Jackson himself in Jackson, Oppy, and Smith (1994), a paper proposing an escape route for noncognitivism, relying on the argument that minimalism about truth need not imply minimalism about truth-aptness, and that it is nonminimalism about truth-aptness that matters for the noncognitivist’s purposes. I’ll return to this issue later. For the moment, I’m arguing that the noncognitivist does not need saving. In important respects, semantic deflationism already represents victory by default. In my view, then, the orthodox view is almost completely wrong-headed. To see why, note first that noncognitivism normally makes two claims about its target discourse, one negative and one positive. The negative claim says that these terms or statements lack some semantically-characterized feature: they are non-referential, non-truth-apt, non-descriptive, non-factual, or something of the kind. The positive claim offers an alternative account of the functions of the language in question—e.g., that it expresses, or projects from, evaluative attitudes. We might say that the negative claim is anti-representational, the positive claim expressivist.⁵ What is the effect on such a combination of views on deflationism about the semantic vocabulary in which the negative claim is couched? Clearly, the negative claim must be abandoned. But this doesn’t imply that, qua theoreticians, we must endorse its negation (i.e., endorse cognitivism). On the contrary, what’s sauce for the goose is sauce for the gander. If semantic terms can’t be used in a ‘thick’ sense, they can’t be used on either side of a (thick) dispute as to whether evaluative claims are representational. Recall the Creationist analogy. If the term ‘God’ has no place in science, then science neither affirms nor denies the sentence ‘God created the universe’. The point of the example was that in cases like this, not affirming is not the same as denying. So what is the effect of deflationism on noncognitivism? It is to deflate the negative claim, while leaving intact the positive claim—the noncognitivist’s expressivist account of the function of the terms in question. Contrary to received wisdom, then, semantic minimalism is a friend and not an enemy of expressivism. If we take it that the expressivist’s core claim is that the key functions of the judgements in question are not representational functions, ⁵ This use of the term ‘expressivist’ is to some extent a term of art. For example, disquotationalism about truth is expressivist in these terms, but doesn’t rest on an idea of expressing a psychological state. The defining feature of an expressivist view, in this sense, is that it theorizes about the use of language without employing semantic properties in its theoretical ontology.

120

huw price

then deflationism about the key semantic notions is a global motivation for expressivism—a global reason for thinking that whatever the interesting theoretical conclusion about the function of a class of judgements turns out to be, it cannot be that they are referential, or truth-conditional. For deflationism amounts to a denial that these notions have an interesting theoretical role. One important point. The kind of expressivism thus supported by semantic minimalism is, crucially, global expressivism. The orthodoxy is right, in my view, to think that deflationism poses a problem for local varieties of noncognitivism, or expressivism—views that are expressivist about some topics but representationalist about others. If representationalist vocabulary has no theoretical role, it can’t be used to characterize the linguistic territory that, from a local expressivist’s point of view, lies on the other side of the fence. The whole point is that deflationism implies that there can be no such territory, and hence no such fence. So deflationism does indeed make local expressivism unstable—but because it implies global expressivism, not because it implies global cognitivism!

4. Does global expressivism avoid location problems? Semantic deflationism thus implies the view I’ve called global expressivism.⁶ In the form outlined in Section 2 it also exemplifies such an expressivism. Recall that a crucial feature of that form of deflationism—the feature that enabled it to side-step Boghossian’s objection—was that its rejection of the representational view of ascriptions of semantic properties was passive, rather than active. As a theoretical view, in other words, it makes no recourse to the semantic notions themselves. At the end of Section 2, I noted that it might be claimed that this version of semantic deflationism does not, after all, avoid the location problem for semantic properties. After all, it does not deny that there are semantic properties, or that such ascriptions represent things as being a certain way. On the contrary, it allows that there are semantic properties, at least in the loose and popular sense. And it says nothing (of a theoretical nature) about whether ascriptions of semantic properties represent things as being a certain way. (No doubt it allows this too, in the loose and popular sense.) ⁶ Again, the term ‘expressivism’ here means nothing more than that the views in question theorize about language in non-representational terms. See also O’Leary-Hawthorne and Price (1996) and Price (2004a, 2004b).

the semantic foundations of metaphysics

121

An analogous suggestion carries over quite generally to my modified form of noncognitivism—the view that keeps the expressivist account of the functions of the part of language in question, while dropping the negative claim that language is not representational. So now we face the question deferred in Section 2. Does this view avoid the location problem? As foreshadowed in Section 2, my claim is not that semantic deflationism provides an all-purpose defeater for location problems. Rather, I claim that deflationism blocks a particular route to such problems—albeit an important route, which otherwise provides a powerful and widely-applicable theoretical motivation for taking such problems seriously. My claim rests on a distinction between what I’ll call ontologically conservative and ontologically non-conservative ways of theorizing about language. Of course, any way of theorizing about language is going to have some ontological commitments—plausibly, at the very least, commitments to speakers, to speech acts of some kind, and probably to various environmental factors (to help to explain, for example, why such speakers produce such speech acts on such occasions). The distinction between ontologically conservative and ontologically non-conservative theories lies elsewhere. At a first pass, putting the matter in Carnapian terminology, it turns on the question as to whether linguistic theory picks up the internal ontological commitments of the linguistic frameworks theorized about. Recall the analogy I used in Section 2 to illustrate the difference between active and passive rejection of a theoretical notion, the use of the notion of creation in biology or cosmology. Assuming for the sake of the example the principle ‘no creation without a creator’, any theory which describes some or all of the entities within its ontology as created will thereby incur an ontological commitment to something (possibly) additional—i.e., to the existence of one or more creators. For any X such that our theory entails that X is created, the theory thus raises the question ‘What is the Y, such that X is created by Y?’ In one sense, this is a trivial point, turning simply on the fact that ‘is created by’ is relational in form. The bite comes from the fact that as exemplified in Creationism itself, this relation is ontologically non-conservative—it relates biological and physical entities to something not itself part of the biological or physical realm. Semantic relations have an analogous effect in linguistic theory. If we say as linguistic theorists that the term X stands for something, then our linguistic theory itself commits us—in general, at least, though we might want to allow for cases of referential failure—to the existence of something, Y, such that X stands for Y. The question ‘What is the Y, such that X stands for Y?’ is thus a question pressed on us by linguistic theory—even if the Y at issue is

122

huw price

something not normally regarded as part of the required ontology of linguistic theory, such as a prime number, or a moral property. This is what I mean by saying that semantic relations make linguistic theory ontologically non-conservative. Ascribing semantic relations to a body of linguistic utterances commits linguistic theorists to an ontology which mirrors, via the semantic relations in question, the internal ontological commitments of the domain in question.⁷ As I put it elsewhere (Price, 2004a), semantic relations thus provide a bridge, or ladder, that leads our theoretical gaze from words to things. It is easy to see how this gives rise to location problems. We find ourselves saying, in our theoretical voice, that terms such as ‘good’, ‘seven’, ‘cause’, ‘belief’ and ‘truth’ do the job of standing for something. Naturally, we want an answer to the question ‘What do they stand for?’ which meets various theoretical desiderata—fitting with the other kinds of things we are inclined to say about speakers, their environments and their capacities, for example. Hence, given our starting point, a pressure towards naturalistic answers—towards location problems. Semantic deflationism generates no such pressure, however. Because it simply doesn’t ascribe to terms and sentences such relational properties as standing for, referring to, or being made true by, it raises no such issues about the items at the ‘world’s end’ of such relations. Now a crucial question. Does ontological conservatism require that we reject the representational view actively, affirming in our theoretical voice that the terms in question are non-representational? Or is passive rejection sufficient? A moment’s reflection shows that passive rejection is all we need. The ontological non-conservatism of representational approaches is generated by their explicit employment of semantic relations. So long as these relations are simply absent from our theoretical vocabulary, no such ‘extra-linguistic’ ontological commitment arises. Thus the use of semantic notions in linguistic theory generates location problems, by rendering the theory in question ontologically non-conservative. Semantic deflationism blocks this path to location problems, even in the passively non-representational form canvassed in Section 2. The kind of expressivism which results from such semantic deflationism provides a global bar to this route to location problems. Moreover, the point is self-applicable, in the way mentioned earlier. That is, this form of semantic deflationism blocks this semantic route to location problems about the semantic properties themselves just as efficiently as it blocks them about other topics. ⁷ Again, this needs to be qualified to allow for the case of systematic referential failure.

the semantic foundations of metaphysics

123

5. A reductio? The argument that semantic deflationism deflates location problems might seem to provide a reductio of semantic deflationism. For isn’t it obvious that there are some legitimate location problems, e.g., for chemical properties in relation to those of physics? Doesn’t the argument imply that if deflationism were true, we wouldn’t need to worry about the relation of chemistry to physics (perhaps, and surely absurdly, because chemical language and physical language ‘are simply in different lines of work’)? But the argument implies no such absurdity. It simply implies (as in any case seems highly plausible) that our reasons for being interested in the relation of chemistry to physics do not depend on a lot of implicit linguistic theory. In exhibiting the sense in which chemistry and physics are in the same line of work, we can’t simply rely on the claim that they are both representing how things are. If we are semantic deflationists, that characterization is both too thin and too broad to do the work that metaphysics requires of it. (Less metaphorically, relying on representational notions in this way would be incompatible with the central tenet of semantic deflationism, that the semantic properties do no significant theoretical work.) In the case of chemical properties, one familiar view as to how the required justification should go is provided by David Lewis’s account of theoretical identification in science (Lewis 1970, 1972). Here, the unifying thread rests on the fact that the theories to which the method is applicable are causal in nature—they are all in the business of ascribing causes and effects. Modulo some (perhaps controversial) assumptions about causation—e.g., that all causation is ultimately physical causation—this provides a theoretically substantial sense in which all such theories are in the same line of work. Lewis’s program applies to theoretical entities characterizable in terms of their causal role. In taking over some of the key ideas of Lewis’s program, however, Jackson is explicit that he wants to generalize beyond the causal realm. As he notes, his proposed ‘moral functionalism’ differs from Lewis’s psychofunctionalism in that in the case of moral functionalism, the ‘principles are not causal principles’ (1998: 131). What, if anything, plays the role of causation in the wider program? It turns out that so long as we are not deflationists, semantic notions will do the trick. To see how this goes, let’s consider an abridged version of the two programs. Consider first Lewis’s model. Let A be a theoretical term in some scientific theory, and let M(A) be a full specification of A’s causal role. (∃x)M(x) is then the Ramsey sentence that results from M(A) by replacing all occurrences of

124

huw price

‘A’ by a bound variable, and (∃!x)M(x) is the version that says that there is a unique realizer ‘of the A role’. Let’s suppose that we have reason to think (∃!x)M(x) true. What do we now know about A? We have the following definite description: A is the unique x such that M(x). How, according to Lewis, do we now address the question ‘What is A?’, or ‘How do we locate A with respect to the entities described by physics?’? We find, or convince ourselves that in principle we could find, some physical entity D that satisfies the Ramsey sentence for A—i.e., such that D is also the unique x such that M(x). A and D then satisfy the same definite description, and are hence the same entity. What grounds our confidence that there is some such D to be found, at least in principle? If we follow Lewis, it is a thesis about causation—the thesis of ‘the explanatory adequacy of physics’, as Lewis puts it at one point. Taken together with the fact that M(A) encodes the causal facts about A, this gives us what we need. Thus our confidence that there is some such D to be found does not rest on a semantic thesis. Although I used a semantic notion a moment ago—I wrote ‘some physical entity D that satisfies the Ramsey sentence for A’—it is clear, I think, that this use is inessential, and thoroughly compatible with a deflationary view of semantic properties. (Indeed, I went on to gloss what I needed to say without using the notion.) But what happens if we want to generalize Lewis’s program beyond the realm of things with causes and effects?⁸ In this case we won’t in general have definite descriptions in causal terms. But we will have them in semantic terms—descriptions of the form B is the (unique) x that satisfies the Ramsey sentence (∃!x)N(x) (where N(B) is our theory of some such notion B). Problems of uniqueness aside, it is (relatively) trivial that we can form the Ramsey sentence for a given term B in this way. And hence there can be no objection to our proceeding to ask the question ‘What is B?’ by asking What is the (unique) x that satisfies the Ramsey sentence (∃!x)N(x)? What is non-trivial, and dependent on a non-deflationary account of satisfaction, is the view that this is anything other than asking ⁸ Or when we want to apply it to the causal terms themselves, for that matter.

the semantic foundations of metaphysics 125 What is the (unique) x such that N(x)? and that this question has anything other than a trivial answer—‘Why, it is B, of course!’⁹ Non-deflationary semantics holds out a promise of a more interesting answer. It gives us a potential handle on the identity question, in precisely the way that causation does in the Lewisean case. That is, it gives us—at least if combined with semantic naturalism of some kind—a new sort of question to investigate: What is the (natural or physical) thing that stands in this semantic relation to the Ramsey sentence? If deflationists are right, however, there can be no new theoretical question of this kind—and no material with which to fashion a semantic principle supporting physicalism, analogous to Lewis’s principle of the explanatory adequacy of physics.¹⁰ For the moment, the main point is that the fact that deflationism provides a global challenge to Jacksonian metaphysics does not imply, absurdly, that the kind of questions such a metaphysical program raises are never appropriate. It simply implies that extra work is needed to show that they are appropriate, in any particular case—work which could otherwise be done by semantic notions. But in the cases that would otherwise generate an obvious absurdity, such as that of chemical properties, we have a reasonably good idea how the extra work should go. Arguably, Lewis has done it all for us.¹¹

6. Is deflationism obviously false? I have argued that if semantic deflationism is true, then there are likely to be fewer genuine location problems than Jackson imagines. One might concede this conditional claim, and yet argue that its antecedent is obviously false. One might argue, in other words, that it is uncontroversial that much of language is ‘substantially’ representational, in the way apparently denied by semantic deflationists. In From Metaphysics to Ethics and elsewhere, Jackson ⁹ Note that to say that the question ‘What is B?’ has no non-trivial answer is not to say that there is no non-trivial theory in the vicinity. There may well be a non-trivial account to be told—in non-representational theoretical vocabulary—about how the speakers in question come to talk in B terms. ¹⁰ In effect, this point simply reiterates the contrast drawn in §4 between ontologically nonconservative and ontologically conservative modes of linguistic theory. ¹¹ In other contexts I would myself be inclined to contest this argument, maintaining that causation is in various ways unsuited to bear this sort of metaphysical weight (being too anthropocentric and too contextual, for example—see Price 2001). But here my interest is in the contrast between this causally-grounded approach to metaphysics, and a more general, semantically-grounded, approach.

126 huw price often makes observations about language which might be taken in this spirit. For example: Why do foreign-language phrase books sell so well? Because they help us find food, shelter, museums and airports when we travel outside our own language communities. Our need for them highlights what is anyway obvious: much of language is a convention-generated system of representation. Although it is obvious that much of language is representational, it is occasionally denied. I have attended conference papers attacking the representational view of language given by speakers who have in their pockets pieces of paper with writing on them that tell them where the conference dinner is and when the taxis leave for the airport. How could this happen? I surmise that it is through conflating the obviously correct view that much of language is representational with various controversial views. (1997: 270)

For my part, I agree with Jackson that there is a sense in which it is obvious that much of language is representational, and a sense in which it is controversial. Roughly, the obvious sense is the loose and popular sense, which simply tracks folk usage. The non-obvious sense is the theoretical sense, in which representational and semantic notions are called on to do various sorts of theoretical work, in philosophy and elsewhere. As we’ll see, Jackson’s own examples establish that what is obvious in the former sense need not be obvious in the latter—there are cases which are clearly representations in the loose and popular sense, about which, as Jackson himself emphasizes, it is a live theoretical issue as to whether they count as representations in the theoretical sense. Once it is conceded that the popular sense and the theoretical sense come apart in this way, and that the latter is non-obvious, it is open for a deflationist to suggest that there might turn out to be no legitimate theoretical sense—no relevant notion of representation that survives into mature theory about language. Jackson’s appeals to what is obvious and uncontroversial about language seem intended to block this possibility. However, once a wedge has been driven between the obvious and the non-obvious issues—between the loose sense in which language is undoubtedly representational, and the controversial theoretical sense—the former has lost its authority as a guide to theory, and cannot exclude thoroughgoing representational deflationism. Or so it seems to me. In order to make it clear that I’m not relying on judicious choice of examples, let me reproduce some more of Jackson’s illustrations of the representational character of language: We use language to tell our community and our later selves how things are. Telling how things are requires representational devices, structures that somehow effect a

the semantic foundations of metaphysics

127

partition in the possibilities. For we say how things are by saying what is ruled in and what is ruled out. (1998: 53) [T]ypically we know something useful ... and are giving voice to this knowledge when we classify happenings as examples of grooming behaviour, pain, rational inference, and so on. For only then can we explain the manifestly useful information we give about what the world is like to each other and to our later selves, through diary entries and notes on fridges, when we use words like ‘pain’, ‘grooming behaviour’, ‘electricity bill’, ‘belief ’, ‘rational’, and so on. (1998: 64–5) Consider what happens when I utter the sentence, ‘There is a land-mine two metres away.’ I tell you something about how things are, and to do that is precisely to tell you which of various possibilities concerning how things are is actual ... The sports section of any newspaper is full of speculations about possible outcomes, conveyed by sentences that discriminate among the outcomes in a way we grasp because we understand the sentences. Again, we find our way around buildings by reading or hearing sentences that we understand like ‘The seminar room is around the corner on the left.’ There are many different places the seminar room might be located, but after seeing or hearing the sentence, and by virtue of understanding it and trusting the person who produces it, we know which of the possibilities is actual. (1998: 71–2)

The representationalist vocabulary in these examples is, as Jackson intends, unremarkable and uncontroversial. But it remains equally uncontroversial, surely, when we substitute examples like these: a) b) c) d) e) f) g)

What Jane told you is true. David Wenham’s new performance is superb. It is probably going to rain on Friday. You should take an umbrella. It’s possible I’ll be able to get to the party. If I get there, then I’ll be a little late. The delay last week was caused by late arrival of the inbound aircraft.

In notes on fridges and in pockets, all of these sentences may be used to ‘partition the possibilities’, provide ‘manifestly useful information’, and so on. All of them provide examples of things we may learn by understanding and trusting people who say these things. On the face of it, in other words, they all have all the features that Jackson takes to characterize representational uses of language. Yet the italicized parts of these claims pick out elements with respect to which expressivist explications are already reasonably well known. In each case, in other words, we have some idea of how a theoretical account of the function of relevant terms might proceed, without employing representationalist vocabulary at the theoretical level.

128

huw price

Jackson himself notes some possibilities of this kind. He notes that it is controversial whether ethical claims and indicative conditionals are genuinely representational—in such cases, noncognitivism is an arguable position. But it seems to me that in admitting this possibility in particular cases, and yet continuing to maintain that in general, our intuitions about the representational character of language are a good guide to theory—a good reason for thinking that much of language is representational, in a theoretical sense—Jackson is guilty of a kind of double standard. ‘Thick’ representational character simply can’t be obvious in one case and yet not obvious in another, without some (obvious and obviously relevant!) difference in the surface phenomena. The tension in Jackson’s position here is masked, I think, because the sort of noncognitivism he has in mind is the orthodox variety, which includes among its theoretical claims the assertion that the language in question is nonrepresentational. Thus his imagined opponents about the status of ethical claims and indicative conditionals are people who agree that much of the rest of language is representational. From this standpoint, it is easy to overlook the more radical opponent, who—as a semantic deflationist—claims that none of language is representational in a substantial theoretical sense. But Jackson himself acknowledges the gap between what theory tells us about language and what we can read off its surface. Once in play, this gap serves the purposes of the more radical opponent just as well as it serves those of the conventional opponent. (If anything, in fact, it serves the radical opponent better, because unlike her more conventional colleagues, she does not have to explain why surface structure is a guide to underlying function in some cases but not others.)¹²

7. Truth-aptness to the rescue? As noted earlier, Jackson, Oppy and Smith (1994) offer noncognitivists a response to the standard charge that minimalism about truth vindicates ¹² When Jackson mentions the claim that indicative conditionals do not have truth-conditions, he offers what might perhaps be read as an objection to deflationism. He considers the suggestion that it is a trivial matter that indicative conditions have truth-conditions, because it is trivial that ‘If it rains then the match will be cancelled’ is true iff it rains then the match will be cancelled. (Clearly, this is the sort of thing deflationists are keen to say.) Jackson responds that ‘this is grammatically fine, but the issue is not about grammar.’ (1998: 117) Jackson is right, of course, that grammar can’t be the end of the story. But it doesn’t follow that semantic properties provide the appropriate theoretical vocabulary to tell the rest of the story—to say what lies behind the grammar, so to speak. Hence it is open to a deflationist to maintain that it is a trivial issue whether indicative conditionals have truth-conditions, or are representational—of course they do, and are. What isn’t open, in this case, is to deny that there are further theoretical questions to be raised about such conditionals, in non-representational terms.

the semantic foundations of metaphysics

129

cognitivism. They argue that minimalism about truth is compatible with non-minimalism about truth-aptness, and that the latter is what the noncognitivist needs. I have suggested that this is a cure for a non-existent disease. In the respects that matter, minimalism about truth is a friend rather than an enemy of noncognitivism. At least, it is a friend of expressivism, in the broad sense identified above—the view that accords the semantic and representational notions no ‘thick’ role in linguistic theory. It might be objected that deflationism about truth-aptness is in any case highly implausible. After all, the application of the terms ‘true’ and ‘false’, and related marks of the declarative or assertoric use of language, are very prominent features of language. How likely is it that they don’t ‘track’ some theoretically significant underlying characteristic of the speech acts concerned? Such a characteristic—call it the truth-aptness basis, or tab, for short—will be what makes those speech acts truth-apt, in one important sense of the term. And once we’ve found it, won’t we have found a substantial distinguishing mark of the representational uses of language? No, because the argument rests on an equivocation—a kind of use–mention confusion, in fact. Does ‘truth-apt’ mean ‘such as to be treated as truthevaluable’—i.e., such that competent speakers apply the terms ‘true’ and ‘false’ to the sentence in question? Or does it mean ‘such as to be capable of being true or false’? From an anthropological perspective it does seem plausible (let us suppose) that there is some property tab underlying the declarative use of language, including the use of the terms ‘true’ and ‘false’ in conjunction with some kinds of utterances but not others. So it is plausible, as we might put it, that being use-of-‘true’-apt is a substantial theoretical property, waiting to be investigated. But being use-of-‘true’-apt does not imply being representational. It merely implies being treated as ‘representational’. (Compare: religious studies, in its anthropological incarnation, does not require theological ontology.) We can confirm this diagnosis by noting that there is no guarantee whatsoever that tab itself be representational in nature. In the next section I’ll sketch an account with this character. The crucial point is that the underlying property that determines truth-aptness—tab, as we put it—need not involve any ‘word-world’ surrogate for a semantic relation, of a kind which would be useful to metaphysics, or ground a ‘thick’ representational view of language.¹³ ¹³ The argument criticized here is not that of Jackson, Oppy and Smith (1994). Their strategy is different. They suggest that sentences acquire truth-aptness by being conventional expressions of truth-apt mental states, that is, beliefs. And they argue ‘that it is not a minimal matter whether or not

130

huw price

8. The expressivist alternative14 Let’s think about the biological functions of the mental states we call beliefs, setting their (apparent) semantic properties explicitly to one side. Or rather, let’s simply take advantage of the fact that the semantic properties do not appear explicitly in an account of the functional roles of the states in question, and neglect the familiar problem of how to fashion representational content from such functional materials. Let’s begin somewhere familiar, in other words, with the idea that there is some functional characterization of our cognitive lives, in which intentional and semantic terms do not explicitly appear. (In fact, this familiar place is just a rough psychological analogue of what, in a passage quoted at the beginning, Jackson asked us to find plausible with respect to language, namely, a scientific account in which semantic terms do not appear.) How did it serve our ancestors to develop a psychology rich enough to contain such mental states? What role did those states play in increasingly complex lives? It is plausible, in my view, that there is no single answer, appropriate for all kinds of beliefs. Perhaps the function of some beliefs can be understood in terms of the idea that it is useful to have mental states designed to covary with certain environmental conditions, but for many, the story is surely more complicated. Consider causal or probabilistic beliefs, for example, which manifest themselves as dispositions to have certain sorts of expectations (and hence to make certain sorts of decisions) in certain sorts of circumstances. Plausibly, there’s an interesting story to be told about the biological value of having an internal functional organization rich enough to contain such dispositions. Or consider some of the other cases in which expressivism has often seemed appealing—universal generalizations, indicative and subjunctive conditionals,

a state is a belief.’ There is a sense in which this is relatively uncontroversial, I think. Roughly, it is a parallel to the sense in which it is uncontroversial, or at least plausible, that there is some property tab that underlies the fact that certain speech acts get cast in declarative form. But as in that case, there is no guarantee that this property is representational in nature (rather than, say, a matter of the place of the state in question in a certain internal architecture). Jackson, Oppy and Smith also defend non-minimalism about beliefs by appealing to certain folk intuitions or platitudes about belief. These platitudes do have a representationalist character. For example, they speak of the ‘truism about belief that it is a state designed to fit the facts’ or ‘designed to fit the way things are.’ (pp. 296–7) As in the case of the various intuitions and platitudes that Jackson offers in support of a representational view of language, it seems to me that these psychological platitudes have two interpretations, one obvious and one controversial. The obvious reading doesn’t challenge deflationism and global expressivism. ¹⁴ This section overlaps to a considerable extent with §9 of Price 2004b.

the semantic foundations of metaphysics

131

modal and moral claims, and so on. In each of these cases we have some sense of what the associated beliefs enable us to do, which we couldn’t do otherwise—some sense of the distinctive role of the states in question in the psychological architecture of creatures like us. Suppose we get to this point without explicitly invoking the idea that states we are talking about have representational contents. Then we have the beginnings of an understanding of what these various kinds of states do for us—why it’s useful to develop a psychological architecture rich enough to contain them—but no understanding of why they manifest themselves as commitments. Why do we take them to be truth-evaluable, for example, or expressible in declarative form? This issue can be approached in the same explanatory spirit, I think. With respect to the various functionally-characterized states we’ve described, we want to consider the question, why do we give voice to those mental states in that form? And notice a prima facie puzzle. If our account up to this point has emphasized the fact that the states in question have various different functions, it may seem odd that states with different functions all get expressed in a similar way, as truth-evaluable declarative judgements.¹⁵ However, it isn’t difficult to find a place to start. Plausibly, many of our psychological states are such that it is often advantageous, with respect to those states, that we tend towards conformity across our linguistic communities. A prime function of assertoric language seems to be that it facilitates and encourages such alignment. It leads us to ‘express’ or give voice to our psychological dispositions in ways which invite challenge by speakers with contrary dispositions. And ‘That’s false’ and ‘That’s true’ are markers of challenge and concession, respectively.¹⁶ As ordinary speakers, we don’t understand that this is what assertoric language is for—we just do it, as it were, and from our point of view, we seem to be ‘saying how things are’. But the function of this practice of ‘saying how things are’ is the one in the background—the function of altering our behaviourally significant commitments much more rapidly than our individual experience in the environment could do, by giving us access to the corresponding states of our fellows (and much else besides). At any rate, that’s one kind of thing that assertion seems especially well-suited to do. Moreover—and for the moment this is the crucial point—it is something it can usefully do in application to ¹⁵ In contrast, a representational view has it easy at this point. The commonality of linguistic expression reflects the commonality of representational function. But the bump in the rug then shows up somewhere else, when functional differences need to be explained in terms of differences in representational content. ¹⁶ Of course, the details of this need to be spelt out rather carefully—especially if we want to avoid simply helping ourselves to the representational notions. See Price (1988, 1990, 2003).

132

huw price

commitments with a wide variety of functional roles of their own (none of them representation as such). The suggestion is thus that assertions are intentional expressions of psychological states with a variety of (not representationally characterized) functions within the internal psychological architecture of the speakers concerned—or better, within the complex network of relations involving both these internal states and the creatures’ external environment—a variety obscured in their expression, when they take on what Wittgenstein calls the ‘clothing’ of assertoric language.¹⁷ Clearly, there has to be common clothing internally, too. Roughly, the states in question come to participate in a belief-like or rational cognitive architecture, and an important set of issues concerns the relations of precedence between thought and talk at this point. Does belief, or at least full-blown rationally-governed belief, rely on internalizing the dialogical structures which come with the development of assertoric language? Leaving these issues of priority to one side, the proposal is thus that representational language and thought is a tool for aligning commitments across a speech community (and, perhaps, within a single head). But though in one sense a single tool, it is a tool with many distinct applications, corresponding to the distinct primary functions of the various kinds of psychological states that take advantage of it—that facilitate their own alignment by expressing themselves in assertoric form.¹⁸ Another of Wittgenstein’s analogies fits this functional architecture particularly nicely, I think: It is like looking into the cabin of a locomotive. We see handles all looking more or less alike. (Naturally, since they are all supposed to be handled.) But one is the handle of a crank which can be moved continuously (it regulates the opening of a valve); another is the handle of a switch, which has only two effective positions, it is either off or on; a third is the handle of a brake-lever, the harder one pulls on it, the harder it brakes; a fourth, the handle of a pump: it has an effect only so long as it is moved to and fro. (1968: #12) ¹⁷ A major theme of the early sections of the Investigations is that philosophy misses important distinctions about the uses of language, distinctions hidden from us by ‘the uniform appearances of words.’ (1968: #11) As Wittgenstein puts it later, ‘We remain unconscious of the prodigious diversity of all the everyday language-games because the clothing of our language makes everything alike.’ (p. 224). ¹⁸ This kind of multifunctional explanatory structure exists in other places in biology. As I have noted elsewhere (Price 2004b), a good example is the human hand. If we say that the function of the hand is manipulation, and leave it at that, we miss something very important: we miss the underlying functional diversity. On my reading, Wittgenstein makes a closely analogous point about assertion—a point in which representation plays the role of manipulation, as the notion whose homogeneity needs to be challenged.

the semantic foundations of metaphysics

133

In my view, assertions are much like handles. Handles connect various functionally-distinct bits of machinery to the muscles of our hands and arms, and to the brains that control them. Assertions connect various functionallydistinct psychological states to the external and internal cognitive machinery that allows the states concerned to be up-dated on the fly. As I’ve stressed in other work, this approach leads to a very different view from Jackson’s of the phenomena that he takes to give rise to location problems. Jackson and I begin in the same place, I think, with an interest in making sense of a range of distinct ‘packets’ of linguistic usage—talk of semantic properties, talk of ethical properties, and so on. And we’ll agree, presumably, in accepting what science tells us about such talk—as Jackson puts it, in the passage I quoted at the beginning, ‘a story about masses, shapes, causal chains, behavioural dispositions of language users, evolutionary history, and the like’. But then Jackson takes a theoretical move that I want to resist. He takes the talk to be representational, and is led by the path we’ve described to an issue about the various worldly relata of the semantic relations thus in question. Having framed the problem in those terms, he is then rightly troubled by the resulting ontological plurality, and faces location problems. By contrast, the expressivist approach stays at the level of the ‘story about ... behavioural dispositions of language users, evolutionary history, and the like’. In particular, it investigates the functions of the linguistic usage in question, and associated psychology, in non-representational terms, and then tries to account for the representational ‘clothing’ in the way just described. There’s no location problem as such. Instead there is an issue of the same cardinality, so to speak, about the various functions of the various groups of terms and concepts—semantic terms, ethical terms, and the rest. However, this is a problem about human psychology and linguistic behaviour—a broadly anthropological problem, as I put it earlier—not a problem in metaphysics. And of course we expect human behaviour to be functionally pluralistic. So we have unsurprising plurality, in a familiar and scientifically-tractable place, in place of the concerns that motivate Jackson’s ‘serious metaphysics’.

9. What if deflationism is false? Let’s recap. So far I’ve been trying to highlight the semantic presuppositions of Jackson’s program, by exhibiting the effects on the program of assuming semantic deflationism. I haven’t argued for semantic deflationism as such, though I have defended it against two objections: first, that in the light

134

huw price

of my own argument, it would lead to the absurd consequence that there are no location problems whatsoever; and second, that its falsity is in any case obvious, something that can be read off from what we all know to be true about language. For all I’ve said so far, however, deflationism may nevertheless be false. Wouldn’t Jackson’s metaphysical program then be out of the woods? No problem with semantic presuppositions, surely, so long as they are true? I want to close by mentioning two reasons for pessimism—two reasons for thinking that metaphysics grounded on semantics isn’t out of the woods, even if we reject deflationism about the semantic properties. These reasons take us back to some concerns I touched on at the beginning. For the first problem, I’ll do little more than allude to an excellent discussion of a closely related issue by Stephen Stich, in the first chapter of Deconstructing the Mind (Stich, 1996). Stich is concerned with eliminativism about the notions of folk psychology—belief, desire, and so on. He notes that many philosophers take the eliminativist thesis to be that the terms ‘belief’ and ‘desire’ do not refer. In other words, they take the thesis to be couched in these semantic terms. But if that’s what eliminativism is, Stich argues, then in order to assess it we need a theory of reference—a theory capable of guiding our judgement about whether these terms actually do succeed in referring to anything. Stich argues that this leaves metaphysics in an unenviable position. For one thing, it leaves it hostage to the almost inevitable indeterminacies in a scientific theory of reference. In other words, it means that we can’t decide whether eliminativism is true until we sort out the issue between competing theories of reference—and that’s likely to mean ‘never’, given the nature of scientific theory. (The threat of deflationism lurks in the background here, of course, but we are now leaving that aside.) Even worse, in would seem that in crucial cases, the metaphysics needs to precede the theory of reference. In order to decide what relation reference is, we need to be able to examine typical cases. In other words, we need to be able to study the various relationships that obtain between words or thoughts on the one side, and the items to which they (supposedly) refer on the other. But how can we do this in the case of ‘belief’ and ‘desire’, while it is up for grabs whether these terms refer to anything? In order to know where to look, we’d have to know not only that they refer, but also to what. To put this in terms of the location problem: If we need reference to locate belief and desire, we’ll find that we need to locate reference, before we can put it to work. Yet we can’t locate reference until we’ve located its worldy relata. Thus we have two problems for eliminativism, if it is to rely on semantic relations such as reference. (Let’s call the first the ‘referential indeterminacy

the semantic foundations of metaphysics

135

problem’ and the second ‘the precedence problem’.) It’s clear, I think, that both problems arise not simply for eliminativism, but for any metaphysical view that relies on reference in this way. In other words, both problems apply just as much if the question is ‘What is belief?’, if this is to be understood as ‘To what does the term ‘‘belief’’ refer?’, as they do to the question ‘Are there beliefs?’, understood as ‘Does the term ‘‘belief’’ refer to anything?’ Quite apart from the threat of deflationism, then, there are reasons for doubting whether metaphysics can rest on semantic foundations. Stich’s own response to the problem is simply to abandon semantics, and ask the questions in material form: ‘Are there beliefs?’, in place of ‘Does the term belief refer to anything?’ However, while this certainly seems the right move in some cases—folk psychology might be more controversial than Stich thinks, perhaps, but chemistry isn’t, for example—it’s clear that Jackson can’t follow Stich down this non-semantic path. To do so would be to abandon the global semantic route to location problems. In my view, then, Jackson’s program remains vulnerable to Stich’s objections, even if we ignore deflationism.

10. The circularity issue Stich’s objection applies to the use of semantic notions in the metaphysics of other topics. It argues, in effect, that even if not deflationary, these semantic notions aren’t up to the job. It’s not that they won’t bear the weight, but rather that they can’t tell us where the weight is born—which is what we want to know in metaphysics—in advance of an answer to that very question. The semantic legs connect thought and talk to the ground. So we can investigate the ground by following the legs to their feet. But we can’t distinguish the real legs from rivals, until we know where the feet are actually located. But what about the legs themselves? Is there an additional problem about using these semantic props to investigate themselves—to ground the metaphysics of the semantic properties themselves? I think that there is an additional problem. Let’s go back to Section 5. There, I compared the role of the semantic notions in Jackson’s program to that of causation in Lewis’s program. Lewis’s technique for theoretical identification proceeds in two steps. The first step constructs a Ramsey sentence, to give us a definite description of the theoretical entity or property in question, couched in causal terms. The second step turns to science, to discover what it is—couched in other terms—that

136

huw price

fulfills that definite description. In setting out to take the second step, we display our confidence in some principle about causation—for example, that all causation is grounded in physical causation. In Section 5 we noted that so long as we can help ourselves to ‘thick’ semantic notions, there is a global variant of this program. In this case, our definite description of our target entity or property tells us that it is the thing or property that satisfies or makes true the description encoded in the non-semantic part of the Ramsey sentence. Formally speaking, this works fine so long as our target entity or property is not itself semantic in nature. Whether it is really any use in metaphysics is another matter, the matter we’ve just raised.¹⁹ Formally, however, there’s no problem. But consider what happens when our targets are the semantic properties and relations themselves. Here the technique is surely disallowed, because in this case the construction of the Ramsey sentence is supposed to eliminate the semantic terms. This brings us back to the circularity noted at the beginning. Whatever the role that semantic properties play in grounding the location problem for other kinds of properties, they surely can’t play that role on their own behalf. Semantic properties cannot bootstrap metaphysics, in other words, but need to be assumed as primitive brute facts. Thus representationalism becomes a kind of surd for Jackson’s program, both essential and inaccessible to its methods. The trajectory from metaphysics to semantic primitives thus goes something like this. Metaphysics begins with an interest in the nature of things. Semantics seems to guarantee a new mode of description of those things, and a new mode of enquiry: describe the things in question as the referents of our terms, the truth-makers of our sentences, and set out to seek these semantic relata. Scaling-down the ladder metaphor, semantic relations thus come to play the role of white canes, for metaphysicians who take us to be the victims of a kind of world-blindness. As we have seen, deflationists argue these semantic canes are not sufficiently substantial to do the job.²⁰ And Stich argues, in effect, that even if they are sufficiently substantial, they are too ill-constrained to be useful to metaphysics—there are too many competing canes, and we don’t know where a given cane leads unless we are already acquainted with what lies at the far end of it. ¹⁹ One reason for being doubtful whether it helps is Quinean. What it relies on, formally speaking, is simply semantic ascent, and as Quine emphasizes, that takes us nowhere new. (Cf. Simon Blackburn on Ramsey’s ladder, in Ruling Passions.) ²⁰ My aim has been not so much to side with deflationists on this point, but to use the fact that this kind of metaphysics is vulnerable to this charge to highlight the role of the semantic presuppositions in question.

the semantic foundations of metaphysics

137

Setting Stich and deflationists aside, however, the circularity concerns remain. These arise, from this perspective, when we try to use our white canes to guide us to themselves—i.e., when we set out to investigate (or ‘locate’) the semantic properties. There are a number of forms the concern might take. The one just mentioned objects that in the case of the semantic terms themselves, we can’t avail ourselves of the generalized Ramsey–Lewis program, in which semantics plays the role that causation does for Lewis himself. As noted, it might be suggested that this is something metaphysics needs to live with. Perhaps semantic properties need to be accepted as ‘brute primitives’. However, this primitivism amounts to philosophy’s imposing a view about the functions of human language. This alone should surely concern naturalists, but all the more so if we agree with the remarks of Jackson I quoted at the beginning, to the effect that semantic properties are likely to be no part of the ‘surface’ scientific story about language. It is bad enough if the semantic categories claim a prior entitlement to a seat at the scientific table, but worse still if they have no other credible reason to be there. We have seen that for Jackson himself, the claim of the semantic properties seems to rest on the proposal that they are implicit in the scientific story. There’s no comfort here for a naturalist bothered by primitivism, however, because in the case of semantic properties, as in other cases, Jackson’s conception of the task assumes a representational view of semantic vocabulary in question. Hence it assumes precisely what such a naturalist is seeking reason to accept. These issues are admittedly very difficult, and it would be premature to insist that there’s no non-vicious way to complete this circle, whether in a form acceptable to naturalists or otherwise. Still, in the absence of further clues as to how the solution might go, I think that there are serious grounds for concern. Representationalism seems to operate as a kind of transcendental foundation for Jackson’s program (and, though perhaps less explicitly, for much of contemporary metaphysics). If so, then at the very least this is something that needs to be clearly acknowledged. (Novices should be told that there are alternative conceptions of the task of philosophy, which don’t depend on such a primitive.) With those alternatives clearly in view, we could proceed to consider what seem to me the hard and important questions in this area. Are these circularities inevitably vicious, from a naturalistic perspective, or should we regard representationalism as a coherent and open question, a matter to be settled by future science? Without alternatives in view, the possibility that representationalism is deeply incoherent looks disastrous, a kind of scepticism

138

huw price

that philosophy cannot afford seriously to entertain. With alternatives, it simply looks like a fascinating (possible) philosophical discovery, namely, that semantic properties are too ‘queer’ to be part of a scientific account of language.

11. Philosophy without ladders Representationalist philosophy has been to this point before, I think. Semantic relations seem an essential ladder for serious metaphysics, but a ladder which is itself unreachable by its own rungs. Serious metaphysics thus seems close to the point that Wittgenstein reaches at the end of the Tractatus, when he offers us a view of the relation between language and the world that can’t be talked about—a view inaccessible by its own lights. (There are other similarities, too, I think. Jackson’s generalized Ramsey–Lewis structuralism seems to offer us at the level of theories, or perhaps conceptual clusters, what Wittgenstein’s picture theory of meaning offers us at the level of sentences: a view in which linguistic structure mirrors ontological structure, via semantic mappings.) Famously, that (early) Wittgenstein urged us to kick away the ladder. Whatever he meant by that, he later advocated something less paradoxical. He turned away from the kind of philosophy that required such a ladder. Rather than climbing it and kicking it away, his later recommendation seems to have been to avoid it altogether: I might say: if the place I want to get to could only be reached by way of a ladder, I would give up trying to get there. For the place I really have to get to is a place I must already be at now. Anything that I might reach by climbing a ladder does not interest me. (1980: 7e)²¹

The nature of Wittgenstein’s later philosophy is a matter for much dispute, of course. Under at least one interpretation (Price 1992, 2004b), however, it is close to what I have called global expressivism—a view that sees the right question, everywhere, as a question posed in non-representationalist terms. What are creatures like us doing with this bit of language? As I have emphasized elsewhere (Price 1992, 1997, 2004a) this is a thoroughly naturalistic viewpoint. Wittgenstein thus becomes a kind of pragmatic naturalist, interested in explaining our use of philosophically puzzling concepts—value, causation, modality, meaning, or whatever—in terms of their role in our cognitive and practical lives. ²¹ I am indebted to Brad Weslake for this reference.

the semantic foundations of metaphysics 139 Another of Wittgenstein’s later themes is that philosophy is a kind of self-help therapy, whose role is to help us to overcome our own philosophical cravings—the craving for metaphysics, for example. In my terms, this therapy takes a specific form. Its goal is to cure the delusion that we are world-blind, to correct the particular form of philosophical hypochondria that leads us to believe that there is a problem that semantic white canes can solve. The key to the cure is to see that the delusion results from imbibing an unhealthy combination of naturalism and representationalism. Some philosophers²² recognize that this combination is unhealthy, and seek relief by abandoning naturalism. Some think that this is Wittgenstein’s prescription. But in my view the right choice, and arguably Wittgenstein’s choice, is the other one. We should keep the naturalism and dispense with the representationalism. Either way, the question of status of the semantic properties well deserves the prominence that Jackson’s book accords it, even if for reasons that he himself does not highlight. For naturalists, in my view, this question is the watershed between two radically different conceptions of the task of a scientifically-grounded philosophy. On one side lies metaphysics; on the one other, scientific anthropology and the study of certain aspects of human linguistic behaviour and psychology. Though Jackson turns the wrong way at the divide, in my view, his book provides the best existing map of the territory on that side. We expressivists—comparatively lonely, these days, on what remains the shady side of the ridge—should surely hope for a Jackson to call our own. University of Sydney

References Boghossian, Paul, 1990. ‘The Status of Content’, Philosophical Review 99: 157–84. Horwich, Paul, 1988. Truth, 2nd edn., Oxford: Oxford University Press. Humberstone, Lloyd, 1991. ‘Critical Notice of F. Jackson, Conditionals’, Philosophy and Phenomenological Research 51: 227–34. Jackson, Frank, 1997. ‘Naturalism and the Fate of the M-Worlds II’, Proceedings of the Aristotelian Society, supp. vol. 71: 269–82. 1998. From Metaphysics to Ethics, Oxford: Clarendon Press. Jackson, Frank, Oppy, Graham and Smith, Michael, 1994. ‘Minimalism and Truth Aptness’, Mind 103: 287–302. ²² e.g., perhaps, Paul Boghossian (1990).

140

huw price

Lewis, David, 1970. ‘How to Define Theoretical Terms’, Journal of Philosophy 67: 427–46. 1972. ‘Psychophysical and Theoretical Identifications’, Australasian Journal of Philosophy 50: 249–58. McDowell, John, 1981. ‘Anti-realism and the Epistemology of Understanding’, in J. Bouveresse and H. Parret, eds., Meaning and Understanding, Berlin: W. de Gruyter, 225–48. O’Leary-Hawthorne, John and Huw Price, 1996. ‘How to stand up for noncognitivists’, Australasian Journal of Philosophy 74: 275–92. Price, Huw, 1988. Facts and the Function of Truth, Oxford: Basil Blackwell. 1990. ‘Why ‘‘Not’’?’, Mind 99: 221–38. 1992. ‘Metaphysical Pluralism,’ The Journal of Philosophy 89: 387–409. 1997. ‘Naturalism and the Fate of the M-Worlds I’, Proceedings of the Aristotelian Society, supp. vol. 71: 247–67. 2001. ‘Causation in the Special Sciences: the Case for Pragmatism’, in Domenico Costantini, Maria Carla Galavotti and Patrick Suppes, eds., Stochastic Causality, Stanford: CSLI Publications, 103–20. 2003. ‘Truth as Convenient Friction’, The Journal of Philosophy 100: 167–90. 2004a. ‘Naturalism Without Representationalism’, in David Macarthur and Mario de Caro, eds., Naturalism in Question. Cambridge, Mass.: Harvard University Press, 71–88. 2004b. ‘Immodesty Without Mirrors—Making Sense of Wittgenstein’s Linguistic Pluralism,’ in Max K¨olbel and Bernhard Weiss, eds., Wittgenstein’s Lasting Significance, Boston: Routledge & Kegan Paul, 179–205. Stich, Stephen, 1996. Deconstructing the Mind, New York: Oxford University Press. Wittgenstein, Ludwig, 1968. Philosophical Investigations, 3rd. Eng. ed., Oxford: Basil Blackwell. 1980. Culture and Value, G. H. von Wright (ed.), P. Winch (trans.), Chicago: University of Chicago Press. Wright, Crispin, 1992. Truth and Objectivity, Cambridge, Mass.: Harvard University Press.

6 The Folk Theory of Colours and the Causes of Colour Experience PETER MENZIES

1. Introduction The subject of colour has long fascinated Frank Jackson. In his book Perception (1977), Jackson advanced a sense datum theory of perception, according to which colours are properties of sense data rather than of material objects. Since then, however, he has changed his mind, and in recent times has affirmed a physical realism about colours, sometimes known as Australian realism about colours, according to which colours are distinctive physical properties. He first stated this position in his (1987) paper with Robert Pargetter. But he has revisited the topic several times (1996a, 1998, 2007), further articulating his view and distinguishing it from various forms of subjectivism that state that colours are essentially dispositions to cause distinctive visual experiences in human subjects. Perhaps one reason that colours have fascinated Jackson is that they represent a striking instance of what he has called a ‘location problem’ (1998). The ‘location problem for colours’ is simply the problem of determining whether colours have any location or role within the metaphysical world picture given by the physical sciences. Colours represent a puzzle for this world picture since the physical sciences do not make any explicit mention of colours in their laws or explanations. To solve this puzzle Jackson argues that it is necessary to examine our commonsense use of the colour concepts. For, after all, what is of interest is not whether colours, as idiosyncratically conceived by some group, fit into the physicalist world picture, but whether colours, as ordinarily conceived by the folk, do so. Jackson assumes that the folk colour concepts are theoretical concepts in the sense that the contents of the concepts are fixed by their role in a theory, in this case a loosely bound together set of platitudes about the colours accepted by the folk. As theoretical concepts of

142

peter menzies

this kind, they lend themselves to definition by the Ramsey–Carnap–Lewis technique. By gathering together the platitudes or axioms of the folk theory and ramsifying over them, one can formulate comprehensive definite descriptions that putatively pick out the colours. Having formulated such descriptions, one can ask whether there is anything in the physicalist world picture that answers to them. Section 2 below sketches Jackson’s solution to ‘the location problem for colours’. Jackson’s methodology is controversial. For example, one might question whether there is a role in philosophy for the kind of a priori conceptual analysis embodied in the Ramsey–Carnap–Lewis approach to concepts; or whether its implicit acceptance of a descriptivist theory of the reference is correct; or whether the technique of ramsifying the axioms of the folk theory of colours can lead to a posteriori, contingent identifications of colours. I shall not, however, question this methodology. Indeed, I have been persuaded by Lewis and Jackson of its considerable merits. (For an application of the methodology to provide a theoretical definition of causation see my paper 1996.) What I will question, on the other hand, is whether Jackson’s account of the folk theory of colours is satisfactory. He accepts two basic axioms of this theory—a causal axiom and a representation axiom to be described below—but his acceptance of these axioms generates a problem for him. On the one hand, he employs the first axiom to argue for his realist identification of the colours with physical properties. On the other hand, he recognizes that this identification of the colours seems to clash with the second, representation axiom. Jackson tries to resolve the apparent tension between the axioms by supposing that there are two kinds of properties that might be identified with colours and that one satisfies the causal axiom and the other satisfies the representation axiom. In Section 3 I argue that Jackson’s attempted resolution of the clash between the two axioms of the folk theory fails because of its inconsistency with other firm folk intuitions about colours. In the following section, after agreeing with Jackson that the folk theory entails both axioms, I argue that these axioms imply a conception of colours as simple, non-physical, intrinsic properties of objects. Such a primitivist conception should, I claim, embrace the additional doctrine that the colours supervene on, without being identical to, physical properties. And when it is supplemented by this doctrine, the folk theory yields a theoretically interesting and completely consistent conception of colours. In Section 5 I take up the issue of why Jackson rejects this primitivist conception of colours and the easy solution it provides to the apparent clash between his two basic axioms. The chief reason comes in the form of an assumption to the effect that physical properties of the kind described in

the folk theory of colours

143

psychophysics are the normal causes of colour experience, and these properties exclude all other properties, including primitive non-physical colour properties, as normal causes. In Section 6 I examine this exclusion assumption and its role in his argument for the conclusion that colours must be physical properties. I contend that the exclusion argument must be mistaken because it leads to absurd consequences about causation. In the following section I outline a conception of causation according to which non-physical properties as well as physical properties can both cause colour experience in ways that do not compete with each other and do not exclude each other.

2. Jackson’s physical realism about colour Jackson advances an appealingly simple argument for his physical realism about colours. It is a transparently valid argument from two premisses to a conclusion. The first premiss is a thesis that he takes to be an axiom of the folk theory of colours. He calls it the causal axiom. Causal axiom: Red is the property of an object that is the normal cause of its looking red. This thesis captures the idea that colours present themselves to visual experience as features of external objects rather than of the experience itself, and that a condition of properties presenting themselves to experience is that they are the normal causes of that experience. The second premiss of the argument consists of a thesis that Jackson claims to be a truth that has been established scientifically. Empirical truth: The only normal cause of an object’s looking red is a complex of physical properties. From these premisses, Jackson straightforwardly infers his physical realist identification of colours: Conclusion: Red is a complex of physical properties of objects. I shall eventually question the soundness of this argument. But it is worthwhile pausing to note some of the good features of Jackson’s account. First, the argument seems to explain the attractiveness of subjectivist accounts that identify colours with dispositions to cause distinctive visual experiences. Like these subjectivist accounts, Jackson’s account says that what counts as the colours depends on what causes human colour experiences. However, unlike these subjectivist accounts, Jackson’s account implies that the properties that

144

peter menzies

count as the colours have some objective, observer-independent nature that goes beyond their power to produce colour experience. Secondly, Jackson’s account explains why colours cannot be dispositions to cause visual experiences. The causal axiom says that colours are the normal causes of colour experiences. As Jackson points out, dispositions cannot be causes in the relevant sense of their manifestations. For example, if a fragile vase is dropped and broken, it is not the vase’s fragility that causes it to break, but rather the property that forms the basis of the disposition—its molecular structure or whatever—that, in conjunction with its dropping, causes the breaking. Similarly, it is not an object’s disposition to cause a colour experience that causes the experience, but rather the intrinsic property of the object that forms the basis of the disposition. Colours, as the normal causes of colour experience, are best thought of as the bases of dispositions to cause such experiences and not the dispositions themselves. Thirdly, Jackson’s account recognizes the implicit relativity of the colour concepts. Jackson’s full account of the colour concepts makes them relative to a kind of subject and a kind of viewing circumstance, since the way properties cause visual experience depends crucially on the kind of experiencing subject and the kind of viewing circumstances. As he puts it, physical realism about colour really combines a causal theory of colour—the view that the colours are the properties that stand in the right causal relationship to colour experiences—with empirical information about what affects the causation of colour experiences. The relativity of colours to subjects and circumstances is of a piece with the ultimately a posteriori, contingent character of their identifications with other properties.

3. The clash between folk theory and Jackson’s physical realism Jackson’s physical realist theory of colour is based on the axiom that colours are the normal causes of colour experience. I do not dispute that this is an axiom of the folk theory of colours. It seems to me, however, that there are other axioms of the folk theory, and that these other axioms clash with and undermine the view that colours are physical properties. In this section I shall consider several such axioms and evaluate Jackson’s response to the fact that they appear to have untoward implications for his physical realism about colour. Several philosophers (notably, Galen Strawson 1989) have proposed, as a crucial platitude of the folk conception of colours, the thesis that colours

the folk theory of colours

145

are properties whose natures are fully revealed, or are transparent, to us in visual experience. Mark Johnston (1992) calls this the revelation thesis and John Campbell (1993) calls it the transparency thesis. This does, indeed, seem to have enough intuitive force to count as a crucial platitude. (However, see the critical remarks in David R. Hilbert (1987): 29–42.) But if it is a crucial platitude of the folk conception, it implies the falsity of the view that colours are physical properties. For colours do not present themselves to visual experience as physical properties. They do not look like spectral reflectance profiles, or any other kind of physical property. So the implied opacity of their true natures to visual experience contradicts the revelation thesis. Jackson discusses the revelation thesis explicitly in only one place and he rejects it for a number of reasons (1998: 102–5). I will not, however, detail his reasons, because in his most recent presentation of his views on colour (2007), it seems to me that he accepts a thesis that is closely related to the revelation thesis. This recent paper advances a representationalist account of the content of colour experience, which highlights the realist character of his position about colour. Jackson argues that it is axiomatic, not just that colours cause our distinctive visual experiences, but that these experiences represent properties in the world. Thus, when we look at a rose, our visual experience represents the rose as having a certain property. In Jackson’s view, the claim that colour experiences have a representational content sets his form of physical realism about colour apart from any kind of subjectivist view. For this claim is inspired, as Jackson puts it, by diaphonousness, ‘‘the view that the properties of experience are putative properties of that which is experienced’’. This suggests that the content of our colour experiences are answerable to the way things are in the world. In contrast, a genuinely subjective property like being poisonous depends ultimately only on the way things are with us: a substance’s being poisonous is answerable at bottom to whether or not it typically causes illness in those who ingest it. Colours are different in this respect: when something looks red to us, the experience does not represent how we or our bodies are, but how the things seen are, so that the conditions on the accuracy of the colour experience are conditions on how the world is. Jackson actually distinguishes a weak and a strong claim endorsed by representationalists. The weak claim says that colour experience is essentially representational. The strong claim says that colour experience is exhaustively representational, so excluding any component to colour experience such as phenomenal feel beyond representational content. Jackson accepts both the weak and the strong claims. He says that it is the strong claim that representational content exhausts the nature of experience that highlights the realist character of his view.

146

peter menzies

On the other hand, the weak claim has some interesting consequences of its own. Jackson points out that the weak claim by itself seems to generate what he calls ‘the opacity problem’, the problem that the true natures of colours are opaque to visual experience. Thus, his physical realism implies that red, for instance, is a physical property discovered by physical science. For the sake of definiteness, let us suppose that it is discovered to be a specific reflectance profile, RRP. But this identification of red with the physical property RRP seems to clash with the weak representationalist claim that colour experience is essentially representational. The clash consists in the fact that when things look red, they do not look as though they have the property RRP. (For the purposes of the argument, it does not matter which physical property red is identified with: any plausible candidate will be sufficiently obscure to generate the opacity problem.) It is not clear to me how Jackson’s weak and strong representationalist claims relate to the revelation thesis accepted by Strawson, Johnston, and Campbell. Is the revelation thesis equivalent to the weak claim, or to the strong claim? Or does the revelation thesis just imply the strong (and so the weak) claim, but not vice versa? Whatever the answers to these questions, it is clear that even the weak claim creates the same kind of opacity problem for physical realism about colour as the revelation thesis. Since Jackson is aware of this problem, it is not surprising that he has a clever answer to it. He distinguishes two ways in which properties might be represented in sensory experience. Our senses, he says, keep track of some properties qua the properties they are, and other properties they keep track track of qua the roles they play. In the first kind of case, an F experience represents that something is F; in the second kind of case, an F experience represents that something has the property playing the F-role. Jackson contrasts the perception of shape with the perception of colour in this regard. The perceptual experience of shape represents the property per se, whereas the perceptual experience of colour represents it qua the role it plays. How does colour experience represent colours qua the roles they play? Jackson draws an analogy with heat. Even though heat is mean kinetic energy, our sensory experiences of heat do not represent it as such, but as the property that is opposite to cold, that comes in degrees, that is harmful to experience for too long; that is, the property that occupies a certain role. And the same is true for colours. Even though red is a physical property, perhaps RRP, our visual experiences do not represent it as such, but as the property that has a certain saturation, brightness and hue, that is more similar to pink than to orange, that is less similar to blue than to crimson, and so on. For short, it is represented as a property that occupies a certain place in the colour solid.

the folk theory of colours

147

When we philosophize about the colour red, Jackson argues, we must distinguish between two kinds of properties. One property might plausibly be identified with a physical property such as RRP. It is this property that satisfies the causal axiom: Causal axiom: Red is the property that is the normal cause of our red visual experiences. (Red as realizer-property.) Jackson calls this property a realizer-property. The other property we need to attend to is a property of the realizer-property: it is the complex relational property of having a realizer-property that occupies the role of red. He calls this second property the role-property. On his view, then, when we have a visual experience of something red, our experience represents the thing as this role-property—the complex relational property of having some property occupy the role of red. It is this role-property that satisfies the axiom that Jackson formulates to capture the weak representationalist claim: Representation axiom: Red is the property that looking red represents things as having. (Red as role-property.) In this way, Jackson resolves the apparent clash between his physical realism about colour, which is implied by the causal axiom, and the weak representational claim, which is expressed by the representation axiom. There is in fact no clash, he argues, because the causal axiom concerns colours as realizer-properties and the representation axiom concerns colours as role-properties. Jackson’s resolution of the clash is far from convincing, in my opinion. I would argue that it fails for several reasons. First, it does not seem to be true to the phenomenology of visual experience. Visual experience does not reveal a significant qualitative difference between the experience of shape and colour: the redness of a ball does not seem to be represented in a different way from its shape. But such a phenomenological difference would have to exist on Jackson’s account, as shape is represented qua the property it actually is whereas colour is represented qua the role it plays. Secondly, the representational content of our visual experiences suggest that colours are simple intrinsic properties of objects. The red of a rose presents itself to our visual experience as a simple intrinsic property of the rose itself. The ‘diaphonousness of colour experience’, which Jackson accepts, implies that the putative features of the experience are features of the things experienced. So, given that we experience the colours as simple intrinsic properties of objects, we may infer that colours in themselves are simple intrinsic properties of objects. But these characteristics are inconsistent with Jackson’s account, which says that the red of the rose is represented as a

148

peter menzies

complex relational property of a realizer-property. This is highly implausible in the light of the apparent simplicity of the actual contents of our colour experiences. Thirdly, it is clear that the folk concept of colour represents it as a unitary phenomenon. Red, for example, is that very property that causes our red visual experiences and is also represented by these visual experiences. It is one and the same property that does both things. In other words, red is the unique property that satisfies both causal and representation axioms. The unitary character of the colours seems to me to be a central feature of the folk theory of colours. It is part of the folk theory, though not a separate axiom, that red, if it exists, is the unique property that satisfies all the axioms. But this is inconsistent with Jackson’s account that states that the red that satisfies the causal axiom is a different kind of property from that of the red that satisfies the representation axiom. On this point, Jackson seems to be mistaken about the nature of the folk theory. For these reasons it seems to me that Jackson’s resolution of the clash between his physical realism and his representationalism about colour fails. A physical realist about colour cannot consistently accept the revelation thesis, nor even anything as strong as the representation axiom. On the other hand, for someone who thinks that representation axiom or revelation thesis is central to the folk theory of colours, there is a way of accepting Jackson’s formulation of the folk theory of colours while rejecting his identification of colours with physical properties, or so I will argue in the next section.

4. Folk primitivism about colours For the sake of simplicity, I shall concentrate henceforth on a formulation of the folk theory that takes it to consist of two axioms: the causal axiom and the representation axiom. (The revelation thesis may be substituted for the latter if it is found to be more intuitive.) It seems to me that the folk theory consisting of these axioms present a very distinctive picture of colours. The first axiom tells us that since colours are the normal causes of colour experiences, they cannot be dispositions to produce those experiences. The second axiom tells us that colours must have certain characteristics by reason of our visual experiences representing them as having those characteristics. Thus, our visual experiences tell us that colours are features of the world and not of our experiences themselves; that they are simple intrinsic properties of objects; and that they are indistinguishable

the folk theory of colours

149

in qualitative kind from primary qualities such as shape; and that they are not physical properties such as spectral reflectance profiles. What kind of objective but non-physical properties of objects are they? I think that folk theory takes colours to be like shapes in a certain respect. Just as the property of being round, for example, is not reducible to other properties, is not identical to the disposition to look round, and is veridically represented as round in our visual experiences, so the property of being red is not reducible to other properties, is not identical to the disposition to look red, and is veridically represented as red in our visual experiences. Like shapes, colours are simple irreducible properties that are sui generis. This view has also been called primitivism about colours (Alex Byrne and David R. Hilbert (1997)). It has been advanced in the past by James Cornman (1975), and more recently by John Campbell (1993), who calls it the simple view of colours. My claim about primitivism about colour is different from Cornman’s and Campbell’s. I do not claim, as they do, that primitivism is the correct view about colour, but merely that the folk theory of colour is committed to primitivism. Primitivism about colour might be developed in a number of ways consistently with what the folk theory says about colour. One plausible thing to say is that colours are supervenient on physical properties. This might be spelled out in terms of global supervenience: any two worlds that agree in the distribution of physical properties and relations, and in their physical laws, must also agree in their distribution of colours. However, it is important to realize that this supervenience claim is not a conceptual truth about colours that follows from the axioms of the folk theory. A philosopher of a prescientific age might, it seems to me, have denied the supervenience claim consistently with endorsing the causal and representational axioms. The plausibility of the claim, such as it is, rests on our a posteriori knowledge of the physical processes involved in the reflection and transmission of light and the neurophysiological processes involved in colour perception. Notwithstanding its a posteriori status, the primitivist can and should endorse it. Philosophers since Galileo have argued that the physical sciences make colours otiose, as they play no role in the laws or explanations of these sciences. This argument proves too much, as I shall try to show in the next section. If it were sound, it would show that all properties employed in the non-physical sciences are otiose as well. The ontological standards implied by the Galilean argument are far too severe. We acknowledge the special status of the physical sciences by affirming the supervenience of colour on the physical. And that is all that is required in the way of acknowledgement. Conversely, the supervenience

150

peter menzies

of colours on the physical does not rob them, as some have thought, of their causal independence. To concede that colours supervene on the physical is, as I try to show in the next section, compatible with allowing that they have causal powers independent of those of the physical. Just as it is common to think that the biological, mental, and social properties invoked by the special sciences are causally autonomous even though supervenient on the physical, so we should think that the same is true for colours. Primitivism about colours raises some intriguing philosophical questions. For example, primitivism takes colours to be mind-independent properties, but ones whose natures are fully transparent to certain mental experiences. Furthermore, primitivism supposes they are objective properties, but ones that can be grasped only from an idiosyncratic perspective. How colours can have these seemingly contradictory characteristics raises serious philosophical questions about the nature of their objectivity or mind-independence. Campbell has begun the exploration of these questions in his (1993), but more work needs to be done. However, rather than pursuing these questions, I shall turn to consider the reasons Jackson gives for rejecting primitivism out of hand.

5. The exclusion argument about colours One argument has been enormously influential in persuading Jackson and others that primitivism about colours is inconsistent with the empirical findings of the physical sciences. The argument receives a very explicit formulation in Jackson’s work (1998, 2007). (Johnston formulates the argument in similar terms in his (1992).) In this section I criticize this argument on the grounds that it trades on an erroneous conception of causation. The argument goes like this. Psychophysics reveals that there are only certain possible candidates for the normal causes of our colour experiences. These candidates are dispositions of light to be reflected or transmitted in certain ways, dispositions of objects to look coloured, and the non-dispositional microphysical properties that underlie these dispositions. This exhausts the list of possible causes. Crucially, the irreducible non-physical properties that the primitivist identifies with colours are not on this list. Moreover, to suppose that these properties exist as overdetermining causes in addition to these physical causes would be to entertain a pre-established harmony between causes that would be miraculous. So barring any such overdetermination, it follows that primitive non-physical colours are not among the causes of our colour experience.

the folk theory of colours

151

As we have seen, the upshot of this argument for Jackson is that the kind of physical property that psychophysics tells us can satisfy the causal axiom is not the kind of property that can satisfy the representation axiom. So acceptance of both axioms involves equivocating on what we mean by colour properties. Johnston also takes psychophysics to imply the same moral (except that he discusses the revelation thesis rather than the representation axiom), but draws a different conclusion. He concludes that it is necessary to give up the revelation thesis in order to ensure the consistency of folk theory with the findings of psychophysics. In my view, both philosophers are mistaken in their conclusions because the argument that purports to demonstrate the incompatibility of the folk theory with psychophysics does no such thing. The key to seeing its error is to recognize that the argument is the exact analogue of an argument used in the mental causation debate to show that irreducible mental properties are not causally efficacious in producing human behaviour. This is the familiar exclusion argument discussed at length by Jaegwon Kim (1998) among others. In its more familiar guise, the argument runs that neurophysiology reveals that the only possible causes of human behaviour are neural or neurophysical properties of brain structures. Irreducible mental properties are not on the list of possible causes. Furthermore, to suppose that such irreducible mental properties are overdetermining causes operating in addition to the known neural causes would be to entertain an incredible pre-established harmony between causes. Rejecting such overdetermining causes, we are left with the conclusion that mental properties must be epiphenomenal, or if they are causally efficacious, they must be identical with physical properties after all. There is not the space here to examine in detail the exclusion argument as it applies to mental properties. Suffice to say, many philosophers will regard argument as mistaken. For many philosophers have been convinced that since mental properties are multiply realized by neurophysical properties they are not identical to them. Nonetheless, these philosophers are reluctant to admit that mental properties are epiphenomenal. Many are attracted to a non-reductive materialism that takes mental properties to supervene on physical properties but to possess genuine causal powers that are independent of their physical supervenience bases. The popularity of this position has meant that there is no shortage of diagnoses of the error of the exclusion argument as it applies to mental properties. My own view is that the central error of this argument lies in its implicit assumption that, barring overdetermination, one set of normal causes for behaviour excludes other sets of normal causes. I argue in my (2003) that a correct understanding of the folk and scientific conception of causation shows that it is not the concept of a single context-invariant relation, but

152

peter menzies

rather a context-sensitive one that picks out underlying processes relative to idealizing models. This fact, I claim, allows for the possibility of multiple causal pathways leading to a given effect, pathways that do not compete with each other and do not count as overdetermining causes. However, it is possible to see that something must be wrong with the exclusion argument, without going into the details of the conceptual analysis of causation. For the exclusion argument proves far too much, as many have pointed out. Consider any of the properties invoked in the special sciences: it may be a biological property like fitness, a psychological property such as preference ordering, an economic property such as gross domestic product. These properties are usually thought to supervene on, without being reducible to, physical base properties. The special sciences employ these properties in causal laws to explain and predict phenomena; and these causal laws are remarkably informative because they operate at appropriate levels of abstraction by virtue of generalizing over the specific ways in which the properties are realized by physical properties. But in each case it is possible to extrapolate the exclusion argument by claiming that such special science properties are excluded from being causally efficacious in producing any effect because the lower-order base properties that realize them can, in principle, be shown to have the causal powers to produce those same effects. Indeed, the same style of exclusion reasoning shows that the microphysical properties of objects that Jackson identifies as colours cannot in fact cause colour experiences: they too supervene on more fundamental quantum-mechanical properties of objects and these lower-order properties can be shown to be the real causes of colour experiences, or at least their neural correlates. Clearly, extrapolating exclusion reasoning in this way will lead to the conclusion that all properties, with the sole exception of the properties of the most fundamental physics, are causally inefficacious. Most philosophers will rightly regard this conclusion as a reductio ad absurdum of exclusion reasoning. However, somewhat surprisingly, Jackson does seem to accept this conclusion. (For example, see his (1996b).) From time to time, he avows that the only genuine causal relations are to be found at the level of fundamental physics. This is surprising since it is barely consistent with his other views about the role of conceptual analysis in philosophical methodology and the importance of ordinary practices in fixing the subject matters of beliefs. For the extreme physicalism he sometimes espouses about causation implies that not only commonsense practice, but also much scientific practice, are somehow mistaken about the nature of causation and systematically err in applying causal concepts in ways unconstrained by considerations of fundamental physics. This is hard to reconcile with the fact that applications

the folk theory of colours

153

of causal concepts in the special sciences seem to be completely in order, even though difficult to analyze. Many causal claims involving biological, psychological, and economic properties pass all the most refined scientific tests of causation; and so, on Jackson’s own methodology, satisfy the conditions of correct application of these concepts. This methodology, moreover, implies that if we are discussing what causation is, our best option is to examine the practices of the scientific community regarding causation rather than the recherch´e beliefs of metaphysicians. Perhaps it may be thought that Jackson’s views are more nuanced than I am acknowledging here. To be sure, it is possible to minimize the distance between Jackson’s extreme physicalist conception of causation and ordinary conceptions by supposing that he actually thinks of causation as coming in different grades. Perhaps he should be seen as thinking that there are different kinds of causation: there is what we might call industrial-strength causation of the kind found in fundamental physics and a lighter grade causation of the kind found in the special sciences and commonsense experience. On this view, ordinary folk and the practitioners of the special sciences are really thinking of the weaker kind of causation—causation-lite, so to speak—in attributing causal powers to non-physical properties rather than the full industrial-strength kind, but they are not too mistaken in their causal beliefs because causation-lite supervenes on industrial-strength causation. It seems to me that such a reconstruction of Jackson’s views does have some merit, especially in view of the fact that in his work with Philip Pettit on program-explanations versus process-explanations Jackson seems to countenance precisely a distinction of grade between two kinds of causal explanation. (See Jackson and Pettit (1988) and (1990).) Such a reconstruction would allow him to speak with the folk and say that colours and mental properties have causal powers, albeit of the lighter grade. But it would also allow him to say that these properties lack the real causal powers enjoyed by the physical properties of fundamental physics. So he could then claim that these properties must be relatively causally inefficacious, by comparison with their underlying physical bases, in producing the effects taken to be characteristic of them. Whatever the merits of this as a reconstruction of Jackson’s position, it does not help to answer the criticism made above. For the same absurd conclusion results from extrapolating this form of exclusion reasoning to the properties of the special sciences. And, as we have seen, it is just not plausible to say that these properties are causally inefficacious, or even that they are relatively inefficacious by comparison with physical properties. Though this reconstruction may conceal its implications by the double talk of different

154

peter menzies

grades of causation, it relies on an error theory about causation that is just as unmotivated as the original. There is another way in which Jackson might seek to block this criticism of his use of the exclusion argument. This would be to concede that the unrestricted application of the exclusion argument leads to trouble; and then argue that a restricted application to colours is justified by virtue of special facts about colours that do not hold true of more generally, and in particular do not hold true of the properties of the special science. In this connection, one popular thought has been that there are no respectable scientific laws governing non-physical colours: the only genuine colour laws are those couched in terms of physical properties. Accordingly, it might be claimed that this is a difference between non-physical colours and the properties of the special sciences that justifies the application of the exclusion argument to the former but not the latter. Unfortunately, this line of thought does not pay dividends. In an interesting article (1992), Justin Broackes has argued that there are genuine scientific laws couched in terms of colour properties that are indifferent to the manner of their physical realization. For example, the laws of additive mixture, of colour contrast, and of change of hue with saturation are generalizations about colours that have all the characteristics of laws; ie they are true, exceptionless in their intended domain, counterfactual-supporting etc. Broackes remarks that these colour laws are causally autonomous from the physical in the sense that they hold regardless of the way in which the colour properties are physically realized: these laws are robust in the sense that the substitution of metamers does not invalidate them. (This is not true for all folk colour ‘laws’. Consider the folk ‘law’ that the colour that an object appears to have is determined by its colour and the colour of the light in which it is seen. This is false since there may be metameric yellow objects that in daylight are indistinguishable in colour, but in violet light no longer match each other.) So, in conclusion, this popular, if erroneous, line of thought cannot be used to justify a limited application of the exclusion argument to colours.

6. Model-relative causation So far I have argued that Jackson’s physical realism about colour relies crucially on an application of the exclusion argument to colours. I have criticized this argument on the grounds that its unrestricted application leads to absurd conclusions; and further that colours do not have a special feature that justifies a limited application of the argument to them.

the folk theory of colours

155

Still, there are many questions that have been left dangling in our discussion. If the application of the exclusion argument to colours is rejected, then we seem committed to the conclusion that colours, as non-physical properties, have powers to cause colour experiences in human subjects that are independent of those of the physical properties that constitute their supervenience base. How exactly is that possible? Furthermore, if it is conceded that primitive colours and their underlying physical base properties can cause colour experience in different ways by different causal pathways, why does this not count as the kind of overdetermination that Jackson discounts as miraculous? Again, if there are multiple causal pathways leading from coloured objects to colour experiences, how are these causal pathways related to each other? Doesn’t any satisfactory account have to acknowledge the priority of the underlying physical pathway? If this is so, doesn’t the exclusion argument re-assert itself as a force to be reckoned with? It will not be possible here to answer these questions in full, as satisfactory answers depend on having a full-scale analysis of causation that can sustain talk of there being different non-competing levels of causation. Sketches of such an analysis are provided in my paper (2003). Here I draw on these sketches to give very brief answers to the first two questions above. Let us start by seeing how it is possible for colour properties to supervene on physical properties while possessing causal powers that are independent of those of their base properties. The crucial observation to make here is that the concept of causation, both within its folk and scientific articulations, is a relative rather than an absolute concept. The concept has to be analysed, I claim in my (2003), as relativized to a causal model, which represents the kind of system in terms of which the causal relations are being conceptualized. In the literature on causal modeling (Judea Pearl 2000, for example), a model is specified in terms of a set of variables whose values represent states of the kind of system being investigated and in terms of a set of laws (typically ceteris paribus laws) that govern the relevant kind of system. The relativization of causation to a model allows us to discuss causal processes at the level of abstraction appropriate for the kind of system being investigated. In the account of causation I have developed in my (2003) and elsewhere, a causal model generates a similarity ordering among possible worlds, with different models generating different orderings. These in turn give rise to different sets of counterfactual dependences, the existence of which are necessary for the existence of causal relations. (It must be emphasized that in the account described there is more to causation than counterfactual dependence. But, at least for our purposes here, I will simplify and take counterfactual dependence between distinct events to constitute

156

peter menzies

causation.) Consequently, different models of roughly the same domain may generate different sets of causal relations. An example will help to illustrate these remarks. Suppose we conceptualize the causal processes occurring in the interaction between coloured objects and human observers in terms of the kinds of variables and laws to be found in folk theories of colour and mental states. Suppose, for example, we take the state variables to be primitive colours of objects and primitive mental properties of persons. And suppose that the laws of this model will include such folk laws as the law that, other things being equal, shades of the same colour cause normal human subjects in standard conditions to have visual experiences of similar shades of colour. This model generates a distinctive similarity ordering of possible worlds and gives rise to distinctive counterfactual dependences. Thus, it may be true in this model that if various objects had not been shades of red, a given observer would not have had a visual experience of them as shades of the same colour. Now contrast this folk model with a model of the kind that might be entertained in psychophysics. Such a model would employ, as state variables, physical properties of objects and neural properties of human subjects. As before, this model will generate a similarity ordering of possible worlds, which is likely to differ significantly from the ordering generated by the folk model, especially in view of the fact that the same colour can be realized by many different physical properties and the fact that different physical properties can cause the same colour experience in many different ways. Indeed, because the similarity orderings generated by the two models are likely to be very different, it is also likely that different counterfactual dependences may obtain in the models. For example, suppose, as seems very likely, that the different kinds of physical properties that can realize shades of red do not bear any physically significant relation of similarity to each other. Then within the psychophysical model, it would not be true that if certain physical properties (corresponding to shades of red) had not been similar to each other, then a person would not have been in a certain neural state (corresponding to the visual experience of the shades being shades of the same colour). Indeed, this counterfactual may not even be well-formed within the psychophysical model, as it makes a presupposition about the existence of a similarity relation among the physical properties that cannot be sustained within the model. In this way, then, causal relations may exist within the folk model that do not answer to causal relations within the psychophysical model. So it would appear that causal relations between colours and colour experiences need not rest on causal relations between the properties that form their respective supervenience bases. Still it is reasonable to think that this will often,

the folk theory of colours

157

even if not always, be the case. Where we find that some colour causes a colour experience relative to the folk model, it will often be the case that the physical property that realizes the colour causes, relative to the psychophysical model, the neural property that realizes the colour experience. Now it might be wondered whether the co-occurrence of such causal relations is not precisely the kind of overdetermination that Jackson rightly discounts as being incredibly miraculous. Why is this kind of co-existence not a questionable form of overdetermination? It is important to distinguish two reasons why Jackson and others believe that it is implausible for there to be both physical and non-physical causes of colour experience. One reason that features a lot in their discussion is that the invariable, or almost invariable, co-occurrence of such causes would be miraculous in a way that strains credulity. It is true in standard cases of overdetermination that the existence of alternative causes is usually taken to be coincidental. Thus, if two assassins fire bullets into their victim at the same time, or two bell-ringers strike a bell at the same time, the overdetermination of the resulting effects rests on the coincidental fact that several people perform similar actions at the same time. It would, indeed, be incredibly miraculous if this kind of coincidental co-occurrence of causes were to happen repeatedly. However, in the situation being envisaged, the joint occurrence of the nonphysical and the physical causes of the colour experiences is not coincidental. For, after all, the non-physical causes supervene on the physical causes; and so there is a principled reason for why the non-physical causes always occur when the physical causes do. So the objection from the miraculousness of the invariable co-occurrences of the alternative causes should not, in the end, carry much conviction. But there is another reason, often lying beneath the surface of these discussions, that might be cited for thinking that the existence of overdetermining causes of colour experience is problematic. It is the thought that in cases of overdetermination there is a redundancy in the causation. For example, if one assassin had fired his bullet but the other had not, the victim would still have been killed; and if one bell-ringer had struck the bell but the other had not, the bell would still have sounded its note. Given that overdetermination implies redundancy in causation, it seems plausible to think that in the case of the causation of colour experiences the physical causes are doing all the work and the non-physical causes are really redundant. But here it is crucial to understand that when the model-relative account of causation posits alternative causes for some effect, it is positing something very different from standard overdetermination. In standard overdetermination examples, the overdetermining causes are posited within the same model.

158

peter menzies

For example, the actions of the two assassins would be represented by variables belonging to the same model. But when the model-relative account posits alternative causes of colour experience, it does so relative to different models. The folk model supports counterfactual dependences between nonphysical colours and colour experiences, underwriting the claim that these colours are causes of the experiences. The psychophysical model supports different counterfactual dependences, underwriting the claim that certain physical properties are the causes of the experiences, or their neural correlates. These models, and the counterfactual dependences they generate, do not compete with one another, nor do they exclude one another. Still it might be asked whether it is possible to combine the two models to provide some kind of unified account of the causes of colour experience. It is, after all, a permissible activity in science to form models that combine, or cut across, several explanatory domains. (For discussion of how this is done in cognitive science see Valerie Hardcastle (1996).) Observe, however, that even when one allows for this kind of cross-domain modelling, the redundancy worry does not show anything problematic about the existence of non-physical and physical causes of colour experience. If anything it suggests that the physical rather than the non-physical causes are redundant. The reason for this is that the formulation of the redundancy worry depends on being able to formulate a counterfactual of the following kind: if one cause had been present and the other absent, then the effect would still have occurred. But notice that it is possible to formulate and affirm the counterfactual that might support the redundancy of the physical causes of colour experience: if a certain object had had its non-physical colour property but lacked the physical property that realized it, a certain observer would have still had the relevant colour experience. The antecedent of this counterfactual is entertainable because the multiple realizability of colour properties allows for the possibility that even if the physical property that actually realized a colour property was absent, another property could have realized that colour property. In contrast, notice that the counterfactual that might support the redundancy of the non-physical colour properties cannot be formulated informatively. The counterfactual ‘if a certain object had its given physical properties but lacked its non-physical colour property, a given observer would have the same colour experience’ is not informative. For the fact that the non-physical colour properties supervene on the physical properties of the object means that the antecedent of the counterfactual is unentertainable. It would seem, in conclusion, that even if the redundancy worry can be reframed in terms of a causal model that combines the folk and psychophysical

the folk theory of colours 159 models, it does not, in fact, show that the non-physical causes are redundant in the causation of colour experience by comparison with the physical causes.

7. Conclusion I fear that more issues are raised than answered by this paper. Doubtless, it will be felt that many issues concerning the tenability of a non-physicalist conception of colours have been swept under the carpet. However, I have not claimed here that the primitivist conception of colours as non-physical properties is correct. My claim is merely that the folk theory of colour is committed to such a conception; and that Jackson’s reason for rejecting this conception, based on the exclusion argument, is without merit. This paper will have served its purpose if it has succeeded in throwing into high relief some obdurate but groundless assumptions about causation that obstruct philosophers’ understanding of colours.∗ Macquarie University

References Broackes, J. 1992. ‘The autonomy of colour’. In D. Charles and K. Lennon (eds.), Reduction, Explanation and Realism (Oxford: Oxford University Press). Byrne, A. and Hilbert, D. 1997. ‘Introduction. Readings on Color: the Philosophy of Color’ (Cambridge, MA: MIT Press). Campbell, J. 1993. ‘A simple view of colour’. In J. Haldane and C. Wright (eds.), Reality, Representation, and Projection (Oxford: Oxford University Press). Cornman, J. 1975. Perception, Common Sense, and Science. (New Haven: Yale University Press). Hardcastle, V. 1996. How to Build a Theory in Cognitive Science. (Albany: State University of New York). Hilbert, D. 1987. Color and Color Perception: A Study in Anthropocentric Realism. (Stanford: CSLI Press). ∗ Past discussions with Frank Jackson, David Lewis, and Philip Pettit have helped me get clearer about the issues of this paper. More recently, I’ve benefited from discussions with Jordi Fernandez and Alex Miller. For different objections to Jackson’s views on colour and causation see Duncan McFarland and Alexander Miller (1998).

160

peter menzies

Jackson, F. 1977. Perception: A Representative Theory. (Cambridge: Cambridge University Press). 1996a. ‘The primary quality view of color’. In J. Tomberlin, (ed.), Philosophical Perspectives, vol x (Cambridge, MA: Basil Blackwell). 1996b. ‘Mental causation’. Mind 105: 377–413. 1998. From Metaphysics to Ethics: A Defence of Conceptual Analysis (Oxford: Oxford University Press). 2007. ‘Colour for representationalists’. Erkenntnis 66: 169–85. Jackson, F. and Pargetter, R. 1987. ‘An objectivist guide to subjectivism about colour’, Revue Internationale de Philosophie 41: 127–41. Jackson, F. and Pettit, P. 1988. ‘Functionalism and broad content’, Mind 97: 381–400. 1990. ‘Program explanation: A general perspective’, Analysis 50: 107–17. Johnston, M. 1992. ‘How to speak of the colors’, Philosophical Studies 68: 221–63. Kim, J. 1998. Mind in a Physical World (Cambridge, MA: MIT Press). McFarland, D. and Miller, A. 1998. ‘Jackson on colour as a primary quality’, Analysis 58: 76–85. Menzies, P. 1996. ‘Probabilistic causation and the pre-emption problem’, Mind 104: 85–117. 2003. ‘The causal efficacy of mental states’. In S. Walter and H. Heckman (eds.), Physicalism and Mental Causation (Charlottesville: Imprint Academic). Pearl, J. 2000. Causality (Cambridge: Cambridge University Press). Strawson, G. 1989. ‘ ‘‘Red’’ and red’, Synthese 78: 193–232.

PA RT I I

The Knowledge Argument

This page intentionally left blank

7 Consciousness and the Frustrations of Physicalism PHILIP PETTIT

Introduction No philosopher can afford to neglect physicalism. Even if the doctrine is not correct, it represents the worst-case scenario—and hence the most challenging hypothesis—under which to think about a variety of phenomena that interest philosophers, ranging from freedom and phenomenal experience and personal identity to meaning and value and social aggregation. Those who are drawn to physicalism will need to show either that we can live with the denial of those phenomena or that the phenomena would be realized, at least to a close approximation, in a physicalist world. And those who go for some form of dualism or emergentism or the like will need to demonstrate why the counterparts that might survive in a physicalist world do not constitute the real thing.¹ My aim here is to sketch what I see as the best interpretation of physicalism, rehearse what I regard as the best way of defending it, and then show that the physicalism forthcoming is still going to be less than fully satisfying; it is going to leave us short of the satisfaction that might be expected from a philosophical theory. We may think that everything is physically constituted and physically governed, and that in principle it is possible to derive the way things are in any other respect from the way they are in physical respects; this is supported by my interpretation of the doctrine. But still physicalism will always seem to short-change us; it will leave us unsatisfied. While I think that the lesson is of more general significance, I shall try to substantiate it here with reference only to the physicalist treatment of phenomenal experience. ¹ On the tricky issues that arise here see Braddon-Mitchell 2003.

164

philip pettit

The problem I see is that no matter how convinced one is of the truth of physicalism, the doctrine is incapable of being absorbed into one’s daily life; the phenomena to which experience testifies cannot be made to present themselves in the shape that they have according to that theory. Thus, one can never come to see what transpires in phenomenal experience—in the experience of colour, taste, feel and so on—as physically derivable and unexceptionable. There will always be a gulf between the theoretical acceptance of its physical character and the experience itself. The problem I allege is not that we are incapable of making physical sense of phenomenal experience, whether because of shortcomings in our cognitive architecture (McGinn 1993, 1999), or because of lacking certain crucial concepts (Stoljar 2001). Nor does the problem arise just from the fact that we are subject to a limitation of perceptual processing such as that whereby one of the two lines in the Mueller-Lyer illusion continues to look longer, even when we realize they are of the same length (Fodor 1983). It is a problem more troublesome than the latter, less dramatic than the former. I describe it, for reasons that will become clear later, as a problem in our simulatory capacities, characterizing it as a simulatory gap that parallels the explanatory gap postulated by Joseph Levine (1993). I believe that the simulatory gap is troublesome in a range of areas—I comment on this in a brief conclusion—but I shall mainly be concerned with it here in relation to the physicalist account of phenomenal experience. My paper is in three sections. In the first section, I give an interpretation of physicalism in the spirit of Frank Jackson’s (1998) treatment; this involves a rich version under which the way things are phenomenally is derivable in principle from physical premisses. In the second I sketch a representationalist or intentionalist argument for physicalism about phenomenal experience, one which also appeals to Jackson (2003). And then in the third section I argue that even if we accept this case for a rich physicalism, still we should not expect to find the doctrine fully satisfying; the simulatory gap will stand between us and a sort of satisfaction that we might have expected to achieve. The existence of the simulatory gap may be bad news for physicalism, so far as it points up the lack of satisfaction in prospect. But I should mention that in another way it is clearly good news for the doctrine. Ignore the inevitability of the simulatory gap and physicalism may be expected to do the impossible, and so may be held to unreachable standards of proof. Admit the simulatory gap, however, and everything changes. The less physicalism is expected to achieve, the more plausible and palatable it may be found.

consciousness and the frustrations of physicalism

165

1. A rich physicalism Going just on our folk understanding of looks, and tastes, and feels—our ordinary sense of phenomenal experiences—we can apparently conceive of the actual world being physically unchanged, while there are variations in such experiences. We can conceive of the experiences assuming a different form, as in red-green inversions of colour appearances; or perhaps of their not materializing at all, as in the colour-blind functioning of zombie counterparts (Chalmers 1996). The conceivability of such independent variation is apparently boosted by the fact that equally we can conceive of someone knowing all that there is physically to know—knowing all the physical facts—and yet not knowing how a colour looks, a food tastes, a burn feels, and so on. Such a physically omniscient but still epistemically deprived individual is exemplified by the figure of Frank Jackson’s Mary, the all-knowing scientist who lives in a wholly black and white world (Jackson 1982, 1986).² The apparent conceivability of phenomenal-physical variation jars with the physicalist ontology that natural science makes attractive. Physicalism holds, intuitively, that the way the world physically is—there are some variations in how ‘physical’ may be understood (Pettit 1993b)—fixes the way that it is in other respects. It fixes what there is and it fixes how what there is behaves or functions. There could not be a world that was exactly like ours in the way it is physically constituted, then, and yet varies from ours in how it otherwise is or behaves. Or at least there could not be such a physically indiscernible, otherwise varying world, short of something new being added there. In Frank Jackson’s (1998) formulation, there could not be a minimal physical duplicate of the actual world that was not also a duplicate in other respects, including the phenomenal respects relating to how things look and feel, smell and taste, and so forth. Under one natural family of views as to what is involved in finding a scenario conceivable, there is going to be a straight inconsistency between the conceivability of phenomenal-physical variation and a physicalist ontology of this kind.³ On the sort of view I have in mind, our use of a phenomenal term is driven by a conception of what is required for the term to apply: a conception of the conditions that anything must satisfy in order to deserve to ² For an argument that the modal intuition of independence and this epistemic intuition of independence are actually in deep tension, not mutually boosting, see McGeer 2003. ³ The best known example of such a view is descriptivism and one well-known exponent of that view is Jackson 1999. But there are other, strictly non-descriptivist versions of the view as well. See Pettit 2004a and Pettit & Stoljar 2004. On different views of conceivability see Gendler and Hawthorne 2002.

166

philip pettit

have the term used of it. To be able to conceive that a phenomenal term might apply differently in a physically indiscernible duplicate of the actual world, then—and, presumptively, to be able to do this in the absence of ignorance or error⁴—will be to reveal a conception of what is required for it to apply that outruns any physical specification. It will be to show that according to the way one thinks of the experience to which the term applies—and so according to the way anyone who shares that conception thinks⁵—the experience does not fit the physicalist bill; it can vary while the physical world remains the same. Put in the first person, it will be to find it a priori knowable—knowable on the basis of the conceptions that drive the terms one uses—that the phenomenal experiences in question can vary independently of how things physically are.⁶ Some philosophers try to resolve the physicalist problematic by rejecting this conception-driven picture of what explains why certain terms—in particular, phenomenal terms—apply to certain items. They say that the ability to conceive of a phenomenal term applying differently in a physically indiscernible world does not reveal a conception of what is required for it to apply; and so not a conception that makes the requirements out to be non-physical. On this view, phenomenal terms refer, not to items that satisfy associated conceptions in the minds of relevant users, but to things that satisfy conditions of which users may not have any conception or inkling: say, to the items that happen to figure in the causal origin of the usage of the terms, whether or not users are aware of the fact, or to items that correlate causally with certain exemplars of continued usage, again without users having to be aware of that fact. The idea is that the terms one learns offer immediately available tags—opponents would say, mysteriously available tags—whereby one can keep empirical track of things. From the point of view of a user, these tags attach directly to the things they refer to, not by virtue of those things satisfying associated conceptions; they attach to their referents sub-cognitively, via a mechanism that operates independently of anything that has to be accessible to a user. On this account of why phenomenal terms attach to certain items, nothing is a priori knowable in the sense in play earlier, since there are no associated ⁴ Stoljar 2001 argues that there is a sort of ignorance, associated with missing a concept, that may well affect our best efforts in this area. ⁵ For all the present argument requires, everyone may have his or her own conception, or different conceptions may be associated with different sub-groups in the linguistic community. See Chalmers and Jackson 2001. ⁶ I assume that the conceptions driving one’s use of the terms will show up in a consideration of possible cases; one will see what one’s conceptions are so far as one can tell apart the cases where the term would apply from the cases where it wouldn’t.

consciousness and the frustrations of physicalism

167

conceptions to be explored for what they leave open and for what they rule out (Stalnaker 2004). Thus we cannot say that it is a priori knowable that the phenomenal can vary independently of the physical and we cannot say, in the associated strict sense, that it is conceivable that the phenomenal should vary in that way. But what we can say, of course, is that it is not a priori knowable that the phenomenal cannot vary independently and in that weaker sense we can say that it is conceivable that the phenomenal should vary. And this is in fact what a number of physicalists do assert, evading the problem raised by the clash of claims with which we began. They hold that while we may be able to conceive of the phenomenal varying independently of the physical, that proves absolutely nothing; it is quite consistent with the phenomenal being fixed, as a matter of empirical fact, by the physical. Indeed it is even consistent with various identities holding, and so holding of necessity, between phenomenal experiences and certain neural processes.⁷ In this paper I am going to assume a conception-driven story as to what makes phenomenal terms pick out corresponding items of experience. With most terms in our language, and certainly with terms of the phenomenal sort, I think that we hold ourselves answerable to antecedent conceptions of what is required to make something the deserver of such a term. The associated conceptions may vary among us and may shift over time but they serve at any time, for any person, as criteria by which he or she is required to navigate. Thus, we are not normally at a loss if we are asked what a term would refer to, according to our guiding conception, were we to inhabit a world that is different in various ways from the actual one. We are able to help ourselves in such an inquiry to a conception of what matters in determining the referent, and we can make a fairly confident judgment on how the term would apply did we inhabit that varying world. Take the term ‘water’ and imagine that the actual world is one in which a substance, XYZ, plays the role played here by H2 O: it falls from the sky, fills the oceans, is potable and transparent and so on. We can be quite confident in judging that were the actual world like that—were it a world where XYZ had all the properties that make H2 O a deserver of the name ‘water’ in our world—then in the scenario envisaged ‘water’ would refer to XYZ. The background idea here is already ⁷ The position sketched here corresponds with what Chalmers 1996 describes as Type B materialism. It is represented with many variations in a range of recent authors. As Daniel Stoljar has made clear to me, however, one may adopt the sort of semantics associated with the position without taking the rather easy line on conceivability mentioned in the text. One may think, like for example Kripke 1980, that there is a much richer sense to finding something conceivable than that of not finding it a priori knowable that it does not hold; and so one may not be able to dismiss the clash of claims so easily after all.

168

philip pettit

in Locke (1975, Bk 3, ch. 2, sect. 2). ‘Words being voluntary signs’, he said of the ordinary speaker, ‘they cannot be voluntary signs imposed by him on things he knows not.’ (See too Jackson and Pettit 1998). In assuming a conception-driven story as to why phenomenal terms have their particular referents, I am taking the side, broadly,⁸ of Frank Jackson and David Chalmers (2001) in a recent debate with Ned Block and Robert Stalnaker (2000). I find the approach attractive, as just mentioned, but in any case there is strategic reason for assuming it here. Under the approach in question physicalism is an intuitively richer doctrine than under the alternative, for it requires us to deny the conceivability of phenomenal-physical variation, not just to deflate it. The claim in this version of physicalism is that it is in principle possible to derive how things phenomenally are from how they are physically: there is an entailment, in principle knowable on an a priori basis, from the physical to the phenomenal. What I want to argue here is that even if such a rich physicalism is vindicated, still physicalism will remain less than a wholly satisfying doctrine; it will still be subject to the problem of the simulatory gap. Those of us who are rich physicalists often complain that the weaker version of the doctrine would be unsatisfying in not being able to invoke a priori conceptions of phenomenal experiences to justify equating the experiences with certain physical processes. But the fact is, so I want to argue, that our own version of physicalism is not going to be fully satisfying either and we should not raise expectations that we cannot hope to meet. In what follows, then, I will be concerned only with rich physicalism, as I have been calling it, though I shall usually leave that to be understood.

2. A case for rich physicalism With rich physicalism characterized, we might go straight to the argument as to why it cannot deliver a certain sort of satisfaction. But that argument will be easier to appreciate if we first sketch out the kind of case that can be made in favour of physicalism. In what I think of as the most persuasive case available, there are two steps. First, the qualitative character of phenomenal experience—the character ⁸ Why only ‘broadly’? Because one can believe that terms are conception-driven and not in the traditional sense count as a descriptivist of the kind that Chalmers and Jackson are taken to be; one can hold, for example, that the availability of the conceptions whereby one’s terms are driven presupposes a causal contact with the world of the kind that non-descriptivists privilege. See Pettit 2004a and Pettit and Stoljar 2004.

consciousness and the frustrations of physicalism

169

that ensures there is ‘something it is like’ to undergo the experience (Nagel 1986)—is said to be fixed by its representational properties. And, second, the representational properties of an experience are said to be fixed by broadly the same sorts of determinants—for example, functional factors—that are taken to fix the representational properties of non-phenomenal representations: say, beliefs and judgments. The approach, pioneered in recent times by Gilbert Harman (1990), and promoted also by Michael Tye (2000) and Frank Jackson (2003), is sometimes described as representationalism or intentionalism. The representational properties of a state include both what is represented to be the case—the content of the state—and the manner in which it is represented to be the case (Chalmers 2004). The manner of representation is what marks off experiences from memories and beliefs and a common view is that it is fixed by the way the state functions. A state will count as experiential so far as it functions in a manner typical of experiences: it generally disposes an agent to come to believe that things are as they are represented to be; it does not control behaviour except when it leads to belief; it may remain in place, continuing to represent things as being thus and so, even when the subject has to come to believe that they are not that way (Fodor 1983); and so on. Assuming that the manner of representation is fixed in these ways, and that we can identify a state as an experience, the focus among intentionalists is, first, how the content of an experience fixes its phenomenal character; and, second, how that content is fixed in turn by physical determinants. I sketch the intentionalist approach in this section, thereby setting up a version of physicalism that will allow us to see, in the next section, where the simulatory gap makes trouble. The gap makes trouble for non-intentionalist counterparts of rich physicalism too, but that point is best made when we have seen how it disturbs the intentionalist variant. The first intentionalist claim The first step in the intentionalist argument for physicalism is to hold that amongst creatures of our kind its content fixes the phenomenal character of an experience, whatever that happens to be. This is best construed as a supervenience thesis, according to which it is impossible for the phenomenal character of an experience of ours to vary over time, or for the phenomenal character of two experiences to vary between them, without a variation in their representational properties (Byrne 2001). The notion of phenomenal character may be left relatively vague for our purposes; it is intended to refer to whatever is involved in there being something it is like to have an experience. But the notion of content requires a number of comments.

170

philip pettit

First comment. The content of an experience is how the experience represents things as being. It consists in the properties that the experience represents as being instantiated, perhaps on or in the subject’s body, perhaps at this or that direction and distance from it. The properties it represents as instantiated in this way may not actually be instantiated, of course. Indeed, they may even be wholly illusory. Those philosophers who think that, in attributing colours and tastes to things in the physical world, sensory experience is necessarily in error can still hold that it is in virtue of representing such non-physical properties as instantiated that the experiences have their particular phenomenal character. Second comment. The content of experience in virtue of which the phenomenal character is said to be fixed involves only narrow properties. They are those properties that will be available to be ascribed in the experience of any members of our kind who are indiscernible from the epidermis in; that is, in the experience of a number of doppelg¨anger. All experience involves the ascription of such narrow properties, so I assume, even if it serves also to ascribe properties of a wide nature (Jackson and Pettit 1992). The reason for the restriction to narrow properties is that it is counter-intuitive to think that phenomenal character might vary, say as between the experiences of two people, without there being any difference within those subjects. Third comment. The properties represented as instantiated must be represented as instantiated in the world on which the agent acts, not in the insulated realm of sense-data imagined in some philosophical theories. This is to say that intentionalism, as I conceive of it, rejects sense-datum theory (pace Byrne 2001). Many of the properties represented by experience will be represented as being instantiated in the external world. That such properties are not represented as instantiated in a realm of sense-data shows up in the fact that if the subject finds that the properties are not instantiated in the external world, then there is no suggestion that therefore they must be instantiated mentally. ‘From the fact that there is no Fountain of Youth’, as Gilbert Harman (1990, 36) puts it, ‘it does not follow that Ponce de Leon was searching for something mental.’ Notice, however, that though I may learn that the external world is not as my experience represents it—the stick in water is not bent, the lemon in the shade is not brown—still, my experience of it as being thus or so may remain: it may still represent the stick as bent, the lemon as brown. Other properties represented in experience will be represented as instantiated, not in the world, but in or on the subject’s body. And here again, if the subject finds that they are not instantiated there, then there is no suggestion

consciousness and the frustrations of physicalism

171

that they must therefore be instantiated somewhere else. From the fact that there is nothing touching my back, it does not follow that I was tickled by something mental. There is a sense in which I may continue to feel tickling, it is true, even as I learn that there is no one tickling me. But that’s just because the experience represents things in a manner that is resistant to correction by belief: this, in the way an external experience can continue to represent a lemon as brown, even when I know it is yellow.⁹ The three comments so far stipulate that for intentionalism the content of an experience is how it represents things to be in the world of the subject’s environment or body, where the way it represents them involves narrow properties only and may be the product of a mistake or an illusion. But there are three further comments that should be added in order to specify more exactly the claim made; they are important, by my lights, for making the intentionalist argument plausible. Fourth comment. The properties represented in experience as being instantiated are always represented at the same time as impacting on the subject: as exercising a causal influence. What my sight represents in the object I apparently see, it represents as, precisely, properties I can see: properties I register only so far as my eyes are open, there is nothing in the way, I am suitably near the object, and so on. I do not find myself representing relevant properties as instantiated, I know not how; I do not find myself inclined to believe that they are instantiated in the way in which blind-sighted subjects might come to believe this. Experiential representation gives the subject information, not just on how things putatively are, but also on their presenting themselves as being that way; it represents them, not just as present, but also as presented.¹⁰ Fifth comment. The content of a typical experience is dense and textured in such a way that it is unlikely to be able to figure as the content of an ⁹ In the tickle case, I can go wrong in locating the tickling (on my back) and in characterizing the tickling (as a tickling); the experience is resistant to correction in those dimensions but we do not treat it as incorrigible. Is the pain experience of a part of my body as hurting equally corrigible, if resistant to correction? I think so. I continue to feel pain even as I recognize that the location—say, a phantom arm—is non-existent and that there cannot be any hurt to ascribe there. I am wrong in experiencing the arm as hurting, then, as I am wrong in experiencing my back as being tickled. But I am still, of course, in pain. It is the fact of having an experience of the arm as hurting, not the fact that there is a real arm that hurts, that means I am in pain. ¹⁰ In Pettit 2003 I express this thought by saying that the experience overtly has various effects on me. Notice that what is involved need not require the subject to have the very sophisticated thought that there is an object present and that object is causally responsible for this independently individuated experience pace Searle 1983. Seeing an object as presented amounts to seeing the presenting of the object, as well as seeing the object itself; it is like seeing the movement of the seconds hand of a clock as well as the hand itself.

172

philip pettit

independent belief or judgment: that is, a belief or judgment that does not consist just in endorsing the experience. Thus, the content of experience is holistically textured, so that of necessity colour is never represented without the simultaneous representation of shape, the volume of a sound is never represented without a representation of its pitch, the quality of a pain I never represented without a representation of its location, and so on (Jackson 2003). And, again, the content of experience is often densely textured, so that it can be exploited to resolve further and further questions of detail: we can go back to the experience to determine previously unnoticed matters such as the exact relations of shade among colours (Dretske 1999). Sixth comment. The notion of content available for intentionalists is not only a content represented in the subject, as that is sometimes put; it is a content also represented for the subject (cf. Cummins 1983). What this means, by my lights, is the following. Not only does an experience with a certain content dispose an agent to act and react in a manner that is appropriate for a world answering to that content: say, in a manner that would satisfy the subject’s desires in such a world. The subject, or at least the human subject, is also enabled to form beliefs about that content, making judgments on whether it is likely to be true, on how exactly it depicts this or that aspect of the scenario involved, on whether it answers to the subject’s wishes, and so on (Pettit 1993a, ch. 2). The subject is able to meta-represent the content of the experience in the non-experiential way associated with belief and judgment. This capacity may require the sort of conceptual sophistication that human beings have and that at least most non-human animals seem to lack (cf. Carruthers 2000). But it is undoubtedly an aspect of what we naturally think of as experiential content. With the notion of experiential content explicated in these ways, we can return to the first claim in the intentionalist argument for physicalism. Assuming that experiences are distinguished functionally, the claim is that the phenomenal character of any experience of ours superveniently depends on its content. And the argument for that thesis is fundamentally an appeal to intuition or thought-experiment. Can you imagine an experience of yours retaining its experiential content, in the rich sense explicated, but varying in its phenomenal character? Can you imagine it continuing to represent things to be a certain way—the same way in all relevant details—while, suddenly, there is something different it is like to have that experience? I am going to assume that the answer to this question is negative and that the first intentionalist claim is sound (Byrne 2001). I think that however quick, the line is fairly persuasive. We can imagine all sorts of differences between experiences that may seem at first to have nothing to do with content: for

consciousness and the frustrations of physicalism

173

example, the difference that consists in something now being sharply presented, now being seen only in a blurry way; now being at the centre of attention, now being unattended to. But in all such cases it appears that the difference can be readily explained in content terms. Thus, the sharp vision of the object represents it as having various properties that the blurry vision fails to do; and similarly for the perception that attends to the object and the perception that does not. In such cases there may sometimes be differences in the other direction as well, though intentionalism does not require this; the blurry vision may represent the object as being at a distance, in bad light, out of focus, or whatever. I think that most candidate counter examples can be rejected on similar grounds and that the first intentionalist claim, properly understood, is very hard to reject. The second intentionalist claim The second intentionalist claim is that just as the content of an experience fixes its phenomenal character, so various familiar sorts of determinants—in particular, physically unexceptionable determinants—can serve to fix that content. If this second claim goes through, of course, then physicalism will be vindicated. For if the physical determinants of content also fix phenomenal character, then there is no possibility of phenomenal-physical variation. The way the actual world is physically configured will fix the way it is configured in all phenomenal respects as well. There are a variety of theories as to what the determinants of content are but by most accounts they include functional determinants, and I shall concentrate exclusively on these. If we can see how functional determinants might plausibly be held to fix content and character, then it will be easier again to see how functional determinants might do so in combination with other factors. On the functionalist account I envisage, the content of an experience will be fixed by what the experience does. It will be fixed by what it does under various inputs, as those inputs give rise to the experience, or cause it to disappear, or make it shift in various ways; by what it does in producing corresponding outputs, where these may range from more or less autonomic adjustments to the judgments and actions it leads us to endorse; and by what it does in interacting in different ways with other experiences, with other memories and fancies and feelings, with other beliefs and desires, and so on. The content will be superveniently dependent on the functional profile displayed by the experience. The idea is that an experience as of something red, for example, will count as an experience with that content just in virtue of what it does in such

174

philip pettit

respects: say, to consider a veridical case, what it does in enabling the subject to track an object in a certain way; to see it as similar in that way to some objects, and different from others; and to do all of this under variations of background, lighting and perspective (for further detail, see Pettit 2003).¹¹ Any experience (as) of an object that does all of these things will count from an external point of view as a perception as of red: it will lead us to say that the subject is perceiving the object as red. And, more importantly, it will count as a perception that from the point of view of the subject is just as a perception of red ought to be; its content and phenomenal character will have the distinctive aspect of such an experience. Or perhaps not quite. As described so far, the functional profile of the perception can be the same in human beings and in less complex animals. But we saw that in human beings experiences represent things for the subject as well as in the subject; they represent things in such a way that the experiential content is available to be interrogated and considered in various non-experiential modes. This points us towards the need for enriching the picture further. The richer picture has to make room for the fact that any human subject who is prompted and primed in the manner that goes with having a red experience will thereby be positioned to recognize the content of the experience, now in this perception, now in that, and now in memory, now in imagination. The subject will be able, in virtue of that prompting and priming, to recognize how the experience represents things to be. So far as each perception as of red is functionally equivalent to other perceptions as of red, and despite any independent, lower-level variation between the experiences, there will be a way that it represents things to be that is available as an item about which the subject is in a position to form judgments, beliefs, and other attitudes. There is nothing mysterious involved in this enriched picture. All that is required is that the functional profile of a phenomenal experience be richer than the profile that would materialize in a conceptually unsophisticated creature. The experience will have to be functionally effective, not just in facilitating the sort of tracking and categorization mentioned earlier, but also ¹¹ As I stress in that paper, the sort of functionalist story I accept is broadly in the spirit of writers in the ‘enactive’ or ‘sensorimotor’ tradition. See Thompson 1995, Hurley 1998, Clark 1999, Myin 2001, Myin and O’Regan 2002, Noe 2002, and O’Regan and Noe 2002. I believe that many of the functional connections that fix content have to be learned as a matter of maturational skill, as those in that tradition stress, and that the subject will normally have a grasp, however implicit or practical in character, of what those connections are. This grasp of the connections involved may be necessary if the content to be fixed is, as I put it in the text, a content for the subject, not just a content in the subject.

consciousness and the frustrations of physicalism

175

the formation of judgments and beliefs, and associated other attitudes, about how things are according to the experience. Even those who would reject the rich physicalism envisaged here admit that such a functional enrichment of experience is unproblematic.¹² Not only will the suitably tuned human subject undergo experiences as of red, then. He or she will also be able to think about the redness that the experiences ascribe to things, attending to the contrast with other colours that might have been represented, to the likelihood that the colour represented is the real colour, and so on. The subject will be able to think about that redness represented in experience as the red look that things have, recognizing that the way such experiences represent things to be—the property-of-theobject-according-to-the-experiences—may not be a way they actually are. And such red looks will have a robust presence for the subject—they will have the status of familiar appearances—since experience, as we noted earlier, often continues to represent things as being a certain way, even when the subject has come to believe that as a matter of fact they are not really that way. Although I know that the stick in water is not really bent, the lemon in the shade is not really brown, the tickling on my back is not really someone tickling me, still the experiences involved persist in representing things that way. Generalizing from the perception of colour, then, the key idea in this physicalist story is that phenomenal experience comes about in a physical subject in virtue of the following factors: • the subject has experiences that represent certain properties as present, and as presented, in the world or in his or her own body; • the experiences play this representational role in virtue of being tuned appropriately to world or body and in virtue of tuning the subject to respond, or be disposed to respond, appropriately; • the experiences represent the corresponding properties for the subject, not just in the subject, so far as he or she is able to think about the properties as experientially represented; and • the experiences often represent those properties in a way that is resistant to correction in the light of belief, so that the properties-as-experientiallyrepresented have the character of robust appearances. What will make an experience distinctively phenomenal, on this account? What will ensure that there is something it is like to undergo the experience? ¹² See Block 2002, where he explicitly countenances this sort of judgmental ‘access’ to experience. The access described there is precisely the sort envisaged in the text, where that is not the case with the sort of access described in an earlier version of the paper (Block 1997).

176

philip pettit

The first thing to say is that whenever an experience represents something in a subject, there will be something the world or its body is like from the viewpoint of that subject (Dretske 1995). The animal or human in question will be positioned to think about the world or body on the basis of how it is represented, where the way it is represented—its content—is highly distinctive. Experiential content, as we have seen, is distinctive so far as properties are represented, not just as present, but also as presented; and so far as it is holistically and densely textured. But it is one thing for there to be something the experienced world is like from the viewpoint of a subject; it is another for there to be something the experience itself is like. On the story told, this will most naturally be identified with the experience’s coming to represent how things are for the subject, not just in the subject: with its enabling the subject, not just to represent things a certain way, but to think about how things are thereby represented. When that point is reached, and only then, it will be appropriate to say that the subject can be aware as such of looks and tastes, and feels and smells. At that point it will be possible for the subject to think about what different experiences as of red have in common; to contrast them with experiences as of green; to savour the charming or eye-catching effect of perceived redness; to wonder about what gives rise to it in the physical world; and so on. The two intentionalist theses point us towards a version of physicalism. They suggest that there is a way I functionally am in virtue of my physical make-up such that it guarantees that I have experiences with a distinctive set of contents and a distinctive phenomenal character. The suggestion is that the way things physically are will necessitate or entail the way they are phenomenally; and this, for reasons that we can understand in virtue of our a priori, intentionalist conception of the phenomenal states involved. I find this rich form of physicalism plausible and attractive and I think that it fits well with what we are learning from psychology and neuroscience (for more in defence, see Pettit 2003).¹³ ¹³ There are two aspects of the fit between the picture sketched and empirical science that are emphasized in that paper. First, the picture fits nicely with the findings of Ivo Kohler 1964, that people compensate for disturbances induced by wearing colour-distorting glasses, and that when they compensate—when they regain their old colour-discriminating skills—they report that things look exactly the way they used to look before. And second, the most surprising aspect of the picture—that the appearances of colour and the like do not cause our responses but materialize in virtue of those responses—connects nicely with the growing neuroscientific emphasis on the fact that the brain often ‘works behind the subject’s back’ (Gazzaniga, Ivry and Mangun 1998), with effects being visually induced in a manner that is independent of visual judgment (Milner & Goodale 1995). On this latter point, see Clark 2001.

consciousness and the frustrations of physicalism

177

3. The simulatory gap Under the rich physicalist doctrine sketched in the last section, the physical facts about any subjects—including perhaps facts about their environment and history—will fix the phenomenal facts that are true of them. Were we able to survey all the physical facts, then, and had we shared the conceptions of phenomenal reality deployed in the story—conceptions relating the phenomenal to the representational to the functional—then presumably we would be able to see that such and such a physical configuration amounts to nothing more or less than such and such a phenomenal profile. We would be able to derive the phenomenal configuration from the physical, given the a priori conceptions of the phenomenal in play and the empirical facts about how the physical is configured. The lesson of the doctrine is that while we cannot actually derive the phenomenal from the physical, not having access to the relevant physical facts, we can see that in principle the phenomenal is derivable from the physical (see Jackson 1998, 2003). This conclusion, if we accept it, ought to drive out any physicalistic mystery attaching to the notion of phenomenal experience. And yet, so it seems to me, it doesn’t quite succeed—at least not at an intuitive level. What I hold of the phenomenal in saying that it is derivable from the physical, is the sort of thing that I hold of the Mona Lisa’s smile in saying that it is derivable—derivable in principle but not in practice—from the exact array of painted points found in Leonardo’s painting. But whereas my belief about the Mona Lisa leaves no mystery in place as to how such and such dots could constitute exactly this representation of a smile, there is a certain mystery still remaining as to why such and such a physical make-up should constitute precisely this sort of phenomenal experience: for example, precisely this experience as of red. Why should the satisfaction of the appropriate physical and functional conditions necessitate the presence of that sui generis experience? Why should it leave open no possibility of things looking otherwise? I devote this last section to a discussion of that question. My hypothesis is that notwithstanding the similarities between the physicalist claim and the claim about the Mona Lisa’s smile, there is an important difference. While I cannot survey all the painted dots in the Mona Lisa’s smile—at least not independently of seeing the smile itself—and so cannot derive the smile from them, I am familiar in other cases with how the addition of dot after dot can suddenly give rise to a figure of a certain kind. Nothing of that sort holds in the physicalist case. Not only can I not survey all the physical facts relevant to having such and

178

philip pettit

such a phenomenal experience, I have no familiarity with other cases where piling up physical facts—or realizing them in one’s own person—suddenly gives rise to a phenomenal experience. It is this difference between the cases that makes for the lack of satisfaction associated with physicalism. So at any rate I wish to argue. Being able to derive a conclusion from certain premisses involves, in the normal situation, satisfying two distinct conditions (Pettit 1998). First, one is able to inspect and understand the premisses. And second, one feels compelled in virtue of that inspection and understanding to affirm the conclusion: one is inferentially moved to endorse it, as a relevant habit of inference is engaged by the premisses. Thus, in the case of a regular modus ponens argument, one recognizes that if p, q, and that p, and the syntactic structure of those premisses activates a primitive compulsion, as Christopher Peacocke (1992) describes it, to go to the conclusion that q. Or, to take a rather different case, one sees that it is beginning to rain and, aware of one’s friend’s habits, one can’t help concluding that she probably won’t want to go for a walk. One does not leap to conclusions in any such normal case of reasoning and inference; one is decidedly pushed. The experience of inference, like the experience of perception, is one of being driven to conclusions, not an experience of willing oneself to accept them. But while this is what happens in the normal case, it does not happen in all. There are many exceptional or marginal cases where one believes that the premisses support the conclusion, but the experience of the inferential push is absent. Consider, for example, cases like the following (McGeer and Pettit 2002): • I recognise that modus ponens is a sound rule of inference but, perhaps because of a cognitive ailment or limitation, I just do not feel pushed by premisses of the appropriate form; I have to exercise brute-force in order to regulate my belief-formation. • I am learning to fly and know that the instrument readings before me imply that the plane is heading towards the ground but, proprioceptive cues still having a grip on my inclinations, I have to fight hard to internalize and act on that lesson. • I recognize that the chance of a fair coin coming up heads on a fair toss is a half but, confronted with a run of tails, I have to struggle to form the correct expectation; the gambler’s fallacy induces the illusion that on the next toss the chance of heads is much higher than a half. These cases illustrate a sort of malaise with which all of us are familiar in some areas. Whether in the sphere of geometrical representations, or X-ray

consciousness and the frustrations of physicalism

179

photographs, or quantificational formulae, or perhaps just facial expressions, most of us know what it is to believe that the fact that things are thus and so has such and such an implication, without that implication having any life or force for us. Psychologists speak of perceptual pop-out in connection with those dot-by-dot drawings where one may or may not see a particular pattern. By analogy we might speak here of inferential pop-out or rather of the lack of inferential pop-out. People who stare blankly at modus ponens premisses and have to drill themselves into seeing what is implied suffer from the lack of inferential pop-out. So do the trainee pilots who fail to attune spontaneously to the readings on the instrument panel. And so do those who, like most of us, are often blocked by gambler’s-fallacy inclinations from seeing the probabilistic indications of a coin’s lack of bias. When people are affected by the lack of inferential pop-out then they may believe or know that a certain conclusion follows from such and such premisses but they will not see the premisses as supporting the conclusion. They may register that the premisses provide support, as we might say, but they will not register the support itself. They may retain inferential judgment, being capable of policing and drilling themselves to draw suitable conclusions, but they will lack inferential perception. Not having access to inferential pop-out, they will be inferentially disabled. Returning now to the case of rich physicalism, what I want to suggest is that the reason it is an unsatisfying sort of doctrine is related to the fact that we have no experience of inferential pop-out in this area. Suppose that we believe in the sort of physicalism sketched and accept that the ways things physically are with any one of us entails, as an a priori matter, that we are phenomenally thus and so. It is true that we will not be able to derive the phenomenal from the physical, not being in a position to survey all the relevant physical facts. But there is also a second problem that is bound to bother us. This is, that we will have no sense whatsoever of what it would be to inspect physical premisses and be driven by them to a phenomenal conclusion. We will have no sense that given an exposure to all the relevant physical facts, therefore, inferential pop-out would spontaneously display the phenomenal configuration that those facts sustain. In the cases mentioned earlier, there are different reasons for the lack of inferential pop-out. In the modus ponens example, the cause of the problem is a cognitive disability of a kind we can imagine but rarely experience. In the example of the trainee pilot, the cause is a lack of training and habituation, where this is compounded by the distraction of rival proprioceptive cues. In the probabilistic example, the cause is a well-known form of intellectual confusion: instead of considering the chance of this toss being a heads, I consider the

180

philip pettit

chance of having a long run of heads in which this is the last toss. But another reason why I might lack inferential pop-out in a certain instance, or lack even a sense of its likelihood or possibility, is the absence of any exposure to the sort of inference involved. And this, I think, is the problem that is more or less bound to affect us in the physicalist case. According to the physicalism sketched, we are invited to imagine a state that functions in the representational manner of an experience and, in particular, that functions in such a way that I, the subject, can form beliefs about the properties-as-experienced that it represents as present and presented. We are invited, more specifically, to consider the physical set-up in virtue of which such a representational state might materialize. And then we are told that could we survey the details of that physical set-up, we would be able to see that the state in question would have such and such a phenomenal profile; there is something in particular it would be like to instantiate it: this, in the way that there is something in particular it is like to see red. In order for this claim to gain a satisfying grip on our imagination, we would have to be acquainted with some examples or analogies where the satisfaction of a physical-cum-functional antecedent of the kind envisaged here leaves us with something akin to the phenomenal experience that it is said to entail. But this, alas, is not something that we are even remotely positioned to enjoy. We may believe with the strongest possible degree of conviction that were we to be physically configured thus and so then that would ensure, just by virtue of what the notion of a phenomenal state involves, that we would be phenomenally primed in this or that fashion. But we can have little or no sense of this being an inferential inevitability. We cannot begin to imagine what the inferential pop-out would look like. The analogy with the Mona Lisa backs up this analysis, for the striking thing in that case is, as mentioned, that we all know what it is for the addition of dot after dot suddenly to give rise to a figure. But a closer analogy is provided by heliocentrism. When Copernicus and Galileo argued for the hypothesis that the earth rotates every twenty-four hours, they did not deny that for all our perceptions tell us at any point on the earth, the sun moves across the sky in the course of the day. On the contrary, of course, they argued that given any particular position on a rotating earth, and given any particular time of year, the sun would be bound to describe such and such a perceptual trajectory across the sky. They did not derive that trajectory for any location on earth, or any time of the year, of course, because the number of physical facts to be enumerated and surveyed in such a derivation would be vast and unmanageable. But still, they had little or no trouble in convincing most people that the heliocentric hypothesis was not

consciousness and the frustrations of physicalism

181

confounded by the sight of the sun moving across the sky; with the help of some auxiliary hypotheses from optics, that sight could be derived from within the hypothesis itself. Galileo had an easy success on this front, if not equal success in actually persuading people of the truth of his hypothesis, because everyone is familiar with an analogous case. Everyone is familiar with the experience of occupying a frame of visual reference that moves indiscernibly in relation to another object, where it is the other object that appears to move, not the person’s own frame. The experience is common in any railway station or, to take an analogy available in Galileo’s own time, on any harbour or river. As the moving train or boat can seem to provide evidence that it is other things that are moving, so it is easy to let people see that a rotating earth might seem to provide evidence that the sun moves across the sky. The problem that I see in the physicalist case is that not only are we unable to survey all the physical premisses that might entail the presence of a phenomenal state. Even worse, we have no inferential sense of how the inspection and appreciation of those premisses would give us, as an irresistible pop-out, the conclusion that in such and such a physical configuration such and such a phenomenal profile would inevitably materialize. The lack of this inferential sense is even worse because, as the heliocentric case suggests, the inability to survey all the relevant physical premisses would not be much of a problem if we could see how we might be inferentially moved by any such premisses to endorse a phenomenal conclusion. In the heliocentric case we have a perceptual sense of the support offered by the hypothesis for the appearance of the sun moving. In the physicalist case, we have absolutely nothing of the sort. What might provide us with the pop-out, or the sense of what a popout would be like, that we are lacking? If the heliocentric case offers a reliable parallel, then we would need experience of the kind that would make it possible to simulate the instantiation of the physical states and see, on that basis, that it is bound to give rise to the corresponding phenomenal profile. The analogous experience is available in the heliocentric case, but not in this. In order to simulate what it would be like to view a relatively stationary sun from an indiscernibly rotating earth, we draw on our experiences with carriages or boats or trains. We know what it is like to take a moving platform as a stable frame of visual reference, given that this happens in such common experiences. And hence we know what to do in order to envisage the analogue with a rotating earth; we merely put the earth imaginatively in the role that trains and boats and carriages sometimes have for us. When we do this, of

182

philip pettit

course, we find as a result of the simulation that in imagination the sun will appear to move across the sky. We cannot do anything of this kind, however, in the physicalist case. We may be convinced that to look red is just to have the look that a suitably functioning experience would enable us to identify and we may be able to argue for this on a variety of grounds. But we will not be able to drive the point home in the manner of heliocentrists. We will not be able to give it imaginative vindication by simulating how it would be to undergo a suitably functioning experience and by finding that it would be to have the things experienced look red in the familiar way. Why can’t we do this? The simulatory move available in the heliocentric case involves being able to simulate an experience under one fairly austere specification and then finding, through the exercise of that ability, that the experience also satisfies a richer specification. We are able to simulate viewing a relatively stationary sun from an indiscernibly moving earth and, simulating the experience under that description, we find that it satisfies another: it is an experience as of a sun that moves in the opposite direction to the direction of the earth’s rotation. The resources that make simulation possible in this case are lacking in the physicalist scenario; say, in the case involving the perception as of something red. Here we would need to be able to simulate having an experience with a certain functional profile, under that relatively austere description, and we would want to be able to find that the simulation delivers the familiar experience as of something red. But we do not know what it would be like, described as such, to have an experience with a certain functional profile. We lack the sorts of analogues provided in the other case by trains and boats and carriages. And so we cannot even begin to get the simulatory exercise under way. The block to our achieving a simulatory, imaginative vindication of physicalism may be fairly contingent. Technological aids may yet make it possible for people to be able to simulate having experiences with certain functional profiles; and such aids may serve in the confirmation of physicalist ideas. The experience of moving pictures—films or ‘movies’—has made us familiar with the fact that as we see scenes succeed one another at smaller and smaller intervals, we begin to see motion and change. Thus we can simulate the effect of more or less static perceptions being registered at smaller and smaller intervals and we can see how, on that physical base, a phenomenal experience as of motion might materialize and ‘akinetopsia’ or motion-blindness be avoided (Pettit 2004c). The problem, however, is that for whatever contingent reasons, this is a very special case and there is no analogous sort of aid to simulation available for phenomenal experiences in general.

consciousness and the frustrations of physicalism

183

If this line of argument is right, then even if we embrace the rich sort of physicalism sketched in the previous section, we should not expect that this will leave us with the sense of intellectual satisfaction that parallel doctrines can provide elsewhere. Our position as physicalists is like that in which heliocentrists would find themselves were we human beings incapable, for whatever reason, of simulating how a relatively stable sun would look from the presumptively unmoving platform of a rotating earth. While being incapable of pulling off that imaginative feat, we might still believe that the way the sun looks is capable in principle of being derived from the heliocentric truths, combined with truths about our perceptual dispositions. But we would have to recognize that for reasons related to our simulatory limitations, we could never give ourselves an imaginative sense of how the derivation goes; we would have to live without any sense of inferential pop-out. One final note. I have been trying to show that the rich physicalism supported by the intentionalist line of argument rehearsed in the last section is likely to be less than fully satisfying, because of the simulatory gap. But the lesson is not limited, of course, to the intentionalist form of physicalism. For even if the phenomenal character of experience is directly fixed by some of its physical features, not just via the fixing of experiential content, still the same sort of problem is bound to arise. We will be unable to simulate what it is to have an experience with those features and so unable to see, at the level of concrete imagination, what it will be to have an experience with such and such a phenomenal character. The problem of the simulatory gap extends to every form of rich physicalism; it is not confined to the intentionalist variety.

Conclusion The nineteenth-century Danish philosopher, Soren Kierkegaard (1980: 43), famously complained that the systematic philosopher is like a prince who builds a palace and then lives beside it in a shack. The argument here suggests that there is a certain justice in his complaint (Pettit 2004b). Although our overall world-view will shift with conversion from a non-physicalist to a physicalist point of view, and although this shift may have an impact on various of our commitments and attitudes, it is not the case that the change will transform the way phenomenal experiences present themselves. Phenomenal experiences will not shift for us so that we can see them as the precipitate of our functional, ultimately physical organization; they will not shift in the way heliocentrism shifts our sense of the sun’s apparent motion. The simulatory gap means that

184 philip pettit they will retain the distinctive, sui generis character that has always seemed to mark a break with the merely physical. The problem that affects phenomenal experience is parallel, so I think, to a problem that arises equally with normative experience and thought (Pettit 2000). Just as the functional organization of some of our experiences makes them phenomenal, by my view, so the organization of other experiences makes them normative: makes them into experiences as of rules imposing normative demands upon us (Pettit 2000, Pettit 2002, Part I). But the story that makes sense of the phenomenology of rule-following, like the story that makes sense of phenomenal experience, is equally subject to the problem of the simulatory gap. I argued in the normative case that if we were to develop such and such dispositions, treat those dispositions in such and such a manner, and engage with one another’s dispositions in that mode of treatment, then we would each have the experience as of rules imposing themselves upon us. But here, as in the phenomenal case, we lack the resources required for simulating what it would be like for the antecedent of that conditional to be fulfilled. And so the claim that is made in the physicalist account of rule-following, like the claim made in the account of phenomenal experience, will not deliver the sort of satisfaction we might have hoped for. Perhaps there is something inherently frustrating, then, about philosophical problems like those that arise with phenomenal and normative experience. And perhaps the frustration with which the best reductive theories are going to leave us is what explains why those problems have such a perennial character. The simulatory gap may not be just a local difficulty that arises for a physicalist theory of consciousness; it may be the source of a central form of philosophical perplexity.¹⁴ Princeton University

References Block, N. 1997. ‘On a Confusion about a function of consciousness’. In G. Guezeldere (ed.), The Nature of Consciousness: Philosophical Debates (Cambridge, MA, MIT Press). ¹⁴ This paper was sketched and written over a period in 2003 when I was co-teaching a graduate seminar on consciousness with Sean Kelly, in Princeton. I owe him and the other participants in the seminar a debt of gratitude for a host of matters that I came to understand better as a result of our discussions. I am also grateful to Sean Kelly, to Frank Jackson, Victoria McGeer, Michael Smith and Daniel Stoljar, for a number of very helpful conversations on the topics covered. Finally, I thank Ian Ravenscroft for very useful and insightful comments on the penultimate draft.

consciousness and the frustrations of physicalism 185 2002. ‘Concepts of consciousness’. In D. Chalmers (ed.), Philosophy of Mind: Classic and Contemporary Readings (Oxford: Oxford University Press). Block, N. and Stalnaker, R. 2000. ‘Conceptual analysis and the explanatory gap’, Philosophical Review 109: 1–46. Braddon-Mitchell, D. 2003. ‘Qualia and analytic conditionals’, Journal of Philosophy 100: 111–35. Byrne, A. 2001. ‘Intentionalism defended’, Philosophical Review 110: 199–240. Carruthers, P. 2000. Phenomenal Consciousness: A Naturalistic Theory (Cambridge: Cambridge University Press). Chalmers, D. 1996. The Conscious Mind: In Search of a Fundamental Theory (New York: Oxford University Press). 2004. ‘The representational character of experience’. In B. Leiter (ed.), The Future for Philosophy (Oxford: Oxford University Press). Chalmers, D. and Jackson, F. 2001. ‘Conceptual analysis and reductive explanation’, Philosophical Review 110: 315–60. Clark, A. 1999. ‘A case where access implies qualia’. Analysis 60: 30–8. 2001. ‘Visual experience and motor action: Are the bonds too tight?’, Philosophical Review 110: 495–519. Cummins, R. 1983. The Nature of Psychological Explanation (Cambridge, MA: MIT Press). Dretske, F. 1995. Naturalizing the Mind (Cambridge, MA: MIT Press). 1999. Knowledge and the Flow of Information (Stanford: CSLI Publications). Fodor, J. 1983. The Modularity of Mind (Cambridge, MA: MIT Press). Gazzaniga, M, Ivry, R., et al. 1998. Cognitive Neuroscience (New York: Norton). Gendler, T. and Hawthorne, J. 2002. Conceivability and Possibility (Oxford: Oxford University Press). Harman, G. 1990. ‘The intrinsic quality of experience’, Philosophical Perspectives 4: 31–52. Hurley, S. 1998. Consciousness in Action (Cambridge, MA: Harvard University Press). Jackson, F. 1982. ‘Epiphenomenal qualia’, Philosophical Quarterly 32: 127–36. 1986. ‘What Mary didn’t know’, Journal of Philosophy 83: 291–5. 1998. From Metaphysics to Ethics: A Defence of Conceptual Analysis (Oxford: Oxford University Press). 1999. ‘Reference and description revisited’. Canberra: Australian National University. 2003. ‘Mind and illusion’, Philosophy supp. vol. 53: 253–73. Jackson, F. and Pettit P. 1992. ‘Some content is narrow’. In A. Meile (ed.), Mental Causation (Oxford: Oxford University Press). 1998. ‘A question for expressivists’, Analysis 58: 239–51. Kierkegaard, S. 1980. The Sickness Unto Death (Princeton: Princeton University Press). Kohler, I. 1964. ‘The formation and transformation of the perceptual world’, Psychological Issues 3: 1–173.

186 philip pettit Kripke, S. 1980. Naming and Necessity (Oxford: Blackwell). Levine, J. 1993. ‘On leaving out what it’s like’. In G. Humphreys (ed.), Consciousness: Psychological and Philosophical Essays (Oxford: Blackwell). Locke, J. 1975. An Essay Concerning Human Understanding (Oxford: Oxford University Press). McGeer, V. 2003. ‘The trouble with Mary’, Pacific Philosophical Quarterly 84: 384–93. McGeer, V. and Pettit, P. 2002. ‘The self-regulating mind’, Language and Communication 22: 281–99. McGinn, C. 1993. Problem of Consciousness: Essays towards a Resolution (Oxford: Blackwell). 1999. The Mysterious Flame: Conscious Minds in a Material World (New York: Basic Books). Milner, A. and Goodale, M. 1995. The Visual Brain in Action (Oxford: Oxford University Press). Myin, E. 2001. ‘Color and the duplication assumption’, Synthese 129: 61–77. Myin, E. and O’Regan, J. 2002. ‘Perceptual consciousness, access to modality and skill theories’, Journal of Consciousness Studies 9: 27–45. Nagel, T. 1986. The View from Nowhere (Oxford: Oxford University Press). Noe, A. 2002. ‘On what we see’, Pacific Philosophical Quarterly 83: 57–80. O’Regan, J. and Noe, A. 2002. ‘What is it like to see? A sensorimotor theory of perceptual experience’, Synthese: 129: 79–103. Peacocke, C. 1992. A Study of Concepts (Cambridge, MA: MIT Press). Pettit, P. 1993a. The Common Mind: An Essay on Psychology, Society and Politics (New York: Oxford University Press). 1993b. ‘A definition of physicalism’, Analysis 53: 213–23. 1998. ‘Practical belief and philosophical theory’, Australasian Journal of Philosophy 76: 15–33. 2000. ‘A sensible perspectivism’. In D. Dunlop (ed.), Dealing with Diversity (London: Routledge). 2002. Rules, Reasons, and Norms: Selected Essays (Oxford: Oxford University Press). 2003. ‘Looks as powers’, Philosophical Issues (supp. to Noˆus) 13: 221–52. 2004a. ‘Descriptivism, rigidified and anchored’, Philosophical Studies 118: 323–38. 2004b. ‘Existentialism, quietism and philosophy’. In B. Leiter (ed.), The Future for Philosophy (Oxford: Oxford University Press). 2004c. ‘Motion blindness and the knowledge argument’. In P. Ludlow, Y. Nagasawa and D. Stoljar (eds.), The Knowledge Argument (Cambridge, MA: MIT Press). Pettit, P. and Stoljar, D. 2004. ‘How to solve permutation problems’. Canberra: Australian National University. Searle, J. 1983. Intentionality (Cambridge: Cambridge University Press). Stalnaker, R. 2004. ‘Assertion revisited: On the interpretation of two-dimensional modal semantics’, Philosophical Studies 118: 299–322.

consciousness and the frustrations of physicalism

187

Stoljar, D. 2001. ‘The conceivability argument and two conceptions of the physical’, Philosophical Perspectives 15: 393–413. Thompson, E. 1995. Colour Vision: A Study in Cognitive Science and the Philosophy of Perception (London, Routledge). Tye, M. 2000. Color, Content and Consciousness (Cambridge, MA: MIT Press).

This page intentionally left blank

8 Jackson’s Change of Mind: Representationalism, a Priorism and the Knowledge Argument ROBERT VAN GULICK

1. Introduction Few arguments in the recent philosophy of mind have generated as much discussion or controversy as Frank Jackson’s Knowledge Argument, and none has had a more surprising history. Having offered the Knowledge Argument (KA) in 1983 and ably defended it for fifteen years against a wide variety of objections, Jackson did an about-face in 1998 and disavowed it without really explaining his reasons for doing so. Only several years later in his paper ‘‘Mind and Illusion’’ did he offer a detailed account of his change of mind and explain where he had come to believe the argument goes wrong. His current rejection of the argument is essentially coupled with the acceptance of another controversial if widely held view, namely the Representational Theory of Consciousness or at least of Conscious Sensory Experience (RTS), which he regards as the sole means to defeat the KA and avoid the cognitive illusion that generates its intuitive appeal. Jackson’s current position thus involves at least three elements: a rejection of the Knowledge Argument, an acceptance of the Representational Theory of Conscious Sensory Experience, and a claim that the RTS provides the only effective grounds for rejecting the KA. My present concern is primarily with the third of these claims and with the question of whether one must accept the representationalist view in order to avoid the anti-physicalist conclusion of the Knowledge Argument? I also have reservations about the representationalism, but I will focus here on its import for the Knowledge Argument rather than on the RTS itself. Determining whether it is the sole means to refute the KA in fact involves answering two component questions: Does the RTS provide a

190

robert van gulick

means to refute the KA? And are there any other ways to refute the KA that do not rely on the RTS? According to Jackson the respective answers are ‘‘yes’’ and ‘‘no’’, but both are open to debate. Many plausible non-RTS objections to the KA have been offered over the past two decades, and some may in fact provide valid means of refuting it without needing to accept the RTS. Moreover, it is not obvious that the RTS in fact suffices to undermine the KA; adoptive defenders of the newly orphaned KA may still have resources to defend it against the line of criticism the RTS supposedly provides. I have elsewhere earlier surveyed the space of critical objections to the Knowledge Argument. Over a decade ago (1993), I described four main lines of objection then found in the literature, and more recently (2004) I have added two others. None of the six relies in any obvious way upon the RTS, so I will begin by briefly reviewing both those old and new options in Sections 2 and 3 respectively. In Section 4, I will consider where, if anywhere, Jackson’s own present critical view fits in the logical geography of options. I will then turn in Section 5 to the question of whether the RTS is in fact either necessary or sufficient for refuting the KA.

2. The space of existing replies The Knowledge Argument appeals to Jackson’s now famous thought experiment of Mary the hypothetical super color scientist who supposedly knows all there is to know physically about having a red visual experience but who has spent her entire life in achromatic isolation and thus herself never had any such experience. The key intuition is that when Mary is finally released and has her first red experience she will learn (or come to know) something that she previously did not. Her epistemic gain is alleged to refute physicalism, which if it were true would have left nothing for Mary to have learned given her prior comprehensive physical knowledge. The master argument can be stated simply as follows. Argument KA P1 P2 Therefore

Mary before her release knows everything physical there is to know about seeing red. Mary before her release does not know everything there is to know about seeing red because she learns something about it on her release.

jackson’s change of mind C1 C2

191

There are some truths about seeing red that escape the physicalist story. Physicalism is false, and phenomenal properties cannot be explained as (or identified with) physical properties.

The argument is simple but bold. It asserts a major metaphysical conclusion with sweeping empirical consequences on the basis of intuitions about a purely hypothetical case. As such it shoulders a heavy burden of proof. The plausibility of the argument and of any assumptions on which it relies must be weighed against the countervailing force of widely held and apparently well-confirmed theory. The Knowledge Argument is also an example of what I have elsewhere (2003) labeled a ‘‘boomerang argument’’. The characteristic feature of such arguments is that they reach over into the epistemic/cognitive/conceptual domain to a supposed fact about our representation of reality and then use it to swing back to reach a conclusion in the metaphysical/ontological realm of facts about reality itself. More specifically, boomerangs often move from supposed gaps or lacks in our representations or concepts of the world to conclusions about objective gaps within the world itself and ontological distinctions among the real things in it. This is clearly the case with the Knowledge Argument as well as with dualist arguments based on the alleged logical possibility of zombies (i.e. beings physically indistinguishable from normal humans but devoid of any conscious mentality). Boomerang arguments have a long history that go back at least as far as the early seventeenth century to Descartes’s argument for mind–body dualism by appeal to the supposed conceivability of the disembodied persistence of his mind. They have an equally long history of criticism, and they seem especially vulnerable to charges of illicitly inflating a subjective or conceptual gap into an objective or ontological one. Nonetheless, each argument deserves consideration on its specific merits; perhaps some boomerangs can be thrown validly and find their target soundly. However, they also merit special scrutiny, and the Knowledge Argument perhaps even more than others since it relies on a doubly epistemic route to reach its result. It invokes intuitions about what Mary would and would not know in highly non-actual conditions, and then uses those supposed epistemic lacks to support its anti-physicalist conclusion. To push the boomerang analogy a bit, the argument has to circle twice within the epistemic realm before landing back on the ontological metaphysical side. Perhaps the Knowledge Argument can fly the course, but it is open to multiple lines of attack by those who believe it cannot.

192

robert van gulick

Q1. Does Mary learn anything, or gain knowledge, when she first experiences red? NO. (Churchland & Hardin)

YES.

Physicalism is not refuted.

Q2. What sort of knowledge does Mary gain? Is it strictly know-how, or does it include new knowledge of facts or propositions? Gains only know-how (Lewis & Nemirow)

Gains new knowledge of propositions.

Physicalism is not refuted.

Q3. Does Mary come to know new facts and new propositions? NO. She only comes to know old propositions in a new way. (Tye, Horgan & Churchland)

YES. She comes to know new propositions.

Physicalism is not refuted.

Q4. On what mode of individuation does Mary learn a new proposition? Only on a fine-grained mode (Lycan & Loar) Physicalism is not refuted.

On a coarse-grained mode (Jackson pre-1998) Physicalism is refuted.

Figure 8.1. Map of Four Critical Options

As I have shown earlier (1993, 2003), most of the critical replies in the literature can be sorted by their answers to a series of diagnostic questions that identify the points at which they take issue with the argument’s line of reasoning. Those questions can be used to generate a tree diagram (see Figure 8.1) that locates the various critical options in logical space. Some challenge the argument early on; others go some distance with it before disagreeing. The first diagnostic question asks whether Mary would in fact learn anything or gain any knowledge when she first experiences red. The argument assumes and requires that she would, but some physicalists have held otherwise and challenged the reliability of our intuitions about such highly counterfactual cases. They note that by stipulation, Mary knows everything physical there is to know about having a red experience. Such comprehensive physical knowledge is so far beyond what we possess today, indeed beyond what a real human being with our inherent cognitive limits could possess, that we should be cautious in evaluating our intuitions about such cases. There is the danger that our intuitions might fail to reflect the absolutely unlimited nature of Mary’s stipulated physical knowledge, and that we might simply imagine someone who knows a lot more of the sort of physical knowledge we possess today but still falls far short of totally comprehensive physical knowledge. Given how

jackson’s change of mind

193

unlike our current physical knowledge Mary’s state would have to be, one might doubt that we can have clear or reliable intuitions about what she would or would not understand from such a physically comprehensive epistemic perspective. Perhaps Mary, or super Mary, when released would exclaim, ‘‘Yes it is just as I knew it would be.’’ Perhaps not, but there seems reason for caution in making any such predictions, given how far from epistemic reality our intuitions are asked to stretch. If one concedes that post-release Mary gains at least some knowledge, the next question in the decision tree (Q2) asks whether she gains any new propositional knowledge or only new know-how. According to the so called ‘‘ability reply’’ Mary gains only new epistemic abilities or skills to recognize, imagine, and perhaps compare red experiences; she gains no new propositional knowledge. If we think about a gain of propositional knowledge, as David Lewis did, as a reduction of uncertainty, then the ability reply denies that post-release Mary eliminates any prior open epistemic possibilities. Her having experienced red does not enable her to locate herself more narrowly or precisely in logical space; her comprehensive physical knowledge already sufficed to identify her location. The differences and enhancements in her epistemic situation are solely matters of acquiring new skills, but these newly acquired abilities to recognize or imagine red experiences supposedly do not constitute the acquisition of any new propositional knowledge. The epistemic additions are all just matters of new practical know-how and not of any new knowledge. Mary no doubt acquires some important new know-how, but whether that is all she acquires is less obvious and more controversial. Lewis and Nemirow see only new abilities, but others including Jackson in the past, insist there is more to Mary’s gain. (Jackson now sides with Lewis and Nemirow, though he claims to have followed a different route to arrive there.) Those who believe Mary gains more than mere know-how, move on to Q3, which initially may sound the same as Q2, though it is importantly different. Q3 asks whether Mary gains any knowledge of new propositions. The difference between Q2 and Q3 concerns the scope of ‘‘new’’. Q2 ask about new propositional knowledge, whereas Q3 asks about knowledge of new propositions. Mary might gain the first—but not the second—if her propositional knowledge could change in ways sufficient to count it as new knowledge, even if it does not involve an extension in the range of propositions she knows. Could there be sufficient changes in the knowledge states themselves, even if she gains no new objects of knowledge? We might for example count Mary as having new propositional knowledge if she gained a new way of conceptualizing, accessing or representing to herself some

194

robert van gulick

propositions that she already knew. The proposition that she knows may be the same, but if her way of knowing it is sufficiently different we may want to count it as a different epistemic state and as a gain in propositional knowledge. Some have disparaged any such gain as trivial. David Lewis argued that if learning a new way of representing a previously known proposition sufficed for a gain in propositional knowledge, one could gain lots of new knowledge about non-linguistic matters just by learning a new language. If I learn Urdu (his example) and gain the ability to represent the proposition that snow is white (‘‘ ’’) I do not increase my propositional knowledge about meteorological matters (as opposed to what I may learn about Urdu itself). Lewis seems right about the example, but its relevance to the Knowledge Argument is less clear. The representational or conceptual changes involved in Mary’s case may not be analogous. The modes of cognitive access and representation made available to post-release Mary likely differ from those afforded by neuroscience far more than Urdu does from English as a means of cognizing meteorological propositions. The distinction to which Q3 points is sometimes couched in terms of modes of presentation; one might know a proposition under (or through) one mode of presentation but not under another. I might know the glass contains water under a perceptual and folk conceptual mode of presentation, but not have access to the same proposition under the more theoretical and abstract chemical mode that presents it in terms of H2 O molecules. Gaining the latter mode of access seems an enhancement of my overall epistemic status even if the proposition and associated property in the latter case is the same as that involved in the former (being water just is the property of being composed of H2 O molecules). Similarly the modes of presentation and access that Mary gains through her experiential, introspective, and empathetic systems when she sees her first tomato would seem to enhance her cognitive and representational access to propositions previously known to her only through theoretical neurobiologial concepts. Thus some critics of the Knowledge Argument have aimed to block its progress at Q3. They concede Mary gains new propositional knowledge, but only in the sense that she gains significantly new modes of representing, conceptualizing, or cognitively accessing propositions that she already knew. Moreover, they argue that such a gain is not inconsistent with physicalism and provides no reason to reject it, except perhaps in extreme reductionist forms that treat physical theories as in themselves sufficient for the practical representation of everything real. Most physicalists are far more pluralistic about the diversity of modes of access and presentation that we need. Few, if any, would advocate doing economics with the resources of physical theory. Nor

jackson’s change of mind

195

should we require neurobiology to duplicate or provide all the useful modes of cognitive access and engagement available through the first person perspective. Mary’s apparent epistemic gain might thus be fully explained in terms of her gaining new modes of access and presentation to facts and propositions she previously knew through the means of physical science. Defenders of the KA may deny that the strategy can be used when it is facts about modes of presentation themselves that are centrally at issue in the argument. It is what Mary learns about the experiential mode of sensing red that supposedly escaped her prior physical knowledge. Physicalists might counter in return that Mary’s prior knowledge did indeed include those very propositions about experience, though her access to them was through non-experiential modes or means. As long as she learns no new propositions, physicalism is not in jeopardy. Others go even further with the argument before making a late swerve away from its anti-physicalist conclusion. They concede not only that Mary gains new propositional knowledge but also that she comes to know propositions she did not previously know, and yet they deny that her gain refutes physicalism. The crucial issue is how one individuates propositions and thus what counts as coming to know a new one. Q4 asks on what mode of individuation does Mary count as learning a new proposition? There are many possible schemes of individuation one might use, some more coarse grained or fine grained than others. The latter distinguish among propositions that are regarded as the same by the former. In a way Q3 and Q4 might be seen as two ways of asking the same question, or two sides of the same critical coin. Where does one draw the line between two ways of knowing the same proposition and knowing two distinct but closely related propositions? Is coming to know that the glass contains H2 O coming to know a new fine grained proposition or merely gaining a new mode of access to a previously known coarse grained one? The fine grained mode distinguishes propositions in terms of conceptual or representation features that more coarse grained schemes disregard as external matters about modes of access rather than the identity of propositions themselves. There may be no ‘‘correct’’ place to draw the line, and its exact location may be of no real importance to the Knowledge Argument. Physicalist critics might try to block it by explaining Mary’s total epistemic change as a matter either of her coming to know an old proposition in a new way or of gaining knowledge of a new proposition in the fine grained sense. What the physicalist critic seemingly cannot concede is that Mary learns a new proposition on a coarse grained mode of individuation. Her doing that would seem to entail the existence of properties cognitively inaccessible to her,

196

robert van gulick

rather than merely concepts or modes of presentation that she could not grasp. And if there were such properties of which she lacked prior knowledge then it would also seem to follow that there are possibilities and uncertainties she could not eliminate until she had had a red experience, possibilities defined in terms of the presence or absence of the relevant property(ies). Since by stipulation she knew about all the physical properties beforehand, it would thus seem to follow that there must be at least some non-physical properties, and physicalism would be refuted. Thus with respect to Q3 and Q4, the key issue dividing supporters and critics of KA seems to be whether or not Mary comes to know a new coarse grained proposition.

3. Two newer critical options Further to that which I once believed and wrote (1993), I have, more recently (2003), come to believe there are at least a couple of further critical options, including one that may even allow the critic of the KA to concede that Mary learns a new coarse grained proposition, while still denying that her doing so refutes physicalism. The two new replies to the KA more or less bracket the earlier four and can be located on an expanded tree diagram (see Figure 8.2) that adds one question before the original Q1–Q4 sequence and another after, call them Q0 and Q5 respectively. Although Q0 and Q5 come at the opposite ends of the diagnostic chain, they turn on a common issue, namely whether or not physicalism can accommodate subjective facts and propositions. Facts are subjective in the relevant epistemic sense (Nagel 1974) if they are fully knowable and understandable only from a particular point of view; that associated with a particular form of experience. It is in that sense that Nagel claimed that facts about what it is like to be a bat are subjective. However, the existence of such facts or propositions refutes physicalism only if one assumes that all physical facts and propositions are objective and capable of being fully understood and known from many different perspectives and by many types of knowers with diverse experiential profiles. Strictly physical facts of the sort represented by physical theory do indeed seem knowable without restriction to any particular experiential point of view, but it is not obvious that all physical facts are of that sort, especially if one defines the domain of the physical to include all facts that are realized by virtue of underlying strictly physical facts. If first person facts about experience are physically realized facts, then there may be some facts and propositions that are both physical and subjective.

jackson’s change of mind

197

Q0. Is it possible for Mary prior to release to know all the physical facts and propositions about seeing red? NO. Not if there are physical subjective facts.

YES

Physicalism is not refuted. to Q1, figure 8.1

Q4. On what mode of individuation does Mary learn a new proposition? Only on a fine-grained mode

On a coarse-grained mode

Q5. Would Mary's learning a new coarse-grained proposition or her coming to better

understand her location in logical space refute physicalism? NO. (Non-reductive physicalism)

YES. (Jackson pre-1998)

Physicalism is not refuted

Physicalism is refuted.

Figure 8.2. Expanded Map of Critical Options

Moreover, some physicalists may be able to give a plausible account of how that might be. If one is a nonreductive physicalist and takes a pragmatic view of understanding and cognition, then it will hardly be surprising that there are physically realized facts that one cannot understand by using physical theory. Ontological realization does not entail that the methods sufficient

198

robert van gulick

for modeling the entities and actions at the level of underlying realization must suffice as well for modeling or understanding all the higher level aspects of the realized systems and their modes of interaction, especially since the relevant modeling needs to be done by particularly structured cognitive agents engaged and situated in a diverse range of contexts of application and interaction. Given the multiple parameters that condition those contexts of understanding and the intentional relations they subserve, it becomes all the less plausible to regard physical theory as a universal tool of understanding, or as an all purpose medium, providing commensurable translations of everything that might be validly represented or expressed by any other engaged and situated cognitive system about any aspect of the physically realized world. This is especially so when the system being understood and the one doing the understanding are one and the same, as they are in consciously selfunderstanding systems, such as a human or a bat. The nonreductive physicalist should not be at all surprised that such systems are able to engage their own physically realized dynamic features in ways that cannot be duplicated by any equivalent external mode of cognitive access. If conscious minds involve a great deal of reflexive self-understanding, as I believe they do, they may well have aspects or properties that cannot be fully understood from an outside perspective. Physicalists of a nonreductive and pragmatic bent would not suppose that third person physical theory must provide us with a means of grasping and understanding everything that physically realized self-understanding systems are able to understand about themselves. A commitment to universal physical realization carries no commitment to physical theory as a universal representational tool adequate for understanding everything real. Indeed I have long argued (1985, 1992) that the existence of subjective facts is not only compatible with physicalism but predicted and entailed by the sort of teleo-pragmatic physicalism that I take to be most plausible. I will not repeat those earlier arguments in detail here, but give just the main idea. According to the teleo-pragmatic view understanding is always a matter of the potential for successful practical engagement, which often involves mutual reciprocal interaction and perhaps what might be called ‘‘causal resonance’’ or ‘‘causal harmonic engagement.’’ The intentional profile or content associated with such understanding is very much a function of the causal profile of that engagement and the sort of access it affords the understander to interact successfully with that which it understands. It is highly unlikely that we could fashion any structure through the use of external third person physical theory that would allow us to come close to duplicating the causal profile associated

jackson’s change of mind

199

with the way in which a complex physically realized self-understanding system, such as a bat or human mind/brain, understands itself. Nor need the teleo-pragmatic physicalist or any nonreductive physicalist, think otherwise. The contexts of application for the respective representations are so dissimilar that it is unlikely we can use the one to replicate the causal interactive profile of the other to a degree that would make them intentionally equivalent. That at least is the main idea, and the details can be found elsewhere by those who may be interested (Van Gulick 1985, 1992). For present purposes, it should suffice to make at least a prima facie case for the possibility that subjective facts might coexist consistently with some plausible versions of physicalism. That possibility provides us with an additional option for parting company with the Knowledge Argument, indeed with an option for ‘‘getting off the train’’ even earlier by disputing premise P1. P1 is generally conceded by the argument’s critics since it functions as a stipulation of the conditions about which we are invited to engage in hypothetical reasoning. Critics may note that no real human could in practice command as much as Mary is imagined to do, but the limits would seem to be merely contingent, such as those on memory or attention capacity. Thus there might seem to be no reason why someone could not at least in principle know all the physical facts as P1 supposes Mary does before her release. However, if some physical facts are subjective (in the sense of being understandable only from a specific experiential perspective), then Mary could not even in principle know them given the stipulated restrictions on her past range of experience. Q0 asks, ‘‘Is it possible for Mary prior to release to know all the physical facts and propositions about seeing red?’’ In reply, a physicalist who believes in physically realized subjective facts should answer ‘‘No’’, and thereby also reject P1 of the KA. Despite its status as a mere stipulation, one might refuse to concede it on the grounds that it embodies a hidden contradiction or a conjunction of conditions that cannot be jointly satisfied. Mary is supposed to have never experienced red and yet to know all the physical facts about doing so. Perhaps she could in principle know all the ‘‘physical science facts.’’ However, if there are subjective physical facts, i.e. facts that are physically realized but cognitively accessible only from the experiential perspective of a certain range of physically realized self-understanding systems (whether bats or normal color experiencing humans), then Mary could not know all such physical facts about seeing red given her stipulated lack of past experience. Thus a physicalist critic who believes in the existence of subjective physical facts, as I believe she should, has good grounds for not taking even the argument’s first step.

200

robert van gulick

Let us turn then to our other new option and the final question in our expanded tree. Q5 asks, ‘‘Would Mary’s learning a new coarse grained proposition or her coming to better understand her location in logical space refute physicalism?’’ I had earlier believed that the answer to Q5 would have to be ‘‘Yes’’, but I now think there are plausible grounds for saying ‘‘No’’, at least if one unpacks the mental-physical relation nonreductively in terms of realization. I have given a fairly complete explanation of those grounds recently elsewhere (2004), but I hope a quick reprise of that discussion here will suffice to indicate the main idea. One needs to consider two crucial and interdependent issues at the outset. First in what sense are higher level facts and propositions about realized mental properties new and different from those that concern their underlying physical realizations? Second, put in epistemic terms directly relevant to the Knowledge Argument, would knowing all the facts or true propositions that hold at the microphysical level suffice for knowing all the true facts and propositions that are realized at every level by those underlying microphysical facts? In what sense would knowing the first set settle every question about the latter? Does knowing all the realizer facts guarantee that one also knows all those that are realized? And would knowing all the microphysical facts leave any room to reduce one’s semantic uncertainty about one’s location in the space of possibilities? If one accepts a principle of microphysical supervenience, then it may seem that no such room could be left. If all the facts supervene on the microphysical facts then once the microphysical facts of the actual world are set, then so too are all the other facts. There are no subregions left for further information to eliminate. In particular the space specified by the microphysical facts does not divide into distinct subregions that differ from each other with respect to any real non-physical facts. That is just what supervenience entails: no real difference without a microphysical difference. However, one must take care in moving from facts about ontology to matters of epistemology. If coming to know a new coarse grained proposition requires locating oneself more narrowly in the logical space of worlds, then being able to locate oneself microphysically would seem to maximize the definiteness of one’s location and thus to preclude learning any new coarse propositions. However, that implication depends crucially on what counts as knowing one’s location in logical space. One may be able to specify a given world as actual and locate it in microphysical space, but still not have a full understanding of the structure of the overall logical space within which it is embedded. One might be like the lost visitor in a cartoon joke who confronts an informational display which announces ‘‘You are here ∗ ’’ but provides no further map to explain where

jackson’s change of mind

201

here ∗ is. Given supervenience, the microphysical facts may fix its location in that space, but in so far as one does not understand the structure of the containing space, one’s knowledge of one’s location is incomplete and may leave room for more than notional additions. And if one’s cognitive mastery of the structure of the space is incomplete, so too may be the range of propositions one can know or even understand. In what sense is the complex proposition that specifies the total realization base RB for some actual world object X a different proposition from that that expresses X’s having the realized higher level property HP? If we think of propositions in the relatively coarse sense as functions from worlds to truth values, it seems that different functions would be associated with propositions that respectively specify objects as having RB or HP. On the assumption that HP, like most higher level properties, is open to a diversity of multiple realizations, the two functions will diverge in cases where objects realize HP in virtue of some base other than RB. The function for the HP proposition will assign T in such cases, but that for the RB will assign F. Given supervenience, everything which gets a T from the RB function will also get one from HP, but the converse does not hold because of multiple realization. Thus even on a coarse mode of individuation, the proposition that X has RB is not the same as the proposition that X has HP. However, their nonidentity does not by itself guarantee the possibility that Mary could know the first but not the second. Even if the propositions are distinct, it still may be that knowing the first in some way suffices for knowing the latter: either because knowing the realizer facts in general suffices for knowing the realized facts, or because it does so for particular reasons in the Mary case. Supervenience might seem to support such an entailment. As a matter of metaphysics or ontology, X’s having RB suffices for X’s having HP (or at least does so modulo the natural laws of the actual world). So knowing the realizer facts and the relevant laws might seem to guarantee that one should also know the necessarily realized facts as well. One need only draw the consequences. Of course, we often fail to explicitly recognize many of the less obvious consequences of our beliefs; mathematics otherwise would be much easier than it is. Moreover, Mary’s lack is worse than that in the mathematical case. Her problem is not that of determining which of a given set of propositions are true (or follow as true), but rather of being able to even grasp or understand them. Mary’s ability to understand all the propositions about the physical base need not seem sufficient to guarantee her mastery of those about the realized properties, nor is there any guarantee that the concepts or modes of

202

robert van gulick

cognitive access needed to grasp propositions about higher-order realized facts are definable in terms of those that suffice for grasping propositions about the realization base. Indeed, according to the nonreductive physicalist, we should commonly expect such failures of definability or translatability between the representational systems of the special sciences and those of underlying physical theory. They follow naturally from the pragmatic nature of representation and the way in which representational content depends upon the causal structures of both the cognitive agent and the object as well as on the larger dynamic context of engagement within which those representations get applied. There is little reason to suppose that we can fashion cognitively equivalent higher level tools out of the resources that suffice for successful engagement at the realization level. As noted above, physics just does not give us the conceptual and representational tools we need for doing economics or introspective psychology. Returning to Mary, we can now see why pre-release she might lack a full understanding of propositions about seeing red, even if such experiences are just higher-order physically realized states. Moreover, we can better appreciate the respect in which she might fail to understand the structure of logical space and thus of her actual location in it. In particular she may fail to adequately understand the structure of similarity relations that hold between the various situations and worlds in which the higher-order property of seeing red is realized by diverse physical substrates. In so far as Mary cannot articulate those relations, her command of the structure of logical space may be inadequate to grasp either the nature of the realized higher-order property or the coarse grained proposition defined in terms of its presence or absence. If we treat coarse grained propositions as functions from worlds to truth values, Mary may lack cognitive command of the relevant function that attributes the higherorder property to particular experiencers. Thus prior to her release there may be coarse-grained propositions concerning physically realized properties that Mary cannot grasp, and in so far as she lacks an adequate cognitive command of the structure of logical space she may fail to fully appreciate her location in it. The physicalist thus might put off his disagreement with the KA until Q5, conceding in response to Q4 that Mary learns a new coarse grained proposition but still denying that her gain refutes physicalism. Thus there are at least six points at which the physicalist might take issue with the KA: (1) If she accepts the existence of physically realized subjective facts, she might reject P1 on the grounds that the stipulated conditions of the hypothetical contain an implicit contradiction. (Q0)

jackson’s change of mind

203

(2) She might deny that Mary learns anything new upon her release. (Q1) (3) She might claim that Mary gains only new abilities and know-how. (Q2) (4) She might concede that Mary gains new propositional knowledge but only in the sense of coming to know a previously known proposition in a new way. (Q3) (5) She might concede that Mary comes to know new propositions but only on a fine grained mode of individuation. (Q4) (6) She might allow that Mary learns a new coarse grained proposition but still deny that her doing so refutes physicalism, if physicalism allows for physically realized subjective facts. If it does, Mary’s gain might be consistently explained as her coming to understand such facts and to appreciate the structure of logical space in the ways needed to grasp the relevant propositions about them. (Q5)

4. Locating Jackson’s rejection Having surveyed the above six options, we can ask where in our tree, Jackson’s own current objections to the argument might best be located. He places himself with Lewis and Nemirow as embracing the Ability Reply and accounting for Mary’s epistemic gains as solely a matter of acquiring new abilities to recognize and imagine red experiences. Thus Jackson is most aptly placed as parting with the KA at Q2. However, there is a respect in which his current view also coincides sympathetically, if not quite logically, with those like Churchland who challenge the KA one step earlier by denying Mary would learn anything upon release. He takes a strong view about the reach of physical theory and claims Mary could, at least in principle, deduce a priori all the truths about seeing red from her comprehensive knowledge of the microphysical facts. He believes she could in principle infer every true proposition about the experience of seeing red from the physical theory facts. Thus his current position might best be described as the conjunction of two claims: (1) Pre-release Mary can deduce all the mental facts or propositions about experiencing red. (2) Mary improves her epistemic status when she herself first experiences red, but only by gaining some new know-how and practical cognitive abilities. Jackson also presents his view as crucially dependent on the representational theory of sensory experience (RTS), but how so? It is invoked to support the

204

robert van gulick

claim that Mary can deduce all the mental facts from the microphysical facts. According to the RTS, experiences of seeing red have no mental properties other than their representational or intentional properties, and Jackson believes all the relevant representational facts are in principle deducible from the microphysical facts. Although he does not endorse any specific account of representational content, Jackson asserts that the extant theories—whether causal, covariational or teleological—allow the a priori deducibility of all the facts about what is represented from a sufficiently complete description of the microphysical facts. If representational intentional facts are fully deducible from microphysical facts, and sensory experiences such as seeing red have no mental properties beyond their representational intentional properties, then it would follow immediately that one can deduce all the mental facts and propositions about seeing red from the microphysical facts. Since Mary supposedly knows all the latter, it would follow that she can know all the mental facts about red experiences before ever having one herself. An important qualification is immediately required. Jackson does not claim that Mary could in practice carry out any such deduction; it may well be beyond her actual calculational or reasoning powers, indeed beyond those of any actual human. Jackson claims only that it is in principle possible for Mary to carry out such an a priori deduction. The role of the RTS is to deflect or rebut a putative objection to the claim of a priori deducibility. If the RTS were false, then conscious sensory states might have further properties whose nature or presence could not be deduced from the microphysical facts. In particular, conscious sensory states might be thought to have phenomenal properties that were inaccessible through the deductive route. Indeed it is the belief in the existence of such properties that constitutes the ‘‘illusion’’ in Jackson’s title ‘‘Mind and Illusion.’’ If one believed, as Jackson apparently did in the past, that sensory experiences themselves had phenomenal color properties (or involved acquaintance with inner items having them), then Mary might not be able to deduce all the relevant propositions about such experiences from the microphysical facts. If phenomenal redness (phred-ness) were an intrinsic property of red experiences (or of red sense data that were accessible only through such experiences), then there would be true propositions about experiences (or inner objects of experience) being phred or having phred-ness. Mary’s prior propositional knowledge would be incomplete unless she could deduce the required propositions about the presence or absence of phred-ness, but her mastery of the microphysical facts need not supply her with adequate cognitive access to phred-ness or its nature. Thus one who believed in the existence of phenomenal properties

jackson’s change of mind

205

would likely deny that Mary could a priori deduce all the propositions true of red experience, and argue instead that Mary upon release would gain knowledge of propositions about phenomenal properties, such as phred-ness, that were previously beyond her ken. (Though we will see in Section 5 that some physicalists might accept the existence of phenomenal properties but still contend that Mary could nonetheless deduce all the facts about experience.) The primary dialectical role of the RTS in Jackson’s current thinking seems to be to block the phenomenal properties objection to the deducibility thesis. According to the representationalist, there are no phenomenal properties such as phred-ness, so there are no propositions about them that pre-release Mary fails to know or is unable to deduce. On the representationalist theory, having a red experience is not a matter of having an experience with the property of phred-ness nor of being acquainted with a phenomenal object or sense datum with the property of phred-ness. It is simply a matter of being in a state that phenomenally represents external objects as being red. The only color property involved is ordinary red-ness, and it occurs as part of the representational or intentional content of the experiential state not as a property of the state or any inner objects. If we use square brackets to indicate contents, the way to think about Mary’s seeing a red tomato according to the RTS is roughly as Mary visually represents [tomato is red]. The only occurrence of ‘‘red’’ is internal to the intentional content, and the relevant property being referred to is red-ness—a property of ordinary external objects. There is no reference to any phenomenal properties such as phred-ness, nor to any color properties of experiences or inner mental objects. Indeed the representationalist denies the existence of phred-ness. If there is no such property as phred-ness, then there are no coarse grained properties concerning which items have it or lack it. Jackson appears to believe that if to the contrary he were to reject the RTS, he would be committed to the existence of phenomenal properties such as phred-ness. And if so, then there would be propositions in the coarse grained sense that escaped Mary’s prior knowledge, namely those concerning which items had and lacked the relevant phenomenal properties. Mary’s pre-release knowledge would thus be incomplete in the strongly propositional sense that there were real possibilities and uncertainties she could not eliminate despite her complete microphysical knowledge. Since Jackson believes any such result would refute physicalism, he regards the RTS with its denial of phenomenal properties as essential for blocking the Knowledge Argument.

206

robert van gulick

5. The RTS and the KA As a matter of intellectual autobiography, we have no reason to doubt the role played by the RTS in Jackson’s change of mind about the KA. However, as a matter of logic, it is less clear that the RTS is either necessary or sufficient to undermine the Knowledge Argument. Though Jackson justifies drawing the line where he does by appeal to the RTS, some critics disagree with the KA at that same location in the question tree but without obviously relying on the RTS to do so. Others challenge the argument at different locations, none of which seem to rely on the RTS. Moreover, looking a bit deeper at the interplay between the KA and the RTS may give us a clearer understanding of just where the original argument may go wrong. At least some physicalists should be happy to have other options for blocking the KA. Though the RTS is currently quite popular and ably defended, it is far from universally accepted, and those who oppose it can marshal some plausible concerns against it, especially against its claim that conscious sensory states have no mental properties other than their intentional representational ones. Inverted spectrum cases, as well as Ned Block’s inverted Earth cases, seem to indicate otherwise. Representationalists have of course replied and claim to have shown that such alleged possibilities pose no threat to their position. As noted at the onset, my current aim is not to adjudicate those disputes nor offer a bottom line judgment on the truth or adequacy of the representationalist view, but only to address its relation to the Knowledge Argument. In that respect, many physicalists who are less than fully convinced about the truth of representationalism, would probably prefer not to have their refutation of the KA held hostage to the fate of the RTS. Thus it is worth getting clear on the link between the KA and the RTS, even if one cannot reach consensus about the RTS itself. Though it may sound odd, I think the best way to see if the RTS is necessary for refuting the KA is to look first at the complementary question of its sufficiency. Doing so should help to clarify in just what way the RTS supposedly undermines the reasoning of the Mary argument. The RTS is of course not by itself sufficient in the strict sense to refute the KA. One must invoke at least some other claims, but Jackson seems to think the other assumptions one needs are all relatively obvious and not open to real challenge. However, some of them may be more problematic than he supposes, particularly the strongly a priorist assumption he relies on to conclude that Mary could deduce all the representational facts from the microphysical facts.

jackson’s change of mind

207

As noted in the previous section, a key conditional in his reasoning has a conjunctive antecedent: If all the representational intentional facts are fully deducible from microphysical facts, and sensory experiences such as seeing red have no mental properties beyond their representational intentional properties, then it would follow immediately that one can deduce all the mental facts and propositions about seeing red from the microphysical facts. The RTS asserts the second conjunct, but one needs independent grounds for the first. What guarantees the a priori deducibility of the representational facts? Another way of putting the need is in terms of the dialectical role of the RTS in Jackson’s rejection of the argument. As also noted in the prior section, the RTS serves to parry an objection that might be raised by supporters of the argument concerning the non-deducibility of propositions about phenomenal properties. The RTS aims to moot that objection by denying the existence of such properties and thus of any facts or propositions about them. But even if that works, it is primarily a defensive maneuver and by itself does not establish the positive truth of the deducibility claim, which requires a prior logical basis. Jackson appears to rely on two sources of support for the deducibility thesis, one more specifically concerned with the nature of intentionality and the other of a more general metaphysical/epistemological sort. On the specific side, he appeals to the kinds of naturalistic theories found in the current literature, such as those that ground and determine matters of intentional content on facts of covariance, causal linkage or teleological function. Jackson believes that the analyses offered by such theories invoke conditions of the sort that could be deduced (at least in principle) from the microphysical facts. Thus such naturalistic analyses would provide an inferential bridge to deduce intentional facts from microphysical ones. There are several potential problems. First it is not obvious that all the determining facts could be deduced from the microphysical facts. Causal and covariational facts might seem safely deducible, but the case for deducing teleological facts is less obvious. Moreover, teleology may infect causal or covariational models as well, in so far as teleological considerations may partly determine which covariance or causal relations are relevant to content. Secondly, the naturalistic theories one actually finds in the literature do not seem as algorithmic and precise as Jackson’s reasoning apparently requires. They explain the sorts of factors that supposedly ground content relations, but they rarely if ever provide formulae which one could use to mechanically compute intentional contents. More often what they offer is a framework within which interpretation can take place, but interpretation is seldom if ever a matter of simple computation or deduction. Jackson might dismiss this as merely a practical shortcoming of our current incomplete theories, a lapse that

208

robert van gulick

could be remedied with further refinement. Perhaps so, but one should be cautious about using our current naturalist theories to support belief in any such future ideal theory. It is not obvious that our current models are of a sort that could ever have the requisite precision and computational rigor. Third and most importantly, the extant naturalistic theories of intentionality may not address the representational facts of most importance to the KA, namely those that concern phenomenal representation. Jackson acknowledges that there is a big difference between representing a given content in thought and phenomenally representing it in visual experience. Thinking the tomato is red is not the same as seeing that it is. Part of the difference is no doubt a difference in total content—the visual experience represents far more than the tomato’s color—but that would seem to be only part of the difference. The structure and organization of phenomenal consciousness almost certainly plays an important role in determining the intentional content of conscious states, yet none of the major current naturalistic theories of content has much if anything to say about its role, at least not explicitly. Of course, it may be possible to expand those theories to accommodate such phenomenal factors, but at present that is more a hope or a promise than a theoretical reality, given their relative silence on the critical factors. Thus confidence in the possibility of constructing such a theory likely depends on more general considerations rather than on the actual specifics of current models. Jackson’s belief in the deducibility thesis relies in part on his general a priorist view about physicalism and the realization relation. If entities, facts or properties at some lower level are sufficient to produce those at a higher level, he believes it must be possible at least in principle to a priori deduce the latter from the former. He takes the metaphysical link of realization in itself to guarantee the epistemological link of a priori deducibility. Jackson acknowledges that many physicalists think otherwise, but he professes to be puzzled about why they regard the claim of a priori deducibility as extreme. He takes it as essential to validating the belief that the higher-order entities or properties are real and physical. As Jackson sees it, if their nature and existence could not be deduced from the totality of the microphysical facts, the physicalist ought simply to deny their reality. Indeed in ‘‘Mind and Illusion’’ Jackson defines physicalism, or what he calls ‘‘bare physicalism’’, as the thesis that the world is exactly as required to make the physical account of it true in each and every detail but nothing more is true of this world in the sense that nothing that fails to follow a priori from the physical account is true of it.

Thus for Jackson the deducibility of the intentional and representational facts from the microphysical facts is just a special case of the general requirement on

jackson’s change of mind

209

the intelligibility of physicalism and the thesis that everything real is physically realized. To put it in a slogan, ‘‘No realization without a priori deduction’’ or perhaps ‘‘No metaphysical determination without logical derivation.’’ These are very large issues and certainly not ones that I could hope to settle here. However, it is important to see the major role they play in the Knowledge Argument. Indeed from the perspective of the nonreductive physicalist, the illusion that drives the KA is more a matter of its commitment to a priorism rather than its failure to accept the RTS or its belief in phenomenal properties. Mary’s learning something upon release provides an argument against physicalism only if the truth of physicalism entails the a priori deducibility of all physically realized facts and propositions from those at the microphysical level. If one thinks of physicalism in terms of a logical empiricist model of reduction and the unity of the sciences involving bridge laws and theoretical derivability, then the truth of physicalism would enable Mary to infer all the physically realized facts. But many physicalists, especially those of a nonreductive bent, reject that account of what physicalism requires. The application of these issues to the KA is complicated by the fact that Jackson appeals to in principle deducibility, while our intuitions about Mary more likely reflect what she or any human agent could infer in practice. Jackson writes, when I talk of being able to move a priori from the physical account to, say, Carter being a one term President, I do not mean being able to move literally, I mean there exists an a priori entailment. ( Jackson 2004)

Thus he couches the thesis of physicalism in terms of a logical relation, but the Mary argument turns on matters of epistemology. It is what Mary supposedly could not know that provides the argument’s basis for rejecting physicalism. The issue is further clouded by Jackson’s use of ‘‘a priori’’ to qualify the nature of the entailment. A prioricity is an epistemic feature not a logical one. One can use the phrase ‘‘a priori entailment’’, as Jackson seems to do, to mean ‘‘logical entailment knowable a priori’’. But in doing so, one mixes logical and epistemological factors which one must take care not to blur together. If facts or propositions of one sort logically entail or determine those of another, it does not automatically follow that any actual cognitive agents could carry out the relevant deduction or derivation, nor does Jackson claim otherwise. But if one gives up the automatic move from matters of logic to those of epistemology, it is not clear why Mary’s apparent lack of knowledge should concern a physicalist. Our intuitions about the gap in her prior knowledge need carry no negative implication regarding the logical dependence of the mental on the physical.

210

robert van gulick

Thus the central conflict behind the KA may not be that between supporters and defenders of the RTS, but that between those who take an a priorist view of physicalism and those like nonreductive physicalists who do not. The latter disagreement is clearly prior to the former, regardless of one’s preferred mode of critical reply. Unless Jackson was already committed to a priori deducibility, the RTS would not suffice to support his claim that pre-release Mary lacked no propositional knowledge. The RTS serves only to counter the phenomenal properties objection to that claim. Indeed it is not obvious that the a priorist even needs the RTS. Perhaps rather than denying the existence of phenomenal properties, the a priorist physicalist might concede their reality but claim that they too are physically realized properties, and thus at least in principle a priori knowable on the basis of the underlying microphysical facts. We may at this point have little intuitive grasp of how such an a priori deduction might go, but refuting the KA supposedly requires only that it be in principle possible. If one is already strongly committed to a priorism, one might argue that deducing propositions about intrinsic phenomenal properties of experiences from their microphysical bases is no less possible in principle than deducing their representational intentional properties. I do not mean to endorse such a claim, especially since my own sympathies are not particularly a priorist, but the mere possibility of making such a move, may further serve to show how relatively little Jackson’s critical stance depends on the RTS as opposed to on his a priorist perspective. Physicalists who reject a priorism can reply to the KA with even less (or no) reliance on the RTS. Their form of ontological physicalism is compatible with Mary’s lacking some propositional knowledge about realized properties despite her comprehensive knowledge of those about their physical realization base. Such a physicalist might part company with the KA at Q0, Q3, Q4, or Q5. Wherever he draws the line, he will do so in part because he rejects the view that physical theory constitutes a universal cognitive tool sufficient—even in principle—for representing and understanding everything that is real in a physically realized world. The disagreement between a priorist and nonreductive physicalists may reflect an even more basic difference in perspective about the nature of understanding and the cognitive role of theories. If one thinks of theories as descriptions or pictures of reality, then it might seem that the difference in theoretical representations at different levels is primarily a matter of resolution. Those at the base level that describe individual microphysical facts might correspond to unit pixels, and every representation at a higher theoretical level would in some sense concern patterns or structures constructed out of those pixels and their relations. Fix the pixels and you’ve represented all there is to

jackson’s change of mind

211

represent; at most one needs to do a bit of deduction to read those further contents off the global base level representation. I do not say any actual a priorist would literally make such a claim, but it provides a useful metaphor for thinking about their perspective. In contrast the nonreductivist stresses the practical role of theories in enabling us to engage the world causally and cognitively in its many aspects and levels of order and dynamic organization. Theories, models and systems of representation are thought of as ‘‘cognitive tools’’ that support modes of engagement that in their respective ways count as successfully understanding real aspects of the world. There is no presumption that the representational structures that are of adaptive value for engaging a given level of organization Li should be decomposable into (or constructible out of) those that suffice for dealing with the lower levels of reality Lb that provide the realization base for those at Li . The sorts of representations and cognitive tools that are apt for a given task of understanding will depend not only on the underlying structure of the objects of understanding but also on that of the understanding subject and the various modes of access and interaction that are afforded by their context of engagement. Of course, even the nonreductivist owes us some account of how the various levels and forms of order fit together and how those at higher levels might be realized by those at lower ones. But that account will likely stop far short of outright deductions or translations between systems of representation at distinct levels. The nonreductivist does not typically expect to find such tight links and constraints between the multiple systems of representations that he regards as pluralistically needed for the dealing with the world in all its diverse complexity (Van Gulick 1992, 2002). Moreover, given the existence of cases, such as those involving conscious mind/brains, in which the understander and the understood are internal to one and the same system, the nonreductivist would expect there to be at least some physically realized subjective facts. As noted above in Section three, some physicalists accept the existence of facts or propositions that are physically realized but nonetheless subjective in the Nagel sense that they are fully knowable or understandable only from a limited range of perspectives, associated with the ability to have particular forms of phenomenal experience. The existence of such facts falls quite naturally out of the nonreductivist view of understanding and representation as contextually situated activities conditioned by the causal and interactive structure of the cognitive engagement between understander and understood. It is hardly surprising that the higher level forms of engagement in self-understanding systems involve modes of access and interaction, as well as the contents supported and determined by them, that

212

robert van gulick

cannot be replicated by any intentionally equivalent representational structures built solely with the resources of microphysical theory or any other third person system of theoretical representation. That is just what a nonreductivist would expect given his view of theories and models as cognitive tools to be applied in particular contexts of pragmatic engagement. Given their differing perspectives on the nature of theories and understanding, a priorist and nonreductivist physicalists part company with the KA at rather different points. Despite both endorsing physicalism in one version or another, they differ in where they respectively agree and disagree with the proponent of the Knowledge Argument. Both the proponent and the a priorist critic agree that physicalism requires the a priori deducibility of all true propositions about conscious mentality from the microphysical facts. And thus they also agree on the following conditional: if physicalism is true then pre-release Mary can know all the facts about seeing red (as well as its transpositive: if she cannot know them all then physicalism is false). They disagree only about the state of Mary’s knowledge. The a priorist critic, such as current Jackson, argues for Mary’s being able to infer her way to complete prior propositional knowledge and uses the RTS to rebut the Phenomenal Properties Objection that aims to show otherwise. The proponent of the KA, such as Jackson pre-1998, accepts the same conditional but denies its consequent and thus claims to refute its physicalist antecedent. Mary cannot know or a priori infer all the facts, and thus physicalism is false. By contrast, the nonreductivist critic rejects the conditional and the deductive thesis on which it relies. The nonreductivist denies that ontological reduction via physical realization carries any epistemic implications about our ability in practice or in principle to deduce every physical fact from a physical theory representation of the base level microphysical facts. Thus the nonreductivist denies that physicalism is refuted or even weakened by the incompleteness of Mary’s prior propositional knowledge or by her increasing or enhancing her propositional knowledge and understanding when she first experiences red. Indeed as noted above, some nonreductivists (I among them) take the existence of physically realized subjective facts to follow naturally from nonreductive physicalism and the pragmatic view of mind and understanding it embodies. Mary’s epistemic gain might be taken to confirm rather than refute any such version of physicalism. The theory itself predicts the existence of limits on what pre-release Mary can know and understand about physically realized experiences. Only after having them herself is she able to engage the relevant aspects of reality in the cognitively required way. Thus a nonreductivist who believes in physically realized subjective facts would be most likely to challenge

jackson’s change of mind

213

the KA at Q0 or Q5, since these critical options appeal in their complementary ways to just such facts. Other nonreductivists might reject the deduction thesis on general grounds, but be less sure about the existence of physically realized subjective facts. They might thus be more inclined to locate their specific disagreement at Q3 or Q4. Nonreductivists of all types agree among themselves in rejecting the deducibility thesis and the epistemic conditional that it implies. Their differences mainly concern the varying respects in which they believe Mary could gain new propositional knowledge without refuting physicalism.

6. Summary Let us take stock of where matters stand on the relation between the RTS and the KA. As we have seen, there are good reasons to think the RTS is neither necessary nor sufficient for refuting the KA. If one is a physicalist and a priorist, as Jackson currently is, then one might use the RTS to block the Phenomenal Properties Objection. But one must still rely upon a strong a priorist assumption about the deducibility of all the representational intentional facts to make one’s case that Mary lacks no prior propositional knowledge. The really ‘‘heavy lifting’’ seems to be done more by the a priorist view than by the RTS. Indeed, as we noted above, if one were strongly enough committed to a priorism, one might be able to forgo the RTS altogether and argue that even if experiences have phenomenal properties they too are deducible in principle from the microphysical facts. If one is not an a priorist and inclines more toward nonreductive physicalism, the RTS will likely play an even smaller role, if any, in one’s critical response to the KA. As has been said, the nonreductivist has several options. If he believes in the existence of physically realized subjective facts, he can draw the line at Q0 or Q5. He can either reject the initial stipulations of the Mary case as inconsistent at the outset, or hold that Mary comes to grasp such subjective facts when she first experiences red and acquires the requisite abilities to articulate the structure of logical space and thus to understand her location in it. Alternatively, the nonreductivist could concede that Mary gains new propositional knowledge, but only by coming to know a previously known proposition in a new way or to know a new fine grained proposition. According to the nonreductivist, neither of the two would pose any threat to physicalism, and neither seems to turn in a special way on the truth of the RTS.

214

robert van gulick

My aim in this paper was not to evaluate the plausibility or truth of the RTS, but only Jackson’s claim that it is essential to defeating the Knowledge Argument and central to seeing where it goes wrong. From Jackson’s perspective the RTS is clearly both, but other physicalists, especially those not committed to his strongly a priorist view will likely see things differently. There appear to be a variety of alternative plausible ways in which physicalists might block the Knowledge Argument without needing to accept the RTS. And nonreductive physicalists are more likely to locate the central flaw of the KA in its a priorist view of what physicalism requires than in its rejection of the RTS, or its belief in phenomenal properties. From the nonreductivist perspective the crucial illusion does not concern phenomenal properties but a priori deducibility. Where one sees illusion depends in part upon the lens through which one looks. Let me end by trying to pull the diversity of critical options into a few coherent overall positions, which may serve to better locate Jackson’s critical stance relative to others. If one is a physicalist of any sort, one must resist the conclusion of the Knowledge Argument, but which means of blocking it seems overall most plausible or attractive will depend largely on how well it fits with the particular version of physicalism one holds. There are different ‘‘package deals’’ from which the physicalist might choose, including at least the following two. #1 The Jackson Package From Jackson’s perspective, physicalism carries an unavoidable commitment to a priori deducibility. He rejects any view of physical necessitation without a priori entailment as failing to adequately explain how the mental could exist as a real part of the physical world and why one should not simply take an eliminativist stance and deny its existence. Jackson discerns no basis on which a physicalist should regard anything as real unless its existence and nature is a priori entailed by the physical facts and deducible from them. Thus no refutation of the KA is acceptable unless it allows Mary at least in principle to deduce every true proposition about conscious experience from the microphysical facts. Jackson finds the representational theory of sensory experience independently attractive, and he offers it to the physicalist as a means to reinforce the deducibility thesis or at least as a shield to defend it from objections that appeal to phenomenal properties such as phred-ness. According to Jackson and other representationalists, sensory experience has fewer mental properties than commonly believed and thus fewer properties that need to be accommodated in the physical scheme or deduced a priori from the microphysical facts. Positive beliefs or intuitions to the contrary about the existence of phenomenal properties are explained away as the result of cognitive illusion.

jackson’s change of mind

215

Physicalist refutation of the Knowledge Argument that includes: Truth of physicalism in its a priorist version. Truth of deducibility thesis: everything real deducible from the micro physical. Truth of representationalism, i.e. the representational theory of sensory experience. Non-existence of phenomenal properties. Phenomenal intuitions are illusory. Mary has complete propositional knowledge before her release. Ability Reply: Mary gains only new know-how.

Figure 8.3. The Jackson Package

Moreover, the difference between representing a given content in thought and phenomenally representing it in sensory experience is explained as partly a difference in the total intentional content of the relevant states, and partly as a matter deploying a different system of representation that gives one some new self-directed know-how but no gain in propositional knowledge about the nature or properties of one’s own mental states, not even about those involved in the use of the relevant representational system. Jackson adopts the Ability Reply and explains Mary’s change in epistemic state as solely a matter of gaining such new know-how; again any intuitions to the contrary are rejected as illusory. #2 The Nonreductivist Package Other physicalists see the options differently. Nonreductive physicalists in particular find principled grounds for resisting the deducibility thesis, and they deny that an ontological commitment to physical realization entails an epistemological commitment to the capacity of physical theory to serve as a universal means for knowing and understanding everything real. The nonreductivist regards representation, understanding and knowledge as all having an essentially pragmatic aspect. Thus he views them as contextual matters that are conditioned in part by the causal structure of the cognitive agent, as well as by that of the object of understanding and their engagement. Given the dependence of content on such pragmatic contextual factors, the nonreductivist does not expect the relations between disparate systems of representation to typically be matters of strict equivalence, translatability or deducibility as more logical empiricist or a priorist models would require. The nonreductivist regards know-how and propositional knowledge as less distinct and more interdependent than do supporters of the Ability Reply. The acquisition of new cognitive skills and new modes of interaction and

216

robert van gulick

engagement can enhance one’s propositional knowledge and give one the ability to represent and understand propositions one could not previously grasp. Indeed, according to the nonreductivist that is just what happens when Mary first experiences red and as a self-understanding physical system first reciprocally engages those aspects of her own dynamic organization that realize her having such experiences. Thus according to the nonreductivist, physicalism does not conflict with the belief that Mary’s newly acquired skills and modes of engagement give her the ability to increase her propositional knowledge, either by understanding old propositions in new ways or even by understanding new propositions about experience. Indeed, the nonreductivist may be able to concede even that Mary learns new coarse grained propositions without contradicting his physicalism. If nonreductivist physicalism allows, or indeed entails, the existence of physically realized subjective facts and propositions, then prior to release Mary could not even in principle understand or know such facts about red experience. Thus the nonreductivist might either deny that Mary could in principle know all the physically realized facts before her release, or he might argue that her learning new propositions about physically realized subjective facts pose no threat to physicalism. They are alternative ways of making the same point. As to the RTS and the existence of phenomenal properties, the nonreductivist is largely free to decide that issue on its individual merits. He need not adopt the RTS or deny the existence of phenomenal properties in order to block the KA. If he finds the specific arguments for representationalism compelling, he can combine it consistently with the rest of his view. However, if he is reluctant to dismiss our positive intuitions about the existence of phenomenal properties as mere illusions, he can try to find a place for them in his model without undercutting his response to the Knowledge Argument. Thus the overall package might be summarized in Figure 8.4. These two packages have their respective strengths, and each will be attractive to some subset of physicalists as a means of refuting the KA. Jackson chooses Package #1 and claims it is the only real option. I prefer #2 and so disagree with his uniqueness claim. Nor is it likely that these two exhaust the range of plausible overall combinations. A popular American TV game show of a few years back offered contestants a choice of prize packages behind one of three doors. The key feature of ‘‘Let’s Make a Deal’’ was that players had to decide whether to trade away their prize already in hand for what might lie behind a yet unopened door. Perhaps an even better package awaits the physicalist behind door #3, but then perhaps he might be wise to stay with

jackson’s change of mind 217 #2 Physicalist refutation of the Knowledge Argument that includes: Truth of ontological physicalism interpreted as a matter of realization: everything real is physically realized. Denial of theoretically/representationally reductive physicalism. (Physical theory is not a universal cognitive means for understanding and representing everything real.) Denial of deducibility thesis. (Not everything real is deducible from the microphysical). Pragmatic/contextual account of representation and understanding. Interdependence of know-how and propositional knowledge. Mary's new skills and modes of engagement enhance her propositional knowledge. Existence of physically realized subjective facts. Intuition that Mary gains new propositional knowledge preserved as consistent with physicalism. Neutral on the RTS and the existence of phenomenal properties.

Figure 8.4. The Nonreductive Package

that available in Package #1 or #2. I leave it up to you to make the best deal that you can. Syracuse University

References Chalmers, D. 1996. The Conscious Mind (New York: Oxford University Press). Churchland, P. 1985. ‘Reduction, qualia, and the direct introspection of brain states’, Journal of Philosophy 82: 8–28. Hardin, C. 1988. Color for Philosophers (Indianapolis: Hackett Publishers). Horgan, T. 1984. ‘Jackson on physical information’, Philosophical Quarterly 34: 147–83. Jackson, F. 1982. ‘Epiphenomenal qualia’, American Philosophical Quarterly 32: 127–36. 1986. ‘What Mary did not know’, Journal of Philosophy 83: 291–5. 1998. ‘Postscript on qualia’. In his Mind, Method and Conditionals (London: Routledge). Lewis, D. 1988. ‘What experience teaches’, Proceedings of the Russellian Society, Sydney, Australia. Reprinted in Lycan 1990a. Loar, B. 1990. ‘Phenomenal states’, Philosophical Perspectives, vol 4: Action Theory and Philosophy of Mind (Atascadero: Ridgeview Publishing). Lycan, W. 1990a. Mind and Cognition: A Reader (Oxford: Blackwell).

218

robert van gulick

Lycan, W. 1990b. ‘What is the subjectivity of the mental?’, Philosophical Perspectives, vol 4: Action Theory and Philosophy of Mind (Atascadero: Ridgeview Publishing). Nagel, T. 1974. ‘What is it like to be a bat?’, Philosophical Review 83, 435–50. Nemirow, L. 1980. ‘Review of T. Nagel Mortal Questions’, Philosophical Review 89: 475–76. 1990. ‘Physicalism and the cognitive role of acquaintance’. In Lycan 1990a. Tye, M. 1986. ‘The subjectivity of experience’, Mind 95: 1–17. Van Gulick, Robert 1985. ‘Physicalism and the subjectivity of the mental,’ Philosophical Topics, 51–70. 1992. ‘Nonreductive materialism and intertheoretical constraint’. In A. Beckermann, H. Flohr and J. Kim (eds.), Reduction and Emergence (Berlin: DeGruyter). 1993. ‘Understanding the phenomenal mind: Are we all just armadillos?’. In M. Davies and G. Humphrey (eds.), Consciousness (Oxford: Blackwell). 2003. ‘Maps, gaps and traps’. In Q. Smith and A. Jokic (eds.), New Essays on Consciousness (Oxford: Oxford University Press).

PA RT I I I

Ethics

This page intentionally left blank

9 Analytical Moral Functionalism Meets Moral Twin Earth TERRY HORGAN AND MARK TIMMONS

In chapters 4 and 5 of his 1998 book From Metaphysics to Ethics: A Defence of Conceptual Analysis, Frank Jackson propounds and defends a form of moral realism that he calls both ‘moral functionalism’ and ‘analytical descriptivism’. Here we argue that this metaethical position, which we will henceforth call ‘analytical moral functionalism’, is untenable. We do so by applying a generic thought-experimental deconstructive recipe that we have used before against other views that posit moral properties and identify them with certain natural properties, a recipe that we believe is applicable to virtually any metaphysically naturalist version of moral realism. The recipe deploys a scenario we call Moral Twin Earth.¹

1. Jackson’s analytical moral functionalism2 We begin by briefly summarizing Jackson’s theory. He proposes to construe moral terms like ‘goodness’ and ‘rightness’ in much the same way that the mentalistic terms of folk psychology are construed by the first-order version of the position in philosophy of mind called analytical functionalism. According to first-order analytical functionalism about the mental—as articulated, for instance, by D. M. Armstrong (1968, 1970) and David Lewis (1966, 1972, 1980)—mental-state terms are functionally definable via the principles of common-sense psychology, and these terms refer not to second-order functional properties but rather to certain neurophysical properties that fill the roles specified by the functional definitions. ¹ See Horgan and Timmons (1991, 1992a, 1992b, 1996a, 1996b, 2000a) and Timmons (1999, ch. 2). ² All citations to Jackson, in this section and throughout the paper, refer to Jackson (1998).

222 terry horgan and mark timmons Jackson’s approach involves four central ideas. First, he posits what he calls ‘folk morality’, something whose (partly implicit) mastery he takes to be required for competence in the use of moral concepts: In the case of the mind, we have a network of interconnected and interdefinable concepts that get their identity through their place in the network ... The network itself is the theory known as folk psychology, a theory we have a partly tacit and partly explicit grasp of ... In the case of ethics, we have folk morality: the network of moral opinions, intuitions, principles, and concepts whose mastery is part and parcel of having a sense of what is right and wrong, and of being able to engage in meaningful debate about what ought to be done ... Moral functionalism, then, is the view that the meanings of the moral terms are given by their place in this network ... (130)

Second, he claims that interpersonal commonality of meaning for moral terms requires that all parties are using moral terms in a way that reflects a mastery (partly implicit) of one and the same folk morality—which he calls mature folk morality. He also assumes that this prerequisite is satisfied in the case of humans, while acknowledging that if it is not then his account would need to be relativized: I have spoken as if there will be, at the end of the day, some sort of convergence of moral opinion in the sense that mature folk morality will be a single network ... accepted by the community as a whole. Indeed, I take it that it is part of current folk morality that convergence will or would occur ... But this may turn out to be, as a matter of fact, false. Indeed, some hold that we know enough now about moral disagreement to know that convergence will (would) not occur. In this case, there will not be a single mature folk morality but rather different mature folk moralities for different groups in the community; and to the extent that they differ, the adherents of the different mature folk moralities will mean something different by the moral vocabulary ... I set this complication aside in what follows. I will assume what I hope and believe is the truth of the matter, namely that there will (would) be convergence. But if this is a mistake, what I say in what follows should be read as having implicit relativization clauses built into it. (137)

Third, he claims that moral terms have conceptual analyses that result by applying the method of defining theoretical terms developed by David Lewis (1970). The idea is first to characterize a system of properties that together conform to the principles of the theory—in this case, the principles of mature folk morality—and then to characterize each member of the system by its place within the whole. Jackson writes: Let M be mature folk morality. Imagine it written out as a long conjunction with the moral predicates written in property name style. For example, ‘Killing someone is typically wrong’ becomes ‘Killing typically has the property of being wrong’. Replace

moral functionalism meets moral twin earth

223

each distinct moral property term by a distinct variable to give M(x1 , x2 , ... ). Then ‘(∃x1 ) ... M(x1 ...)’ is the Ramsey sentence of M, and (∃x1 ) ... (y1 ) ... (M(y1 ...) iff x1 = y1 & x2 = y2 ...) is the modified Ramsey sentence of M which says that there is a unique realization of M. If moral functionalism is true, M and the modified Ramsey sentence of M say the same thing. For that is what holding that the ethical concepts are fixed by their place in the network comes to. Fairness is what fills the fairness role; rightness is what fills the rightness role; and so on. We can now say what it is for some action A to be, say, right, as follows: (R) A is right iff (∃x1 ) ... (A has xr & (y1 ) ... (M(y1 , ... iff x1 = y1 & ...))) where ‘xr ’ replaced ‘being right’ in M. We now have our account of when A is right: it is right just in case it has the property that plays the rightness role as specified by the right hand side of (R) ... (140–1)

Fourth, he maintains that certain first-order natural properties fill the respective roles that define the respective moral terms, and hence that moral terms denote these role-filling natural properties. (He remains neutral, and he says that mature folk morality is itself neutral, as to whether moral terms denote such role-filling natural properties rigidly or non-rigidly.) Here the key line of thought is this: by virtue of the a priori supervenience of the ethical on the descriptive, each moral term is necessarily coextensive with some natural property, namely, the disjunction of all natural supervenience-base properties for the moral term. That property is the role-filler, and hence is the referent-property of the moral term. (Jackson holds that necessarily-coextensive properties are identical, but he also offers independent arguments, including parsimony considerations, in support of the proposed property-identities; see pp. 125–8.)

2. Trouble: Moral Twin Earth3 Competent wielders of language and concepts have substantial intuitive mastery of the semantic norms governing the terms of the language they employ and the concepts those terms express—just as they have substantial intuitive mastery of the syntactic norms governing their language. Implicit mastery of the semantic workings of the term ‘water’ and the concept it expresses, for instance, ³ This section is largely adapted from section IV of Horgan and Timmons (1992a), with minor modifications to make the discussion directly applicable to Jackson’s position.

224

terry horgan and mark timmons

presumably is reflected in people’s strong intuitions about Putnam’s Twin Earth scenario: e.g., the intuition that Twin Earthers do not mean by their Twin English term ‘water’ what English speakers on Earth mean by ‘water’, and the intuition that the Twin English term ‘water’ is not translatable by the orthographically identical term of Earth English. Such intuitions constitute strong (though of course defeasible) empirical evidence for the hypothesis that ‘water’ rigidly designates the specific stuff called water here on Earth, namely, H2 O. For this hypothesis nicely provides a plausible non-debunking explanation of the intuitions themselves—i.e., an explanation that treats the intuitions as veridical, and thus as reflective of people’s semantic competence with the concept water. Presumably, competent wielders of moral terms and concepts have a comparable intuitive mastery of the semantic workings of ‘rightness’, ‘fairness’, and moral terms and concepts more generally. So if indeed these terms have conceptual analyses of the kind Jackson claims they do, then it should be possible to construct a suitable Twin Earth scenario with these features: (i) reflection on this scenario generates intuitive judgments that are comparable to those concerning Putnam’s original scenario, and (ii) a plausible non-debunking explanation for these judgments is provided by analytical moral functionalism (henceforth, AMF). Conversely, if the appropriate Twin Earth scenario does not possess feature (i)—i.e., if the semantic intuitions of competent speakers turn out not to be what they should be if AMF is true—then this will mean that AMF is probably false. We say ‘probably’ false because the inference to AMF’s falsity would be nondemonstrative, an inference to the best explanation. Semantic intuitions about Twin Earth scenarios are empirical evidence about matters of semantics (just as syntactic intuitions about grammaticality are empirical evidence about matters of syntax).⁴ We will now mount an argument against AMF by arguing that things go the latter way—i.e., that one’s intuitive judgments concerning a suitable Twin Earth scenario go contrary to AMF. For present purposes let us provisionally suppose that Jackson is right in his assumption that there is a single mature folk morality M to which all Earthly persons would converge under suitably ideal ⁴ For more on philosophical appeals to intuition as being, in effect, semantic-competence arguments that provide empirical support for philosophical hypotheses about the semantics of concepts and terms (even though philosophers often do not appreciate this fact about their own intuition-based reasoning), see Horgan (1993) and Graham and Horgan (1994). For treatment of such arguments as conforming to part, but not all, of the traditional conception of the a priori, see Henderson and Horgan (2000, 2001), where these arguments are dubbed ‘‘low-grade a priori’’ because they rest on data that is armchair-accessible, such as one’s own semantic intuitions.

moral functionalism meets moral twin earth

225

reflection. (We do not for a moment believe this supposition, and we will have more to say about it below. But applying our deconstructive recipe to any given version of naturalist moral realism involves granting any such optimistic assumption figuring in the view under consideration, and then arguing that the view would be mistaken even if the assumption were correct.) What is wanted, then, is a Twin Earth where things are as similar to earth as possible, consistent with the stipulation that there is some different mature folk morality M∗ , distinct from M, to which all Twin Earthly persons would converge under suitably ideal reflection. So let us begin by supposing that, as a matter of empirical fact, Earthly mature folk morality is consequentialist in nature, and is best systematized by some specific consequentialist normative theory; call this theory Tc . Let us further suppose that there is some reliable method of moral inquiry which, if properly and thoroughly employed, would lead Earth folks to discover this fact about their uses of moral terms and concepts. Now consider Moral Twin Earth, which, as you might expect, is very much like good old Earth: same geography and natural surroundings, with people who live in Twin Australia and by and large speak Twin English, etc. Of particular importance here is that Moral Twin Earthers have a vocabulary that works very much like human moral vocabulary: they use the terms ‘good’ and ‘bad’, ‘right’ and ‘wrong’ to evaluate actions, persons, institutions, and so forth (at least those who speak Twin English use these terms, whereas those who speak some other Twin language use terms orthographically identical to the terms for good, etc., in the corresponding Earth language). In fact, were a group of explorers from Earth ever to visit Moral Twin Earth they would be strongly inclined to translate the Moral Twin Earth terms ‘good’, ‘right’, and the rest as identical to their own orthographically identical English terms. After all, the uses of these terms on Moral Twin Earth bear all the formal marks that are usually taken to characterize moral vocabulary and moral practice. In particular, the terms are used to reason about considerations bearing on the well-being of persons on Moral Twin Earth; Moral Twin Earth people are normally disposed to act in certain ways corresponding to judgments about what is ‘good’ and ‘right’; they normally take considerations about what is ‘good’ and ‘right’ to be especially important, even of overriding importance in most cases, in deciding what to do; and so on. Let us suppose that investigation into Twin English twin-moral discourse and associated practice reveals that Twin Earthers all would converge, under ideal reflective inquiry, to a mature folk morality that is nonconsequentialist, and thus is distinct from the consequentialist mature folk morality to which (we are supposing) Earthers would all converge. Suppose too that Twin Earthly

226

terry horgan and mark timmons

mature folk morality is best systematized by some specific deontological normative theory; call this Td . The theory Td , although importantly different from Tc , nonetheless is similar enough to Tc to account for the fact that twin-moral discourse operates in Twin Earth society and culture in much the same manner that moral discourse operates on Earth. (We have already noted that if explorers from Earth ever visit Moral Twin Earth, they will be inclined, at least initially, to construe persons on Moral Twin Earth as having beliefs about good and right, and to translate Twin English uses of these terms into orthographically identical English terms.) The differences in the respective mature folk moralities of Earthers and Twin Earthers, we may suppose, are due at least in part to certain species-wide differences in psychological temperament that distinguish Earthers from Twin Earthers. (For instance, perhaps Twin Earthers tend to experience the sentiment of guilt more readily and more intensively, and tend to experience sympathy less readily and less intensively, than do Earthers.⁵) Given all these assumptions and stipulations about Earth and Moral Twin Earth, what is the appropriate way to describe the differences between moral and twin-moral uses of ‘good’, ‘right’, ‘fair’, etc.? Two hermeneutic options are available. On one hand, one could say that the differences are analogous to those between Earth and Twin Earth in Putnam’s original example, to wit: the moral terms used by Earthers designate the unique natural properties that respectively satisfy the respective Lewis-style conceptual analyses of those terms obtainable from theory Tc , whereas the twin-moral terms used by Twin Earthers designate distinct unique natural properties that respectively satisfy the respective conceptual analyses obtainable from Td ; hence, because corresponding moral and twin-moral terms have different, incompatible, conceptual analyses, moral and twin-moral terms differ in meaning, and are not ⁵ In order to forestall any attempt to parlay this postulated difference into a basis for resisting the argument we are about to give, let us further stipulate (i) that the difference is merely a matter of initial psychological tendencies within Twin Earthers and Earthers respectively, (ii) that these tendencies are psychologically malleable in both groups, and thus (iii) that both groups are plastic with respect to how their moral sensibilities get molded, rather than being ‘‘hard-wired’’. Thus, for both groups it is true that if certain cultural developments were to transpire, then the members of the group would develop an altered moral sensibility and would sustain this change via alterations in moral education. For instance, if someone like Peter Singer were to exert widespread influence on Moral Twin Earth, then the Moral Twin Earthers would develop and sustain a utilitarian moral sensibility. Or, if the concept of sin were to become ubiquitously influential on Earth, then Earthers would develop and sustain a deontological moral sensibility. (We ourselves would argue not only that human moral psychology is indeed malleable in this way, but also that such differences in moral sensibilities and in associated modes of moral education are abundantly present right here on Earth. But remember that we are currently granting, for the sake of argument, Jackson’s optimistic assumption that there is a single mature folk morality to which all Earthers would converge under ideal reflection.)

moral functionalism meets moral twin earth

227

intertranslatable. On the other hand, one could say instead that moral and twin-moral terms do not differ in meaning or reference, and hence that any apparent moral disagreements that might arise between Earthers and Twin Earthers would be genuine disagreements—i.e., disagreements in moral belief and in normative moral theory, rather than differences in meaning. We submit that by far the more natural and plausible mode of description, when one considers the Moral Twin Earth scenario, is the second. Reflection on the scenario just does not generate hermeneutical pressure to construe Moral Twin Earth uses of ‘good’ and ‘right’ as not translatable by the orthographically identical terms of English. But if AMF were true, and moral terms had the kinds of conceptual analyses that Jackson claims they do, then reflection on this scenario ought to generate intuitions analogous to those generated in Putnam’s original Twin Earth scenario. That is, it should seem intuitively natural to say that we have here a difference in meaning and that the twin-moral terms of Twin English are not translatable by English moral terms. But when it comes to characterizing the differences between Earthers and Moral Twin Earthers on this matter, by far the more natural-seeming thing to say is that the differences involve belief and theory, not meaning. One’s intuitions work the same way if, instead of considering the Moral Twin Earth scenario from the outside looking in, one considers how things would strike Earthers and Twin Earthers who have encountered each other. Suppose that Earthers visit Twin Earth (or vice versa), and both groups come to realize that members of their respective species would converge to different mature folk moralities that conform respectively to the consequentialist theory Tc and the deontological theory Td . If AMF were true, then recognition of these differences ought to result in its seeming rather silly, to members of each group, to engage in inter-group debate about goodness—about whether it conforms to normative theory Tc or to Td . (If, in Putnam’s original scenario, the two groups learn that they have respectively been using ‘water’ to refer to two different physical kind-properties, it would be silly for them to think they have differing views about the real nature of water.) But such inter-group debate would surely strike both groups not as silly but as quite appropriate, because they would regard one another as differing in moral beliefs and moral theory, not in meaning. Since semantic norms are tapped by human linguistic and conceptual competence, and since the relevant competence is presumably reflected in one’s intuitive judgments concerning Twin Earth scenarios, these intuitions about Moral Twin Earth constitute strong empirical evidence against AMF. Barring overwhelmingly strong reasons to think otherwise, the best explanation of the intuitive judgments is a non-debunking explanation that treats them as

228

terry horgan and mark timmons

the products of semantic competence, and hence as veridical. And if indeed they are veridical, then analytical moral functionalism is false.

3. An unappealing fallback6 Briefly stated, the problem with Jackson’s position is that it is guilty of chauvinistic conceptual relativism—that is, it is committed to the claiming that actual or possible agents who have a mature folk morality different from that of humans would not possess the concepts of goodness, rightness, etc. at all. This is objectionably human-centered, because it chauvinistically builds the folk morality supposedly shared by all of humankind directly into moral concepts themselves. In fact, Moral Twin Earthers would share moral concepts with Earthers, despite having somewhat different moral beliefs by virtue of their different mature folk morality. (By contrast, the claim that ‘water’ rigidly designates H2 O is a non-chauvinistic form of conceptual relativism, because it does not wrongly exclude Putnam’s non-human Twin Earthers from possessing a concept that they really do possess. On the contrary, the human concept water does indeed rigidly refer to the particular kind of clear liquid, whatever kind this is, that fills the lakes and streams in the local Earthly environment occupied by us humans. Likewise, the Twin Earthers’ corresponding concept is indeed different, since it rigidly refers instead to the kind of clear liquid that fills the lakes and streams in their local environment—namely, XYZ.) The view in philosophy of mind often called psychofunctionalism is chauvinistic in a way similar to Jackson’s metaethical position. According to psychofunctionalism, common-sense mentalistic concepts are functionally definable via the ideally complete and correct empirical psychological theory that is true of human beings—which we will call mature empirical psychology. But consider a race of possible creatures—say, Martians—who undergo internal states that (i) conform to all the principles of folk psychology just as much as the internal states of humans do, but (ii) conform to a somewhat different mature empirical psychology than do the internal states of humans. The trouble with psychofunctionalism is that it chauvinistically denies that such Martians have beliefs, desires, and other such mental states; it denies this despite the fact that Martians undergo internal states that jointly conform to the principles of folk psychology just as well as do the relevant internal states of humans themselves. ⁶ Parts of this section are adapted, with minor modifications, from section 5 of Horgan and Timmons (1996a) and from section 2 of Horgan and Timmons (2000a).

moral functionalism meets moral twin earth

229

Analytical functionalism in philosophy of mind, on the other hand, avoids this form of conceptual chauvinism about mental states exhibited by psychofunctionalism. Analytical functionalism claims that the mentalistic concepts of folk psychology should be construed as being functionally definable not by the specific mature empirical psychology that happens to be true of humans, but rather by folk psychology itself. Analytical functionalism thus would count the recently-mentioned Martians as possessing beliefs, desires, and other folk-psychological mental states, by virtue of undergoing first-order states that collectively conform to the generalizations of folk psychology.⁷ Analytical functionalism avoids the mistake of chauvinistically building into folk-psychological concepts the specific empirical psychological principles that happen to be true of humans. In light of this comparative advantage of analytical functionalism vis-`avis psychofunctionalism in the philosophy of mind, a fallback metaethical position that Jackson might consider would be to construe moral terms and concepts as functionally definable not by the specific mature folk morality supposedly possessed by all humankind, but rather by a set of uncontroversial moral platitudes: principles that are common currency within all possible mature folk moralities. This approach would avoid the problem of conceptual chauvinism. But the trouble with this idea is that the kinds of platitudinous, nontendentious, generalizations that clearly count as constitutive of people’s common-sense understanding of moral terms and concepts are simply not sufficient to pin down determinate referents for them. One can distinguish between formal and substantive moral platitudes. Formal moral platitudes would include those generalizations that link moral terms and concepts to one another and thus express definitional connections among them—for instance, ‘‘If an action is wrong, all things considered, then one ought not, all things considered, perform that action’’ and ‘‘If an action is morally permissible, all things considered, then it is not morally wrong, all things considered, to perform that action.’’ There are also those formal moral platitudes that represent features of the so-called ‘logic of moral discourse’, like the principle of universalizability: ‘‘If an action is right (or wrong) for one agent to perform in certain circumstances, then it is right (or wrong) for any similar agent in similar circumstances.’’ Substantive moral platitudes would be ones that apparently link moral concepts more directly to non-moral ones. Many philosophers ⁷ Assuming that the relevant first-order states are different in Martians than in humans (say, because Martians are composed of silicon rather than organic molecules), a first-order version of analytical functionalism will need to construe mental state-names as population-specific nonrigid designators in order to accommodate Martian mentality. David Lewis explicitly took this tack in Lewis (1980).

230

terry horgan and mark timmons

have claimed that there are such substantive platitudes, for instance, ‘‘Right actions are concerned to promote or sustain or contribute in some way to human flourishing’’ and ‘‘Right actions are expressive of equal concern and respect.’’ But formal considerations alone clearly are not enough to secure determinate referents for moral terms and concepts; in general, such a priori constraints are compatible with any of a great variety of normative moral theories that deliver incompatible verdicts about numerous specific moral issues. Nor will appeal to substantive moral platitudes (together with the formal ones) suffice to produce referential determinacy. Consider, for example, the recently mentioned generalizations involving flourishing and impartiality. The generic notions of flourishing and impartiality are quite vague, and thereby can be construed very differently within competing, incompatible, moral theories. Let us focus for a moment on the notion of impartiality—the idea that everyone is to be accorded equal respect. The problem of appealing to this notion is that it lacks sufficient determinacy to serve as an anchor to uniquely pin down referents for moral terms. James Griffin brings out the point nicely: Every moral theory has the notion of equal respect at its heart: regarding each person as, on some sense, on an equal footing with every other one. Different moral theories parlay this vague notion into different conceptions. Ideas such as the Ideal Observer or the Ideal Contractor specify the notion a little further, but then they too are very vague and allow quite different moral theories to be got out of them. And the moral theories are not simply derivations from these vague notions, because the notions are too vague to allow anything as tight as a derivation. Too vague, but not totally empty; although the moral theories that we end up with put content into all these notions, the notions themselves also do something toward shaping the theories. (Griffin 1986: 208)

Talk of flourishing is vague in just the same way. Moreover, the same will be true of other notions that, like equal respect and flourishing, might plausibly be understood as part of the very concept of moral thought and discourse. So it does not appear that moral platitudes alone can collectively generate Lewis-style functional definitions that fix determinate referents for terms like ‘goodness’, ‘rightness’, and ‘fairness’. As Michael Smith (1994) remarks, ‘‘These platitudes need not and should not be thought of fixing a unique content or substance for moral reasons all by themselves, rather they simply serve to tell us when we are in the ballpark of moral reasons, as opposed to the ballpark of non-moral reasons’’ (p. 184); for further substantiation of this claim, see Smith’s own discussion. One might try maintaining (i) that the indeterminacy here described involves only relatively few borderline hard cases about which competing moral theories

moral functionalism meets moral twin earth

231

would disagree, and (ii) that these cases can be comfortably relegated to the category ‘‘no moral fact of the matter.’’ But actually the resulting indeterminacy of truth-value would be massive, since it would extend to virtually any kind of case about which there is actual or potential moral disagreement. Acceptance of similar-looking, superficially substantive, platitudinous principles by people with differing moral values does very little to secure agreement about concrete cases, because the concepts that feature in such principles—equal respect, for instance—are apt to be applied to specific acts and situations so very differently by the different parties. (For a powerful elaboration and defense of this claim, see Snare 1980.) Thus, the fallback retreat that replaces the appeal to mature folk morality by an appeal to uncontroversial moral platitudes is not viable, because it immediately encounters—with a vengeance—the problem of radical indeterminacy of reference for moral terms and radical indeterminacy of truth-value for moral statements. Out of the frying pan of chauvinistic conceptual relativism, into the fire of radical moral indeterminacy! This is an instance of a generic dilemma posed by Moral Twin Earth for naturalist versions of moral realism: objectionable relativism on one hand, or objectionable indeterminacy on the other.

4. The full extent of Jackson’s chauvinistic conceptual relativism Earlier we granted, for the sake of argument, Jackson’s optimistic assumption that there is a single mature folk morality to which all humans would converge under ideal reflection. In adapting our generic recipe for cooking up specific Moral Twin Earth counterexamples against specific versions of naturalist moral realism, we argued that even if the optimistic assumption is true, Jackson’s analytic moral functionalism is untenable anyway; for, it chauvinistically builds into moral concepts the specific mature folk morality that supposedly would be converged upon by all Earthers, thereby wrongly entailing that Moral Twin Earthers lack the moral concepts that we Earthers possess. But it is entirely possible—we think likely—that different humans would converge to different mature folk moralities under ideal reflection. Prima facie, this is the most plausible explanation of the persistent, recalcitrant-looking, actual moral disagreements that commonly exist within humankind. As we pointed out in section 1, Jackson is unflinching in the face of this possibility: he

232

terry horgan and mark timmons

is prepared to extend his conceptual relativism to different human subgroups, if necessary. As he says (in a passage quoted earlier) about the possibility that convergence to a single folk morality would not occur, ‘‘In this case, there will be not a single mature folk morality but rather different mature folk moralities for different groups in the community; and to the extent that they differ, the adherents of the different mature folk moralities will mean something different by the moral vocabulary’’ (137). This is a very large bullet to bite. Jackson’s position entails that apparent moral disagreements among humans with deeply differing moral values are merely apparent: the different parties are expressing different concepts with their moral terms, are talking past one another rather than disagreeing, and often are both right given what they respectively mean by their moral terms. This conceptual-relativist construal of such apparent moral disagreements is wildly contrary to the common-sense, intuitive, way of understanding such situations. Common sense, and ordinary discursive practice, construe the appearances as veridical: the parties in such a dispute are employing common moral concepts, are using moral terms with common meaning, and are engaged in a deep and genuine disagreement in moral belief. Barring some overwhelmingly strong reason to think that this common-sense construal of such cases is mistaken, the enormous size of the bullet Jackson is biting constitutes a further strong consideration against his position. (This is an extension of, and a further strengthening of, the lesson of Moral Twin Earth; the point is that deep moral disagreements of the kind described in the Moral Twin Earth scenario very probably exist right here on Earth.)

5. Non-descriptivist cognitivism vs. analytical moral functionalism We do not claim to have conclusively refuted Jackson’s metaethical position; conclusive arguments are rare in philosophy. Philosophical theories, like scientific theories, should be evaluated in terms of their overall theoretical benefits and costs—and so should be evaluated comparatively, with an eye on benefits and costs of the competing philosophical theories on the conceptual landscape. Bullet-biting can be appropriate, if the advocates of competing theories all must bite even bigger bullets. Thus, how telling our negative arguments are against Jackson ultimately depends in part upon what available alternative metaethical theories exist, and upon the viability of those alternatives. In this section we locate our case against AMF within a wider dialectical setting,

moral functionalism meets moral twin earth

233

by briefly comparing it with the metaethical position we ourselves favor—a version of non-descriptivism. Non-descriptivists maintain that the overall declarative content of a moral judgment or a moral sentence is not descriptive content: such a judgment or sentence does not represent the world as being a certain way. Jackson, following the usual tradition in metaethics, uses the term ‘non-cognitivism’ for non-descriptivism. He says this about non-descriptivism, thus labeled: It is only under the assumption of cognitivism that ethics presents a location problem. If the non-cognitivists are right and ethical sentences do not represent things as being a certain way, there is no question of how to locate the way they represent things as being in relation to how accounts told in other terms—descriptive, physical, social, or whatever—represent things as being ... Although I cannot rule out non-cognitivism simply by noting that ethical sentences are meaningful and syntactically right for truth, I do think it is very much a ‘last resort’ position. (117)

Elsewhere (Timmons 1999 chapter 4, Horgan and Timmons 2000b, 2006a, 2006b, 2006c), we ourselves have articulated and defended a version of nondescriptivism that we maintain is not a ‘last resort’ position at all, but rather is more plausible than the various other metathical positions currently on offer. We call this view both ‘non-descriptivist cognitivism’ and ‘cognitivist expressivism’; here we adopt the former name. We now present an extremely truncated sketch of this position. Non-descriptivist cognitivism makes the following claims. (1) Contrary to non-cognitivist views like emotivism and prescriptivism, moral judgments are genuine beliefs, and utterances of moral sentences are genuine assertions. (The label ‘non-cognitivism’ fits emotivism and prescriptivism because these views deny that moral judgments are beliefs, and instead treat them as noncognitive states—for instance, as conative states of approval or disapproval.) (2) Contrary to descriptivist views, the overall declarative content of moral beliefs and assertions is not descriptive content: these beliefs and assertions do not represent things as being a certain way. (3) Beliefs with the most basic kinds of declarative content are psychological commitment-states with respect to a logically atomic descriptive content. (4) Such commitment-states are of two fundamental kinds: is-commitments and ought-commitments; each kind has both affirmative and negative versions. Thus, there are four basic kinds of belief: affirmative is-commitments (e.g., the belief that Bush is U.S. President), affirmative ought-commitments (e.g., the belief that it ought to be that Gore is U.S. President), negative is-commitments (e.g., the belief that it’s not the case that Gore is U.S. President), and negative ought-commitments (e.g., the belief that it ought not to be the case that Bush is U.S. President). (5) The constitutive features of ought-commitments include both (a) certain distinctive typical roles played

234

terry horgan and mark timmons

by these states in the psychological economy of the morally-judging agent, including distinctively motivational roles in the case of first-person oughtcommitments, and (b) certain distinctive typical phenomenological features exhibited by these states, including an experiential aspect of ‘‘reasons-based fittingness.’’ (6) The constitutive features of logically complex beliefs with component moral content—e.g., the belief that if Jones promised his wife to pick up the children from school then Jones ought to pick up the children from school—involve certain distinctive inferential roles played by such states in an agent’s psychological economy, roles in which various kinds of logically basic is-commitments and/or ought-commitments are ‘‘in the offing.’’ (7) A sincere assertion that p expresses the belief that p but does not describe that belief; thus, when the belief expressed is an ought-commitment, the declarative content of the corresponding assertion is not descriptive content. (8) Moral assertions typically have action-guiding roles in social intercourse that are similar to the typical action-guiding roles of moral beliefs (especially firstperson moral beliefs) in a morally judging agent’s own psychological economy. (9) A truth ascription to a moral belief or statement normally conforms to the Tarski T-schema, and thus normally constitutes a fusion of moral and semantic evaluation; such a morally engaged truth ascription is a meta-level expression of a moral belief (i.e., an ought-commitment), and hence is itself non-descriptive in overall declarative content (as is the first-order moral judgment or statement to which truth is ascribed).⁸ If indeed non-descriptivist cognitivism is a viable and independently plausible metaethical position, as we argue in the papers recently cited, then this fact considerably strengthens the force of Moral Twin Earth scenarios as evidence ⁸ Non-descriptivist cognitivism does not claim, however, that this morally engaged way of using the truth predicate is the only legitimate use. We ourselves maintain that the concept of truth is governed by implicit, contextually variable, semantic parameters, and that in the case of moral beliefs and assertions, any of three distinct uses of the truth predicate can be semantically sanctioned in a specific context: (1) a morally engaged disquotational use, expressive of one’s own moral beliefs, (2) a morally disengaged, nondisquotational, correspondence use, under which only beliefs and assertions whose overall declarative content is descriptive can be true or false, and (3) a morally disengaged, non-disquotational, overtly relativistic use, under which truth ascriptions get explicitly relativized to the moral standards of some specific person or group. When non-descriptivists assert that moral judgments and statements are neither true nor false (as we ourselves sometimes do), the truth predicate is being employed in manner (2) rather than manner (1). This is not inconsistent with using the truth predicate disquotationally vis-`a-vis moral claims, although one cannot use it both ways in one breath. Also, when the truth predicate is used in the third, overtly relativized, manner, typically one is simply making a descriptive remark about what the standards of some person or group imply about the moral status of a type of action or an act token. Again, using the truth predicate this way on one occasion is not inconsistent with using it disquotationally (or correspondence-wise) on another. For further related discussion, see Timmons (1999 chapter 4, especially pp. 149–52), Horgan (2001 section 5), Horgan and Timmons (2000b section VII.1; 2002 notes 10 and 19; and 2006c).

moral functionalism meets moral twin earth 235 against naturalist moral realism in its various versions, including Jackson’s version. Non-descriptivism is not a ‘last resort’ position that is worth avoiding even at the cost of embracing chauvinistic conceptual relativism. One final point. The availability of non-descriptivist cognitivism as a credible theoretical option also calls into question a dialectical move that Jackson makes in an effort to fend off a classic argument against naturalist moral realism, G. E. Moore’s famous ‘‘open question argument.’’ Jackson says: It may be objected that even when all the negotiation and critical reflection is over and we have arrived at mature folk morality, it would still make perfect sense to doubt that the right is what occupies the rightness role. But now I think that we analytical descriptivists are entitled to dig in our heels and insist that the idea that what fits the bill that well might still fail to be rightness, is nothing more than a hangover from the platonist conception that the meaning of the term ‘right’ is somehow a matter of its picking out, or being somehow mysteriously attached to, the form of the right. (151)

But of course the open question argument has long been employed by nondescriptivists, and not merely by non-naturalist moral realists, against naturalist moral realism. The argument has considerable intuitive force, and indeed is closely related in spirit to our own Moral Twin Earth argument; see Horgan and Timmons (1992a). Even granting the (dubious) assumption that all humans would converge upon a single mature folk morality, there is nothing especially platonistic about the claim that some possible agent who employs the same moral concepts that humans do—e.g., a Moral Twin Earther—could intelligibly doubt, of a natural property that the agent knows fits the rightness role that is functionally defined by the mature folk morality of humans, whether this natural property is identical to rightness. On the contrary, to insist that there could be no such moral agent is to be guilty of conceptual chauvinism. T. Horgan, University of Arizona M. Timmons, University of Memphis

References Armstrong, D. 1968. A Materialist Theory of the Mind (London: Routledge). 1970. ‘The Nature of the mind’. In C. Borst (ed.), The Mind/Brain Identity Theory (London: Macmillan). Griffin, J. 1986. Well-Being (Oxford: Clarendon Press). Graham, G. and Horgan, T. 1994. ‘Southern fundamentalism and the end of philosophy’, Philosophical Issues 5: 219–47.

236

terry horgan and mark timmons

Henderson, D. and Horgan, T. 2000. ‘What is a priori and what is it good for?’, Southern Journal of Philosophy 38 (Spindel Supplement): 51–86. 2001. ‘The a priori isn’t all that it is cracked up to be, but it is something’, Philosophical Topics 29 ( The Philosophy of Alvin Goldman): 219–50. Horgan, T. 1993. ‘The austere ideology of folk psychology’, Mind and Language 8: 282–97. 2001. ‘Contextual Semantics and Metaphysical Realism: Truth as Indirect Correspondence’. In M. Lynch (ed.), The Nature of Truth: Classic and Contemporary Perspectives (Cambridge, MA: MIT Press). Horgan, T. and Timmons, M. 1991. ‘New wave moral realism meets moral twin earth’, Journal of Philosophical Research 16: 447–65. Reprinted in J. Heil (ed.), Rationality, Morality, and Self-Interest (New York: Rowman & Littlefield, 1993). 1992a. ‘Trouble for New Wave Moral Semantics: The ‘‘Open Question Argument’’ Revived’, Philosophical Papers 21, 153–75. 1992b. ‘Troubles on moral Twin Earth: Moral queerness revived’, Synthese 92: 221–60. 1996a. ‘From moral realism to moral relativism in one easy step’, Critica 28: 3–39. 1996b. ‘Troubles for Michael Smith’s metaethical rationalism’, Philosophical Papers 25: 203–31. 2000a. ‘Copping out on moral Twin Earth’, Synthese 124: 139–52. 2000b. ‘Nondescriptivist cognitivism: Prolegomenon to a new metaethic’, Philosophical Papers 29: 121–53. 2002. ‘Conceptual Relativity and Metaphysical Realism’, Philosophical Issues (Realism and Relativism) 12: 74–96. 2006a. ‘Morality without moral facts’. In J. Drier (ed.), Contemporary Debates in Moral Theory (Oxford: Blackwell). 2006b. ‘Cognitivist expressivism’. In T. Horgan and M. Timmons (eds.), Metaethics after Moore (Oxford: Oxford University Press). 2006c. ‘Expressivism, yes! Relativism, no!’. In R. Shafer-Landau (ed.), Oxford Studies in Metaethics, vol 1. (Oxford: Clarendon Press). Jackson, F. 1998. From Metaphysics to Ethics: A Defence of Conceptual Analysis. (Oxford: Clarendon Press.) Lewis, D. 1966. ‘An argument for the identity theory’, Journal of Philosophy 63: 17–25. 1970. ‘How to define theoretical terms’, Journal of Philosophy 67: 427–46. 1972. ‘Psychophysical and theoretical identifications’, Australasian Journal of Philosophy 50: 249–58. 1980. ‘Mad pain and martian pain’. In N. Block (ed.), Readings in the Philosophy of Psychology, vol. 1 (Cambridge, MA: Harvard). Smith, M. 1994. The Moral Problem (Oxford: Blackwell). Snare, F. 1980. ‘The diversity of morals’, Mind 89: 353–69. Timmons, M. 1999. Morality without Foundations: A Defense of Ethical Contextualism (Oxford: Oxford University Press).

10 Consequentialism and the Nearest and Dearest Objection MICHAEL SMITH

1. Bloggs’s story Imagine that Bloggs is faced with a choice between giving a benefit to his child, or a slightly greater benefit to a complete stranger. The benefit is whatever the child or the stranger can buy for $100—Bloggs has $100 to give away—and it just so happens that the stranger would buy something from which he would gain a slightly greater benefit than would Bloggs’s child. Let’s stipulate that Bloggs believes this to be, and let’s stipulate, as well, that he believes that the consequences of his actions are otherwise identical. He chooses to give the benefit to his child. What do we learn about Bloggs from his choice? We learn that Bloggs cares more about his child than he does about complete strangers. Nor is anyone likely to be surprised by this, for it just goes to show that he is much like the rest of us. He gives preferential treatment to his nearest and dearest when he acts, those with whom he has a special relationship, much as we do. Now imagine that we ask Bloggs to justify his choice. Suppose he says that he did what was best, and that it was the best thing for him to do because the benefit went to his child. What do we learn about Bloggs from his attempt to provide a justification for giving the benefit to his child? Assuming that this is supposed to be the most basic value relevant to his choice we learn that, in addition to caring more about his child than he does about complete strangers, he thinks that, in so doing, he cares about what is of fundamental value. As he sees things, there is a distinctive form of value—‘relative’ value, as it has come to be called (Parfit 1984: 27)—realized in his giving the smaller benefit to his child. That is what is signalled by his citing the fact that the benefit accrues to his child: this is the special relationship just mentioned. Moreover Bloggs

238

michael smith

thinks, correctly, that this kind of value wouldn’t be realized if he instead gave the greater benefit to the stranger. We can represent Blogg’s universalized belief, the belief from which he derives his belief about the value of giving the benefit to his child, as follows (Smith 2003). R: ∀x (x’s child’s enjoying a benefit is goodx ) R entails that there is valueBloggs in Bloggs’s children enjoying benefits, and it also entails that there is valuesomeone else in someone else’s children’s enjoying benefits. But it leaves it open that there is no valueBloggs in someone else’s children’s enjoying benefits, and no valuesomeone else in Bloggs’s children enjoying benefits. Assuming that R is true, and assuming that the question relevant to the justification of choice is what has valuechooser , Bloggs’s choice would thus seem to be justified. Finally, imagine we challenge Bloggs’s justification. ‘You must surely agree’, we say, ‘that there is value of another form, namely neutral value, that attaches to the benefits enjoyed by your child and the stranger. For just imagine that you didn’t bear any special relationship to either your child or to the complete stranger. Wouldn’t you then assign at least some value to the benefit each receives, and wouldn’t the amount of value you assign covary with the level of the benefit? If so, then you will surely agree that there is more value of this form in giving the greater benefit to the complete stranger. So how can you think that the value that attaches to giving the benefit to your child justifies your giving the smaller benefit to him when you must also agree that there is more value in giving the greater benefit to the complete stranger?’ Bloggs’s response, let’s suppose, is to concede that there is value of another form—‘neutral’ value, as it has come to be called (Parfit 1984: 27)—realized in giving the larger benefit to the complete stranger, and that there is more value of this form realized in giving the larger benefit to the complete stranger than there is in giving the smaller benefit to his child. But he insists that the additional relative value realized in giving the smaller benefit to his child more than compensates for that loss of neutral value. Indeed, we can suppose that Bloggs thinks that if the loss in neutral value had been much larger, but the gain in relative value had remained much the same, then the additional relative value would not have compensated for the loss. His justification is thus specific to the circumstances in which he makes his choice. His claim is simply that, in these circumstances, the gain in relative value compensates for the loss of neutral value.

nearest and dearest objection

239

We can represent the content of the universalized belief from which Bloggs derives his beliefs about the neutral value of the benefits that accrue to both the stranger and his child as follows (Smith 2003). N: ∀x ∀y (Benefits enjoyed by the maximal aggregate composed inter alia of y are goodx ) N entails that benefits enjoyed by the aggregate composed of Bloggs’s child and the complete stranger are both goodBloggs , and it presumably entails that the greater the level of the benefit that that aggregate enjoys, the betterBloggs . In other words, unlike R, N doesn’t discriminate between a benefit enjoyed by Bloggs’s child and a benefit enjoyed by a complete stranger. Each of these benefits contribute to the level of benefit enjoyed by aggregates of which they are both a part, and so each has valueBloggs . In these terms we can now represent the way in which Bloggs weighs the values posited by R and N. Though, in virtue of the truth of N, Bloggs reasons that it would be betterBloggs to give the benefit to the complete stranger, he also reasons that there is, in virtue of the truth of R, additional goodnessBloggs in the benefit enjoyed by his child. In the circumstances, he thinks that this additional goodnessBloggs tips the balance, making it the case that there is more goodnessBloggs overall—that is, more goodnessBloggs in virtue of the truth of both R and N—in the benefit enjoyed by his child than there is in the benefit enjoyed by the complete stranger. As a result, he concludes that it is betterBloggs overall for him to give the benefit to his child. Finally, suppose we tell Bloggs that he really shouldn’t give the smaller benefit to his child; that he should give the larger benefit to the complete stranger none the less. What might Bloggs say in response? One obvious response would be that we are mistaken, as our view of what he should do seems to be based solely on the neutral values at stake, whereas there are relative values at stake too, values of which we are either ignorant or which we are willfully ignoring. In either case, he might say, we should think again. This would be to object to what we say he should do on the grounds that it is false. However suppose we convince him that we aren’t willfully ignoring relative values, and that it is he who is mistaken: there are no relative values, and hence only neutral values are at stake. Another perhaps less obvious response, but one which he might well give at this point, is that it would ruin his life, and perhaps his child’s as well, if he had to systematically ignore his special relationship with his child in making such choices: that consistently doing what we say he should do is inconsistent with his living a worthwhile life, and perhaps with his child’s living a worthwhile life too. This would not be to object to what

240

michael smith

we say he should do on the grounds that it is false. Indeed, it isn’t clear that this is any sort of objection to what we say he should do. Bloggs seems rather to be taking it for granted that he should do what we say he should do. He is simply putting on record the personal cost to him of his doing so, and the cost to his child.

2. Big ‘C’ consequentialism and the nearest and dearest objection Bloggs’s story brings into sharp relief the debate over big ‘C’ consequentialism: that’s big ‘C’, as opposed to small ‘c’, consequentialism. Big ‘C’ consequentialism makes two crucial claims. First, it makes a substantive claim about the nature of value. It says that all values are neutral (Parfit 1984). Second, it makes a conceptual claim about the nature of obligation. It says that facts about what we ought to do can be analysed in terms of facts about which of the various things that we can do will maximize value (we will return to the details of this analysis in a moment). Unsurprisingly, since big ‘C’ consequentialism entails that there are no relative values, it follows that, in Bloggs’s choice situation, there is no additional relative value realized in his giving the smaller benefit to his child to justify his doing that rather than giving the greater benefit to the complete stranger. Big ‘C’ consequentialism thus holds that Bloggs acts wrongly when he gives the $100 to his child. Having said that, however, we must immediately add that what Bloggs does isn’t at odds with small ‘c’ consequentialism. Small ‘c’ consequentialism agrees with big ‘C’ consequentialism’s conceptual claim about the nature of obligation. It simply denies big ‘C’ consequentialism’s substantive claim about the nature of value. Whereas big ‘C’ consequentialism holds that all values are neutral, small ‘c’ consequentialism is silent about the nature of values. Values might all be neutral, or they might all be relative, or some might be neutral while others are relative. To repeat, small ‘c’ consequentialism, unlike big ‘C’ consequentialism, takes no stand on this substantive issue about the form of the values. Bloggs implicitly committed himself to small ‘c’ consequentialism when he attempted to justify giving $100 to his child. For he argued that, when we take into account the value that derives from both N and R, we see that his giving the smaller benefit to his child is actually betterBloggs than giving the larger benefit to the complete stranger. What Bloggs appeals to is thus the following principle connecting indexed value with obligation:

nearest and dearest objection

241

(O) ∀x (x ought to φ iff φ-ing is bestx ) This is, in effect, the claim that the justification of choice is a matter of what has valuechooser alluded to above. Frank Jackson joins this debate over big ‘C’ consequentialism in his ‘Decision-Theoretic Consequentialism and the Nearest and Dearest Objection’ (Jackson 1991). Here is the opening passage of Jackson’s paper. Our lives are given shape, meaning and value by what we hold dear, by those persons and life projects to which we are especially committed. This implies that when we act we must give a special place to those persons (typically our family and friends) and those projects. But, according to consequentialism classically conceived, the rightness and wrongness of an action is determined by the action’s consequences considered impartially, without reference to the agent whose actions they are consequences of. It is the nature of any particular consequence that matters, not the identity of the agent responsible for the consequence. It seems then that consequentialism is in conflict with what makes life worth living. ( Jackson 1991: 461)

Jackson’s suggestion is thus that big ‘C’ consequentialism ‘is in conflict with what makes life worth living’ because it ‘would, given the way things more or less are, render the morally good life not worth living’. Unfortunately, Jackson doesn’t ever spell out what exactly this conflict is supposed to amount to. Indeed, the only other time he explicitly mentions the nature of the conflict is right at the end of the paper when he sums up. My concern ... has been to reply to the objection that consequentialism would, given the way things more or less are, render the morally good life not worth living. I take this to be the really disturbing aspect of the nearest and dearest objection. Consequentialists ... cannot live with the conflict with a life worth living, given the way things more or less are. That would be to invite the challenge that their conception of what ought to be done had lost touch with human morality. ( Jackson 1991: 482)

Before considering the details of his response, we therefore need to bring out what, as he sees things, the conflict between big ‘C’ consequentialism and living a life worth living is supposed to be. The reason for doing this is that, at first blush at any rate, Jackson looks to be making a version of the second response that we imagined Bloggs making above. But, much as we said in that case, this response doesn’t seem to constitute an objection to big ‘C’ consequentialism at all. It seems rather to be an observation about the personal cost of doing what the theory tells an agent to do. Indeed, it is hard to see how the mere possibility of a conflict between an agent’s living up to the obligations posited by a moral theory and that agent’s living a life worth living could constitute an objection to that moral theory. This is because every moral theory, or anyway every

242 michael smith plausible moral theory, will tell people that they should sometimes act in ways that will make their own lives not worth living. All that is required is that they have the bad luck to find themselves in the right (wrong?) kind of circumstances. Jackson in fact considers this reply to the objection with which he is concerned at the very beginning of his paper. One way to reply ... would be to break the implicit connection between acting morally and living a life worth living. Doing what is morally right or morally required is one thing; doing what makes life worth living is another. Hence, runs the reply, it is no refutation of a moral theory that doing as it enjoins would rob life of its shape and meaning. This is a chilling reply and I will say no more about it. ( Jackson 1991: 461)

But there is nothing especially chilling about a reply that points out an obvious consequence of every plausible moral theory. Imagine a case in which an agent finds himself faced with a choice between either submitting to torture and subsequent death, so making his life not worth living, or his bringing about ... and here substitute whatever your favourite moral theory deems to be an even worse outcome than this agent’s torture and subsequent death. It doesn’t matter whether your favourite moral theory says that what makes outcomes worse is their realizing less neutral value, or their realizing less relative value, or their realizing less of some weighted sum of value of both kinds. So long as the agent has to choose between making his own life not worth living and bringing about something that is even worse, by the theory’s own lights—and, to repeat, every plausible moral theory will say that there is a worse outcome than one agent’s living a life that is not worth living—then every plausible moral theory will tell that agent to make his own life not worth living. This will simply be an application of O. The upshot is that, if there is to be an objection to big ‘C’ consequentialism based on the fact that it is in conflict with an agent’s living a worthwhile life, then the objection will have to be that the conflict with big ‘C’ consequentialism is in some way much more direct and systematic than the conflict we have just noted. I take it that Jackson thinks that the conflict between big ‘C’ consequentialism and living a life worth living is indeed more direct and systematic, and that this explains why he finds the biting the bullet on the objection to be such a chilling prospect. But is it plausible to suppose that there is such a direct and systematic conflict? Derek Parfit famously argues that it is indeed plausible. He suggests that big ‘C’ Consequentialism—or ‘C’, as he calls it—is ‘indirectly collectively self-defeating’. Call [a moral theory] T

nearest and dearest objection

243

indirectly collectively self-defeating when it is true that, if several people try to achieve their T-given aims, these aims would be worse achieved. On all or most of its different versions, this may be true of C. C implies that, whenever we can, we should try to do what would make the outcome as good as possible. If we are disposed to act in this way, we are pure do-gooders. If we were all pure do-gooders, this might make the outcome worse. This might be true even if we always did what, of the acts that are possible for us, would make the outcome best. The bad effects would come, not from our acts, but from our disposition. (Parfit 1984: 27)

As I understand it, the nearest and dearest objection that worries Jackson is best understood as a particular manifestation of big ‘C’ consequentialism’s being indirectly collectively self-defeating. The objection can be brought out by imagining a third response Bloggs might have made to our suggestion that he is mistaken and should give the $100 to the complete stranger. Bloggs might say, ‘Suppose we all consistently try and succeed in doing what big ‘‘C’’ consequentialism tells us to do, which is to maximize neutral value. This would require that the desire for what is neutrally valuable—that is, the desire for benefits, independently of who receives them—is the strongest of our desires. But if this was the strongest of our desires then the lives we would each end up leading would be much worse—that is, there would be less neutral value in each of our lives, less benefits, considered as a whole—than there would have been if we had had different desires, and so acted differently. This is because our each having, as our strongest desire from time to time, the desire that benefits accrue to our nearest and dearest, is a crucial factor in causing benefits both to ourselves and to our nearest and dearest. Yet if we sometimes have, as our strongest desire, the desire to give benefits to our nearest and dearest, then it isn’t the case that we always have, as our strongest desire, the desire that there be as many benefits as possible for people. So our leading lives that are worth leading—that is, our leading lives in which neutral value is maximized—is plausibly inconsistent with our always performing acts that maximize neutral value. So big ‘‘C’’ consequentialism’s account of which acts we should perform must be mistaken.’ The crucial assumption Bloggs makes in giving this response is that, as Parfit puts it, effects can ‘come, not from our acts, but from our disposition.’ This seems to be what Jackson has in mind too when he says, in the passage quoted at the outset, that our ‘lives are given shape, meaning and value by what we hold dear, by those persons and life projects to which we are especially committed.’ The fact that we sometimes have, as our strongest desire, the desire that benefits accrue to our nearest and dearest, is an important causal factor in producing benefits both to ourselves and to our nearest and dearest independently of the

244

michael smith

acts that such a desire produces (Adams 1976). But in that case it follows that if we all try and succeed in doing what big ‘C’ consequentialism tells us that we should do, which is to perform acts that maximize neutral value, then the outcome may well be worse in big ‘C’ consequentialism’s own terms. The outcome would have been better if we had been differently motivated, and so had failed to do what big ‘C’ consequentialism tells us to do. Big ‘C’ consequentialism thus appears to condemn its own account of what we ought to do. For it tells us that things may be worse when we try and succeed in making them better.

3. Jackson’s reply to the nearest and dearest objection Jackson’s initial reply to the nearest and dearest objection, so understood, takes issue with the premiss that big ‘C’ consequentialism tells us to act in ways which are such that, if we consistently acted in those ways, we would lead lives that are not worth living. His reason for rejecting this premiss turns on his view that, according to most plausible formulation of big ‘C’ consequentialism, what agents should do is not act so as to maximize neutral value—that’s what we assumed in constructing the argument against big ‘C’ consequentialism—but rather act so as to maximize expected neutral value. I will comment on the plausibility of this view presently. For the moment, however, let me spell out how, as Jackson sees things, accepting this alternative formulation of big ‘C’ consequentialism provides us with (the beginnings of) a reply to the nearest and dearest objection. As Jackson points out, we generally have a much better idea of what will benefit those who are known to us, and only the vaguest idea of what will benefit complete strangers. For example, if I give my child $100, I have a pretty good idea of what he will spend it on, but I have relatively little idea of what a complete stranger would spend $100 on. Moreover, even when we do know what a complete stranger will spend $100 on, the causal pathway from any action that we perform to the delivery of that benefit is usually somewhat more unpredictable in the case of a complete stranger than it is in the case of one well known to me. I place $100 in my child’s hand, and, because I know him and his habits, I am confident that he will keep it and spend it in ways that will benefit him. But I am rarely in a position to identify a complete stranger to whom I could provide a substantial benefit by putting $100 in his hand. I rely on an aid agency to identify such strangers for me, and then it becomes somewhat obscure how much of what I give to the aid agency, if anything, ends up in the hands of the complete stranger.

nearest and dearest objection

245

Generally speaking, then, it follows that limited creatures like ourselves will have much higher confidence that our actions will produce benefits for those well known to us, and much lower confidence that our actions will produce benefits for complete strangers. But if this is right then, even assuming that benefits are of neutral value only, it follows that acting in ways that produce smaller benefits for those well known to us, rather than greater benefits for complete strangers, will, generally speaking, have greater expected neutral value. To be sure, on each such occasion there will be an alternative action available to us that would produce more neutral value in fact, namely, the action of providing the greater benefit to a complete stranger. But since we have only a very low expectation that any particular act we could perform is one which would provide a greater benefit to some complete stranger, the expected neutral value of providing a smaller benefit to those well known to us will, in most cases, be much greater than the expected neutral value of providing a greater benefit to a complete stranger. Jackson notes a potential problem with this initial reply to the nearest and dearest objection. It might well be objected that we can distinguish two nearest and dearest objections, and that I have replied to only one of them. One objection is, ‘‘How can consequentialists make sense of the fact that there is a relatively small group of people whose welfare plays a special role in our lives, given the agent-neutral nature of consequentialism’s value function?’’ Our reply was that ... right value translates into right action ... through an agent’s beliefs, and that when this is appreciated, empirical facts about our cognitive powers and situation make it plausible that our actions should be highly focussed much of the time. The other objection is, ‘‘How can consequentialists make sense of it being the particular small group of people that it mostly is?’’ Perhaps consequentialism can make sense of there being a small group, but why the small group of family, friends, fellow citizens, and the like that it so often is? ( Jackson 1991: 478)

As I said, Jackson’s initial reply to the nearest and dearest objection is that it presupposes an implausible conception of what, according to big ‘C’ consequentialism, we should do. But what the potential problem shows is this isn’t the case. For suppose that we substitute his preferred conception. This makes no difference at all. Someone who consistently maximizes expected neutral value, even someone with limits on their knowledge like those mentioned, is, for all we’ve been told, still someone whose strongest desire is the desire for what is of neutral value. Someone who consistently maximizes expected neutral value is therefore still someone who fails to have the desires that make their life worth living. For they do not have, as their strongest desire from time to time, the desire to provide benefits to their family, friends, fellow citizens, and the like.

246

michael smith

Jackson thus goes on to supplement his initial reply with the further suggestion that additional empirical facts explain why the particular small group on which we focus is the small group of family, friends, fellow citizens, and the like. [I]n deciding what to do here and now an agent must take account of what he or she will do in the future, and that involves taking very seriously questions of character. Do I have the persistence that will be called for, will I remain sufficiently enthusiastic about the project to put in the time required, will I be able to retain a sufficiently impartial outlook, will I be able to avoid the various temptations that will arise, and so on and so forth? ... [A]s a rule we do better for reasons of character (that no doubt have an evolutionary explanation) with projects that involve family and friends rather than strangers. This is simply because we are much less likely to lose the enthusiasm required to see the project through to a successful conclusion when the project benefits people we have a particular affection for. ( Jackson 1991: 480)

As Jackson sees things, given that we are enjoined to maximize expected neutral value, these two empirical considerations—the limits of our knowledge, and the limits of our self-control—combine to focus our attention on the benefits that we can provide for our nearest and dearest. Jackson’s reply suggests that when we told Bloggs’s story at the outset we went badly wrong when we stipulated that he believes that, were the complete stranger to receive the $100, he would buy something from which he would gain a slightly greater benefit than would Bloggs’s own child. Jackson’s point is, in essence, that we thereby stipulated something empirically implausible. To be empirically plausible we would have to imagine that Bloggs is far less confident about the level of benefit a complete stranger will gain from giving $100 to him than he is about the level of benefit that would accrue to his child from being given $100. So when we ask which of the actions available to Bloggs will maximize expected benefit the answer is going to turn out to be giving the $100 to his child. If Jackson is right then we see that Bloggs’s third response to the suggestion that he should give $100 to the complete stranger turns on a false claim. Bloggs’s response was supposed to be that big ‘C’ consequentialism condemns itself: by big ‘C’ consequentialism’s own lights, things may be worse when we try and succeed in making them better. The false premiss in Bloggs’s argument for this conclusion is the premiss that if we sometimes have, as our strongest desire, the desire to give benefits to our nearest and dearest, then it isn’t the case that our strongest desire is always the desire that there be as many benefits as possible for people. For, given the limits of our knowledge and self-control, it turns out that the acts that maximize expected benefits for

nearest and dearest objection

247

people generally are the acts that maximize expected benefits for our nearest and dearest. There is therefore no inconsistency in having a pair of desires tied as our strongest desires. Our actions can all be overdetermined. For these desires—the desire that our nearest and dearest benefit, and the desire that the people in general benefit—don’t require us to act in different ways, not given our limited knowledge and self-control. So even if we all try and succeed in doing what big ‘C’ consequentialism tells us to do we can still have, within ourselves, the crucial desire whose possession is necessary for our living a worthwhile life. What should we make of this reply? Big ‘C’ consequentialism is a moral theory. I take it that this means that it is supposed to tell us what makes acts right and wrong, not just in the actual world, but in every possible world. Now what the nearest and dearest objection purports to show is that there is something wrong with big ‘C’ consequentialism by the theory’s own lights; for, to repeat, the theory itself seems to tell us that the outcome is worse if we all try and succeed in making it better. But it is hard to see the relevance of the empirical considerations Jackson adduces to this objection. To be sure, if the people mentioned in the crucial premiss of the nearest and dearest objection are people like you and me, people with limited knowledge and self-control, then the premiss of the objection is false. But in that case we should just assume that the people mentioned in the premiss of the objection don’t have (say) such limited self-control. For the premiss of the argument is in that case true, and the conclusion then apparently follows. Nor is this a fanciful reply. Jackson himself admits that there are people in the actual world who are much more self-controlled than the rest of us, people like Mother Theresa and Ralph Nader (Jackson 1991, p.481). They seem both willing and able to act in ways that increase expected neutral value for a group that isn’t the small group of their family, friends, fellow citizens, and the like, and they seem willing and able to do so notwithstanding the fact that they make their own lives worse than they could have been as a result. So let’s just imagine a community of people who are all more like Ralph Nader and Mother Theresa. When the people in this community all try and succeed in making the outcome better then, by big ‘C’ consequentialism’s own lights, the outcome is worse than it would have been if they had been differently motivated, and so had acted differently. If this is an objection to big ‘C’ consequentialism at all, then the mere possibility of such a community seems to be all that’s required to bring that objection to the fore. It therefore seems to me that Jackson’s reply to the nearest and dearest objection misses its mark. It does not help with that objection to reformulate

248

michael smith

big ‘C’ consequentialism as the doctrine that we should maximize, not neutral value simpliciter, but expected neutral value.

4. Parfit’s reply to the nearest and dearest objection So far we have simply taken it for granted that Blogg’s third response to our suggestion that he should give the $100 to the complete stranger constituted an objection to big ‘C’ consequentialism. If the members of a community could all try and succeed in making the outcome better, by big ‘C’ consequentialism’s own lights, and yet the outcome is worse, again by the theory’s own lights, than it would have been if they had been differently motivated, and so had acted differently, then, we have assumed, this does indeed constitute an objection to big ‘C’ consequentialism. But does it really? Like Mill, Sidgwick and others, Parfit insists that we distinguish sharply between the different things that can be evaluated in big ‘C’ consequentialist terms (Mill 1861, Sidgwick 1907, Adams 1976, Hare 1981, Railton 1984). One question we can ask is which act an agent should perform. In big ‘C’ consequentialist terms, this amounts to the question, ‘Of the various acts that an agent could perform, which is the act that will produce the greatest neutral value?’ Another question we can ask is which desires an agent should possess. In big ‘C’ consequentialist terms, this amounts to the question, ‘Of the various desires that an agent could have, which are the desires whose possession by him will produce the greatest neutral value?’ And yet another question we can ask is which life an agent should live. In big ‘C’ consequentialist terms, this amounts to the question, ‘Of the various lives an agent could live, which is the life the living of which by him will produce the greatest neutral value?’ And so we could go on (Parfit 1984 pp.28–9; see also Railton 1988, Brink 1989, Pettit and Smith 2000). Parfit points out that so long as the items up for evaluation have different consequences from each other, big ‘C’ consequentialism’s answers to these different questions will be logically independent of each other. Notwithstanding the intimate connection between the desires people have and the acts that those desires produce, it therefore follows that, since their desires can have effects independently of these acts, big ‘C’ consequentialism may tell people to perform certain acts, but also tell them to have desires which are quite different from those that they would have if they were to perform all of those acts. The same is true of the lives that people lead and the acts that they perform in leading their life. Since too these have different effects, it follows that big ‘C’

nearest and dearest objection

249

consequentialism may tell people to perform certain acts, but also tell them to lead a life which is quite different from the life that they would lead if they did perform all of those acts. Far from condemning itself, big ‘C’ consequentialism thus simply applies its principle of evaluation perfectly consistently to anything and everything that we might want to evaluate. To my mind this provides us with a powerful and decisive reply to the version of the nearest and dearest objection that worries Jackson. The reply concedes the premiss of the objection. Big ‘C’ consequentialism does indeed tell us to act in ways which are such that, if we consistently acted in those ways, we would lead lives that are not worth living. For what the reply denies is that it follows from this that big ‘C’ consequentialism tells us to lead such lives. On the contrary, the reply goes, big ‘C’ consequentialism tells each of us to live that life, of those we could live, which is such that, by living that life, we maximize neutral value. The premiss of the nearest and dearest objection thus doesn’t support the conclusion. It gives us no reason to believe that big ‘C’ consequentialism condemns itself. Big ‘C’ consequentialism doesn’t condemn itself. Rather, to repeat, it requires us to apply its principle of evaluation consistently to everything that we might want to evaluate—actions, desires, lives—independently of our application of it to anything else. At one point Jackson considers this line of reply to the nearest and dearest objection, but rejects it. I am not here denying the correct and important point that some particular action may be wrong in consequentialist terms and yet spring from a character which is right in consequentialist terms. I am denying that the point helps with the essentials of the nearest and dearest objection. For the consequences of having a character which gives a special place in one’s affections and concerns to those persons who are closest to one are, in the main, consequences of the manifestations of such a character, that is, of the actions which are especially directed to the needs of those closest to us. Hence, a consequentialist justification of such a character presupposes a consequentialist justification of those actions—which returns us to the very question raised by the nearest and dearest objection. ( Jackson 1991: 479)

But I am not sure how Jackson can say what he goes on to say, given the qualifier ‘in the main’. The crucial point is that, since the consequences of our actions and our character are non-identical, no conclusion about the character we ought to have, or the life we ought to lead, can be drawn from big ‘C’ consequentialism’s answer to the question ‘Which actions ought we to perform?’ This completely undermines the nearest and dearest objection because that objection, in effect, attempts to draw just such a conclusion.

250 michael smith

5. Jackson’s preferred formulation of big ‘C’ consequentialism As we saw, Jackson makes a good deal of the fact that big ‘C’ consequentialism requires us to act so as to maximize not neutral value simpliciter, but rather expected neutral value. Though, as we have seen, formulating the doctrine in these terms doesn’t help with the nearest and dearest objection, the argument he gives for formulating big ‘C’ consequentialism in this way is worth considering on its own merits. The argument proceeds by way of a discussion of two examples which are supposed to help us choose between the competing formulations. There are many formulations to consider, but Jackson restricts himself to considering three. According to one, the formulation which we gave initially, big ‘C’ consequentialism tells us that we ought to act so as maximize neutral value. According to the second, it tells us that we ought to act so as to have the best chance of maximizing neutral value. And according to the third, the formulation which Jackson prefers, big ‘C’ consequentialism tells us that we ought to act so as maximize expected neutral value. As I said, Jackson’s argument for formulating big ‘C’ consequentialism in terms of expected neutral value proceeds by way of a discussion of examples. Here is the first example. Jill is a physician who has to decide on the correct treatment for her patient, John, who has a minor but not trivial skin complaint. She has three drugs to choose from: drug A, drug B, and drug C. Careful consideration of the literature has led her to the following opinions. Drug A is very likely to relieve the condition but will not completely cure it. One of drugs B and C will completely cure the skin condition; the other though will kill the patient, and there is no way she can tell which of the two is the perfect cure and which the killer drug. What should Jill do? The possible outcomes we need to consider are: a complete cure for John, a partial cure, and death. It is clear how to rank them: a complete cure is best, followed by a partial cure, and worst is John’s death ... But how do we move from that ranking to a resolution concerning what Jill ought to do? The obvious answer is to take a leaf out of decision theory’s book and take the results of multiplying the value of each possible outcome given that the action is performed, summing these for each action, and then designating the action with the greatest sum as what ought to be done. In our example there will be three sums to consider, namely: Pr(partial cure/drug A taken) x V (partial cure) + Pr(no change/drug A taken) x V (no change); Pr(complete cure/drug B taken) x V (complete cure) + Pr(death/drug B taken) x V (death); and

nearest and dearest objection

251

Pr(complete cure/drug C taken) x V (complete cure) + Pr(death/drug C taken) x V (death). Obviously, in the situation as described, the first will take the highest value, and so we get the answer that Jill should prescribe drug A. ( Jackson 1991: 462–3)

What this example is supposed to show is that it would be implausible to formulate big ‘C’ consequentialism as the doctrine that we ought to act so as to maximize neutral value. For neutral value is at a maximum when John is completely cured, so Jill’s prescribing drug A, which will partially but not completely cure him, is, by this criterion, definitely not the thing that we ought to do. This leaves us with only two alternatives: prescribing either drug B or drug C. But, Jackson says, ‘We would be horrified if she prescribed drug B, and horrified if she prescribed drug C.’ (Jackson 1991, p.466) So, Jackson concludes, it can hardly be plausible to suppose that big ‘C’ consequentialism tells us that we ought to maximize neutral value. Here is Jackson’s second example. As before, Jill is the doctor and John is the patient with the skin problem. But this time Jill has only two drugs, drug X and drug Y, at her disposal which have any chance of effecting a cure. Drug X has a 90% chance of curing the patient but also has a 10% chance of killing him; drug Y has a 50% chance of curing the patient but has no bad side effects. Jill’s choice is between prescribing X or prescribing Y. It is clear that she should prescribe Y, and yet that course of action is not the course of action most likely to have the best results. ( Jackson 1991: 467)

What this second example is supposed to show is that it would be implausible to formulate big ‘C’ consequentialism as the doctrine that we ought to act so as to have the best chance of maximizing neutral value. For Jill’s prescribing drug X has a 90% chance of maximizing neutral value—that is, of bringing about a complete cure—whereas prescribing drug Y only has a 50% chance of having this result. Yet what Jill plainly should do is prescribe drug Y. Jill should prescribe drug Y, notwithstanding the fact that it does not have the greatest chance of producing the best result. Taken together, what the two examples are supposed to show is that big ‘C’ consequentialism is best formulated as the doctrine that we ought to act so as to maximize expected neutral value. For this formulation supposedly gives us the right answers in both examples. It tells us that Jill ought to prescribe drug A in the three drugs example, and it tells us that she ought to prescribe drug Y in the two drugs example. But while I admit that this sounds superficially plausible, it seems to me that, on further reflection, we see that Jackson’s use of the two examples to support the expected value formulation of big ‘C’ consequentialism is flawed in a quite fundamental way. Insofar as it is a doctrine

252

michael smith

about what we ought to do, big ‘C’ consequentialism is best formulated as the doctrine that we ought to act so as to maximize neutral value (see also Parfit 1984, Railton 1984, Brink 1989). This means that I disagree with what Jackson says about the three drugs example. We get a hint of the problem with Jackson’s argument very early on when he explains how we can represent what big ‘C’ consequentialism tells agents to do in terms of what a variant on standard decision theory tells them that they ought to do. [W]e can think of consequentialism’s value function as telling us what, according to consequentialism, we ought to desire. For a person’s desires can be represented—with, of course, a fair degree of idealization—by a preference function which ranks states of affairs in terms of how much the person would like the state of affairs to obtain, and we can think of consequentialism as saying that the desires a person ought to have are those which would be represented by a preference function which coincided with consequentialism’s value function. The other ingredient in the decision-theoretic account of what consequentialism says a person ought to do, the agent’s subjective probability function, is an idealization of the agent’s beliefs. Hence, the decisiontheoretic account is one in terms of what the person ought to desire and in fact believes. ( Jackson 1991: 464)

The suggestion is thus that what big ‘C’ consequentialism tells agents to do is what they ought to do according to a variant on standard decision theory, the variant in which we combine what they in fact believe with the desires which big ‘C’ consequentialism tells them they ought to have. There is a major problem with this suggestion, however. For we need to provide an interpretation of the ‘ought’ that makes the crucial sentence in this passage—‘we can think of consequentialism as saying that the desires a person ought to have are those which would be represented by a preference function which coincided with consequentialism’s value function’—come out true. There are three interpretations to be considered: I will call these the moral, the formal and the critical interpretations of the ‘ought’. Let me begin with the moral interpretation, just to set it to one side. What the quoted sentence says, on the moral interpretation, is that big ‘C’ consequentialism tells us that the desires a person ought (morally) to have are those that coincide with big ‘C’ consequentialism’s value function. But this is evidently false. As we saw in our discussion of Parfit’s response to the nearest and dearest objection, just as big ‘C’ consequentialism says that people ought (morally) to perform those acts, of those they could perform, that maximize (expected) neutral value, it says that people ought (morally) to have those desires, of those they could have, which are such that their possession of those

nearest and dearest objection

253

desires maximizes (expected) neutral value. This straightforwardly conflicts with Jackson’s claim in the quoted sentence. Suppose big ‘C’ consequentialism tells us that only one thing is of value, namely pleasure. What Jackson says in the quoted sentence, on the moral interpretation of the ‘ought’, is that big ‘C’ consequentialism tells us that people ought (morally) to have the desire for pleasure, whereas, to repeat, what big ‘C’ consequentialism in fact tells us is that people ought (morally) to have that set of desires whose possession maximizes (expected) pleasure. This may be a set of desires that consists of one member, the desire for pleasure, but, equally, it may not be. Everything depends on whether possession of the desire for pleasure maximizes (expected) pleasure. So on the moral interpretation of the ‘ought’, the quoted sentence isn’t obviously true at all. In any event, the moral interpretation of the sentence is plainly wrongheaded. What Jackson is trying to do is to formulate big ‘C’ consequentialism’s account of what we ought (morally) to do, so we can hardly use the big ‘C’ consequentialist principle itself in formulating that account. If we already have a big ‘C’ consequentialist principle, why not just apply it directly to acts? As Jackson himself puts it in discussing a related idea: [P]erhaps we are being offered a variant on consequentialism according to which an action is to be judged not directly but via the status, judged consequentially, of the character which gives rise to it, but then we appear to be landed with a dubious compromise reminiscent of rule utilitarianism. If consequences are the key in one place, why not across the board? ( Jackson 1991: 479)

Let’s therefore put the moral interpretation to one side. This leaves us with two alternative interpretations: what I’ve called the formal, and the critical interpretations. The idea behind the formal interpretation of the ‘ought’ is that, just as belief has the formal aim of truth, from which it follows that people ought (in this formal sense) to have true beliefs or knowledge, so desire has, as its formal aim, the good, from which it follows that people ought (in this formal sense) to desire what is, in fact, good. If, for example, pleasure is good, then, the suggestion goes, people ought (formally) to desire pleasure. So, just as big ‘C’ consequentialism’s claim that people ought (morally) to have those beliefs whose possession maximizes neutral value is no objection to the claim that people ought (formally) to have true beliefs, so, according to this alternative interpretation of the quoted sentence, big ‘C’ consequentialism’s claim that people ought (morally) to have those desires whose possession maximizes neutral value is no objection to the claim that people ought (formally) to desire

254

michael smith

what is, in fact, good. It is the latter interpretation that is crucial in formulating big ‘C’ consequentialism decision-theoretically. Or so the suggestion goes. People ought (morally) to do what they ought to do according to the variant on decision theory in which we replace their actual desires with the desires that they ought (formally) to have. The problem with this formal interpretation of the ‘ought’ in the quoted sentence, however, is that though it makes the quoted sentence come out true, it also makes much more glaring Jackson’s lack of even-handedness in explaining how we can represent what big ‘C’ consequentialism tells people to do in terms of what they ought to do according to a variant on standard decision theory. For if, in formulating the variant on standard decision theory, regular desires are to be replaced by desires which meet their formal aim, then shouldn’t regular beliefs also be replaced by beliefs that meet their formal aim? In other words, shouldn’t regular beliefs be replaced with knowledge? Jackson notes this possibility himself at one point: [I]n addition to distinguishing what a person in fact desires from what he or she ought to desire, we also distinguish what a person in fact believes from what a person ought to believe ... Hence, it might well be suggested that we should recover the consequentialist answer to what a person ought to do from the value function via what that person ought to believe rather than from what he or she in fact believes. ( Jackson 1991: 464)

But if we were to do this then, given that what a person ought (formally) to believe is what is true—since a person ought (formally) to have knowledge—it would follow that, according to big ‘C’ consequentialism, people ought to maximize neutral value, not maximize expected neutral value. For this is what agents ought to do according to the variant on standard decision theory that replaces an agent’s ordinary desires and beliefs with desires and beliefs that meet their formal aim. Yet this is the formulation of big ‘C’ consequentialism that was supposedly refuted by the three drugs example. Something therefore seems to have gone badly wrong. To anticipate what is to come, if the three drugs example does indeed tell against this formulation of big ‘C’ consequentialism, then surely we must suppose that it tells equally against both the suggestion that we replace ordinary beliefs by beliefs that meet their formal aim, and ordinary desires with desires that meet their formal aim. As I said, Jackson himself notes the possibility that we should formulate big ‘C’ consequentialism decision-theoretically by replacing an agent’s desires and beliefs with desires and beliefs that both meet their formal aim. He doesn’t, however, think that this idea is very plausible. In other words, he denies that we should be even-handed in our treatment of belief and desire when we

nearest and dearest objection

255

formulate big ‘C’ consequentialism decision-theoretically. He explains why in the course of giving a response to what he calls an ‘annoying complication’. ... I need to note an annoying complication. I have been arguing for an interpretation of consequentialism which makes what an agent ought to do the act which has the greatest expected moral utility, and so is a function of the consequentialist value function and the agent’s probability function at the time. But an agent’s probability function at the time of action may differ from her function at other times, and form the probability function of other persons at the same or other times. What happens if we substitute one of these other functions in place of the agent’s probability function at the time of action? The answer is that we get an annoying profusion of ‘oughts’ ... I think that we have no alternative but to recognize a whole range of oughts—what she ought to do by the lights of her beliefs at the time of action, what she ought to do by the lights of what she later establishes ... , what she ought to do by the lights of one or another onlooker who has different information on the subject, and, what is more, what she ought to do by God’s lights, that is, by the lights of one who knows what will and would happen for each and every course of action ... I hereby stipulate that what I mean from here-on by ‘ought,’ and what I meant, and hope and expect you implicitly took me to mean when we were discussing the examples, was the ought most immediately relevant to action, the ought which I urged to be the primary business of ethical theory to deliver. When we act we must perforce use what is available to us at the time, not what may be available to us in the future or what is available to someone else, and least of all what is available to a God-like being who knows everything about what would, will and did happen. ... ( Jackson 1991: 471–2)

Jackson’s explanation of why we shouldn’t be even-handed in our treatment of belief and desire is that we can’t be even-handed in this way if we are trying to elucidate the ‘ought most immediately relevant to action’. This suggests the third possible reading of the ‘ought’ in the crucial sentence that appears in Jackson’s account of how big ‘C’ consequentialism can be represented decision-theoretically. When he says ‘we can think of consequentialism as saying that the desires a person ought to have are those which would be represented by a preference function which coincided with consequentialism’s value function’, perhaps he intends the ‘ought’ to be read as the ‘ought most immediately relevant to action.’ Now, as I understand it, the ‘ought’ that is most immediately relevant to action is the ‘ought’ that underwrites rational criticism of what agents do: it is, as I shall say, what they ought (critically) to do. When we act, our aim is to live up to our responsibilities as rational agents, thereby avoiding rational criticism. In order to do this we must exercise all of those rational capacities we possess whose exercise is required for rational action. For, so long as we do this, we do all that we can; and, since it cannot be supposed that we ought (critically) to do something that we cannot do, we do all that we ought (critically) to do.

256

michael smith

I take it that this is why Jackson is so scathing about the idea of an ‘ought’ that is based on God-like knowledge. Since no one could have such knowledge, given that we are trying to elucidate what agents ought to do in the sense of ‘ought’ that is ‘most immediately relevant to action’—that is, given that we are trying to elucidate what agents ought (critically) to do—it is irrelevant what agents would do if they had God-like knowledge. What is relevant is rather what they would do if they were to fully exercise such, admittedly limited, belief-forming capacities as they have. Let me illustrate the ways in which this idea can be developed with some examples. Suppose an agent desires to walk into a room, and he has two beliefs about how he might achieve this. He can walk through door 1, which he is certain will get him into the room, or he can walk through door 2, which he is fairly confident will get him into the room, though he is not absolutely certain: he thinks it might just be a door to a cupboard. Given that the strength of agents’ instrumental desires should be a function of the strength of their desires for ends and their level confidence about the means to their ends, it follows that this agent should have a stronger instrumental desire to open door 1 than to open door 2. In decision-theoretic terms, he will maximize expected utility by opening door 1. This is therefore what he would do if he fully exercised his capacity for instrumental rationality. In this sense, it is what he ought (critically) to do. But of course, for all that, the agent might not exercise these capacities fully. For example, he might put his desire to walk into the room together with his belief about door 2, but simply not put it together with his belief about door 1. If this is what he does then he will presumably have an instrumental desire to open door 2, and no instrumental desire to open door 1. What he will do is open door 2, and, in so doing, he will act intentionally, but irrationally. He will fail to do what he ought (critically) to do. If this is the right way to understand the ‘ought’ that Jackson is trying to elucidate, then we can generalize. When an agent acts, and in so doing fully exercises all of the rational capacities he possesses whose exercise is required for rational action—that is to say, when, in acting, he does everything that he ought (critically) to do—then he will presumably exercise other rational capacities as well. For example, just as he will act on instrumental desires that elude criticism in terms of norms of instrumental rationality, he will act on means-ends beliefs that elude criticism in terms of epistemic norms. His means-end beliefs will be formed on the basis of sound reasoning from evidence, rather than on the basis of unsound reasoning, or whimsy, at least to the extent that he has the capacity to engage in such reasoning. The mere

nearest and dearest objection

257

fact that an agent fully exercises such rational capacities as he possesses in the formation of his means-end beliefs will, of course, be no guarantee that his means-end beliefs are true, for he may lack certain crucial rational capacities, or his evidence may be misleading. Human agents are not God-like. All it guarantees is that, if the agent’s means-end beliefs are false, then at least it will not be his fault. For in acting he will have done all that he can, and all, therefore, that he ought (critically) to do. Moreover when, in acting, an agent does everything that he ought (critically) to do, he will also have desires for ends that elude rational criticism. Of course, radical Humeans deny that desires for ends are fit objects of rational criticism. But this certainly isn’t the view of those, like Jackson, who defend the modest form of internalism which holds that judgements of value and desires for ends stand in the following rational relation (Jackson and Pettit 1995): MI: Reason requires that ∀x (If x judges that p is goodx , then x desires that p) For, according to MI, those who possess the capacity to have desires for ends for the things that they judge to be good, but who fail to exercise that capacity, are also rationally criticizable (Smith 1994). They fail either to have the desires for ends that they ought (critically) to have, or they fail to make the value judgements that they ought (critically) to make. Someone who, in acting, does everything that he ought (critically) to do thus doesn’t just act so as to realize their desires for ends, but also acts so as realize what he judges to be good. Finally, as this example suggests, an agent who acts having done everything that he ought (critically) to do, will also be an agent whose value judgements elude rational criticism. That is to say, much like his means-end beliefs, his value judgements will be formed on the basis of sound reasoning from evidence, rather than on the basis of unsound reasoning, or whimsy, at least to the extent that he is capable of such reasoning. Again, as with his means-end beliefs, the mere fact that he fully exercises such rational capacities as he possesses in making his value judgements will be no guarantee that his value judgements are true. He may lack certain crucial capacities, or his evidence may be misleading. All it guarantees is that, if they are false, then it is not his fault. In acting he will have done all that he can, and all, therefore, that he ought (critically) to do. To sum up, when agents act, they are liable for rational criticism if, in acting, they fail to exercise all of the rational capacities they have whose exercise is required for rational action. If we diagram the various elements potentially

258

michael smith

facts about means to desired ends

available evidence about means to desired ends

means-ends beliefs

bodily movements

facts about what is good

available evidence about what is good

judgements about what is good

desired ends

desires for ends

Figure 10.1. Rational Action

involved in rational action, then the places where these capacities can be exercised are represented by the arrows and plus signs relating the elements in bold (see also Smith 2004). As rational agents, most of us have capacities of some sort to respond to available evidence in the formation of both our means-end beliefs and our value judgements; to respond to our value judgements in the formation of our desires for ends; and to respond to our desires for ends and means-end beliefs when we subsequently move our bodies. An agent who does everything that he ought (critically) to do when he acts is thus one who fully exercises these capacities. But of course, since an agent cannot guarantee that the evidence available to him isn’t misleading, there is no reason to suppose that, merely by exercising his rational capacities, he can guarantee that his means-end beliefs or value judgements are true. Nor, for that same reason, can an agent guarantee that the ends he desires will result from his moving his body. There is thus no general presumption that agents ought (critically) to act in accordance with the true and the good, because their doing so is not, in general, under their rational control.

nearest and dearest objection

259

I have laboured this point because it shows the extent to which Jackson misunderstands what is required if his goal is indeed to elucidate the ‘ought most immediately relevant to action’—that is, what we ought (critically) to do. True enough, we rationally criticize agents if they fail to act on appropriate means-end beliefs: that is, if they fail to exercise their capacity for instrumental rationality. What they ought (critically) to do is to act having exercised this capacity. But we equally rationally criticize agents if they act on means-end beliefs that are formed on the basis of sloppy reasoning, or whimsy. In other words, since expectations are rationally criticizable, so too are actions performed on the basis of expectations. What agents ought (critically) to do is to act having exercised their capacity for epistemic rationality. And we equally rationally criticize agents if they act on desires for ends that do not accord with their judgements of value, and agents who act on value judgements that are formed on the basis of sloppy reasoning or whimsy. Glaringly, however, we no more rationally criticize agents for failing to do what is in fact good than we rationally criticize them for failing to act on means-end beliefs that are in fact true. This helps us understand what is really going on in Jackson’s examples. Consider again Jill, faced with a choice between giving John drugs A, B, and C, but this time let’s suppose that Jill disagrees with us about the relative value of the outcomes, and let’s also suppose that this disagreement in no way reflects badly on her as a rational subject. Perhaps she was simply raised in a context in which all the evidence supported value judgements that we don’t ourselves agree with. As Jackson tells the story, we are supposed to be horrified if she gives John drug B, and equally horrified if she gives John drug C, because in doing so she fails to maximize expected value. To maximize expected value she must give John drug A. But what his discussion of the example buries is the fact that, to the extent that our horror reflects our assessment of Jill as a rational agent—that is, to the extent that our horror is based on our belief that Jill fails to exercise such rational capacities as she has—the values we take to be relevant will have to be Jill’s values, not our own. If, relative to her own assignment of values, giving John drug B, or drug C, maximizes expected value then we would rationally criticize her for failing to give John either drug B, or drug C. This is what she ought (critically) to do. Conversely, if we suppose that in assessing Jill’s conduct we are allowed to replace Jill’s values with some other system of values—suppose, for example, that in the spirit of Jackson’s preferred decision-theoretic formulation of big ‘C’ consequentialism, we replace her values with what big ‘C’ consequentialism

260

michael smith

tells us to be good—then, when we assess Jill’s conduct, we are plainly no longer interested in assessing her as a rational agent. The ‘ought’ claims we make are not claims about what Jill ought (critically) to do; they are not ‘ought’ claims that are most immediately relevant to action. They are ‘ought’ claims of the other kind, claims about what Jill ought (formally) to do. The upshot is that, to the extent that we feel horrified merely by Jill’s failure to act in accordance with what big ‘C’ consequentialism deems to be really good—to the extent that we find something to lament about that, independently of whether her own judgements of value accord with big ‘C’ consequentialism, and, if they do, whether this can be traced to sloppy reasoning or whimsy on her behalf—so, by parity of reasoning, we should feel equally horrified by her failure to act on means-end beliefs that are true. In assessing whether she ought to give drug A, or drug B, or drug C, we therefore have no choice but to replace her desires by desires for what big ‘C’ consequentialism deems to be good, and her means-end beliefs with beliefs about what really are means to the ends that are good. We have no choice but to adopt a God-like point of view. But if we do this then, of course, we reach the conclusion that Jill ought to give John that drug, whichever it is, that in fact effects a complete cure: drug B, or drug C, whichever it is. In other words, we reach the conclusion that what Jill ought to do is maximize neutral value. To sum up, there are two salient ways in which we can represent what an agent ought (morally) to do in terms of what they ought to do according to a variant on decision theory. According to one, what we are interested in is what the agent ought (critically) to do. This is the ‘ought’ that is most immediately relevant to action. What we do in this case is ask what the agent ought to do according to the variant on decision theory in which she has the desires and beliefs that she would have if she fully exercised her rational capacities. Such an agent will maximize expected value, but the beliefs and values in question will both be closely related to her own. There will be no general presumption that such an agent will value what is good, or believe what is true. According to the other, what we are interested in is what the agent ought (formally) to do. What we do in this case is ask what the agent ought to do according to the variant on decision theory in which her desires and beliefs are both replaced by desires and beliefs that meet their formal aim. The agent desires what is good and believes what is true. Such an agent will therefore simply maximize value. But it is hard to see what ‘ought’ we would be elucidating by lacking even-handedness in the way Jackson suggests, that is, by asking what the agent ought to do according to that variant on decision theory in which an agent has his actual beliefs, but his desires are

nearest and dearest objection

261

replaced by desires for what is good. It therefore seems to me that Jackson fails to give a convincing argument for the expected neutral value formulation of big ‘C’ consequentialism. We should suppose that big ‘C’ consequentialism tells agents to maximize neutral value simpliciter, not maximize expected neutral value.

6. The epistemological version of the nearest and dearest objection In this final section, I would like to return to consider another version of the nearest and dearest objection, a version I mentioned at the beginning, but put to one side. The objection I have in mind is epistemological (Smith 2001). It is the version of the objection that Bloggs has in mind when he gives the first of the responses we considered to our suggestion that he should give the $100 to the stranger. According to John Rawls, we all come to moral philosophy with various firmly held convictions about what is value and what is not, and about which actions are right and which are not, and what we want to know is whether these convictions are justified (Rawls 1951). Rawls’s famous suggestion is that there is a common-sense procedure by which we test these convictions. This is the reflective equilibrium procedure. We test our convictions by trying, as best we can, to bring them into some sort of system. Justification is a matter of surviving in such a systematization. More precisely, his suggestion is that we test our convictions by first coming up with a hypothesis about what is of fundamental value, or what the fundamental principles are. In doing this, we invoke the standard criteria for constructing theories. We formulate a hypothesis that is simple, elegant, powerful, and adequate to the task of explaining why the convictions with which we began are true. A moral theory is, essentially, just such a hypothesis. Sometimes, when we formulate such a hypothesis we find that it is sufficiently simple, elegant, powerful and adequate to the task of explaining why some subset of our firmly held convictions are true that it gives us the confidence to reject, as mistaken or misguided, those convictions with which we began with which it is inconsistent. But sometimes the reverse is true. A hypothesis might have the wrong mix of theoretical virtues—it might be simple, say, but insufficiently adequate to the task of explaining our firmly held convictions—and so not inspire the confidence required to get us to reject the convictions with which it is inconsistent. In the former case we

262 michael smith reject our convictions as unjustified. In the latter case we reject our hypothesis as incredible. As I understand it, there is an epistemological version of the nearest and dearest objection to big ‘C’ consequentialism, a version that is best understood in terms of this Rawlsian reflective equilibrium procedure. Big ‘C’ consequentialism is a hypothesis that is supposed to give system to our various firmly held convictions about what is of fundamental value and what the fundamental principles are. According to big ‘C’ consequentialism, stating the fundamental principles is easy, as there is only one such principle, and it is analytic: we ought to act so as to maximize value. The hard part, the part where big ‘C’ consequentialism really does amount to a substantive hypothesis, is to give an account of the values that are to be maximized. Though big ‘C’ consequentialism is consistent with a range of possibilities in this regard, it places one significant global constraint on the form of all the values: values are one and all neutral in form, which is to say that they are expressed in principles like N, rather than in principles like R. Big ‘C’ consequentialism’s hypothesis that all values are neutral is very simple. But what the nearest and dearest objection brings out is that the simplicity of the hypothesis doesn’t inspire sufficient confidence in us to reject our firmly held conviction that some values are relative. We hear a story like Bloggs’s and we are supposed to find ourselves dismissing his attempt to give a justification of his conduct, but we can’t dismiss it. We share Bloggs’s conviction that there are relative values at stake and that they can be realized by his giving the $100 to his child. Our confidence in the claim that there are relative values is thus greater than our confidence in big ‘C’ consequentialism’s hypothesis that all values are neutral, notwithstanding the fact that that hypothesis purchases an abundance of the theoretical virtue of simplicity. In this way what the nearest and dearest objection brings out is that big ‘C’ consequentialism is simply incredible. This, as I understand it, is the epistemological version of the nearest and dearest objection. Right at the very end of his paper Jackson briefly considers, and rejects, the nearest and dearest objection in its more epistemological cast. One objection to consequentialism is that it conflicts with firmly held moral convictions, in particular concerning our obligations toward our nearest and dearest. It may be urged that my reply to this objection is seriously incomplete. For we can reasonably easily describe a possible case where the factors I mentioned as providing a justification in consequentialist terms for favouring one’s nearest and dearest do not apply, and yet, according to commonsense morality, one should favour, or at the least it is permissible to favour, one’s nearest and dearest. My concern, though, has been to reply to the

nearest and dearest objection

263

objection that consequentialism would, given the way things more or less are, render the morally good life not worth living. I take this to be the really disturbing aspect of the nearest and dearest objection. Consequentialists can perhaps live with the conflict with commonsense morality, drawing for instance on the notorious difficulties attending giving a rationale for its central features. But it seems to me that they cannot live with the conflict with a life worth living, given the way things more or less are. That would be to invite the challenge that their conception of what ought to be done had lost touch with human morality. ( Jackson 1991: 482)

Jackson thus doesn’t seem to be much impressed with the epistemological version of the nearest and dearest objection. But though only mentioned in passing, it is important to note that Jackson’s response to the epistemological version of the nearest and dearest objection is less than convincing. Jackson tells us that there are ‘notorious difficulties attending giving a rationale’ for the ‘central features’ of a non-big ‘C’ consequentialist hypothesis. He provides a hint as to nature of these ‘notorious difficulties’ by footnoting Shelley Kagan’s The Limits of Morality (1989). But what Kagan argues, and argues decisively, is that it is difficult to make sense of absolute constraints on behaviour, constraints of the kind that are supposed to be required by deontological restrictions. In small ‘c’ consequentialist terms, this requires not just the hypothesis that some values are relative—for example, the restriction on my harming people requires the disvalue of the outcome of someone’s being harmed by me, which does indeed require formulation in a principle of the same form as R—but also that this outcome has infinite disvalue. Kagan’s argument is, in essence, that such infinite disvalues cannot be combined in any plausible way with uncertainty about their realization. But conceding that Kagan is right about this, as it seems to me we should, doesn’t tell in favour of the big ‘C’ consequentialism’s hypothesis that all value is neutral. There is an alternative hypothesis in between these two, namely, the hypothesis that in addition to neutral values there are relative values, but that all such values are finite. As I understand it, this is all that the nearest and dearest objection assumes. The objection is not that the relative value that attaches to a benefit to Bloggs’s child, in virtue of its being a benefit to his child, is such as to place an absolute prohibition on his bringing about neutral value. The objection is rather that, though in circumstances in which the gain in relative value is small enough and the loss of neutral value large enough, the gain in relative value would not outweigh the loss of neutral value, in circumstances like those described in Bloggs’s story where the gain in relative value is large and the loss of neutral value small, the gain in relative value is great enough to

264

michael smith

outweigh the loss in neutral value. This is not the hypothesis that the relative values that attach to the weal and woe of our nearest and dearest are such as to give rise to deontological restrictions. But it is still a hypothesis that is inconsistent with the big ‘C’ consequentialist hypothesis that all value is neutral. It is perhaps worth adding that there is supposedly another, and much more well-known, difficulty with giving a rationale for the central features of a non-big ‘C’ consequentialist hypothesis. The difficulty is that brought out by G. E. Moore in his famous argument against ethical egoism (Moore 1903). Moore argues that since the concept of value is not an indexed concept—in other words, since there is no such thing as goodnessBloggs , but only goodness (unsubscripted)—we must reformulate both N and R. Principles that ascribe agent neutral value, principles like N, must all be reformulated using an unindexed concept of goodness, and principles that purport to ascribe relative value, principles like R, must all be abandoned. They can be given no coherent formulation at all once we reject the assumption that the concept of value is an indexed concept. But if the concept of relative value is, quite literally, incoherent, then that provides us with a decisive reason to reject the epistemological version of the nearest and dearest objection. As I said, Moore’s argument turns on whether, according to the best analysis of the concept of value, that concept is indexed or unindexed. Unsurprisingly, given that he takes the concept of value to be simple and unanalysable, Moore thinks that the concept is unindexed. But these days no one takes seriously the idea of there being a metaphysically simple property of goodness. Much the most widely held theory of value is rather some version of the dispositional theory (Smith 1994). According to the dispositional theory, to say that p is good is to say (roughly speaking) that we would desire that p if we had an ideal desire set. However, if something along these lines is the right analysis of the concept of value then it turns out that that concept is indexed after all (Smith 2002). Goodness is indexed to that group of individuals whose ideal desires are the truth-makers of evaluative claims: all goodness is goodnessthat group of individuals . Contrary to Moore, it is therefore at least coherent to suppose that there are both neutral values and relative values (Smith 2003). For if we had an ideal desire set then it is possible we would desire that all people benefit, independently of whether or not they bear any special relationship to us, and it is also possible that we would have an independent desire that our own children benefit. In this way we can see that it is at least coherent to suppose that both N and R are true.

nearest and dearest objection 265 The upshot is that Jackson’s response to the epistemological version of the nearest and dearest objection, the response that follows Kagan, misses its mark, and so too does Moore’s more well-known response to the epistemological version of the objection. The epistemological version of the nearest and dearest objection is therefore still very much on the table. This is not, of course, a serious objection to any of Jackson’s principle concerns in ‘Decision Theoretic Consequentialism and the Nearest and Dearest Objection’. But it does mean that, at the end of the day, we are yet to be told why we shouldn’t dismiss big ‘C’ consequentialism on the grounds that it is simply incredible. Princeton University

References Adams, R. 1976. ‘Motive utilitarianism’. In Journal of Philosophy 73: 467–81. Brink, D. 1989. Moral Realism and the Foundations of Ethics (Cambridge: Cambridge University Press). Hare, R. 1981. Moral Thinking (Oxford: Oxford University Press). Jackson, F. 1991. ‘Decision theoretic consequentialism and the nearest and dearest objection’, Ethics 101: 461–82. Jackson, F. and Pettit, P. 1995. ‘Moral functionalism and moral motivation’, Philosophical Quarterly 45: 20–40. Kagan, S. 1989. The Limits of Morality (Oxford: Clarendon Press). Mill, J. S. 1861: Utilitarianism (London: Fontana Library, 1962). Moore, G. E. 1903: Principia Ethica (Cambridge: Cambridge University Press). Parfit, D. 1984: Reasons and Persons (Oxford: Oxford University Press). Pettit, Philip and Michael Smith 2000. ‘Global consequentialism’. In B. Hooker, E. Mason and D. Miller (eds.), Morality, Rules, and Consequences: A Critical Reader (Edinburgh: Edinburgh University Press). Railton, P. 1984. ‘Alienation, consequentialism, and the demands of morality’. Reprinted in S. l. Scheffler (ed.), Consequentialism and its Critics (Oxford: Oxford University Press, 1988). 1988. ‘How thinking about character and utilitarianism might lead to rethinking the character of utilitarianism’. In P. French, T. Uehling, and H. Wettstein (eds.), Midwest Studies in Philosophy: Volume XIII. Ethical Theory: Character and Virtue (Notre Dame: University of Notre Dame Press). Rawls, J. 1951. ‘Outline of a decision procedure for ethics’. Philosophical Review 42: 177–97. Sidgwick, H. 1907. The Methods of Ethics (London: Macmillan).

266

michael smith

Smith, M. 1994: The Moral Problem (Oxford: Basil Blackwell). 2001. ‘Immodest consequentialism and character’, Utilitas (Special Issue on Consequentialism and Character) 13: 173–94. 2002. ‘Exploring the implications of the dispositional theory of value’, Philosophical Issues: Realism and Relativism 12: 329–47. 2003. ‘Neutral and relative value after Moore’, Ethics (Centenary Symposium on G. E. Moore’s Principia Ethica) 113: 576–98. 2004. ‘The structure of orthonomy’. Philosophy 55: 165–93.

11 The ‘Actual’ in Actualism1 JULIA DRIVER

The work of Frank Jackson has been important to at least two central debates in consequentialist ethical theory—those are the debates (1) between possibilism and actualism and (2) between objective consequentialism and expectabilism (or a variety of subjective consequentialism).² Suppose that we define the right action as that action which maximizes the good. Some writers, such as Michael Slote, have argued that this straightforward criterion is underdetermined. (See Slote 1992: 239–48.) Are we to maximize ‘actual’ good or ‘expected’ good? That is the debate between the objective consequentialist and the subjective consequentialist (with a caveat to be discussed later). There is also the issue of whether or not the agent who is deliberating considers what would be best given what will happen as opposed to what could or can happen. That is the debate between the actualist and the possibilist in determining relevant options for the moral agent to consider in deliberation. This essay explores differing answers to both of these questions, and then explores one strategy for answering both—a strategy which has been very much influenced by the work of Frank Jackson. Though I agree with Jackson on actualism I will disagree with him on expectabilism. My claim is that the definition of ‘right action’ is clear and not at all underdetermined—the right action just is the action that maximizes the good, the actual good. But what is often confusing is that the semantics of right is confused with an issue in the epistemology of right—that is the issue of determining how we are to go about doing the best that we can properly. ¹ I thank Roger Crisp, Mike Ridge, Walter Sinnott-Armstrong, and Roy Sorensen for helpful comments on an earlier draft of this paper. ² It’s worth pointing out, however, that the actualism/possibilism debate has significance for moral theory far beyond consequentialism. Any theory which holds that weighing options is crucial to moral decision-making will need to confront the issue.

268

julia driver

I. Jackson and Pargetter’s version of actualism The view that we are to maximize the actual good consequences is termed ‘actualism,’ and is contrasted with ‘possibilism’. Jackson and Robert Pargetter put the basic distinction the following way: By Actualism we will mean the view that the values that should figure in determining which option is the best and so ought to be done out of a set of options are the values of what would be the case were the agent to adopt or carry out the option, where what would be the case includes of course what the agent would simultaneously or subsequently in fact do: the (relevant) value of an option is the value of what would in fact be the case were the agent to perform it. We will call the alternative view that it is only necessary to attend to what is possible for the agent, Possibilism. ( Jackson and Pargetter 1986: 233)

To illustrate the distinction between these two positions, they ask us to consider the case of a procrastinator who is asked to write a book review. ‘Procrastinate’ knows that, ideally, the best thing for him to do is to say ‘yes’ and write the review on time; however, he also knows that he procrastinates, and will not write the review on time, in which case the best thing is to say ‘no’. The question is—what is the right thing for him to do, or what ought he to do? The Actualist will argue that he should say ‘no’, but on the possibilist position he ought to say ‘yes’. The actualist position is the one advocated by Jackson and Pargetter. There is a good deal of intuitive support for the view—denying it would seem to endorse the worse outcome.³ How does this connect with some of Jackson’s other published views on consequentialism? In his paper ‘‘Decision-theoretic Consequentialism and the Nearest and Dearest Objection’’ Jackson argues in favor of a form of expectabilism in that he argues that what the agent ought to do is determined by the agent’s own expectations of what will generate the most utility, and in this way he can explain agent-relative attitudes in consequentialist terms. On this view, the consequentialist ought to maximize ‘expected moral utility’. Jackson refers to this
as the ‘decision-theoretic’ approach: ‘‘... the rule of action is to maximize i Pr (Oi /Aj ) x V (Oi ), where Pr is the agent’s probability function at the time, V is consequentialism’s value function, Oi are the ³ Roger Crisp has pointed out to me that we could actually distinguish two further questions here—an agent-independent question and an agent-centered one. The first is the question of what one ought to do—the action that would or the action that could produce the best outcome? The second is the question of what one ought to do—ought one do what she believes will be best, or believes could be best? My take on this, to be argued for later in the paper, is that the actualism I favor be viewed as a realistic commitment on the part of the agent deliberating. So the focus is on the second question when it comes to discussions of agent deliberation.

the ‘actual’ in actualism

269

possible outcomes, and Aj are the possible actions.’’ (Jackson 1991, pp. 463–4) The agent does not necessarily aim for the best—and, indeed, Jackson argues that this view will frequently be incompatible with aiming for the best. In addition to what’s best, the agent will also need to consider the risks associated with each possible outcome and factor that into her calculations. If the ‘best’ is too risky, then we ought to avoid it. One case Jackson uses to motivate his view is the following: Consider Jill, a doctor, who has a patient, John, with a skin condition. She can prescribe either drug X, which has a 90 percent cure rate but a 10 percent fatality rate, or drug Y that has a 50 percent cure rate, but no bad side effects. What should she do? Well, she should not go for the best, since that is too risky. Instead, she should clearly prescribe drug Y even if that means forgoing the best outcome (Jackson 1991: 465). Because of the ambiguity in the initial characterization of actualism, what we term ‘actualism’ has often been contrasted with ‘expectabilism’ however. Michael Slote, for example, argues that the fact that Utilitarianism can be formulated these two distinct ways (i.e. as either actualist or expectabilist) shows that the theory is crucially underdetermined. Should Utilitarianism be understood as ‘‘... commending acts, motives, etc., whose overall effects or consequences are actually favorable to human happiness or as commending acts, motives, etc., whose overall effects are ... expectably favorable to such happiness?’’ (Slote 1992: 239). The first he terms ‘actualist’ the second ‘expectabilist’. I believe it likely that Slote is using ‘actualist’ differently than Jackson, and this may be the source of some unclarity in the literature. The contrast might best be understood differently. Actualism and expectabilism are compatible, and Jackson’s view is a form of actualist expectabilism. There are two ways in which ‘actual’ gets used in spelling out consequentialism. One contrast is between actual consequences vs. expected consequences, and the other contrast is between what will in fact, or will actually, happen and what, ideally, will or could happen (or, again more precisely, what the agent ought to do given what she will do versus what the agent ought to do given what she could do). If we look at the above case, then, the actualist says ‘no’ to the book review because the expected utility of saying ‘yes’, given what he knows about himself, and what he will actually do or what will actually happen, is less than the expected utility of saying ‘no’. The actual outcome of saying ‘no’ then is most likely better (that is, is expected by the agent to be better). These two points of contrast need to be kept distinct. One way to look at it is to note that actualism/possibilism marks a disagreement about how to determine the relevant options or alternatives open to the moral agent, whereas actualism/expectabilism marks a disagreement about what counts as the morally good outcome. Jackson is perfectly clear on this issue, though in the literature

270

julia driver

on Utilitarianism—as the quote from Slote’s discussion illustrates—there has been some unclarity caused by differing uses of ‘actualism’. All of this is by way of background. I intend to stick with Jackson’s usage. Jackson’s overall view is that the right action or what the agent ought to perform is the action with the highest expected moral value (by the agent’s own lights). Further, one must consider how one will in fact, or actually, act, in determining the expected value. There is another distinction we should be aware of, and how it relates to the earlier ones. That is the distinction between objective and subjective consequentialism. Jackson is quite critical of objective consequentialism—at least the sort advocated by Peter Railton. Railton puts the distinction this way: Subjective consequentialism is the view that whenever one faces a choice of actions, one should attempt to determine which act of those available would most promote the good, and should then try to act accordingly ... Objective consequentialism is the view that the criterion of the rightness of an act or course of action is whether it in fact would most promote the good of those acts available to the agent. (Railton 1984: 105)

Objective consequentialists believe that the right action is the one that brings about the best outcome; subjective consequentialists believe that the right action is the one that the agent tries to perform with the belief (or expectation) that it will bring about the best outcome. Note that this distinction marks a distinction between criteria of rightness, not decision procedures, though the subjective consequentialist believes rightness is determined by the consequentialist decision procedure, in effect. Railton also notes that it is only the subjective consequentialist that advocates the agent follow a consequentialist mode of deliberation. Objective consequentialism is neutral on this issue since it is an empirical issue as to what types of deliberation (if any) get us the best outcomes. Just as the objective hedonist can avoid the paradox of hedonism, the objective consequentialist can avoid the problem of alienation. Jackson clearly believes that the privileged ‘ought’—the appropriate focus of normative ethical theory—is what the agent ought to do ‘by her own lights’, and thus it is very tempting to view him simply as a subjective consequentialist. While he is opposed to objective consequentialism, the subjective/objective distinction may just be too crude to really capture the nature of the opposition. His decision-theoretic approach is one in which what the agent ought to do is understood ‘‘... in terms of what he or she ought to desire and does in fact believe ... ’’ as opposed to ‘‘... what ought to be believed.’’ The decisiontheoretic approach can be spelled out in a variety of ways, and Jackson’s way can probably be viewed as quasi-subjective. Jackson believes that we need to heed what we ought to desire, which is a function of the agent’s idealized

the ‘actual’ in actualism

271

preference function coinciding with the consequentialist value function. But, in addition to this, the agent’s actual, non-idealized, beliefs figure in—the agent determines what she ought to do based on her actual beliefs about what generates moral utility. To my mind this makes his view a form of subjective consequentialism, since what determines the moral quality of the action are the agent’s subjective states—what she believes. These are her actual beliefs, not the beliefs she could have if she were well informed, etc. But it is not really subjective consequentialism in the sense that Railton articulates. Jackson is not claiming that we need to always try to aim for the best. Indeed, quite the contrary. Jackson is aware that his approach needs to deal with a fairly standard worry. That is the worry about culpable ignorance. A critic of his proposal might argue that the agent’s ‘own lights’ might be rather dim. There’s a difference between what the agent believes to be best and what she would believe best if she attended responsibly to the evidence available at the time. Jackson thinks he can handle it by arguing that these cases simply illustrate that one sometimes also ought to get more information. I’m not sure this handles the worry, at least intuitively. It isn’t just that sometimes people don’t bother to get information in a responsible manner—some people sometimes don’t use the information responsibly either. For example, a person might be so prejudiced that he systematically misinterprets disconfirming evidence. Not only might the agent be dim or misinformed, but actually corrupt as well. This sort of consideration has led other expectabilists to qualify their claim: it isn’t just what the agent expects, but what the agent ‘reasonably’ expects which provides the criterion of rightness.⁴ If Slote’s sense of actualism is interpreted as a criterion of right action then it is a form of objective consequentialism, and expectabilism is one form of subjective consequentialism. Again, Jackson’s sense of actualism, however, is best understood as simply specifying the relevant sorts of factors restricting the options for the agent to consider. Jackson views his criterion of rightness as conforming to a decision-procedure. Indeed, he views that as a real strength of his view as against a weakness of objective consequentialism. Without the decision-procedure itself being basic, the moral theory lacks connection to the agent faced with moral decisions. I disagree with this and will be arguing that objective consequentialism is quite compatible with the aim of providing ⁴ In Contemporary Ethics: Taking Account of Utilitarianism William Shaw claims that most expectabilists are in favor of holding that it is reasonable, or corrected, expectations that figure into the criterion: ... most expected-outcome utilitarians believe that the utilitarian standard is expected happiness as calculated by a reasonable and well-informed agent based on the available evidence. (p. 30)

272 julia driver a decision procedure in moral theory. However, I agree that in making the appropriate sorts of decisions the agent should factor in what he sincerely believes he will do, not what he thinks he could do (unless, of course, those happen to coincide). Thus, there is the right act and then there is what the agent ought to do. These may come apart on my analysis. Again, this is because the semantics of right action should be distinguished from the epistemological issue of determining right.

II. The actualism/possibilism debate Again, it would seem that actualism has a good deal of intuitive support because the alternative seems to endorse the worse outcome. However, a number of philosophers disagree with the actualist position. Michael Zimmerman, for example, has argued in favor of possibilism, using the following case to generate possibilist intuitions: I have been invited to attend a wedding. The bride-to-be is a former girlfriend of mine; it was she who did the dumping. Everyone, including me in my better moments, recognizes that she was quite right to end our relationship; we were not well suited to one another, and the prospects were bleak. Her present situation is very different; she and her fianc´e sparkle in one another’s company, spreading joy wherever they go. This irks me no end, and I tend to behave badly whenever I see them together. I shouldn’t, of course, and I know this; I could easily do otherwise, but I don’t. The wedding will be an opportunity for me to put this sort of boorishness behind me, to grow up and move on. The best thing for me to do would be to accept the invitation, show up on the day in question, and behave myself. The worst thing would be to show up and misbehave; better would be to decline the invitation and not show up at all. (Zimmerman 2003)

The actualist would maintain that he ought to decline the invitation, whereas the possibilist would hold he should accept it. Zimmerman’s intuitions are that the possibilist is correct in this case. I don’t share those intuitions, but Zimmerman’s case rests on the view that the actualist allows people to get off the hook, morally, by appealing to their deficiencies.⁵ Further, he (as well as ⁵ According to Zimmerman (2003), one could characterize the positions in the following way. Actualism: Someone, S, ought to do some act A, if and only if S can do A and what would happen if S did A is better than what would happen if S did not do A. Possibilism: Someone, S, ought to do some act, A, if and only if S can do A and what could happen if S did A is better than what could happen if S did not do A. But these formulations have a problem with risk. So he suggests an account that might be able to avoid the problem for possibilism, a modification of World Possibilism. The modification holds: S

the ‘actual’ in actualism

273

many other writers) notes that actualism seems to be committed to the denial of: If S ought to do both A and B, then S ought to do A and S ought to do B. This he thinks is absurd, and he uses the following dialogue to illustrate: You, being an actualist, say to me, ‘‘What you ought to do is accept the Invitation, show up at the wedding, and behave yourself.’’ ‘‘OK, then,’’ I reply, ‘‘I’ll send in my acceptance.’’ ‘‘Oh, no,’’ you say, ‘‘you mustn’t do that!’’ What on earth am I to think? (Zimmerman 2003)

While there does seem to be a superficial oddity to this, I don’t think this oddity survives reflection. It will not work in any case where the goodness of B depends on A. S ought to do B, given A. Note that we are working within the framework of non-idealized expectabilism still, or subjective consequentialism. There may be other problems with this approach, but at this point my hope is to defend the actualist portion of Jackson’s position. Note that in combination with the decision-theoretic approach the picture is something like this: The best outcome is the one where one shows up at the wedding and behaves oneself, just like when one is in Las Vegas the best outcome is the one where one gambles and wins. However, the expected utility of the ‘best’ under ideal circumstances outcome is pretty low in the gambling case, and if, similarly, it is pretty low in the wedding case then one ought not to go to the wedding since one will, in fact, not achieve the ‘best outcome under ideal circumstances’. Let’s consider two different comparisons: 1. Compare the cases where I factor in what I can expect to actually happen. For example, I plan on throwing a party and it would be best if I could throw it in my beautiful garden on a sunny day. However, this would be worst in the event of a rainstorm—under those circumstances it would be far better to have the party inside. If I know it will rain, then, should I not arrange for the party inside (even though the best is having it outside under sunny skies)? Clearly, yes. However, of course possibilists will say this sort of calculation is entirely different from the sort they are concerned with, as in the Procrastinate case. They will argue that it may be fine to calculate the ‘behavior’ of objects, and the occurrence of certain events—but I cannot treat myself as an object or event—after all, I can control my behavior. I am an agent, not an object.⁶ ought at T to do A at T if and only if there is a possible world open to S at T that includes his doing A at T and whose ‘adjusted core prospective value’, relative to S at T, is greater than that of any possible world open to S at T that includes his not doing A at T . ⁶ For a discussion of this consideration see Thomason 1981.

274

julia driver

But now consider the second comparison: 2. Compare this with cases where the goodness depends on what others do. Sally ought to give 2% of her income to Oxfam, given everyone else does. If few other people do, she ought to give 10% of her income to Oxfam. The rightness of giving her 2% depends upon what others are doing or going to do, and of course, I can’t treat others as mere objects in my calculations either, can I? Well, the possibilist will respond that to the extent that you don’t have appropriate control over them, they are relevantly like events. The point, they will say, is that you are an agent, and you certainly have control over what you will do (with the appropriate caveats). However, this line of argument can be countered by a factor Jackson and Pargetter bring out in support of their claim. Consider the sort of advice that ought to be given to Procrastinate. Given that you know he is not going to write the review on time, you ought to advise him not to accept the invitation to write it, instead to say ‘no’. Given that you ought to advise him to say ‘no’ then it follows that he ought to say ‘no’. But this line of argument has been attacked. For example, Erik Carlson argues that suppose S has 3 options, A, B, C—A is the best, B second-best, C the worst. Suppose also that if you give any advice other than the advice to do B, S will instead do C. The possibilist would argue that S still ought to do A, though you ought to tell him to do B. To deny this is simply to beg the question against the possibilist. (Carlson 1995: 127). In a way these observations offer a way to slip agent-relativity into the account. The possibilist wants us to view the moral agent making his own decision as quite different from the agent advising others, or deciding what to do on the basis of how he thinks others will act. However, on my view this raises an interesting theoretical issue: it looks like we should, in effect on the possibilist view, be actualists of a sort when it comes to considering other events and the behavior of others, and possibilist when it comes to our own behavior (though possibilists would deny this by definition of the views). After all, this is what would allow Carlson’s argument against Jackson and Pargetter to have any force. The agent ought to do A, nevertheless he should be advised to do B. We are asked, then, to be realistic when it comes to factoring in what others do, or what other things happen, into our calculations; but not to do this when it comes to our own behavior over which we have some control. One way to view the divide is to note that possibilism seems to capture the perspective of the agent choosing and actualism the perspective of the observer advising. If this is right, then actualism does seem poorly suited to combination with expectabilism, which is supposed to privilege the perspective of the deliberating agent, the agent making the

the ‘actual’ in actualism 275 choice, in determinations of right. There’s no logical incompatibility. Jackson would simply note that the agent considers what she believes she will do in determinations of rightness. But this just exacerbates the problems that the possibilist raises for the actualist view. Actualism with expectabilism really does seem to let the agent off the hook. I am going to argue later that we just drop expectabilism, or the decision-theoretic approach, as a criterion of right. There is a good deal of intuitive support for the possibilist view, too, provided by even more extreme cases: for example, suppose that Roberta needs money and is thinking of robbing a bank. Roberta ought not to rob the bank at all, but if she does she ought not to use a gun that could harm innocent persons, or even lead to their death. If Roberta decides that she is just too weak and will rob the bank anyway (though she won’t use a gun)—well, then the actualist seems committed to the view that robbing the bank (albeit without a gun) is the right thing for her to do. This seems to allow way too much consideration for an agent’s weaknesses and imperfections. My position will reject this implication while admitting that given she is going to rob a bank she ought not to use a gun. That is because I will be arguing for a separation between right and ought.

III. Rejecting expectabilism as a criterion of rightness Is there any way for the actualist to avoid this problem? Well, one suggestion would be to abandon the task of formulating a decision-procedure according to which we would define right action, or determine right action. This is what the objective consequentialist does, claiming that what will be the best decision-procedure for the moral agent is just an empirical matter, and we should even expect it to vary from context to context. The distinction between evaluation and decision-procedure is at the heart of this issue. Jackson believes that the theory must provide a decisionprocedure—something the objective consequentialists have seemingly given up on. Of the two major problems he cites for objective consequentialism, this is one: ... we are dealing with an ethical theory when we deal with consequentialism, a theory about action, about what to do. In consequence we have to see consequentialism as containing as a constitutive part prescriptions for action. ( Jackson 1991: 466)

However, as Jackson himself notes, Railton and other objective consequentialists have not utterly abandoned the task of providing a decision-procedure

276

julia driver

for right action. Rather, they view the decision-procedure issue to be empirical: ... objective consequentialism sets a definite and distinctive criterion of right action, and it becomes an empirical question (though not an easy one) which modes of decision making should be employed and when. It would be a mistake for an objective consequentialist to attempt to tighten the connection between his criterion of rightness and any particular mode of decision making: someone who recommended a particular mode of decision making regardless of consequences would not be a hardnosed, non-evasive objective consequentialist, but a self-contradicting one. (Railton 1984: 117)

But Jackson dismisses this maneuver, noting that ... Railton’s proposal is ... that the moral decision problem should be approached by setting oneself the goal of doing what is objectively right ... and then performing the action which the empirical evidence suggests is most likely to have this property. ( Jackson 1991: 467)

But this, he argues, will give us the wrong answers. Consider again one of Jackson’s examples, that of Jill the doctor who must decide which medication to prescribe John who has a skin problem. She can treat him with medicine X that has a 90% cure rate, but a 10% chance of killing the patient. Or, she can treat him with medicine Y that has a 50% cure rate, but no bad side effects. Even though X has the highest probability of the best outcome, Y is clearly what ought to be prescribed. The play-it-safe option is the best, though it is guaranteed not to have the best consequences. Jackson regards this as a counter-example to Railton’s model. Thus, he regards the decision theoretic approach as a vast improvement. But the cases that Jackson cites are particular cases. One could hold, for example, a form of objective consequentialism that endorses something like the decision theoretic approach to making the right moral decisions. There is the right action, and then there’s what one ought to do under the circumstances. Ideally, we would like congruence. What is the best way to achieve that? It seems quite plausible that it is through the decision-theoretic approach. But this approach will only be justified if we believe it to produce the objectively right result overall. It might be helpful to think of this on analogy with debates on the nature of ‘justice’—is real justice a matter of outcome or procedure? Well, we can note a distinction between procedural and outcome justice, as long as the correct judicial process is followed we have procedural justice, though that may not always result in the just outcome. For example, a criminal may go free on a procedural flaw with the prosecutor’s case. So, clearly there is a distinction to be made.

the ‘actual’ in actualism

277

I don’t believe the objective consequentialist is committed to the view that Jackson attributes to Railton. He is misconstruing the objective consequentialist’s criterion, since the criterion of rightness he attributes to the objective consequentialist would apply no matter how good the ‘‘best’’ outcome is, and no matter how bad the side-effects are. But it is open to the objective consequentialist to consider degrees of good and bad in addition to raw probabilities.⁷ As I see it there are basically two ways for the objective consequentialist to go: (1) argue that Jackson has stuck the objective consequentialist with an overly narrow conception of the ‘best’, and/or (2) reject the decision theoretic approach as determining right action. In pursuing (1), for example, one might note that Jackson is himself opting to choose one kind of ‘rightness’ as central to ethical theory. He argues that what the agent ought to do (the right action) is the act that has the greatest expected moral utility, ‘‘... and so is a function of the consequentialist value function and the agent’s probability function at the time’’ (Jackson 1991, p. 471). But he notes that the probability function can change for that person over time and, further, will vary between individuals. He writes ‘‘I think that we have no alternative but to recognize a whole range of oughts—what she ought to do by the light of her beliefs at the time of action, what she ought to do by the lights of what she later establishes, ... what she ought to do by the lights of one or another onlooker who has different information ...’’ (Jackson 1991: 471). He believes that in spite of this array of ‘oughts’ what we typically mean by ‘ought’ is the ought which is ‘‘most immediately relevant to action’’ (Jackson 1991: 472). This privileged sense of ‘ought’ makes ‘right’. Objective consequentialists could also grant a profusion of ‘oughts’—‘oughts’ relative to different probability calculations and background information. So, the objective consequentialist could argue that what is ‘best’ and what the agent ought to do and what is right is to follow a decision-procedure which is the one which maximizes expected utility—given risk factors are taken into consideration in a calculation of utility and what qualifies as ‘best’—as that maximizes utility overall. Prescribing drug X leads to 10 deaths out of 100 prescriptions. Our decision-procedure ought not to be risk neutral. Thus, drug Y is the way to go because it really is the best. Prescribing this drug as a policy gives us the best aggregate outcome. However, this is not the way that I will go (though I do believe this is a promising strategy and worthy of further consideration). Here, I’d like to go with (2) instead. There will always be the need to sacrifice the prospect of the perfect for the sake of the good. That doesn’t change the fact that the perfect ⁷ I thank Walter Sinnott-Armstrong for pointing this out.

278

julia driver

is the best, and if it would certainly be achieved, it would be the way to go. Objective consequentialism gives us an account of the right action which is correct and points us in the direction of a decision-procedure, which may well turn out to be the one proposed by Jackson. Jackson’s skin rash case is simply another case that shows that sometimes we ought not to aim for what looks like the right thing. But this holds in non-moral contexts as well. Suppose one has the following choices: 1. Betting $10 in a $1 million lottery where the chances of winning are 1 in a million. 2. Betting $10 in a $999,999 lottery where the chances of winning are 1 in 2. Obviously the prudent choice is 2 though it means forgoing a chance at winning more money, or a chance at the maximally good outcome. But even though 2 is the best choice in this situation, this doesn’t mean that if I knew that I could get the million I should (prudentially) forgo that. That million is the best. From the prudential point of view, if I knew I would win the lottery I should go for the million. But, under the circumstances described in the choice situation above, I’d be crazy to go for the ‘best’ under those idealized circumstances. That’s because the real ones are not ideal. This simply highlights the need to have a theory that provides a criterion of evaluation that can serve as the basis for articulating a moral decisionprocedure, which is exactly what the objective consequentialist tries to do. Indeed, as I will argue at the end, I suspect that Jackson’s actualism goes rather well with the objective consequentialist project. But given that we stick to actualism we still seem to have the problem of Roberta’s robbing the bank as the right action for Roberta to perform, and this does seem counterintuitive. But this problem can be solved for the consequentialist, I claim, by rejecting expectabilism as an account of right action. The right action is the action that brings about the best state of affairs. We ought to go about aiming at the best prospect. The decision-procedure itself does not determine rightness, but it does track patterns of appropriate praise and blame. Now, compare the cases we have looked at so far to cases where the goodness depends on what others do. Sally ought to give 2% of her income to Oxfam, given everyone else does. If few other people do, she ought to give 10% of her income to Oxfam. The rightness of giving the 2% depends upon what others are doing or going to do. If they aren’t going to give (though they ought to) the right thing for her to do is give 10%. That’s the best she can do. She ought to give 10% given she realizes that others will not be giving their

the ‘actual’ in actualism

279

fair share—the share that they could reasonably give. Doesn’t it make sense, then, to suppose that whether or not the spurned boyfriend ought to go to the wedding depends upon what he will do? As in handling Jackson’s advice argument, a possibilist, however, might note that there is a crucial difference between first and third person perspectives, the perspective of the agent as opposed to the observer. However, both of these standards are relevant only to the epistemological issue of what criterion to use in determining how best to proceed. Given he will misbehave if he goes, he ought not to go to the wedding and it would be the wrong thing for him to do. This is just a reminder that this issue needs to be separated from the issue of understanding what is the right action, or, rather, what ‘right action’ means. Under the circumstances as described, the right thing for the spurned boyfriend to do is not go to the wedding. In determining what he ought to do he should again try to think in terms of what he believes will happen—if he believes he will misbehave, that’s an important reason not to go. The actualist decision-procedure simply incorporates more realism into the moral decision-making process.

IV. Actualist decision-procedure, objective criterion of rightness My suggestion is that we define right action as that action which produces or results in the best state of affairs (given the relevant alternatives). However, what we ought to do is to be guided by a decision-theoretic procedure committed to actualism in specifying the relevant alternatives. My arguments for the first have primarily been negative—simply defending the view against Jackson’s claims. I have also tried to provide similar support for the latter claim, but also would like to provide a positive argument that speaks to the advantage of this ‘‘hybrid’’ approach, combining actualism in a decision-procedure with an objective consequentialist account of ‘right’. ‘Best’ is central to our understanding of ‘right’. The right thing will be the best thing among available options. But what someone ought to do is (quite likely) what she believes has the best pay-off relative to what she believes, sincerely, will happen. It is this latter sense of ‘ought’ which invites praise and blame. But the former sense is, nevertheless, instructive. Awareness of the best option serves some guiding function even if one is not aiming for the best. It is a mistake to think that the only way one is guided by some end is through aiming at the end. Here is just one of many possible cases to illustrate

280

julia driver

this: Maria realizes that, though she did fairly well by deciding to major in communication, the best choice for her would have actually been psychology. She wishes now that she had decided to major in psychology. This motivates her to make further career choices sensitive to that preference. Though it is too late to re-do her major, she can still seek employment in areas more amenable to her interest in psychology. This approach has the advantage of making sense of ‘contrary-to-duty’ imperatives. These are imperatives of the form ‘‘Don’t do x, but if you do x, don’t also do y.’’ Remember the case of Roberta, a case that posed a major intuitive difficulty for the actualist approach? Well, this hybrid approach would hold that it would be wrong for Roberta to steal, it would not be the right thing for her to do. However, given that she is going to steal, she ought to follow a decision-procedure that minimizes harm. Thus, she ought not to use a gun, given that she is going to steal. In this context, it makes sense to say ‘‘She ought not to use a gun.’’ But, given my account, this does not make stealing the right thing for her to do even if she does not use a gun. Thus, I believe this hybrid approach can avoid Zimmerman’s problem, which holds that it lets the moral agent off the hook. It doesn’t. It just offers a nuanced evaluation. Dartmouth College

References Carlson, E. 1995. Consequentialism Reconsidered (Dordrecht: Kluwer Academic Publishers). Jackson, F. 1991. ‘Decision-theoretic consequentialism and the nearest and dearest objection’, Ethics 101: 461–82. Jackson, F. and Pargetter, R. 1986. ‘Oughts, options, and actualism’, Philosophical Review 95: 233–55. Railton, P. 1984. ‘Alienation, consequentialism, and the demands of morality’, Philosophy and Public Affairs 13: 134–71. Shaw, W. 1999. Contemporary Ethics: Taking Account of Utilitarianism (Malden, MA: Blackwell). Slote, M. 1992. From Morality to Virtue (Oxford: Oxford University Press). Thomason, R. 1981. ‘Deontic logic and the role of freedom in moral deliberation’. In R. Hilpinen (ed.), New Studies in Deontic Logic (Dordrecht: Reidel). Zimmerman M. 2003. ‘The relevance of risk to wrongdoing’. ISUS, Lisbon, April 2003.

PA RT I V

Conditionals and the Purposes of Arguing

This page intentionally left blank

12 Conditionals, Truth, and Assertion DOROTHY EDGINGTON

1. Introduction When is an indicative conditional, ‘If A, C’, true? It will be agreed that it is false if A is true and C is false. Beyond that, there is no consensus. The truth-functional theory says that it is true in all other circumstances, that is, true provided that it is not the case both that A is true and that C is false. Other theories say that more is required for truth than that: it may or may not be true when it’s not the case both that A is true and that C is false. There are apparently compelling arguments for each of these two alternatives. For the first and against the second: if you know merely that A and B are not both true (nothing stronger), this is enough to conclude that if A is true, B is not; and if you know merely that at least one of them is true, that is enough to conclude that if A is not true, B is. These facts are (on the face of it) incompatible with the claim that something stronger than the satisfaction of the truth-functional truth condition is required for the truth of the conditional. For the second and against the first: suppose you and I agree in thinking that the Labour Party won’t win. That leaves room for disagreement between us about what will happen if they do win. We do not take our common opinion that the antecedent is false to commit us to thinking that any conditional with that antecedent is true, as we should if the truth-functional account of the truth conditions of indicative conditionals were correct.¹ Frank Jackson advanced a novel, subtle and ingenious theory of indicative conditionals, the ‘Supplemented Equivalence Theory’, designed to escape this quandary. The theory appeared first in 1979, in subsequent articles, and most ¹ Much more could be said, of course, on each side; but these cases are central, and the source of most of the other difficulties.

284

dorothy edgington

comprehensively in his book (1987)². On his view, all that’s required for the truth of a conditional is that the truth-functional truth condition be satisfied. But more is required for one to be justified in asserting ‘If A, C’, and this is a matter of the meaning of ‘if’. There is more to meaning than truth conditions. Analogy: ‘A but B’ has the same truth condition as ‘A and B’, yet they do not mean the same. It is a fact about the meaning of ‘but’, stated in dictionaries, that it indicates a contrast between the propositions it conjoins. Similarly, says Jackson, it is a fact about the meaning of ‘if’ that when you say ‘If A, C’ you indicate not only that you believe the truth-functional conditional (A ⊃ C), but also that your belief is ‘robust’ with respect to the antecedent. Exactly what robustness is will be discussed shortly, but it is meant typically to imply that you wouldn’t give up your belief that (A ⊃ C) if you learned that A. Therefore, you don’t believe (A ⊃ C) just because you think the antecedent is false; for if you did, and then learned that A, you would change your mind about the conditional, rather than deduce the consequent. Robustness makes asserted conditionals fit to be used in inferring by modus ponens, for they survive the learning of A, and hence the inference to C. Jackson respects the tradition which associates conditional judgements with judgements of conditional probability. As Ramsey put it (1929), to assess a conditional, you add the antecedent hypothetically to your stock of knowledge, and then consider, under that supposition, what you think about the consequent: in doing this people are ‘fixing their degrees of belief that [C] given [A]’ (Ramsey, 1990, 155). By ‘degree of belief in C given A’ Ramsey meant the conditional probability one gives to C, on the supposition that A, p(C|A), read ‘the probability of C given A’, which obeys the law: p(A & C) = p(A) × p(C|A).³ For Jackson, this demonstrably gives the right ‘assertibility’ condition for a conditional: it is assertible to the extent that one’s conditional probability of consequent given antecedent is high. In Jackson’s first article, the notion of a belief’s being robust with respect to a proposition A, is introduced as meaning that the belief would survive the learning of A. A second characterization of robustness followed: he said ‘If we accept Conditionalization, the plausible thesis that the impact of new information is given by the relevant conditional probability, then ‘‘B is robust with respect to A’’ will be true just when p(B) and p(B|A) are close and high.’ (1979: 115–16; notation changed slightly). By the time he wrote his book, ² When not otherwise indicated page references are to this book. ³ The thought experiment is as follows: hypothetically eliminate the ¬A-possibilities, and assess the probability of C in the hypothetical distribution of probabilities consisting of just the A-possibilities. This amounts to comparing p(A & C) and p(A & ¬C). If you think that A & C is about 10 times more likely than A & ¬C, you think it is about 10 to 1 that C on the supposition that A.

conditionals, truth, and assertion

285

however, David Lewis had pointed out that these two ways of characterizing robustness can come apart (see Lewis 1986: 155–6); and while much of the argumentation continues to use the first notion (which gives point to the label ‘robust’), the second becomes the official definition of robustness. I discuss the divergence below in Section 3. Suppose, then, that you judge B to be highly probable, so count as believing that B. This belief is robust with respect to A if you also think B highly probable given A. Thus, a high degree of confidence in a material conditional, A ⊃ C, is robust with respect to its antecedent, A, if p((A ⊃ C) given A) is high. But p((A ⊃ C) given A) = p((¬A ∨ C) given A) = p(C given A). So the material conditional is robust with respect to its antecedent, hence assertible, if and only if one’s p(C given A) is high. Moreover, a high p(C given A) guarantees a high p(A ⊃ C)⁴. So this assertibility rule captures both high probability of truth and robustness with respect to the antecedent. This assertibility condition he calls ‘(Adams)’, after Ernest Adams, who has done most to develop a theory of conditionals based on conditional probabilities. The name is not entirely appropriate. Adams did use ‘assertibility’ in this context in his earliest writings on conditionals (1965, 1966), but by the time he wrote his book (1975), he had abandoned talk of assertibility, as not the right way to formulate his thesis. Adams’s mature formulation of the thesis rather concerns degrees of belief, or degrees of closeness to certainty. Idealizing by making these degrees precise enough to be represented by numbers, Adams’s thesis, and Ramsey’s, is this: just as your degree of closeness to certainty in an unconditional proposition, A, is your subjective probability for A, so your degree of closeness to certainty that if A, C is your conditional subjective probability for C given A. Jackson cannot accept this Ramsey–Adams thesis. To believe a proposition is to believe that it is true, to judge it to be probable is to judge that it is probably true. A conditional ‘If A, C’, Jackson holds, is true iff A ⊃ C; so one should believe it or judge it to be probable to the extent that p(A ⊃ C) is high. But p(A ⊃ C) can be high while p(C given A) is low. For instance, I’m 90% certain that Jane won’t be offered the job, 9% certain that she will be offered it and accept, 1% certain that she will be offered it and decline. p(offered ⊃ declines) = p(not offered, or (offered and declines)) = 91%. p(declines given offered) = 10%. ⁴ p(A ⊃ C) is high to the extent that p(A & ¬C) is low. p(C|A) is high to the extent that p(A & ¬C) is low relative to p(A). Except in two special cases where they are equal (when p(A) = 1 or when p(A & ¬C) = 0), p(A & ¬C) is a larger proportion of p(A) than it is of 1. So p(C|A) < p(A ⊃ C) except in the two special cases when they are equal.

286

dorothy edgington

It is not only the truth-functional conditional the probability of whose truth can part company with the conditional probability of C given A. Lewis (1976) showed that a conditional probability does not measure the probability of the truth of any proposition at all: there is no proposition X such that we can systematically relate a person’s degree of belief in C given A to their degree of belief that X is true. (We shall see later (Section 4) how the probability of a ‘close-possible-worlds’ conditional also systematically diverges from a conditional probability.) The discovery that a conditional probability is not the probability of the truth of a proposition, together with the conviction that a conditional probability is the best measure of one’s degree of confidence that if A, C, led Adams and his followers to deny that conditionals express propositions, or state facts, or are evaluable in terms of truth and falsity. The ‘A’ and the ‘C’ in ‘If A, C’ play different roles. ‘A’ expresses a supposition. ‘C’ expresses a belief, or makes a statement, within the scope of that supposition. And conditional beliefs or statements, Lewis’s proof shows, do not reduce to unconditional beliefs or statements about how things are. Still, if one ties conditional probabilities to assertibility, one may consistently hold that conditionals have truth conditions, but, for special reasons, they are not always assertible when these truth conditions are believed to be satisfied. This is the shape of Jackson’s theory. As a gentle reminder that the label is not entirely appropriate, I shall use double quotation marks: ‘‘Adams’’ is the thesis that a conditional, ‘If A, C’ is assertible to the extent that one’s conditional probability of C given A is high.

2. The Gricean strategy The illustrious predecessor of a theory of the shape of Jackson’s is that of H. P. Grice (1967, published much later (see 1989) but influential and much discussed in the intervening decades); and the more immediate predecessor was David Lewis’s (1976: 86–9) attempt to derive ‘‘Adams’’ from Grice’s ideas. Grice also defended the truth-functional truth condition for indicative conditionals, and also offered an explanation of its awkward features in terms of the fact that we would not assert ‘If A, C’ just on the ground that ¬A, or just on the ground that C. But Grice’s explanation did not appeal to a special feature of the meaning of ‘if’: for Grice ‘If A, C’ means the same as ‘A ⊃ C’. Rather, his explanation is in terms of general principles to which people are expected

conditionals, truth, and assertion

287

to conform in conversation, the ‘conversational implicatures’ to which they give rise, and the consequent fact that one can speak truly but misleadingly by violating these principles. The principle which is relevant to the present issue is (∗ ) Make your contribution as informative as is required (for the current purposes of the exchange). (1989: 26) As Jackson glosses it, ‘Assert the stronger rather than the weaker (when you have sufficient grounds for the stronger, and when the stronger is relevant)’. As you are expected to conform to (∗ ), you will mislead your audience if you do not, even when you speak truly. I am asked how many were at my lecture today and I say ‘Less than 100’. Three people attended. I spoke truly but left you with the wrong impression. Similarly if I say ‘I’ve just seen part of the butler’s body in the cellar’, when I saw the butler going about his everyday business there. Closer to our topic: I am asked where John is. I know he is in the pub, and that he never goes near libraries. I say ‘He’s either in the pub or the library’. I speak truly but mislead you into thinking this is all the information I have. And an example of Lewis’s (1976: 87): ‘You won’t eat those and live’, I say of some wholesome and delicious mushrooms, knowing that you will now leave them alone, deferring to my expertise. You don’t eat them, so what I said was true. Similarly for conditionals, according to Grice: if I say ‘If he’s not in the pub he’s in the library’ knowing that he is in the pub, and knowing that he never goes to libraries, or if I say ‘If you eat those you will die’, knowing that you won’t eat them, and that they are harmless, I speak truly (and knowledgably) but mislead; for why say something equivalent to ‘A or B’ when all I know of relevance is A, and I could just as well have said A? My audience will assume that I am not violating (∗ ), hence that I could not just as well have made the stronger statement, and so they will be misled. Conversational implicatures are inferences one makes not solely from the content of what is said, but also from the fact that someone says it, given the presumption that they are observing the principles of conversation, i.e. being helpful. Grice also had the notion of conventional implicature: what is suggested, indicated, but not strictly said by the use of such words as ‘but’, ‘although’, ‘nevertheless’; and it is to these words that Jackson assimilates ‘if’. He has several criticisms of the Gricean defence of the truth-functional truth condition. First, the Gricean story would seem to militate against the assertion of any conditional when you are pretty sure the antecedent is false, for in that case you could just as well have said ‘¬A’. But some conditionals whose antecedents are disbelieved are assertible, others not. Jackson’s example: ‘If the sun goes out of existence in ten minutes’ time, the earth will be plunged into darkness

288

dorothy edgington

in about eighteen minutes’ time’. In this case, ¬A is very probable, hence ‘¬A or C’ gets almost all its probability from ¬A; so ‘assert the stronger rather than the weaker’, pronounces in favour of asserting ¬A rather than ‘If A, C’. Yet the conditional is assertible. I think Gricean principles can handle these cases, and that it is only an over-simplistic and too formulaic reading of Grice on which they cast doubt. In such cases, although you judge the antecedent to be very improbable, this isn’t your only reason for believing that A ⊃ C i.e. ¬A ∨ C. You have other reasons, which are independent of your attitude to the antecedent. And this makes the conditional worth asserting. Your audience will assume correctly that your belief is not solely due to disbelief in the antecedent, and will not be misled. Put differently: even if you didn’t disbelieve the antecedent you would still have grounds for believing the conditional. This is very like Jackson’s own view, but it does not seem to require a departure from the Gricean story. Second, conditionals whose high probability is almost entirely due to the probability of the consequent may be assertible, says Jackson. If I’m sure that the match will be cancelled because all the players are ill, I can say that the match will be cancelled whether or not it rains, i.e., if it rains it will be cancelled and if it doesn’t rain it will be cancelled. Well, yes, but Grice still has a point here: in this situation it would be misleading in precisely the way Grice highlights, in an enquiry about the likelihood of the match being played, simply to assert ‘If it rains, the match will be cancelled’; that is, unless you make clear that you’re trying to get across that C whether or not A, in a context in which your hearer might wrongly think that C depends on whether A, it is misleading to assert ‘If A, C’ when you’re in a position to assert C. There is a related point which seems to tell in Grice’s favour against Jackson. Suppose you believe A and C. Then it will usually be misleading for you to assert ‘If A, C’, for this is weaker than what you are in a position to say. For instance, you are asked who is coming to the party. You know that John is coming, and you know that Mary is coming, and you also know that there is no relevant connection between these two facts. It is misleading to say ‘Mary is coming if John is’. But for Jackson, this is assertible: a high p(A & C) guarantees a high p(C|A). And given that you already believe both conjuncts, trivially, you would believe the second if you were to believe the first. This point is decisive against the derivability of ‘‘Adams’’ from Gricean considerations, which Lewis attempted. If, on the other hand, conditional probabilities measure your confidence in a conditional (rather than its assertibility), we merely find, as we should expect to find, that there are cases (very boring cases) in which you can be confident that if A, C, but it would be misleading, for Gricean reasons, to say this.

conditionals, truth, and assertion

289

At first sight, this case looks likes a refutation of Jackson’s account of assertibility. But it emerges that it is not. Jackson is not aiming to explain the same data as Grice: he is operating with a different notion of assertibility. (He introduces two spellings, ‘assertability’ (for Grice’s notion) and ‘assertibility’, to distinguish the two senses.) Jackson abstracts away from considerations of what would be misleading to assert in conversation. If you judge A to be highly probable, and you are not breaking any linguistic rules concerning terms carrying conventional implicature, A is assertible in Jackson’s sense, though it may not be ‘assertable’ in Grice’s sense. So Jackson will accept that if I’m sure that A & B, ‘If A, B’ is assertible, in his sense of the term. Jackson’s third objection concerns the fact that some statements that are tautologies with ‘if’ interpreted as ‘⊃’, strike us as nothing of the sort, for instance many instances of ‘(If A, B) or (if ¬A, C)’. Since all tautologies are maximally weak, the difference between those that strike us as correct and those that don’t, can’t be explained by the principle ‘Assert the stronger rather than the weaker’. Now, as both Grice’s and Jackson’s accounts focus on when it is correct to assert a conditional, neither has any direct consequences for the role of conditionals as they appear, unasserted, as parts of longer sentences. Grice might nevertheless reply that it is a natural corollary of the fact that one cannot properly assert a conditional on the sole ground that its antecedent is false, that one cannot properly assert a disjunction of conditionals on the sole ground that one disjunct has a false antecedent. Grice might also reply that outside a special context such as a logic class, tautologies tend not to be asserted at all, and if one is asserted, we tend to look for something non-tautologous which is implicated: ‘Business is business’. That is, tautologies are not straightforwardly assertible (in Grice’s sense of the term), for reasons Grice can explain. This third objection is a special case of his fourth, which is that Gricean principles cannot explain why, among logically equivalent sentences, some are assertible and some are not (for logical equivalents are equal in strength). For example, sentences of the form ‘¬A, and if A, B’ are logically equivalent to ‘¬A’; yet we assert some conjunctions of this form: ‘He won’t take the job—and if he does, he’ll regret it’; and of course there are others which we would not assert. Grice can reply that the assertion of a conjunction implicates your willingness to assert both conjuncts, and you shouldn’t assert the second conjunct just on the grounds that ¬A. But there are other cases which are more difficult for Grice. For instance, we say things like ‘If it rains, it won’t rain much’, but not its contrapositive: ‘If it rains much, it won’t rain’. What can Grice say about this? The second is plain unassertible, because it can only be true by having a false antecedent. Nobody could accept it except on the grounds that the

290

dorothy edgington

antecedent is false, and that is not a good reason to say it. The puzzle is why we should assert the first, rather than the simpler and equivalent ‘It won’t rain much’. Well, an assertion of ‘It won’t rain much’ perhaps implicates that it will rain, though not much. Compare ‘I won’t spend much at the sale’, which carries the suggestion that I will spend a little. You want to indicate that you are not committed to rain. You could say ‘Either it won’t rain at all, or it won’t rain much’; that the first disjunct does not strike us as otiose confirms the implicature from ‘It won’t rain much’ to ‘It will rain’. Equivalently, you could say ‘If it rains, it won’t rain much’. Note that the implicature which makes this a sensible thing to say has nothing to do with conditionals. Again, independently of conditionals, Grice can allow that logically equivalent sentences may have different implicatures. ‘Some tigers are dangerous’ suggests, but does not say that some tigers are not dangerous. ‘Some dangerous things are tigers’ does not suggest that some tigers are not dangerous. (These examples are Jackson’s own.) So far, I have argued that Jackson’s criticisms of the Gricean strategy are not compelling. They are compelling only against a rigid and formulaic application of ‘Assert the stronger rather than the weaker’ as the whole truth about the assertibility of conditionals. This was used by Lewis in his attempt to derive ‘‘Adams’’ from Gricean principles. But this is an unsympathetic reading of Grice’s ideas. Grice had a battery of principles to draw upon, which can interact, conflict and be overridden, with which to tell an intelligible story about why a particular sentence is assertible or not. Jackson’s final objection, however, is in my view convincing. He observes that if Grice were correct, the inference: ‘A; so A or C’ should strike us as just as bad as the inference ‘¬A; so if A, C’. And it does not. Grice’s story rings very true for many examples including disjunctions, yet seems far-fetched for conditionals. One cannot help thinking that Grice is pushing an excellent idea too far. To put this asymmetry differently: if I say ‘He is either in Oxford or London’ and it turns out that he is in Oxford, we agree that my assertion has been verified; if I say ‘If you touch that wire you will get a shock’ no one takes the fact that you don’t touch the wire as establishing that what I said was true. Grice’s general thesis about conversational principles is correct and valuable: believing that something is true (on good grounds) is not always a good reason to say so (in fact it is never, by itself, a good reason to say so); and you can seriously mislead your audience by ignoring maxims of conversational helpfulness, for they will assume that you did have good reason to say so. But it cannot be stretched to defend the truth-functional conditional, simply because there is no convincing case that you have good grounds for believing that it is

conditionals, truth, and assertion

291

true, when your sole evidence is that the antecedent is false. In locating the problem as one about whether it is a reasonable thing to say, in conversation, Grice places it at too superficial a level. No one takes as grounds for believing a conditional, or for establishing it as correct, that its antecedent is false.

3. The two concepts of robustness Having stressed the importance of robustness, Jackson argues that in the case of conditionals this must be a conventional, rather than a conversational matter: Why did the theory have to be given in terms of a conventional signal of a material conditional’s robustness with respect to its antecedent, instead of in terms of the presumption that conversational exchanges are directed at the transfer of useful, interesting information? The answer turns on how much we can properly lay at the door of this presumption ... [I]f we worked from the presumption alone, we should equally expect someone who asserted ‘If A, B’ to have evidence for ‘A ⊃ B’ that stood up to learning A, and to have evidence that stood up to learning ¬B. [For ‘A ⊃ B’ is equivalent to ‘¬A ∨ B’; disjunction is symmetric, and it is a conversational presumption that disjunctions are typically robust with respect to the negation of each disjunct.] Now very often we do have such expectations, nevertheless, as we have emphasised a number of times, the situation is not symmetrical. It is never (well, hardly ever) right to assert ‘If A, B’ when your evidence for (A ⊃ B) would not survive learning A, but sometimes right to assert ‘If A, B’ when your evidence would not survive learning ¬B ... witness ‘If it rains, it won’t rain much’. The conversational presumption cannot explain this asymmetry. We perforce must lay it at the door of linguistic convention, and to do that is ipso facto to put it in the category of conventional implicature. (96)

Note the ‘well, hardly ever’ a few sentences back. The alleged asymmetry is between what it is hardly ever right to assert and what it is sometimes but not often right to assert. This is not a very pronounced asymmetry! And I shall argue that the ‘hardly ever’ underestimates the exceptions. ‘Hardly ever’ is an acknowledgement of a difficulty pointed out by Lewis (1986, 154–6), for Jackson’s (1979) assumption that two ways of characterizing robustness coincide. Lewis distinguishes ‘robust1 ’ and ‘robust2 ’: B is robust1 with respect to A iff p(B) and p(B|A) are close, and both are high. B is robust2 with respect to A iff p(B) is high, and would remain high even if one were to learn that A.

The most famous kind of counterexample to the alleged equivalence of robustness1 and robustness2 is due to Richmond Thomason (reported by Bas van Fraassen 1980, 503). There are many things we are likely never to know;

292

dorothy edgington

consequently, there are many things I’m prepared to assert of the form ‘If A, I’ll never find out that A (or, no one will ever know that A)’. For example, if there was a chaffinch on this lawn a hundred years ago today, I’ll never find out (or: no one will ever know). I have a high probability for the consequent given the antecedent. But were I, surprisingly, to learn that the antecedent is true (say, by coming across a bird-watcher’s diary that inspires total confidence), I would not come to believe that I will never find out that it is true! On the contrary, I would reject my previous conditional belief. In a footnote attached to the ‘well, hardly ever’, Jackson describes these ‘odd exceptions’ as counterexamples to Conditionalization. They are not counterexamples to Conditionalization. Conditionalization is a principle which says that if at a given time your p(C|A) = x, and then you learn that A and nothing else of relevance to C, your new p(C) should be x. In this case, as well as learning that there was a chaffinch on the lawn, you learn something else of relevance to the consequent, namely that you have learned that there is a chaffinch on the lawn, so Conditionalization is inapplicable. For many conditionals, this particular extra information, that you have learned the antecedent, is irrelevant to the consequent. But it is not irrelevant when the consequent is ‘I’ll never find out’! If the above were a counterexample to Conditionalization, so would this be: I am about to pick a card and look at it. I know that the pack contains three kings, one red and two black. My degree of confidence that I will pick a red card, given that I pick a king, is 1/3. But if I were to pick a king, I would not have a degree of confidence of 1/3 that it is red, but a degree of confidence of either 1 or 0 (or close enough) that it is red: I am not colour blind. The only difference between the two examples is that in this case it is glaringly obvious that I have gained extra relevant information, in addition to the fact that I picked a king, while in the first it is easy to overlook the fact that I have gained extra relevant information. But now, when someone has a high conditional probability for C given A, and subsequently learns that A, it’s not going to be that uncommon to acquire extra information relevant to C, along with A, which may prevent them from believing that C. Two features of these two concepts of robustness, both integral to Jackson’s proposal, exacerbate the difficulty: the facts that ‘robust2 ’ is defined in terms of a subjunctive conditional, and ‘robust1 ’ requires only a high probability for C given A. Jackson insists that ‘robust2 ’ be explicated in terms of a subjunctive conditional, when considering an imagined objection that his account is circular, explaining conditionals in terms of conditionals. He replies that he is explaining indicative conditionals in terms of subjunctive conditionals, which we know to

conditionals, truth, and assertion

293

be different (33). And he explicitly adopts a Lewis-style account of subjunctive conditionals: ‘(A
B) is true at w if and only if, among the worlds where A is true, all the most similar or closest to w are worlds where B is true’ (63). Note that for it to be false that B is robust2 relative to A, we do not need that you would not believe B if you learned that A, but merely that you might not believe B if you learned that A, i.e., among the closest learning-A-worlds are some in which you do not believe that B. Combine that with the fact that ‘robust1 ’ is a probabilistic notion. The probability of B given A is high, but not necessarily 1. Someone with a high probability for B given A typically has a non-zero probability for A & ¬B: it may be the case that A & ¬B, though they judge this to be much less likely than A & B. But if they think it may be the case that A & ¬B, they may very well think that there are some close ways they might learn A which are A & ¬B-ways; and hence, that if they were to learn A, they might not believe B. So they are not confident that if they were to learn A, they would believe B (on Lewis’s interpretation of the subjunctive). Two examples: 95% of the balls with black spots are red. I am about to pick a ball and look at it. I have a high conditional probability for red given a black spot. But I think there are some close worlds in which I learn that I pick a black spot, and see that it is not red. I do not believe that in all close worlds in which I come to believe that I picked a black spot, I will believe that it is red. So again, the conditional is robust1 but not robust2 with respect to its antecedent. Or: I think that if John enters this room, he will be accompanied by Mary. Interpreted as a judgement of conditional probability, I am about 95% confident of this. A John-without-Mary entering of the room is epistemically possible, but unlikely. But, as in the previous case, there are, as far as I know, some close worlds in which the unlikely happens: if I were to learn that John is entering the room, it might be by seeing John entering the room alone. So I do not accept the subjunctive conditional (read according to Lewis). The trouble here lies with Lewis’s theory of subjunctive conditionals, applied to cases involving uncertainty. Intuitively, in the above examples, I am highly confident (though not certain) that I would come to believe the consequent, if I were to learn that the antecedent is true. Judgements expressed as subjunctive conditionals also need to be treated in terms of conditional probability, in my view. These matters are discussed further in the next section. Two further points are worth mentioning. Jackson and Lewis write as if what matters for robustness2 is that you would still believe B, if you acquired the belief that A. But as this is something that you signal when you assert the conditional, what really seems to matter is whether you think you would. The point generalises to other cases of conventional implicature: suppose I’m under the misapprehension that Fred almost invariably hates parties and say ‘Even

294

dorothy edgington

Fred enjoyed the party’. I am not making a linguistic error, an error which casts doubt on my understanding of the word ‘even’; or, having the misguided opinion that there is a contrast between poverty and honesty, I do not manifest a misunderstanding of ‘but’ by saying ‘She is poor but honest’. What matters for the correct use of the English word ‘but’ is that you think there is a contrast between the conjuncts; similarly for conditionals and robustness. Put in terms of whether I would, rather than whether I think I would, believe B if I learned that A, there will be many more counterexamples of conditionals which are robust1 but not robust2 , because the world is not as I think it is: ‘If you press that switch, the light will go on’, when unbeknownst to me, the power is off. Finally, Jackson does not distinguish between a normative and a descriptive reading of Conditionalization. Let us grant that if one has a high conditional probability for C given A, and then learns that A and nothing else of relevance to C, one should then have a high probability for C. But as a matter of fact, people sometimes change their beliefs in irrational or non-rational ways. They have optimistic days and pessimistic days. Or consider Mary, who is an atheist. For pretty well any proposition A, which she takes to be epistemically possible, she accepts ‘(Even) if A, God does not exist’. But if certain tragic misfortunes were to befall her or her loved ones, Mary would acquire a belief in God (and she may even have enough self-knowledge to have some suspicion that this might be true of her). As Ramsey put it when introducing the idea of conditional degrees of belief, ‘This is not the same as the degree to which [one] would believe [C], if he believed [A] for certain; for knowledge of [A] might for psychological reasons profoundly alter his whole system of beliefs’ (1926, in 1990: 76). In sum, the relation between robustness1 and robustness2 is far from tight, primarily because it is normally at least possible to learn other things relevant to the consequent, along with the antecedent. So the alleged asymmetry which Jackson gives as the reason for ‘going conventional’ is hard to discern. When someone believes that if A, C, then normally, should they acquire the belief that A, they will conclude that C; and normally, should they acquire the belief that ¬C, they will conclude that ¬A. But there are exceptions in both directions. There is, however, an asymmetry in conditional probabilities: supposing that A, and judging C to be probable under that supposition, is not equivalent to supposing that ¬C, and judging ¬A to be probable under that supposition. Contraposition can fail for conditional probabilities. All we need is p(A & B) > p(A & ¬B) > p(¬A & ¬B), and we have p(B|A) > 1/2, p(¬A|¬B) < 1/2. For instance, Jane is to choose at random one of ten bottles of wine. Nine are French, one is German. Of the nine French, six are red and three are white, and the one German is white. The probability that she picks red,

conditionals, truth, and assertion

295

given that she picks French, is 2/3; the probability that she doesn’t pick French, given that she doesn’t pick red, is only 1/4.⁵ Now suppose that ‘If A, C’ means the same as ‘¬A ∨ C’, but is assertible iff the probability of the second disjunct, given the negation of the first, is high, i.e. iff p(C|A) is high. Disjunction is symmetric—which disjunct is ‘first’ and which ‘second’ is immaterial—‘¬A ∨ C’ means the same as ‘C ∨ ¬A’. So we must also say that ‘If A, C’ is assertible iff p(¬A|¬C) is high. It follows that p(C|A) is high iff p(¬A|¬C) is high. But we have just seen that this is false. Therefore, ‘‘Adams’’ is incompatible with the thesis that ‘If A, C’ means the same as ‘¬A ∨ C’. This is the second decisive argument against the derivation of ‘‘Adams’’ from Gricean considerations. I agree with Jackson’s conclusion. But it cannot be convincingly argued for in terms of robustness2 . Note that without the alleged explanatory role of robustness2 , Jackson’s thesis is simply that an indicative conditional ‘If A, C’ is true iff ¬A ∨ C, and, as a matter of convention, assertible iff p((¬A ∨ C)|A), i.e. p(C|A), is high.

4. The relation between indicative and subjunctive conditionals In a symposium with Jackson (1984), Brian Ellis argued for a unified account of indicative and subjunctive conditionals, against Lewis’s and Jackson’s bipartite account. Jackson replied that he could explain the considerable extent to which indicatives and subjunctives agree. This issue occupies Chapter 4 of his book. I consider the matter now because it already arose in the last section. Two kinds of case make it hard to believe that there is much difference between a conditional in the indicative and one in the subjunctive mood: for future-looking indicatives, it is standardly a matter of indifference whether you say, e.g. ‘If you touch that wire, you will get a shock’ or ‘If you were to touch that wire, you would get a shock’. Secondly, when time passes and it turns out that the antecedent is false—you don’t touch it—the same thought as was first expressed in the indicative becomes ‘If you had touched it, you would have got a shock’.⁶ ⁵ We can explain why contraposition is nevertheless often reliable. First, when p(C|A) is 1, p(¬A|¬C) is 1, so it is a safe rule in a priori reasoning. Second, when p(A) and p(C) are both middling, and p(C|A) is close to 1, p(¬A|¬C) is close to 1, and this is a common case. Note that contraposition also fails on the Stalnaker–Lewis semantics. ⁶ There are exceptions. Suppose I have strong grounds for believing that either A or B will come about. For instance, I receive a note from someone who intends to visit me this afternoon, but can’t

296

dorothy edgington

‘I accept the basic approach to subjunctive and counterfactual conditionals developed by Stalnaker and Lewis’ says Jackson, and gives the Lewis-style truth condition I quoted above. He then continues, ‘The possible-worlds account of subjunctive conditionals needs no special rule of assertibility ... the assertibility of (A
C) for a speaker is simply and precisely the probability the speaker gives to the closest A-worlds being C-worlds ... Hence, we can straightforwardly identify their assertibility with their probability of truth.’ (62–3). Jackson accepts that indicatives and subjunctives are closely related: ‘I might say ‘‘If it rains, the match will be cancelled’’, or I might say instead ‘‘If it should rain, the match would be cancelled’’; before the last presidential election commentators said ‘‘If Reagan loses, the opinion polls will be totally discredited’’, afterwards they said ‘‘If Reagan had lost, the opinion polls would have been totally discredited’’, and this switch from indicative to subjunctive counterfactual did not count as a change of mind. They were expressing the same opinion both times. Examples like that could be multiplied’ (66). It is hard to see how one can express the same opinion, once with a true sentence, and once with a false sentence—when it didn’t rain and the match would not have been cancelled if it had, or when someone misguidedly said ‘If Reagan loses, the opinion polls will be vindicated’, and later reiterated his counterfactual opinion. Nevertheless, Jackson argues that indicatives and subjunctives are typically assertible in the same situations, i.e. p(A
C) typically equals p(C|A). He relies on this argument. Take any conditional A ⇒ C, for which modus ponens is valid, and which is such that A & C entails A ⇒ C. Then ‘A & (A ⇒ C)’ is logically equivalent to ‘A & C’, and so these should get the same probability. Note also that trivially, p(C|A) = p((A & C)|A). By the equivalence, this last is equal to p((A & (A ⇒ C))|A), which, again trivially, equals p((A ⇒ C)|A). If, in addition, A ⇒ C is judged to be probabilistically independent of its antecedent, i.e. p((A ⇒ C)|A) = p(A ⇒ C), it follows that p(A ⇒ C) = p(C|A). The Lewis conditional satisfies the conditions mentioned above. The closest world to the actual world is the actual world itself, so if A and C are actually make out the signature—it is either Ted or Fred. I’m strongly inclined to think it is Ted who is coming. But if Ted doesn’t come, Fred will come, I think. I’m not inclined to say ‘If Ted were not to come, Fred would come’; and when Ted duly turns up, I don’t think: ‘If Ted hadn’t come, Fred would have’. ( These are Oswald-Kennedy cases transposed into the future: they are not the standard sort of grounds for future-looking indicatives.)

conditionals, truth, and assertion

297

true, C is true at the closest world to the actual world at which A is true. (It will be helpful to express the Lewis truth condition thus: ‘A
C’ is true iff either A & C, or ¬A and all the closest A-worlds are C-worlds.) So p((A
C)|A) = p(C|A). If, in addition, we have probabilistic independence, p(A
C) = p(C|A). Note that probabilistic independence is not a relation between a pair of propositions, but a relation between judgements about a pair of propositions. No two contingent propositions are probabilistically independent in all probability distributions (one always might acquire the information that they are not both true, or at least one of them is true, by testimony, or in a general knowledge quiz, say). This could be seen as the crux of Lewis’s (1976) result: if there were a proposition such that of necessity its probability equaled p(C|A), it would have to be independent of A in all probability distributions. But there is no such proposition. Still, it could be held that typically, subjunctive conditionals are judged to be independent of their antecedents—that only in rather special cases do they come apart. Jackson maintains something like this: ‘subjunctive conditionals whose probability is independent of their antecedent’s should have the same assertibility as the corresponding indicative conditional. Now, as this probabilistic independence often obtains, the assertibility of the two conditionals will often be the same, and we have an explanation of why (A → B) and (A
B) are so often co-assertible.’ (77). It may be intuitively right to say that a subjunctive conditional is often judged to be probabilistically independent of its antecedent. But give the subjunctive conditional the Lewis truth conditions, and it is typically not independent of its antecedent. Its probability diverges from the conditional probability of C given A, just as systematically as does that of the truth-functional conditional, but in the other direction: p(A ⊃ C) is typically greater than p(C|A); p(A
C) is typically less than p(C|A); that is, A ⊃ C is typically too weak, A
C is typically too strong, where uncertainty is involved. To illustrate with the simplest of examples, consider the probability that if you toss this fair coin, it will land heads. The conditional probability of heads given toss is, of course, 1/2: it’s half as likely that (you toss it and it lands heads) as it is that you toss it. Suppose that it’s also 50% likely that you toss it. Then the probability of (toss ⊃ heads) is the probability that (either you don’t toss it, or you do and it lands heads) which is 3/4. Now consider the probability of the Lewis conditional (toss
heads)—the probability that, were you to toss it, it would land heads. That is the probability that either (you toss it and

298

dorothy edgington

it lands heads), or (you don’t toss it and it lands heads in all the closest worlds in which you do toss it). The second disjunct gets 0: if in the actual world you don’t toss it, it lands heads in some but not all of the close worlds in which you do. The first disjunct gets 1/4: it’s 50% likely that you toss it, and so 25% likely that you toss it and it lands heads. So p(toss
heads) is just 1/4: the only way it can be true that in all closest worlds in which you toss it, it lands heads, is that you do toss it, and it does land heads; if you don’t toss it, it’s false.⁷ Note the sensitivity to the probability of the antecedent: make it only 10% likely that you will toss it, and p(toss ⊃ heads) = 95%, p(toss
heads) = 5%. The truth-functional conditional demands too little when the antecedent is false, the Lewis conditional demands too much when the antecedent is false—when uncertainty is involved. Conditional probabilities agree with intuition, and agree with neither kind of probability of truth. (Conditional probabilities concern just the A-worlds. They are indifferent to the ¬A-worlds. In asking how likely it is that C if A, you have no need to consider what is true if ¬A.) The numerical example of the coin is just for illustration. Very many uncertain conditional judgements exhibit this phenomenon. A dog almost always, but not quite always, attacks and bites when strangers approach. We can detect no relevant difference between the occasions when it does and when it does not. Either there is some indeterminism involved, or it depends on some subtle and undetectable feature of the manner of approach. I say ‘It’s very likely that you will be bitten if you approach’. The conditional probability of consequent given antecedent is high. Consider instead ‘It’s very likely that you would be bitten, if you were to approach’. If you actually don’t approach, it’s false that in all close worlds in which you approach, you are bitten. So the Lewis conditional is true only if you approach and are bitten. If it’s 10% likely that you approach, it’s just under 10% likely that you will approach and be bitten, so just under 10% likely that the Lewis conditional is true. You wisely decide not to approach. ‘It’s very likely that I would have been bitten if I had approached’, you say. You express the belief that the conditional probability of bitten given approach was high. But now, the probability that you actually approached is 0, and the probability that in all closest worlds in which you approach, you are bitten, is 0. The Lewis conditional gets things badly wrong. ⁷ According to Stalnaker (1981), the conditional has no determinate truth value when there is no unique closest A-world and amongst the closest A-worlds some are C-worlds and some are not. But the probability that the conditional is true, in such a case, as for Lewis, is just p(A & C).

conditionals, truth, and assertion

299

You ask your doctor, ‘Would I be cured if I had this operation?’ She replies: ‘It’s very likely, but not certain, that you would be cured if you had the operation. I’m about 90% confident that you would be cured if you had the operation.’ Her uncertainty concerns imponderables about the small details of what will happen in the operating theatre, if you go ahead. Her conditional probability of cure given operation is 90%. She thinks that in most but not all close ways in which the operation takes place, you will be cured. Again, her probability for the Lewis conditional is simply the probability of A & C. In defence of typical independence, particularly with future-looking subjunctives, Jackson says that when (A
C) is probable, it typically remains probable on the assumption of A (79). He is right. But the problem lies in the other direction: C can be highly probable on the assumption of A, (hence A & C can be highly probable on the assumption that A, hence (A
C) can be highly probable on the assumption that A), while (A
C) is highly improbable simpliciter, on the Lewis interpretation. Conditional judgements of importance are often uncertain judgements. That is why probability theory enters the picture. The probability of a material implication and the probability of a Lewis-conditional are about equally bad as measures of our uncertainty in conditional judgements. The first overestimates because the conditional is always true when the antecedent is false. The second underestimates because often, when the antecedent is false, the most we can say is that in most close A-worlds, C is true. Fortunately, when we turn to probability theory, we find a better measure of conditional uncertainty.⁸ Jackson has not succeeded in explaining why indicatives and subjunctives often agree. Perhaps—if we stick with the Lewis truth conditions—subjunctives need a special assertibility rule too?

5. The nature of conventional implicature ‘Why does our language have words with assertibility conditions separate from truth conditions ... why conventional implicature?’, Jackson asks (91). His answer, for at least the subset of such words in which he is interested—‘but’, ‘nevertheless’, ‘however’ and the like—is that they aid the transfer of information from speaker to hearer; they make the assimilation of a given content ⁸ Adams treats subjunctive conditionals in terms of conditional probabilities—typically the judgement that it was probable that C given A—in chapter 4 of his book (1975). See also Edgington(1995, §10), and (2004).

300

dorothy edgington

easier; they do not change the content to be assimilated. I make it easier for you to accept that Fred can solve this problem by saying ‘Even Fred can solve it’. ‘We want a way of aiding the transfer of beliefs which does not change what is up for transfer. ... If you are unconvinced of the importance of words like ‘but’, ‘yet’, ‘nevertheless’, ‘anyhow’, ‘even’, ‘however’ and so on in aiding the transfer of information and furthering understanding, I suggest the following experiment. Take a philosophical paper with which you are unfamiliar; get someone to blank out all those words and their kin; then try to follow it. It is not easy.’ (94–5). This is plausible: these words help with the transfer of information; the same content could be expressed, but less helpfully, without them. But how does ‘if’ fit into this picture? Note that, in Jackson’s suggested experiment, the words in question are to be deleted, leaving the informational content intact. ‘But’ can be replaced by ‘and’, or a semi-colon, or a full stop. What are we to do with the ‘if’s (and associated ‘then’s) in this experiment? Let the sentence be ‘If John is in London, Mary is in London’. Just deleting the ‘if’ clearly alters the sense! (This point is made by Jonathan Bennett (2003: 41).) I presume Jackson’s answer is that we must replace ‘if’ by ‘⊃’ to convey the information neat, but less helpfully. An obvious difficulty is that no natural language has a word for ‘⊃’ (unless ‘A ⊃ C’ and ‘If A, C’ mean the same, which Jackson denies). Only those who have studied logic know what ‘⊃’ means. We can express the same truth conditions, of course, by saying ‘Either not A, or C’, or by saying ‘It’s not the case that both A and not C’: ‘Either John isn’t in London or Mary is in London’, ‘It’s not the case both that John is in London and that Mary isn’t in London’. These sentences will carry their own conversational implicatures, but that doesn’t affect their meaning. It could be objected that it is theoretically possible that someone could understand ‘if’ yet not understand negation, disjunction or conjunction; but this is far fetched, for it is hard to imagine any speaker not having mastery of these concepts. However, their combinations—a disjunction with one negated disjunct, a negated conjunction with one negated conjunct—are relatively hard to get one’s mind round. I can imagine people who understand ‘if’ perfectly well but have a hard time getting the message in these forms. There’s another problem (pointed out by Anthony Appiah (1984: 78)): Jackson’s assertibility condition for conditionals was supposed to ‘prepare the way for a justified use of modus ponens’ (98). But if the literal content needs to be expressed as a disjunction or negated conjunction, there is no modus ponens in the offing, but some different rule of inference such as disjunctive syllogism.

conditionals, truth, and assertion

301

The story about aiding the transfer of information and removing obstacles to being understood properly and believed, doesn’t have any obvious application to conditional statements.

6. Conditionals which are not asserted ‘What about compound sentences that have conditionals as constituents?’ Thus wrote Lewis (1976: 85), citing this as the reason for insisting on truth conditions, despite the attractiveness of ‘‘Adams’’⁹ and the impossibility of finding a proposition the probability of whose truth matches the conditional probability: I have no conclusive objection to the hypothesis that indicative conditionals are nontruth-valued sentences ... I have an inconclusive objection, however: the hypothesis requires too much of a fresh start. It burdens us with too much work still to be done, and wastes too much that has been done already ... We think we know how the truth conditions for compound sentences of various kinds are determined by the truth conditions of constituent subsentences, but this knowledge would be useless if any of these sentences lacked truth conditions. Either we need new semantic rules for many familiar connectives and operators when applied to indicative conditionals ... or else we need to explain away all seeming examples of compound sentences with conditional constituents. (1976: 85)

(And this is one of the commonest complaints against the no-truth-conditions view.) So Lewis tried to derive ‘‘Adams’’ from the Gricean defence of the truth-functional truth conditions. But, first, the truth-functional truth conditions give many bizarre results for embedded conditionals. Second, the Gricean defence focuses exclusively on what more is needed to justify the assertion of a conditional, beyond the belief that the truth condition is satisfied. It is silent about the occurrence of conditionals, unasserted, as constituents of longer sentences. We are still burdened with more work to be done, and neither Lewis nor anyone else has done it. Third, Lewis accepted Jackson’s criticisms of the Gricean defence, and adopted Jackson’s position instead (1986: 152–6). But Jackson, stressing that the assertibility condition is part of the meaning of ‘if’, argues that his theory has no implications for occurrences of ‘if’ in embedded contexts. We know what ‘A ⊃ C’ means as a constituent of longer sentences. But ‘A ⊃ C’ does not mean the same as ‘If A, C’. It ⁹ ‘Adams has convinced me. I shall take it as established that the assertability of an ordinary indicative conditional A ⊃ C does indeed go by the conditional subjective probability p(C|A).’ (Lewis 1976: 77)

302

dorothy edgington

is compatible with his thesis that there are no meaningful sentences with unasserted conditional constituents: ‘It simply does not follow from the fact that I give (A ⊃ B) truth conditions, that I must find, say, [(A ⊃ B) ⊃ C] ... a meaningful sentence’ (129). In fact he adopts the ‘explaining away’ strategy: whenever we have a sentence we understand which seems to have an embedded indicative conditional, unasserted, within it, we can give a paraphrase, sometimes in a systematic way, sometimes more or less ad hoc, in which a conditional occurs only as the main connective. This is the strategy of Adams and his followers, who deny that conditionals can be understood in terms of truth conditions. Note that the analogy with ‘but’, ‘even’, etc. fails. These can occur in unasserted clauses: ‘Either he arrived on time but didn’t wait for us, or he never arrived at all’ (Woods 1997: 61); ‘If even Fred can’t reach it, we’ll need a ladder’. Note also that ‘but’ et al. can occur in speech acts other than assertions: ‘Shut the door but leave the window open’; ‘Does anyone want eggs but no ham?’. ‘But’ means ‘in contrast’. No doubt its meaning has implications for when you should use it in assertions, but its meaning is not given by an ‘assertibility condition’. Like ‘but’, ‘if’ can occur in commands and questions as well as in statements. The conditional clause, ‘If he phones’ surely plays the same role in ‘If he phones, what shall I say?’, ‘If he phones, hang up immediately’, and ‘If he phones, Mary will be pleased’. A conditional command is not happily construed as a command to make the truth-functional conditional true. The doctor’s order, ‘If the patient is still alive in the morning, change the dressing’ would be equivalent to ‘Make it the case that either the patient is not still alive in the morning, or you change the dressing’. The nurse kills the patient. On the truthfunctional interpretation of the conditional command, the nurse can claim that he was carrying out an order. On the analogue of Jackson’s account, the doctor said ‘Make it the case that either the patient is not alive in the morning or you change the dressing’ and indicated that she would still command this if she knew that the patient would be alive. This doesn’t help. The nurse who kills the patient still carried out an order. Why should the nurse be concerned with what the doctor would have commanded in a counterfactual situation? The truth-functionalist could reply that any reasonable person would realize that this was not the manner in which the doctor intended the order to be carried out. But this is weak. Take a less extreme example: at Fred’s request, the Head, Mary, agrees that Fred can give the Kant lectures if his appointment is extended. Mary then puts every effort into ensuring that Fred’s appointment is not extended. Is it plausible to say that this is doing what Fred asked her to do, albeit not in the way Fred intended?