chapter 8. 18 This conception is to be found in McDowell: see footnote 7 above.

314

Mind, World and Value posed problem of other minds, but in the space it leaves for the explanation of behaviour in terms of states of mind. To say that it is in principle possible simply to perceive what someone thinks and feels is not to say that it is always in fact possible, or that it is impossible to be mistaken about someone's motives. We can still find room for the idea that someone's thoughts may be hidden from us: what we cannot do is think that they are altogether behind their behaviour. And, of course, there is room for informative explanation of another's behaviour in terms of her states of mind only insofar as we can be ignorant of them. In the cases where we offer an interpretation of another, what is generally perceived is just that the person has behaved - that is, that she has some motive or other. What we supply is just an account of what the motive is. How is this explanatory? One requirement of explanation in general is that it meet something like the following counterfactual condition: if what we offer as explanation had not been true, there would not have been the thing we are trying to explain.19 In our case, we want something like this to be true: if she hadn't believed that, she wouldn't have done that. (I say 'something like this', because there is dispute about exactly which counterfactuals we should insist on.) There are two ways in which we can achieve this within a non-causal, non-naturalistic conception of the mind. First, and most simply, if someone does something because she believes something, the belief is intrinsic to the behaviour. There would not have been this behaviour without that belief, because without that belief whatever was done would not have been this behaviour. When we explain why someone has done something, we are revealing the character of the behaviour itself. The explicitly evaluative account I have been sketching makes it easy to see how this can be. It is essential to what people do, if it is to be behaviour at all, rather than mere happenings, that there is something at stake in it. That is, it is essential to it, that there is an issue of whether it is foolish or bad. For it to be foolish is for it to have been done on the basis of a false belief; for it to be bad is for it to have been done on the basis of a bad desire. 'On the basis of: this marks the asymmetry between state of mind and behaviour which secures the direction of explanation (and gives the causal theory its opening). For although there could not have been the behaviour without the belief, there could have been " We should therefore doubt the adequacy of counterfactual analyses of causation, if we accept that there are non-causal explanations. But this does not undermine the main purpose of counterfactual analyses, which is to distinguish between explanatory and non-explanatory relations in cases where causation is the only realistic candidate. 315

Michael Morris the belief without the behaviour. All that is required for someone to have a belief at a particular time is that it should be legitimate at that time to count what she does as foolish or not, according to whether or not what is believed is false. No particular piece of behaviour is required for this, even if it is perhaps plausible that some appropriate behaviour or other is necessary. And this now makes room for the contrast which Davidson was concerned to emphasize: someone can have a particular belief or desire, and do something of a kind which would be appropriate for that belief or desire, without doing it because of that belief or desire. The right explanatory connection is only in place when the behaviour is itself properly assessed or foolish or bad, or otherwise, in terms of that belief or desire. There is another way in which something like the claim that she would not have done that if she had not believed that can be true. I have suggested that the particular behaviour could not have been done without that belief because the belief is essential to the behaviour. But often, in asserting such counterfactuals, we are concerned not with particular pieces of behaviour, but with behaviour of a certain type. When I say, 'He would not have run if he hadn't seen the bus', I am not concerned to insist just that there would not have been that running if he had not seen the bus. I am now explaining the fact that he ran at all. Does this provide any basis for a causal theory? Not at all. The claim that he would not have run if he hadn't seen the bus rests on two underlying claims: 1. He would not have run if he hadn't had reason to; 2. He would not have had reason to run if he hadn't seen the bus. Presumably (2) is true in virtue of the nature of the circumstances and the person's general state of mind. It is (1) which contains the link between reason and types of behaviour. The causal theory, of course, holds that it is a causal counterfactual. But a non-naturalistic account can suggest, on the contrary, that it is true simply because running is a type of behaviour. It is distinctive of behaviour, as opposed to mere occurrences, that it is done for a reason. Which is to say, on the explicitly evaluative conception, it is intrinsic to it that there is an issue of non-foolishness and virtue in it. The evaluative approach provides an altogether different account of the kind of understanding we get from explanations of behaviour from that offered within the dominant naturalistic tradition. In the first place, it does not suppose that explaining another's behaviour has merely instrumental value. It is difficult to see anything but instrumental value in the kind of explanation which the naturalistic 316

Mind, World and Value tradition thinks is involved. The emphasis is often explicitly on prediction (notice how often the real importance of explaining other people's behaviour is said to reside in the possibility of such things as making appointments). And the point of prediction is generally to be able to do something about the predicted outcome. In contrast with this, the point of the explanation of behaviour on the alternative account which I am outlining is just to understand other people. The evaluative account makes a distinctive claim about what it is to understand others: it is a matter of getting clear about the basis on which it is legitimate to evaluate what they do. But this is not to make understanding others a matter of instrumental value for the larger purpose of commending or criticizing them. For the values we bring to the understanding of others are themselves liable to be revised in the light of that understanding: for example, we can come to see a point in doing something which we had not previously seen. In general, the evaluative account has the Kantian virtue of treating people as people, rather than as objects whose movements need to be negotiated in the course of our own independent projects in 'getting around the world'.20 There is no difficulty at all in understanding how it is that the content of states of mind can be explanatory. What is explained when we explain behaviour is just what has to be the case for the behaviour to be decent and not foolish. This is to understand the nature of the behaviour. And it is also to give the content of the states of mind which inform the behaviour. It is not just the content of states of mind in any old sense of 'content' which is essential to the explanation of behaviour on the evaluative conception of the mind: truth conditions are essential. There is therefore no tendency at all for considerations to do with the explanation of behaviour to draw us towards 'internalism' about content, or towards a bifurcation of content into the 'broad' and the 'narrow', or the universal as opposed to the singular.21 On a naturalistic view of the explanation of behaviour, by contrast, some such tendency seems almost irresistible. The issue can be clarified by considering the counterfactuals 20 Although this conception of the fundamental task for human beings (or any cognitive creature) is widespread, we should note how contentious it is: it seems already to presuppose naturalism. 21 For a classic statement of a view which distinguishes between 'broad' and 'narrow' content, see C. McGinn, 'The Structure of Content', in Thought and Object, ed. A. Woodfield (Oxford: Oxford University Press, 1982). For the view that thoughts are essentially individuated by universal features, see S. Blackburn, Spreading the Word (Oxford: Oxford University Press, 1984), chapter 9. 317

Michael Morris which are held to be presupposed by a claim such as that he ran because he saw the bus. How are we to understand 'He would not have run if he had not seen the bus'? If our concern is with prediction, we cannot restrict our attention just to whether this person would have run in these circumstances if he had not seen the bus. We will want to be able to generalize: we will hope to be able to predict running in general. In pursuit of this goal, we will notice that is not strictly necessary for such running that the person be motivated by a belief with precisely the content that the bus was there already. Similar beliefs about taxis, trains or any other mode of transport would motivate very similar running. And although the belief that the relevant transporter was there then is no doubt relevant to explaining why the person ran in that direction and at that time, we can imagine very similar beliefs motivating similar runnings at different times. The pressure towards generalizing to a uniform abstraction from specific content, or narrowing down to a basic common core of content, seems irresistible. But this is only because a general concern with prediction makes us interested in similar behaviour in similar circumstances. If we are interested just in understanding people in a way which privots on evaluation, there is no pressure to widen our gaze to include all other circumstances: what we want to know is precisely why that person did that then. We therefore understand the counterfactual, 'He would not have run if he had not seen the bus', as being to be evaluated against the background of this person's state of mind at the time. And this means that the explanatoriness of truth-conditional content can remain integral to the evaluative theory's account of states of mind. The explicitly evaluative conception of the mind turns out to be thoroughly anti-Cartesian in two respects. First, in making states of mind intrinsic to behaviour, and therefore in principle, even if not inevitably, perceptible in behaviour, it prevents the traditional problem of other minds from getting going at all. Secondly, in making states of mind essentially world-involving, it prevents traditional scepticism about the 'external' world from arising. (I refer to these problems in their 'traditional' form, with reason: it is another question whether every version of other-minds or external-world scepticism is undercut by these moves.) But although it brings with it these epistemological benefits, the reasons I have been advancing for considering an evaluative conception of the mind have had nothing to do with epistemology: the task has simply been to provide a non-naturalistic account of the explanation of behaviour, using the typical awkwardness of naturalistic accounts as a clue to their replacement. 318

Mind, World and Value What is it to have a mind, then? The suggestion is that it is for there to be an issue of right and wrong, strictly and literally, in what one does. There is an obvious worry about this: it is that the view is behaviouristic, since it requires that there be something one does if one is to have a mind. The charge of behaviourism is worrying only to the extent that behaviourism is a bad thing. Whether behaviourism is a bad thing depends chiefly on the conception of behaviour which informs it. The principal tradition of philosophical behaviourism arose alongside logical positivism. Its concern was to replace an unscientific or pseudo-scientific psychology. What seemed to be needed was an approach which treated the field of psychology as consisting of unquestionably natural phenomena, and which eradicated mysterymongering by insisting on empirical testability. These two requirements define the conception of behaviour which informs traditional behaviourism. On the one hand, behaviour is thought of as essentially a type of occurrence: it is therefore not something which is intrinsically mental at all. On the other hand, behaviour is thought of as essentially publicly observable. Any behaviourism which supposes that states of mind require at least a disposition to produce behaviour which meets either of these descriptions seems to be denying what is most important about the mind. Neither condition is required for the evaluative conception of the mind which I have been developing. The rejection of the first condition is integral to its opposition to naturalism. And adherence to the second condition is simply of no interest to the evaluative conception. It is true that states of mind are counted intrinsic to behaviour. And since some behaviour is observable, there is no reason in principle why states of mind should not be directly perceptible. But the fact that some behaviour is observable does not mean that behaviour is essentially observable; or that there is a special relationship between states of mind and publicly observable behaviour. The crucial relationship is between states of mind and what someone may do because of her states of mind. Anything which is properly explicable by appeal to states of mind can count as behaviour, as far as the evaluative conception is concerned. This means that mental acts are included in behaviour. No doubt these can sometimes be observed: it is easy to think of visible decisions, for example. But there is no reason why they should be open to public view, and often they are not.22 22 The resulting view of behaviour is very like that advocated by G. Strawson in Mental Reality (Cambridge, MA: MIT Press, 1994), chapter 10, though Strawson's general account of mind is very different from mine. 319

Michael Morris Is the evaluative conception behaviourist, then? Not as behaviourism is traditionally understood; so not in any way which permits there to be an easy objection. There will remain those, however, who will find even mental behaviour too active to be essential to having a mind. The core of mentality, it might be said, is experience, where experience is supposed to be not itself active.23 Experiences, someone might say, just happen to one.24 The difficulty with any such view is to understand why what I experience must be something to me. The evaluative conception is the result of accepting a fundamentally Kantian answer to that question. What I experience is something to me only insofar as I take it in some way or other. But to take it in some way is already to do something which might be admired or regretted, as an object of pity or congratulation. This is the deep way in which the evaluative conception differs from naturalism. To think of the world as a world of natural objects and natural phenomena is to think of it as the counterpart to Hume's faculty or 'reason'. Hume's 'reason' is an essentially disengaged and dispassionate faculty. It is supposed to 'discover objects as they stand in nature, without addition or diminution'. It is supposed to be the fundamental, and truly revealing, stance towards reality; the deliverances of 'taste' merely 'gild' or 'stain' what it shows.25 According to the evaluative account, there can be no such thing as Hume's 'reason'. Experience is never disengaged or dispassionate. The evaluatively neutral view is not basic, but reached by abstraction from our fundamental engagement with the world. Once we see this, we have undermined the basic presumption of naturalism. 23 This is Strawson's view: see Mental Reality, p. 315 (but passim too). 24 Strawson seems to be tempted by this: see ibid., p. 303. 25 D. Hume, Enquiry concerning the Principles of Morals, ed. L. SelbyBigge; 3rd edition, ed. P. Nidditch (Oxford: Oxford University Press, 1975), p. 294. 320

Mind, Knowledge and Reality: Themes from Kant QUASSIM CASSAM According to what might be described as 'humanist' approaches to epistemology, the fundamental task of epistemology is to investigate the nature, scope and origins of human knowledge. Evidently, what we can know depends upon the nature of our cognitive faculties, including our senses and our understanding.1 Since there may be significant differences between human cognitive faculties and those of other beings, it would seem that an investigation of the nature, scope and origins of human knowledge must therefore concern itself, in the first instance, with uncovering the structure and operations of the human cognitive apparatus. The most influential versions of humanism in epistemology have also been inclined to insist both that it is contingent that our cognitive faculties are as they are, and that an investigation of these faculties must be largely empirical. An empirical investigation is to be understood, very roughly, as one which relies upon observation and experiment, and to describe such an investigation as naturalistic is to draw attention to the fact that it is presupposed by humanism that the faculties being investigated are a part of the natural world, the world of space, time and causal law. Locke's Essay2 and Hume's Treatise3 are arguably the most distinguished examples of works whose approaches to epistemology are, in these terms, 'humanist'. To begin with, the Essay is An Essay Concerning Human Understanding, and the Treatise is A Treatise of Human Nature. As Nidditch remarks in connection with the Essay, the epithet 'human' is intended by Locke to make it clear that the work is about 'man and not about the understanding belonging to I am grateful to P. F. Strawson for his comments on an earlier version of this paper. ' I take it that a cognitive faculty is one the proper exercise of which is necessary for the acquisition of knowledge. 1 John Locke, An Essay Concerning Human Understanding, ed. P. H. Nidditch (Oxford: Clarendon Press, 1975) 3 David Hume, A Treatise of Human Nature, ed. L. A. Selby-Bigge (Oxford: Clarendon Press, 1978). 321

Quassim Cassatn God, angelic spirits, or intellectual corporeal beings, infinitely different from those of our little spot of Earth elsewhere in the universe'.4 Thus, Locke writes in the opening chapter of the Essay that his purpose is to 'inquire into the Original, Certainty, and Extent of humane Knowledge; together, with the Grounds and Degrees of Belief, Opinion, and Assent', and that it will suffice for this purpose to 'consider the discerning Faculties of a Man, as they are employ'd about the Objects, which they have to do with'.5 In a related passage in his introduction to the Treatise, Hume maintains that all the sciences have a greater or lesser relation to human nature since 'they lie under the cognizance of men, and are judged by their powers and faculties'.6 Given that 'the science of man is the only solid foundation for the other sciences... the only solid foundation we can give to this science itself must be laid on experience and observation'.7 What is the alternative to regarding an empirical investigation of the discerning faculties of human beings as foundational for epistemology? The obvious alternative is what might be called 'universalism'. For the universalist, it is the proper business of epistemology to give an account of the conditions of any knowledge or, perhaps, any empirical knowledge of reality. There is nothing in the concept of empirical knowledge which would justify the stipulation that only human beings are capable of such knowledge, so arriving at an understanding of the peculiarities of the human cognitive apparatus cannot be as important for the universalist as it is for the humanist. Instead, universalism might hold that an account of the conditions of empirical knowledge must be grounded in an analysis of concepts such as those of knowledge, belief and justification. For example, if it is a conceptual truth that fulfilment of some condition C is necessary for empirical knowledge of reality, then the cognitive faculties of any beings, including human beings, who are capable of empirical knowledge of reality must be so structured as to enable them to fulfil condition C. If it turns out that only beings whose cognitive faculties are similar to ours can fulfil condition C, this would be an important discovery, but it would still not be a reason to accept a definition of the concept of empirical knowledge which has built into it the stipulation that only members of a particular animal species are capable of such knowledge. The key to universalism, therefore, is its insistence that epistemology must concern itself with the uncovering of conditions of knowledge which are 4 Foreword to Locke's Essay, p. xxiii. 5 Essay, I.i.2. 6 Treatise, p. xv. 7 Ibid., p. xvi. 322

Mind, Knowledge and Reality: Themes from Kant universal in scope, conditions which must be fulfilled by any knowing subject, human or otherwise. It is far from obvious that humanism and universalism are mutually exclusive.8 Indeed one of my aims here will be to look in some detail at a position which combines elements of humanism and universalism in its account of the conditions of empirical knowledge. Before saying more about this 'mixed' position, however, it would be worth remarking that the distinction between humanism and universalism also has a bearing on our understanding of the subject matter of the philosophy of mind. The philosophy of mind is, presumably, concerned with the nature of mind, but it would be natural to ask whose mind or what kind of mind is at issue here.9 The humanist in the philosophy of mind is someone who thinks that the primary concern of what is generally known as the philosophy of mind is, and ought to be, with the nature of the human mind. Human minds are not the only minds that there are, but the philosophy of mind is only concerned with non-human minds to the extent that they are like human minds. In contrast, the universalist is someone who thinks that the philosophy of mind ought to concern itself with what might be called 'mind in general'10 or 'mind as such'.11 For the universalist, human and non-human minds are all instances of mind in general, and this means that theories of mentality which only address the human case are guilty of an unwarranted parochialism.12 These remarks might prompt the thought that the contrast between humanism and universalism in the philosophy of mind is a false one since, as Wittgenstein puts it, 'it is only of a living human being and what resembles (behaves like) a living human being one can say: it has sensations; it sees; is blind; hears; is deaf; is conscious or unconscious'.13 This anthropocentric approach to the mind mainH In his Introduction to Epistemology and Cognition (Cambridge, MA: Harvard University Press, 1986), Alvin I. Goldman defends a conception of epistemology which combines elements of what I am calling 'universalism' and 'humanism', though Goldman's label for humanist epistemology is 'psychologistic epistemology'. '' Colin McGinn presses these questions in The Character of Mind (Oxford: Oxford University Press, 1982), pp. 1-4. "' McGinn, The Character of Mind, p. 2. " W. H. Walsh, Kant's Criticism of Metaphysics (Edinburgh: Edinburgh University Press, 1975), p. 89. 12 For example, McGinn claims in The Character of Mind, p. 1, that 'it is frequently a good test of a theory of some mental phenomenon to ask whether the proposed theory would be applicable to all actual and possible creatures exemplifying that phenomenon'. 13 L. Wittgenstein, Philosophical Investigations, translated by G. E. M. Anscombe (Oxford: Basil Blackwell, 1978), section 281. 323

Quassim Cassam tains that what the universalist calls 'mind as such' is simply an abstraction from certain abilities and dispositions possessed by live human beings,14 so one should not follow the universalist in representing the humanist's investigation of the human cognitive apparatus as something which falls short of the proper business of giving an account of the nature of mind as such. Rather, once it is recognized that all minds must be conceived of on an analogy with our own, an investigation of 'our' cognitive faculties may be seen as deepening our understanding of mind as such. By the same token, if all minds must be conceived of on an analogy with our own, then one should also refuse to follow Locke and other humanists in contrasting the human understanding with other types of understanding that might be fundamentally different from ours. I will have more to say in due course about anthropocentrism, but the next section will be concerned with an epistemological position which combines elements of humanism and universalism, while apparently distancing itself from anything recognizable as anthropocentrism. The position which I have in mind is one which is suggested by Kant's discussion in the Critique of Pure Reason of what he refers to as 'a priori conditions of the possibility of experience' (A94/B126).15 It is true that Kant himself does not employ the terminology of humanism and universalism. Instead, he sets out to defend transcendental idealism, the view that space and time are merely the forms of our sensibility rather than conditions of objects as they are in themselves, and to distinguish his conception of the a priori or necessary conditions of the possibility of experience from the subjectivist thesis that such conditions are merely subjectively necessary. Nevertheless, it will emerge that there is a close connection between, on the one hand, Kant's humanism and his transcendental idealism, and, on the other, between his universalism and his opposition to subjectivism, a position which he associates with Hume. One of my aims in what follows will be to explore these connections. While transcendental idealism may be a position for which few late twentieth-century epistemologists and philosophers of mind have much sympathy, it would be a mistake to conclude that Kant's position is of merely historical interest. Quite apart from the intrinsic interest of the proposal that humanism and idealism are linked in the way that Kant suggests, questions about the appropriate balance between universalism and humanism remain of fundamental 14 Cf. Walsh, Kant's Criticism of Metaphysics, pp. 251—2, and Q. Cassam, Self and World (Oxford: Clarendon Press, 1997), pp. 12-21. 15 All references in this form are to Norman Kemp Smith's translation of Kant's Critique of Pure Reason (London: Macmillan, 1929). 324

Mind, Knowledge and Reality: Themes from Kant importance for our understanding of knowledge and mind. Whether or not Kant's deep and subtle account of these matters is ultimately acceptable, it is not one which one can afford to ignore if one has ever been struck by the difficulty of giving a straight answer to the question of what the philosophy of mind is about. Thus, the issues which I propose to address are these: first, does Kant's conception of the status and origins of his a priori conditions of the possibility of experience constitute a stable and coherent combination of humanism and universalism? Secondly, what is the bearing of Kant's theory of a priori conditions on the dispute between transcendental idealism and what he calls transcendental realism? My own view is that the first of these questions should be answered in the negative, and that the most attractive position in connection with the second question is one which combines a form of universalism with transcendental realism, the view that space and time are conditions of the possibility of things in themselves as well as conditions of our sensible awareness of empirical reality. II A priori conditions of the possibility of experience are necessary conditions of the possibility of experience. Experience, for Kant, is a form of empirical knowledge. So a priori conditions of the possibility of experience may be understood as necessary conditions of the possibility of empirical knowledge. Empirical knowledge of an object is represented by Kant as involving two factors, 'first, the concept, through which an object in general is thought ... and secondly, the intuition, through which it is given' (B146). At this stage, an 'object' may simply be understood as a particular instance of a general concept."' The concepts which the understanding must employ in thinking or conceptualizing objects of (sensible) intuition are the Kantian 'categories', which include those of substance and causality. Intuitions are singular and immediate representations of particulars which are given to us by means of sensibility. Sensibility is 'the capacity (receptivity) for receiving representations through the mode in which we are affected by objects' (A19/B33). In the Transcendental Aesthetic, Kant argues that space and time are the 'two original forms of sensibility' (A41/B58), that is, 'necessary conditions under which alone objects can be for us objects of the senses' (A28-9). Thus, space and time might be described as the "' This is the less 'weighty' of the two senses of 'object' which P. F. Strawson distinguishes in The Bounds of Sense: An Essay on Kant's Critique of Pure Reason (London: Methuen & Co. Ltd., 1966), pp. 72-3. 325

Quassim Cassam sensible conditions of empirical knowledge and the categories as its intellectual conditions.17 While this way of representing Kant's position is undoubtedly along the right lines, it is incomplete in at least one important respect, for it fails to make anything of the fact that Kant is not even-handed in his treatment of the sensible and intellectual conditions of our knowledge. With regard to space, Kant insists that it is 'solely from the human standpoint that we can speak of space, of extended things, etc' (A26/B42). There are two closely related points being made here. The first is that space does not represent any property of things in themselves; it is merely a subjective condition of sensibility. The second is that 'we cannot judge in regard to the intuition of other thinking beings, whether they are bound by the same conditions as those which limit our intuition and which for us are universally valid' (A27/B43). In other words, we are only entitled to assert that space and time are the necessary conditions of human awareness of particularity. To the extent that Kant is only concerned in the Transcendental Aesthetic with the conditions of human awareness of particularity, his account of the original forms of sensibility is humanist rather than universalist, though he would have insisted that his investigation of these forms is a priori rather empirical or naturalistic. A universalist about the forms of sensibility would be someone who maintains that spatio-temporal intuition is a necessary condition of any empirical knowledge of objects, but Kant leaves open the possibility that spatio-temporal awareness is a mode of perceiving which is 'peculiar to us, and not necessarily shared in by every being, though, certainly, by every human being' (A42/B59). As for the link between humanism and transcendental idealism, the very fact that Kant expressed his idealism by saying that it is solely from the human standpoint that we can speak of space suggests that he saw the two as closely connected, though the precise nature of this connection has yet to be explained. Kant's account of the status of the categories is very different. It is true that the categories are said to have their 'first seeds and dispositions in the human understanding' (A66/B91), but he also insists that they relate 'to objects of intuition in general, whether that intuition be our own or any other, provided only it be sensible' (B150). In other words, the categories are not just concepts which we humans must employ in order to conceptualize the objects of our spatio-temporal sensible intuition, but also concepts which must be employed by any discursive intellect in thinking the objects of its 17 This way of putting things is suggested by Henry Allison in Kant's Transcendental Idealism: An Interpretation and Defense (New Haven: Yale University Press, 1983), chapters 5 and 6. 326

Mind, Knowledge and Reality: Themes from Kant sensibility, spatio-temporal or otherwise.18 The primordial being - God - would have no use for the categories, since divine intuition would be 'intellectual' rather than sensible, but this is compatible with maintaining that the categories are the universal and necessary conditions of any empirical knowledge of objects, any knowledge the acquisition of which involves the exercise of sensibility. So the most striking difference between Kant's theory of the categories and his theory of the forms of our sensibility is that the former is universalist in its orientation,19 whereas the latter has much more in common with the humanist tradition. My question is whether Kant's conception of the status and origins of his a priori conditions of the possibility of experience constitutes a stable and coherent combination of humanism and universalism. In the light of what has just been said about the respects in which Kant is a humanist and the respects in which he is a universalist, it should now be clear that this extremely general question can now be broken down into a series of rather more precise subquestions. With regard to the theory of sensibility, the most pressing sub-questions would appear to be these: 51. Are there good grounds for thinking that space and time are necessary conditions of human awareness of objects which would not also be good grounds for thinking that space and time are necessary conditions of any sensible awareness of objects? 52. To the extent that the Transcendental Aesthetic is an investigation of our mode of perceiving objects, how can it be anything other than an empirical investigation of contingent aspects of the functioning of the human cognitive apparatus? 53. What exactly is the relationship between the claim that it is solely from the human standpoint that we can speak of space and the thesis that space is transcendentally ideal? The point of these questions is this: if, in connection with (SI), it turns out that the only legitimate grounds for thinking that space 18 A discursive understanding is characterized by Kant as one whose knowledge must be 'by means of concepts' (A68/B93). He adds that 'the only use which the understanding can make of these concepts is to judge by means of them' (ibid.). As W. H. Walsh remarks, it was Kant's consistent doctrine that 'the categories were by no means peculiar to human nature, but were involved in discursive thinking as such' {Reason and Experience (Oxford: Clarendon Press, 1947), pp. 163-4). " To the extent that Kant's account of the role of the categories is an important element of his 'transcendental psychology', its universalist orientation is not brought out by Patricia Kitcher's characterization of transcendental psychology as seeking 'to determine the necessary and universal elements of human cognition' {Kant's Transcendental Psychology (Oxford: Oxford University Press 1990), p. 19). 327

Quassim Cassam and time are necessary conditions of human awareness of particularity would also be grounds for thinking that they are necessary conditions of any sensible awareness of particularity, then the thesis that it is solely from the human standpoint that we can speak of space is open to question. Yet, as is suggested by Kant's response to (S3), the loss of this thesis would amount to the loss of an important element of the case for transcendental idealism. As for (S2), Kant is apparently prepared to grant that it is contingent that space and time are the forms of our sensibility.20 In that case, it might be wondered whether he is right to be so resistant to the idea that his investigation of these contingent aspects of the human cognitive apparatus can only properly be read as an empirical investigation. Yet, Kant's conception of the relationship between his theory of mind and the theories of more straightforward humanists such as Locke and Hume makes it very important for him to resist any such account of the nature of his investigation. With regard to Kant's theory of understanding, it is the humanist rather than the universalist who may have doubts on this score. Among the questions which will need to be addressed in this connection are these: Ul. If, as Kant insists, we know nothing but our mode of perceiving objects, then should he not also have insisted that we know nothing but our mode of thinking objects of sensible intuition, and that the categories can only be known by us to represent the intellectual conditions of human knowledge of objects rather than conditions of all discursive thinking? U2. What kind of investigation would be required to demonstrate that the categories relate to objects of sensible intuition in general, whether that intuition be our own or any other? U3. How can concepts which, by Kant's own lights, have their first seeds and dispositions in the human understanding have the universal scope which Kant claims for the categories? The point of these questions, then, is to suggest that Kant should have been as modest about the status of his theory of understanding as he was about the status of his theory of sensibility. Like Locke, he should have refrained from making claims about the intellectual faculties of finite thinking beings other than man. One way of working up to a detailed consideration of these ques20 Strawson reads Kant in this way in 'Kant's New Foundations of Metaphysics', in Entity and Identity and Other Essays (Oxford: Clarendon Press, 1997), pp. 237-8. He quotes B145-6 in support of this reading. For a more detailed discussion on Kant's position on this question, see Lome Falkenstein, Kant's Intuitionism: A Commentary on the Transcendental Aesthetic (Toronto: University of Toronto Press, 1995), pp. 193—200. 328

Mind, Knowledge and Reality: Themes from Kant tions would be to look more closely at the notion of an a priori condition. So far, I have said that a priori conditions are necessary conditions of the possibility of experience, but it might be wondered what kind of necessity this is supposed to be. On one view, a priori conditions are subjectively necessary conditions. I will refer to this view as subjectivism. On another view, they are what Henry Allison calls 'epistemic conditions'.21 I will refer to this view as idealism about a priori conditions. A third view, which I will refer to as realism, would be that a priori conditions are objectively necessary conditions of the possibility of experience. Subjectivism and idealism are both humanist rather than universalist, whereas realism's conception of a priori conditions is universalist rather than humanist. With regard to the dispute between transcendental realism and transcendental idealism, it should come as no surprise that idealism about a priori conditions goes with transcendental idealism and that realism about a priori conditions goes with transcendental realism. The position of the subjectivist on this question is less clear, though it will eventually emerge that subjectivism is at least compatible with transcendental realism. Kant's own conception of a priori conditions is idealist rather than subjectivist or realist. My claim will be that idealism fails to provide satisfactory answers to the questions outlined above. I will also argue that most of Kant's objections to subjectivism and realism are misguided. The failure of these objections is due, in part, to the fact that Kant's attempt to combine elements of humanism and universalism results in a position which is fundamentally unstable. To this extent, it remains unclear how the Kantian programme in epistemology is to be understood. While Kant's objections to subjectivism are largely misguided, my own position is not subjectivist. Instead, I will be arguing that the best position in this area is realism, and that this has important consequences for our understanding of the nature of epistemology and the philosophy of mind. Ill What would it be for a condition of human knowledge to be merely subjectively necessary? To regard a condition as merely subjectively necessary is to regard it as grounded in contingent, empirically discoverable facts about the structure or constitution of our — that is, human - cognitive faculties. This is the sense in which subjectivism is a form of humanism. Suppose, for example, that our senses are so 21 See Kant's Transcendental Idealism, 10-13. 329

Quassim Cassam constituted as to restrict us to the perception of objects with spatial properties. If this is a contingent, empirically discoverable fact about the constitution of what Kant refers to as our 'faculty of sensible intuition', then it is subjectively necessary that for an object O to be perceivable, and hence knowable, by us, O must be spatial. This might be described as an intuition version of the subjectivist view.22 An example of a concept version of subjectivism would be this: suppose that our understanding is so constituted as to make it impossible for us to think of objects of experience unless they are, and are thought of by us as being, causally ordered. In that case, it is subjectively necessary that for objects to be thinkable, and hence knowable, by us, they must be causally ordered and must be represented as such. This would be an example of a subjective necessity, as long as it is assumed that this fact about the structure of our faculty of understanding is empirically discoverable. If there is one thing that is clear about Kant's position, it is that he did not accept that a priori conditions are merely subjectively necessary. His explicit objection to subjectivism is that it amounts to a form of scepticism. The sense in which the subjectivist is a sceptic is supposed to be this: if all we can say is that our senses are so constituted as to restrict us to the perception of objects in space, then this is simply a fact about us, from which nothing follows about the nature of the objects themselves. By the same token, if all we can say is that we are so constituted as to be incapable of thinking of given representations other than as causally connected, we would not be able to say that 'the effect is connected with the cause in the object'. This, Kant continues, 'is exactly what the sceptic most desires' (B168). On the face of it, this objection to subjectivism is mistaken. To say that we are limited to perceiving objects with some property P is not just to make a claim about the perceiving subject; it is also to make a claim about how objects must be if they are to be perceivable by us. So if O is an object that is perceivable by us, then O must have the property P. On this reading, the judgement that any possible object of human sense-perception must have the property P is, in Broad's terminology, 'transcendentally a priori', one which is 'entailed by certain very general facts about the way in which human minds work'.23 The fact that transcendentally a priorijudgements are grounded in such facts would arguably be one good rea22 Paul Guyer appears to attribute this version of subjectivism to Kant in Kant and the Claims of Knowledge (Cambridge: Cambridge University Press, 1987), p. 367. 23 C. D. Broad, Kant: An Introduction, edited by C. Lewy (Cambridge: Cambridge University Press, 1978), p. 7. 330

Mind, Knowledge and Reality: Themes from Kant son for insisting, in response to (SI), that O's possession of P might be a necessary condition of our awareness of O without being a necessary condition of any sensible awareness of it. Moreover, the claim that O would not be perceivable by us if it lacked the property P is independent of any commitment to the idea that O's possession of P is something for which 'we' are responsible; it can be true both that P is a mind-independent property of O, and that our ability to perceive () is conditional upon its possession of P.24 So the subjectivist should argue, in response to (S3), that even if space and time are specifically human conditions of object-awareness, it does not follow that 'if the subject, or even the subjective constitution of the senses in general, be removed ... space and time themselves, would vanish' (A42/B59). This might prompt the thought that the sense in which subjectivism is what the sceptic most desires is not that it creates an unbridgeable gap between mind and world, but that it fails to establish the 'objective validity' of the categories and of the concepts of space and time. To establish the objective validity of the categories would be to show that they 'furnish conditions of the possibility of all knowledge of objects' (A89-90/B122). If the need for us to employ categories in thinking objects of sensible intuition is simply a reflection of the peculiar constitution of the human understanding, on what basis can the subjectivist claim that they furnish conditions of the possibility of all knowledge of objects? The subjectivist will not be troubled by this question. To begin with, Kant himself describes the concepts of space and time as objectively valid (A87/B120) despite the fact that he does not regard himself as entitled to assert that spatio-temporal awareness is a necessary condition of any sensible awareness of objects. For the subjectivist, the moral is that Kant was not entitled to assert that the categories are conditions of any discursive thinking, any more than he would have been within his rights to claim that space and time are conditions of any sensible awareness of objects. This is the essence of subjectivism's response to (Ul). This response to (Ul) means that (U2) lapses. As for (U3), the subjectivist will maintain that Kant simply fails to recognize the tension between his own conception of the human origins of the categories and his universalist conception of their scope. A somewhat different interpretation of Kant's objection to subjectivism would be that while it is perfectly legitimate to represent 24 This is the moral of Ross Harrison's important paper on 'Transcendental Arguments and Idealism', in Idealism Past and Present, ed. G. Vesey, Royal Institute of Philosophy Lecture Series 13 (Cambridge: Cambridge University Press, 1982). 331

Quassim Cassam very general facts about the way in which human minds work as the basis of a priori conditions, it is not legitimate for the subjectivist to represent these facts as empirically discoverable. The underlying point here is presumably that an investigation of a priori conditions must itself be a priori. The difficulty with this objection is brought to light by (S2). As has already been remarked, Kant is in agreement with the subjectivist that it is contingent that our cognitive faculties are as they are. In other words, while it is necessary that space and time are the conditions of human awareness of particularity, it is not necessary that they have this status.25 If, however, the truths about our cognitive faculties upon which Kantian a priori conditions are grounded are in themselves contingent, then it is far from obvious how our knowledge of such truths could be anything other than empirical. For Kant, 'any knowledge that professes to hold a priori lays claim to be regarded as absolutely necessary' (Axv), but his own assertions about the constitution of our faculty of intuition do not lay claim to be regarded as absolutely necessary. In that case, it can scarcely be a good objection to subjectivism to point out that it represents subjectively necessary conditions as a reflection of facts about our cognitive apparatus that are empirically knowable, since, by Kant's own lights, we cannot have a priori knowledge of a contingent truth. So much, then, for Kant's objections to subjectivism. In the light of these difficulties, it would be worth considering the possibility that a priori conditions are what Allison calls 'epistemic conditions'. An epistemic condition is an 'objectivating' condition, 'one that is necessary for the representation of an object or objective state of affairs'.26 What makes this position idealist is its distinctive conception of the scope and origins of epistemic conditions. With regard to the scope of epistemic conditions, Allison's thesis is that they 'express the universal and necessary conditions in terms of which alone the human mind is capable of recognizing something as an object at all'.27 To the extent that epistemic conditions are specifically human conditions of object-awareness, they might be described as being species-specific in scope. With regard to the origins or basis of epistemic conditions, the proposal is that they 25 As Falkenstein puts it, Kant's view is that it is a 'contingent truth that, for us, space is a necessary ground of outer appearances' (Kant's Intuitionism, 199). The related idea that some necessary or 'eternal' truths are only contingently necessary has also been attributed to Descartes. See Edwin Curley, 'Descartes on the Creation of Eternal Truths', Philosophical Review 93/4 (October 1984), 569-97. 26 Kant's Transcendental Idealism, 10. 27 Ibid., p. 9. 332

Mind, Knowledge and Reality: Themes from Kant 'reflect the structure and operations of the human mind'28 rather than the nature of objects as they are in themselves. In this sense, epistemic conditions are species-specific in origin. Since epistemic conditions are species-specific both in scope and in origin, Allison's account is humanist rather than universalist. On the other hand, it is also supposed to be quite different from subjectivism. A subjectively necessary condition is merely psychological, that is, 'a propensity or mechanism of the mind which governs belief and belief acquisition'.21' The alleged difference between epistemic conditions and psychological conditions is that the latter have no objectivating function and so lack objective validity. Does idealism provide a satisfactory response to (SI)? In defence of Kant's theory of space, Allison argues that in order to be aware of things as numerically distinct from one another, it is necessary to be aware, not only of their qualitative differences, but also of the fact that they are located in different places. In other words, the representation of place, and therefore of space, functions within human experience as a necessary condition of the possibility of distinguishing objects from one another ... it is not a logically necessary condition. There is no contradiction in the thought that there might be some other, nonspatial mode of awareness of numerical diversity; we simply do not know what such a mode of awareness would be like.30 The difficulty with this passage is that it is extremely unclear what work the qualification 'human' is supposed to be doing in its second sentence. On the face of it, the best possible case for saying that spatial awareness is a necessary condition of awareness of things as numerically distinct from one another is that we can make nothing of the idea of non-spatial awareness of numerical diversity. To the extent, however, that nothing can be made of the idea of non-spatial awareness of numerical diversity, this would appear to constitute a case for saying that spatial awareness must function within any sensible experience as the form of awareness of particularity. Perhaps the 'must' in this formulation is not, as Allison insists, a strictly logical 'must', but it is still plausible that the connection between spatial awareness and awareness of numerical diversity is either valid for all sensible cognition or it is not genuinely necessary. Unless what is at issue is a merely psychological necessity, it makes 28 Henry Allison, 'Transcendental Idealism: A Retrospective', in Idealism and Freedom: Essays on Kant's Theoretical and Practical Philosophy (Cambridge: Cambridge University Press, 1996), pp. 4—5. 2" Ibid., p. 4. 3" Kant's Transcendental Idealism, pp. 83—4. 333

Quassim Cassam no sense to suppose that a genuinely necessary condition of the possibility of experience might turn out only to be necessary for human beings, qua members of a particular animal species. To the extent, therefore, that the point of the reference to human experience in Allison's second sentence is to imply that spatial awareness is, or might be, a species-specific condition of awareness of particularity, it is not justified by the best case for his opening sentence. So idealism fails to provide a satisfactory response to (SI). It might be objected that this line of argument ignores the simplest and most obvious case for insisting that the sensible conditions of human knowledge are species-specific in scope. It has just been argued that the best way of explaining the peculiarly intimate connection that obtains between spatial awareness and awareness of particularity would be to point out that nothing can be made of the idea of non-spatial awareness of numerical diversity. To say this, however, is simply to say that we can make nothing of this idea, and this only goes to show, at best, that we cannot imagine being aware of particularity other than in spatial form. To conclude that spatial awareness is a necessary condition of any awareness of numerical diversity would be grossly to exaggerate the importance of our imaginative limitations. Later I will argue that this attempt to explain the force of the thesis that epistemic conditions are species-specific in scopes is unsuccessful. First, there are other matters to consider. With regard to (S2), the idealist must insist, on pain of undermining the distinction between idealism and subjectivism, that his investigation of the forms of human sensibility is not an empirical investigation. The difficulty with this, however, is that it is far from obvious that the idealist has a coherent alternative to the subjectivist's conception of the status of Kant's inquiry. The natural alternative would be to think of Kant's investigation as broadly conceptual or analytical, but if spatial awareness is a conceptually necessary condition of the possibility of awareness of particularity, then it would seem that it must be a conceptually necessary condition of any sensible awareness of particularity. Since it is quite mysterious how a conceptually necessary condition of awareness of numerical diversity could only be valid for human cognition, it is the thesis that space is a species-specific conditions of empirical knowledge which is one against under threat. What is the relationship between the thesis that space is a speciesspecific condition of human knowledge, and the transcendental idealist thesis that space does not represent any property of things in themselves? For the idealist, the connection might be something like the following: if space is only a necessary condition of human 334

Mind, Knowledge and Reality: Themes from Kant awareness rather than a necessary condition of any sensible awareness of objects, then it cannot also be a property of objects or things as they are in themselves. The spatial perspective would simply be our perspective on the world, a mode of representing objects which is peculiar to us, and properties which are simply a reflection of the constitution of our faculty of intuition cannot properly be attributed to the world as it is in itself. Things in themselves are things considered independently of the subjective conditions of human sensibility, so if space and time are such conditions, then spatial and temporal predicates cannot meaningfully be ascribed to things as they are in themselves. This response to (S3) raises the following question: in arguing from the premise that space is only a condition of human awareness of objects to the conclusion that spatial predicates cannot properly be ascribed to things as they are in themselves, is the idealist not guilty of begging the question against transcendental realism? According to transcendental realism, things in themselves are spatial, and this means that spatial awareness is not just a subjective condition of human sensibility but a necessary condition of any empirical knowledge of objects. Thus, according to transcendental realism, one would only have grounds for thinking that space is merely a condition of human awareness rather than of any sensible awareness of objects if one is already persuaded of the non-spatiality of things as they are in themselves. It would appear, therefore, that the thesis that space is a species-specific condition of awareness of particularity cannot be regarded as a non-question-begging premise in an argument for transcendental idealism. By the same token, however, the thesis that things as they are in themselves are spatial can hardly be regarded as a non-question-begging premise in an argument against the idealist's conception of the scope of epistemic conditions. The idealist is hardly likely to accept that his humanist conception of the scope of epistemic conditions is inadequate because things as they are in themselves are spatial, any more than the realist is likely to accept that spatial predicates cannot meaningfully be ascribed to things as they are in themselves because epistemic conditions are merely subjective. At this point, it might seem that the idealist and realist have reached a stand-off, with each accusing the other of begging the question. I will have more to say about this apparent stand-off when I give a more detailed account of realism about a priori conditions, but first there is another objection to idealism to discuss. So far, I have represented transcendental idealism as relying upon the thesis that epistemic conditions are species-specific in scope, but it is not clear that this thesis is faithful to Kant. To begin with, I have 335

Quassim Cassam already remarked that Kant does not represent the categories as species-specific conditions of knowledge. One reaction to this observation would be to argue, in response to (Ul), that Kant was wrong to be a universalist about the categories, but this attempt to bring Kant's theory of understanding into line with his theory of sensibility is arguably guilty of misunderstanding the latter. For while Kant does say that it is solely from the human standpoint that we can speak of space, he adds that spatio-temporal intuition 'need not be limited to human sensibility. It may be that all finite, thinking beings necessarily agree with man in this respect, although we are not in a position to judge whether this is actually so' (B72). If all finite, thinking beings necessarily agree with man in this respect, then spatio-temporal intuition is not a species-specific condition of sensibility. To the extent that Kant is agnostic on the question of whether space and time are species-specific conditions of sensibility, it would not bring his theory of understanding into line with his theory of sensibility to insist upon the species-specificity of the categories as conditions of knowledge. By the same token, it might be held to be a misunderstanding of Allison to represent him as maintaining that epistemic conditions are species-specific in scope. It might be argued, instead, that the point of the qualification 'human' in his account of epistemic conditions is to mark a distinction between finite cognition and the 'standard of cognition theoretically achievable by an "absolute" or "infinite" intellect'.31 With regard to the representation of space, therefore, the claim is that it constitutes a universal and necessary condition in terms of which alone any finite intelligent being is capable of recognizing something as an object. A finite intelligent being is one whose intuition is sensible and whose understanding is discursive. Since any knowledge that involves the exercise of sensibility is empirical knowledge, finite cognition is empirical cognition. So the claim that a given condition is a necessary condition of all finite cognition amounts to the claim that it is a necessary condition of any empirical knowledge of objects, and this is precisely what the universalist wishes to maintain. Since humanism about the conditions of sensibility has so far been represented as holding the key to transcendental idealism, what would remain of idealism once it is conceded that space, time, and the categories are not just conditions of human knowledge but conditions of any empirical knowledge of objects? One response to this question would be to argue that idealism remains a distinctive position as along as it continues to insist that a priori or epistemic conditions are species-specific in origin. In other words, the fact 31 Ibid., p. 19. 336

Mind, Knowledge and Reality: Themes from Kant that space, time and the categories are universal conditions of empirical knowledge is to be seen as a reflection of the 'cognitive structure of the human mind'32 rather than the nature of things as they are in themselves. Presumably, the cognitive structure of the human mind consists in the fact that space and time are the forms of human sensibility and that the categories are the rules of our understanding. On this reading, Kant's agnosticism about the scope of space and time as conditions of sensibility is not essential to his idealism; what is essential is his conception of the human origins of a priori conditions. At this point, however, a generalized version of (U3) becomes especially pressing, for it might be wondered how conditions which are universal in scope can still be species-specific in origin. The difficulty is that unless all finite minds are assumed to be like human minds, it cannot coherently be supposed that the cognitive structure of the human mind is the source of conditions which bind all finite minds that are capable of empirical knowledge of objects. Given the assumption that finite intelligent minds might be of very many different types, the cognitive structure of the human mind can surely only be the source of epistemic conditions understood as the universal and necessary conditions of human cognition. So if, as Kant says, the categories have their first seeds and dispositions in the human understanding, then he is not entitled to regard them as universal in scope. If they are universal scope, then it is not the human understanding that is the source of the categories, but the Understanding in a more generic sense. More generally, a priori conditions which are truly universal in scope must be seen as reflecting the structure of 'mind as such' rather than the cognitive structure of the human mind. There are several difficulties with this proposal. The first is that it remains to be seen whether anything can be made of the notion of 'mind as such'. The second is that very little remains of idealism about a priori conditions once it is accepted that such conditions are not only universal in scope but also that they are not species-specific in origin. In the light of these difficulties, it might be worth exploring the anthropocentric position outlined above. Anthropocentrism, it will be recalled, is the view that all minds must be conceived of on an analogy with our own, and that what the universalist calls 'mind as such' is nothing more than an abstraction from certain abilities and dispositions that are possessed by live human beings. As Jonathan Lear puts it, in connection with what he takes to be Wittgenstein's conception of mind, anthropocentrism in its strongest form maintains that 'the concept of being minded in 32 Ibid., p. 29. 337

Quassim Cassam any way at all is that of being minded as we are'." Given this conception of mindedness, it is no longer mysterious how the structure of the human mind can be the source of a priori conditions which are universal in scope. If the concept of being minded in any way at all is that of being minded as we are, then the conditions of empirical knowledge which bind human minds must also bind all minds which must be conceived of on an analogy with human minds, that is to say, all minds. This attempt to make something of the suggestion that nonspecies-specific a priori conditions might still be grounded in the structure of the human mind is only as good as the case for anthropocentrism. A non-anthropocentric, 'objective' conception of mind would be one which insists, with Nagel, that 'we must think of mind as a phenomenon to which the human case is not necessarily central, even though our minds are at the center of our world'.34 Human minds are instances of mind in general, but what entitles us to regard ourselves as the central instances? To this it might be replied, on behalf of anthropocentrism, that for something to be a mind at all it must be interpretable by us, and that we can only interpret minds which resemble our own. Presumably, however, someone who is opposed to anthropocentrism will reject this argument on the grounds that its assumption that mindedness requires interpretability by us presupposes, and so is not an argument for, the notion that mind is a phenomenon to which the human case is necessarily central. Fortunately, it is not necessary to resolve this dispute here since it is clear that Kant's conception of mind is not anthropocentric. Far from accepting that all minds must be like human minds, Kant repeatedly contrasts our minds and the distinctively human perspective on reality with the minds and perspectives of beings who might be very different from ourselves.'5 One contrast is that between human mindedness and the mindedness of the primordial being, whose intuition would be intellectual rather than sensible. Another contrast is that between ourselves and other finite thinking beings whose sensibility might be, for all we know, nonspatio-temporal. If Kant were to accept the anthropocentric proposal that these contrasts are in some way illegitimate, he would hardly have chosen to express his idealism by saying that space is 33 'The Disappearing "We"', Proceedings of the Aristotelian Society, Supplementary Volume 58 (1984), 233. 34 Thomas Nagel, The View From Nowhere (Oxford: Oxford University Press, 1986), p. 18. 35 This aspect of Kant's position is rightly emphasized by Lear in 'The Disappearing "We"', 232. 338

Mind, Knowledge and Reality: Themes from Kant something which we can only speak of from the human standpoint, and that we cannot judge in regard to the intuition of other thinking beings whether they are bound by the same conditions as those which limit our intuition. Unlike Wittgenstein, therefore, Kant does not accept that the concept of being minded in any way at all is that of being minded as we are. So while it is arguable that anthropocentrism is itself a form of idealism, it is not idealism in Kant's sense. The position, then, is this: idealism was introduced as the view that a priori or epistemic conditions are species-specific in both scope and origin. It was then argued that idealism fails to give a satisfactory explanation of how a priori conditions can be species-specific in scope, since the considerations which support the view that space, time and the categories are the universal and necessary conditions of human cognition also support the view that they are among the universal and necessary conditions of all empirical knowledge. An attempt was then made to preserve something of the spirit of idealism about epistemic conditions by arguing that even if such conditions are universal in scope, they might nevertheless reflect the cognitive structure of the human mind. It was then argued that this hybrid position is unsatisfactory, and that Kantian idealism about a priori conditions is not helped by an anthropocentric conception of mind. In the light of these difficulties with idealism, the time has come to explore the prospects for realism. In the course of this exploration, I will return to what was referred to above as the simplest and most obvious idealist case for insisting that the conditions of human knowledge are species-specific in scope, namely, the idea that so-called epistemic conditions can only be a reflection of human imaginative limitations. Realism is the view that a priori conditions of the possibility of experience or empirical knowledge are objectively necessary conditions of the possibility of experience or empirical knowledge. To regard a priori conditions as objectively necessary is to regard them as universal in scope - that is, as necessary conditions of any empirical knowledge of objects - and as reflecting the nature of the objects of empirical knowledge as they are in themselves. These two aspects of realism are connected in the following way: a priori conditions are universal in scope because they must be faithful to the character of things in themselves. The underlying point here is that conditions which have no objective basis, in the sense that they do not accurately reflect the character or constitution of reality as it is in itself, could not properly be described as conditions under which knowledge of reality is possible. On the other hand, conditions which are objectively based must be universal in scope, since conditions 339

Quassim Cassam which reflect the nature of things as they are in themselves are, precisely, conditions which are not grounded in the peculiarities of the human cognitive apparatus. The attractions of realism, together with some of its limitations, may be brought out by examining P. F. Strawson's writings on Kant. Strawson represents Kant as pushing to the limit the distinction between intuitions and concepts, 'trying to extract as much as he can of the a priori conditions of empirical knowledge or experience from a consideration merely of one half of the distinction, namely, the necessity of bringing particular objects of experience — whatever the forms of particularity may be - under general concepts'.36 The bringing of objects under concepts is identical with the making of judgements about objects, and it is 'an analytic truth that the judgement involves concepts, that concepts are such as to be applicable or inapplicable to one or more instances, that judgements or propositions are capable of truth or falsity'." The problem with Kant's position is that reflection on the conditions of making objective judgements seems unlikely to 'yield much of a harvest in the way of categories',38 but there is a relatively simple way of dealing with this difficulty. For we are creatures whose intellects are discursive and whose intuition is sensible; such creatures must in judgement, employ and apply general concepts to the objects of sensible intuition; the very notion of the generality of a concept implies the possibility of numerically distinguishable individual objects falling under one and the same concept; and once granted that objects are themselves spatio-temporal, then space and time provide the uniquely necessary media for the realization of this possibility in sensible intuition of objects.3" This account has a two important consequences. The first, which distinguishes realism from subjectivism as well as idealism, is that spatio-temporal intuition now appears as 'a uniquely fundamental and necessary condition of any empirical knowledge of objects'.4" The second is that once objects of experience are assumed to be spatio-temporal, there are indeed particular ways in which we must conceive of them if we are to be able to make objective judgements '" The Bounds of Sense, pp. 77-8. " Strawson, 'Sensibility, Understanding, and the Doctrine of Synthesis: Comments on Henrich and Guyer', in Kant's Transcendental Deductions: The Three 'Critiques' and the 'Opus postumum', ed. E. Forster (Stanford: Stanford University Press, 1989), p. 71. 18 The Bounds of Sense, p. 81. " Strawson, 'Kant's New Foundations of Metaphysics', pp. 239-40. 40 Ibid., p. 240. 340

Mind, Knowledge and Reality: Themes from Kant about them; specifically, such objects must be conceived of as spaceoccupying substances that are subject to causal law, and this is the best that can be done to explain the status of the categories of substance and causality as a priori conditions of empirical knowledge.41 There can be little doubt that Strawson's account if a priori conditions is realist rather than idealist or subjectivist. Strawsonian a priori conditions are universal in scope, and they are grounded in the supposed nature of things in themselves, as well as certain analytic truths about the nature of judgement and of concepts. The crucial difference between Strawson and Kant is that Strawson is a transcendental realist about space and time, for it is the assumption that objects are in themselves spatio-temporal which underpins the Strawsonian derivation of the categories of substance and causality and its universalism about the necessary conditions of empirical knowledge. This version of realism yields the following responses to the questions outlined above. With regard to (SI), Strawson's universalism and his transcendental realism clearly commit him to answering this question in the negative. This also means that (S3) lapses, since it is not solely from the human standpoint that we can speak of space. In connection with (S2) and (U2), Strawsonian a priori conditions are supposedly analytically or conceptually necessary conditions, so the realist's investigation of sensibility and understanding must be conceptual rather than empirical or psychological. This is the essence of realism's disagreement with subjectivism. Given certain assumptions about the nature of things in themselves, it is not contingent that we have 'just the forms and functions of judgement and just the spatial and temporal forms of intuition that we do have'.42 Unlike Kant, therefore, Strawson is not faced with the difficulty of explaining how it is possible to have a priori knowledge of allegedly contingent aspects of the functioning of the human cognitive apparatus. With respect to (Ul), the fact that the employment of categories such as those of substance and causality is a conceptually necessary condition of the possibility of making judgements about spatiotemporal objects is not in itself a reason for maintaining that the categories are necessary conditions of all discursive thinking, unless it is also assumed that the making of judgements about spatio-temporal objects is a necessary condition of discursive thinking. Whether or not such an assumption would be defensible, it might be enough for the purposes of realism about a priori conditions that the categories can be known, on broadly conceptual grounds, to represent the universal intellectual conditions of empirical knowledge 41 Cf. The Bounds of Sense, pp. 82-5. 42 Strawson, 'Kant's New Foundations of Metaphysics', p. 237. 341

Quassim Cassam of a unified spatio-temporal world. Since the categories are, in this sense, universal in scope, there is no reason to suppose, in connection with (U3), that they have their first seeds and dispositions in the human understanding rather than in reflection on the conceptually necessary conditions of the possibility of experience of an objective and spatio-temporally unitary world. The most serious objection to realism, at least in its Strawsonian form, is that it begs the central question in its dispute with idealism. Once granted that objects are themselves spatio-temporal, there might indeed be a case for maintaining that spatio-temporal intuition is a uniquely fundamental and necessary condition of any empirical knowledge of objects, but in a dispute between realism and transcendental idealism, such an assumption about things as they are in themselves is surely not one to which the realist is entitled without a great deal of further work. The question, then, is whether realism has an argument for the thesis that objects are in themselves spatio-temporal, or whether this thesis functions as an unargued premise in realism's defence of its distinctive conception of the scope and basis of a priori conditions. In the absence of an argument for transcendental realism about space and time, idealism about a priori conditions is still in play. At this point, the realist might proceed as follows: 'objects', it will be recalled, are to be understood as particular instances of general concepts, and the basis of the assumption that objects are in themselves spatio-temporal is that 'spatio-temporal position provides the fundamental ground of distinction between one particular item and another of the same general type, hence the fundamental ground of identity of particular items'.43 Once granted that spatio-temporal position provides the fundamental ground of identity of particular items, it must also be granted that such items are in themselves spatio-temporal. So far from functioning as an unargued premise in the context of realism about a priori conditions, this thesis about the nature of objects as they are in themselves is supported by realism's theory of identity. As well as calling transcendental idealism into question, this theory also undermines subjectivism about a priori conditions. If spatio-temporal position provides the fundamental ground of identity of particular items, then the need for us to perceive the world as spatio-temporal cannot properly be seen as nothing more than a reflection of contingent, empirically discoverable facts about the constitution of our faculty of intuition. This argument for realism about a priori conditions is unlikely to convince the transcendental idealist. Kant had no difficulty accepting that 'difference of locations, without any further conditions, 43 The Bounds of Sense, p. 49. 342

Mind, Knowledge and Reality: Themes from Kant makes the plurality and distinction of objects, as appearances' (A272/B328), but he disputed the assertion that this tells us anything about the nature of objects as they are in themselves. For Strawson, the thought (T) that there might be general concepts such that 'we could encounter and distinguish in experience different particular instances of those concepts, and yet such that their instances were not spatio-temporal things at all', is a thought which 'leaves us quite blank',44 but this still leaves the idealist with a certain amount of room for manoeuvre. In the first place, it might be claimed that it fails to exclude the possibility that 'the things which we intuit are not in themselves what we intuit them as being; (A42/B59). In the second place, to say that (T) leaves us blank is to make a point about our imaginative limitations, and there is a quite general question about the legitimacy of taking our imaginative limitations as a reliable guide to the nature of reality as it is in itself. The second of these points leads back to the simplest and most obvious case for idealism about a priori conditions. For when realism claims that a priori conditions are grounded in the nature of reality as it is in itself, the idealist will want to know the basis upon which the realist claims an insight into the character of things in themselves. The fact that (T) leaves us blank does not entitle us to draw any conclusions about things in themselves since the world as it is in itself is not limited by what we humans can conceive of. Realism is only entitled to regard the conditions which it identifies as objectively necessary given certain assumptions about reality as it is in itself, but the mind-independence of reality as it is in itself means that any attempt to argue for these assumptions on the basis of our alleged imaginative limitations is mistaken in principle. Since the thesis that reality as it is in itself is spatio-temporal is one for which the realist lacks an adequate defence, the most that can be concluded from the discussion of (T) is that spatio-temporal intuition is for us a condition of empirical knowledge. In response to this argument, the realist should concede that there is, in principle, a gap between claims about what we can and cannot conceive of and claims about how the world is in itself, but insist that it would be inappropriate to regard our inability to make anything of (T) simply as a reflection of how things are with us. According to the realist, to suppose that this inability has no bearing at all on the nature of things in themselves is to be committed to what can only be described as a form of scepticism about reason. Just as scepticism about the senses questions the ability of our senses to deliver knowledge of mind-independent reality, so 44 P. F. Strawson, Subject and Predicate in Logic and Grammar (London: Methuen & Co. Ltd., 1974), p. 16. 343

Quassim Cassam scepticism about reason questions the ability of rational reflection concerning the fundamental ground of identity of particular items to deliver knowledge of the nature of such items as they are in themselves. In reply, it needs to be pointed out that while there may be no absolute guarantee that rational reflection is a reliable guide to how the world is in itself, the validity of assumption that we are estranged from reality to the extent implied by the sceptic is, to say the least, no more obvious than that of the assumption that reality is, in Craig's words, 'transparent to our faculties'.45 The initial idealist argument against realism was that the latter's conception of the nature and status of a priori conditions rests upon an unargued realism about space and time, but it should now be clear this is not a charge that can be made to stick. It is one thing to say that the considerations to which realism appeals in support of its conception of reality as it is in itself are not scepticproof, but this is not to say that these considerations do not even amount to an argument. The current state of play might be summarized as follows: when it is claimed that a given condition (C) is an objectively necessary condition of empirical knowledge, it needs to be explained both what makes it the case that (C) is such a condition and how we can know that (C) has this status. If (C) is an a priori condition in virtue of certain truths about the nature of the objects of our knowledge as they are in themselves, or in virtue of analytic truths concerning the nature of judgement or concepts, then it would seem that knowledge that (C) is an a priori condition requires knowledge of those truths in virtue of which (C) is an a priori condition. According to one version of realism, knowledge of these truths is made possible by rational reflection. The idealist is, in effect, someone who questions the ability of rational reflection to account for knowledge of a priori conditions that are objectively necessary and universal in scope, but there are now two things to be said in response to idealism. First, even if it has succeeded in identifying an epistemological problem for realism, the moral is not that genuinely a priori conditions are species-specific in scope; the moral is that we do not know which conditions are genuinely a priori. Secondly, by associating itself with scepticism about reason, idealism lays itself open to the charge of paying an unacceptably high price for its anti-realism. To mistrust reason is to mistrust a basic cognitive faculty, and if reason or rational reflection cannot even provide us with knowledge of the objectively or analytically necessary conditions of judgement, 45 Edward Craig, 'Arithmetic and Fact', in Exercises in Analysis: Essays by Students of Casimer Lewy, ed. I. Hacking (Cambridge: Cambridge University Press, 1985), p. 91. 344

Mind, Knowledge and Reality: Themes from Kant then there seems no reason to trust any of our basic cognitive faculties.46 While these considerations help to bring out the unattractiveness of some anti-realist arguments, they fail to address several residual anxieties about realism. One has to do with the role of realism in Kantian epistemology. On one interpretation, a major element of Kantian epistemology is the devising of anti-sceptical transcendental arguments. Suppose that P is a proposition about mind-independent reality, and that it is in question whether we can know that P is true. A transcendental argument responds to the sceptic about P by arguing that the truth of P is a necessary condition of something which is not and cannot coherently be doubted by the sceptic, namely, experience. On the face of it, the necessary conditions of the possibility of experience which figure in the minor premises of transcendental arguments are precisely the a priori conditions that are at issue between realism, idealism and subjectivism. Suppose, then, that one's interest in the nature and basis of a priori conditions is the result of an interest in the anti-sceptical potential of transcendental arguments. Realism would then face the objection that in conceiving of a priori conditions as objectively necessary conditions, it renders transcendental arguments wholly superfluous in its dispute with scepticism. For, to the extent that objectively necessary conditions are grounded in propositions about mind-independent reality which we can know to be true on the basis of rational reflection, it would seem that the same rational reflection ought to be capable of providing us with knowledge of the truth of those propositions about mind-independent reality which are disputed by the sceptic. The difficulty, in other words, is that as long as rational reflection is a self-sufficient source of knowledge of things as they are in themselves, there is no need for the epistemological reassurance allegedly promised by transcendental arguments. If, on the other hand, rational reflection cannot tell us about mind-independent reality, then it cannot tell us which conditions of the possibility of experience are objectively necessary. In that case, it is difficult to understand how transcendental arguments can carry much weight against scepticism, since the sceptic will claim that we have no insight into the necessary conditions of the possibility of experience. There are several things that might be said in response to this anxiety about realism. It might be held, for example, that it is 46 These remarks are only intended to give a very rough indication of how an effective response to scepticism about reason might go. On their own, they are unlikely to persuade those who question the ability of rational reflection to deliver knowledge of things in themselves. 345

Quassim Cassam enough for the purposes of transcendental arguments that the a priori conditions to which they appeal are in fact faithful to how things are in themselves. Once it is conceded that transcendental arguments do not presuppose knowledge of those truths about things in themselves in virtue of which their minor premises are objectively necessary, there is no need to try to account for the possibility of such knowledge in terms of rational reflection. It would also be worth remarking that if the minor premise of a transcendental argument is analytic, then its epistemological standing ought to be no more problematic than the epistemological standing of any analytic judgement.47 While it may not be analytic that things in themselves are spatio-temporal or that spatio-temporal intuition is a necessary condition of empirical knowledge, this is not a reason for the realist not to attempt to construct transcendental arguments which exploit a priori conditions that are objectively necessary in virtue of being analytic.48 Finally, the capacity of rational reflection to deliver knowledge of the objectively necessary condition that figures in the minor premise of a given transcendental argument is no guarantee of its ability to disarm the specific version of scepticism that is the target of that argument. So even if the minor premise of a transcendental argument presupposes the kind of knowledge of how things are in themselves that is the product of rational reflection, this need not render the argument superfluous as long as the proposition about things in themselves which underpins the realist's a priori condition is not the very proposition about mind-independent reality that is under sceptical attack. Each of these responses to the first residual anxiety about realism deserves to be examined in greater detail than is possible here, but it is important not to exaggerate the significance of this discussion. The most that the first anxiety shows is that a certain kind of realism about a priori conditions raises a question about the role of transcendental arguments in Kantian epistemology, but it is not obvious that an interest in the nature and scope of a priori conditions needs to be motivated by an interest in the anti-sceptical potential of transcendental arguments. Nevertheless, the kind of realist who is not preoccupied by scepticism, and who thinks that rational reflection is potentially a source of knowledge of things as they are in themselves, is still under an obligation to say more about 47 This is, of course, not to suggest that the epistemological standing of analytic knowledge is wholly unproblematic. 48 One question for someone who argues in this way is whether, as has so far been assumed, analytically necessary conditions are, in the realist's sense, objectively necessary conditions. I will not pursue this difficult issue here. 346

Mind, Knowledge and Reality: Themes from Kant the nature of rational reflection and its alleged ability to provide an insight into the objectively necessary conditions of the possibility of experience. The second residual anxiety about realism, then, is that its epistemology is incomplete in certain important respects. Despite these reservations, realism remains the most attractive of the three positions considered above. The capacity of rational reflection to yield knowledge of the objectively necessary conditions of the possibility of experience is certainly something which needs explaining, but it would be unwise for the idealist to attempt to make too much of this gap in realism's epistemology. For acceptance of the thesis that a priori conditions are epistemic conditions rather than objectively necessary conditions would still leave one with the task of explaining how knowledge of epistemic conditions is possible, and this is arguably as difficult a task for idealism as the corresponding task for realism. If epistemic conditions reflect the cognitive structure of the mind, then knowledge of epistemic conditions is presumably a form of self-knowledge, but it would be a mistake to suppose that the relevant form of self-knowledge is any easier to account for than our knowledge of objectively necessary conditions. When it comes to developing a credible epistemology of a priori conditions, both realism and idealism have more work to do, but the reason for preferring realism is that it is the only position which does justice to the point that the conditions of empirical knowledge must reflect the nature of the objects of empirical knowledge as they are in themselves, and cannot simply be a reflection of the structure of our cognitive faculties. Idealism is unacceptable because it fails to respect this fundamental point, and fails to demonstrate that this is a point which does not deserve to be respected. IV I began by contrasting humanism and universalism in epistemology, and by suggesting that this contrast also has a bearing on our understanding of the subject matter of the philosophy of mind. In epistemology, the universalist is someone who thinks that the theory of knowledge must concern itself with the uncovering of conditions of empirical knowledge which are universal in scope, conditions which must be fulfilled by any knowing subject, human or otherwise. Similarly, a universalist philosopher of mind is one who maintains that the philosophy of mind must concern itself not just with the human mind but with the nature of mind as such. How are these two versions of universalism related, and what is to be made of the 347

Quassim Cassam notion of 'mind as such'? Realism suggests the following response to those questions: if 'mind as such' is understood as the knowing mind, then one way of arriving at an understanding of its nature would be to arrive at an understanding of the conditions that must be fulfilled by any mind, human or otherwise, which is capable of empirical knowledge. These are precisely the conditions which interest the universalist epistemologist. Suppose, next, that realism is right to insist that it is in virtue of their objective basis that the a priori conditions of empirical knowledge are universal in scope. In that case, an account of mind as such must, as Locke might have said, take into consideration the inherent nature of the objects which knowing minds have to do with. This is not to suggest that the peculiarities of the human cognitive apparatus should be of no interest to the universalist. It is to suggest, however, that knowledge of the structure of mind as such cannot be detached from knowledge of the world in which knowing minds, including human minds, are embedded. 348

The Modality of Freedom CHRISTOPHER PEACOCKE I The Problem as an Instance of the Integration Challenge The classical problem of free will is one instance of the Integration Challenge. The Integration Challenge in its general form is that of reconciling our metaphysics of any given area with our epistemology for that same area. In the case of free will, the challenge is that of reconciling our seeming first-person knowledge of our exercise of free thought, deliberation, choice and action with a description of what is really going on in the world as characterized in terms of causation, determination, explanation and causal possibility. There are at least six general theoretical options to be considered whenever we are faced with a philosophical problem which is an instance of the Integration Challenge. These options divide into two groups. Each of the two groups comprises three of the six options. The first group contains the conservative options of (i) providing an improved metaphysics which meets the challenge; (ii) providing an improved epistemology; and (iii) providing an improved conception of the relations between the appropriate metaphysics and epistemology. These options are evidently not exclusive of each other. Each of these options aims at head-on reconciliation. The options are conservative in the sense that proper, successful development of one of these options will attribute some truth condition to the problematic sentences, and will explain how we can come to know that they obtain, when we do. In the case of free will, development of one of these options will involve attributing to a sentence such as 'He chose freely' a truth condition which aims to be a correct elucidation of what that sentence states, and which is knowably true pretty much in the cases, and in the ways, we normally take it to be. Of course development of these options may not be conservative of philosophical theory. Highly revisionary, innovative theory may be required to develop the options in this first group. It is rather our pretheoretical views of which options in this first set aim to be conservative. Versions of this material were also presented in colloquia at Columbia University and the University of Connecticut, Storrs. I thank Akeel Bilgrami, Paul Boghossian, Tyler Burge, John Collins, Paul Horwich, Kirstie Laird, Michael Martin, Ruth Millikan and Stephen Schiffer for comments and discussion, and the Leverhulme Trust for time. 349

Christopher Peacocke The second, revisionary group of options includes the option (iv) of offering some cleansed or reduced truth condition for the problematic sentences, a truth condition which captures some but not all of what we ordinarily mean by the problematic sentences, but which is purged of those features which were preventing integration. The next member of the second group is option (v), under which no truth condition is assigned to the problematic sentences: rather, under this option, some role of the problematic sentences in our assertoric, inferential or social practices is described, a role which is made rational or at least explained by something other than that role's being justified by the sentences' truth conditions. Finally in this group is option (vi), which claims that the relevant concepts expressed in the problematic sentences are quite incoherent, in such a way that not even options (iv) and (v) are available. The options in this group are each radical in the sense that they can only be rationally endorsed by someone who thinks that it is impossible to give a truth condition for the problematic sentences which both captures their pretheoretical meaning, and for which the Integration Challenge can be met. Such has been the recalcitrance of the problem of free will that the topic is perhaps the only one on which each of the more radical revisionary options have been endorsed by some recent thinker or other. The views of Peter Strawson1 and Christine Korsgaard2 are examples of the fifth option. Galen Strawson is one of those endorsing the sixth in print,3 and others endorse it in conversation. It is also of particular pertinence to the problem of free will that the distinction between conservative options and the radical option (iv) may be very hard to apply in some examples. If a concept is widely and firmly associated with some misconception or incorrect articulation, there may be some indeterminacy on the issue of whether some theoretical proposal about an underlying truth condition is or is not conservative of our pretheoretical conceptions - that is, on the issue of whether the proposal is really a case of (i) or a case of (iv). My plan of action in this paper is first to articulate some aspects of our intuitive notion of freedom in decision, actions and thoughts. Then I will attempt an account of 'could have done otherwise' which seems to meet the Integration Challenge by building a theory 1 P. Strawson, 'Freedom and Resentment', in Freedom and Resentment and Other Essays (London: Methuen, 1974). 2 C. Korsgaard, 'Creating Kingdom of Ends: Reciprocity and Responsibility in Personal Relations', in Creating the Kingdom of Ends (Cambridge: Cambridge University Press, 1996). 3 G. Strawson, 'The Impossibility of Moral Responsibility', Philosophical Studies 75 (1994), 5-24. 350

The Modality of Freedom which is either conservative or at worst indeterminate as between conservatism and a form of case (iv). I will try to respond to those arguments which have pressured others, in the case of free will, into the revisionary options (v) or (vi), or into competing conservative conceptions. In terms of the traditional labels, the conservative conception I will be advocating is compatibilist, but it is of a rather different stripe of compatibilism than that made familiar by G.E. Moore.4 My own view is that we should aim, if at all possible, for an account which gives a genuine and satisfiable truth condition for an attribution of freedom. Though the point would be disputed by some distinguished writers — to whose views I will turn later on — it seems to me implausible that an attribution of freedom involves no factual as opposed to practical commitments; or that it involves on the factual side nothing going beyond the immediate phenomenology of decision-taking. Consider one of Penfield's experiments, in which Penfield inserts an electrode into patient's brain, and fires it. As a result, the patient says he has spontaneously taken the decision to do something. I doubt anyone would happily classify this as a free decision, even though it may have the phenomenology of one. In more ordinary cases, we can make sense of suggestion that our decisions in some area are really the result of some neurosis, and that our decisions are thus not freely made. It needs empirical investigation to confirm or refute any such hypothesis. It cannot be ruled out just by first-person consciousness of the decision taker. It may be said the existence of factual commitments in an attribution of freedom is consistent with a non-truth-conditional view (that is, with option (v)). It is just that there are certain conditions - manipulation by others, neuroses, which are on a list sufficient for legitimate assertion that the agent is not freely thinking, deciding or acting. It seems, though, prima facie implausible that our understanding of a predication 'x is deciding freely' involves tacit knowledge of a mere list, and nothing more. We seem to have an openended ability to classify new examples as free or unfree, in a way going beyond anything which could be captured by a list of cases. We had a general conception of free decision prior to any theory of neuroses, and then applied that conception in classifying a new case, once neurosis had been identified. I doubt whether something so important to us as freedom could ever be captured by a list of conditions which do not have some deeper unifying characteristic. I am then aiming to give a metaphysical account of freedom, in the belief that this is the first step we need to take if we are ever to meet the Integration Challenge for attributions of freedom. It is 4 G. E. Moore, Ethics (London: Oxford University Press, 1947), chapter 6. 351

Christopher Peacocke impossible within the ambit of a paper to address all of the issues which could be raised for the position I will be developing. To identify a potentially occupiable position in logical space will be my main task. Of some of those who attempted to give metaphysical accounts, Nietzsche wrote: The desire for 'freedom of the will' in the superlative, metaphysical sense, such as still holds sway, unfortunately, in the minds of the half-educated, the desire to bear the entire and ultimate responsibility for one's actions oneself, and to absolve God, the world, ancestors, chance, and society, therefrom, involves nothing less than to be precisely this causa sui, and, with more than Miinchausen daring, to pull oneself up into existence by the hair, out of the slough of nothingness5. The metaphysical account I will be offering will be less than a 'superlative' metaphysical sense of the sort Nietzsche has in his sights. Something less than the superlative may suffice, however, consistently with recognition of those truths Nietzsche is emphasizing. II An Intuitive Characterization of Freedom The concept of freedom is organized around the notion of a person being free with respect to a factor. Such a factor might be either a prima facie reason for decision one way rather than another, or it might be some other factor which may influence one's decision in some wholly nonrational fashion. To be a thinker who is free with respect to a factor is to have the capacity to reflect on that factor and to decide effectively whether to let that factor influence one, and if so in what way. The claim that the heroin addict is not free, without assistance, not to act on his desire for the substance is the claim that he does not, unaided, have the capacity not to act on that desire. There are nonrational influences on choices which cannot be regarded as the operation of any kind of prima facie reason. Consider an agent who has to appoint one out of several candidates to some post. When comparing two candidates for a position, one of whom has been interviewed, and given a favourable impression, and the other of whom is known only from his CV, many people give, and indeed are aware that they give, undue weight to the favourable impression left by the interviewed candidate. It is an empirical question whether the appointing person has the capacity to over5 F. Nietzsche, Beyond Good and Evil, trans. H. Zimmern (Buffalo, NY: Prometheus Books, 1989), section 2. 352

The Modality of Freedom come this tendency, that is, whether he is free with respect to it. When it is operative, this tendency need not be operating as any kind of prima facie reason within the agent's deliberation. Given this characterization of freedom with respect to a factor, we can go on to say that a person is free with respect to a range of factors if he is free with respect to each one of them. Similarly, one person is more free than a second person if the range of factors with respect to which the first is free properly includes those with respect to which the second is free. One feature of this initial, intuitive characterization of freedom with respect to a factor is that it makes clear one of the links between conceptualization and freedom. We can deliberate only about what we can think about. A thinker must conceptualize some factor before it can enter his rational - or irrational - deliberations about whether it should influence him, and if so in which way. New theoretical knowledge, framed in concepts which are also new, may identify new factors in our decision-taking. It will again be an empirical question which of these newly identified factors are ones with respect to which we are free. In some cases the new theory may entail that we are not free in some range of our activities. But only when we have properly conceptualized the factors influencing us will there be any chance of our being, or becoming, free with respect to them. Another feature of the characterization of freedom with respect to a factor is that it applies to thought too. On this characterization, a thinker can be free with respect to the factors which may be influential when he is making up his mind what to think or judge. The capacity effectively to give weight to some prima facie reasons, and none to others, or to block otherwise influential nonrational factors, is one that can be present in making judgements just as it is present in taking practical decisions. My own view is that this is no accident, for judging just is one species of action. There are necessarily certain limits to freedom in the case of thought. The rational thinker who accepts that all Fs are Gs is not free to judge that something is both F and not-G. This point cannot be dismissed by the observation that we should distinguish causal from rational determination. We should of course distinguish them. Many theorists of concepts would, however, insist that a proper theory of possession of, for instance, logical concepts entails that certain states of acceptance stand in causal relations to other content-involving states, including relations of production and exclusion. It is partially constitutive of a concept's being that of universal quantification that a thinker rejects the thought that all Fs are Gs when he comes to accept that something is both F and not353

Christopher Peacocke G. However, the apparent loss of freedom such theories entail should not be thought of as something to be regretted. If certain combinations of thoughts cannot be simultaneously accepted, as a consequence of a philosophical account of the possession of particular concepts involved, then there is, on a priori philosophical grounds, no possibility of still employing those concepts whilst not being subject to those constraints. Losing the capacity to think certain thoughts is not a way of increasing freedom. In fact one paragraph back, we were noting just the contrary. It is a consequence of this intuitive characterization of freedom that an animal without the capacity to think of itself as influenced by a range of factors will not be free with respect to them. Nor does the definition make it suffice for freedom with respect to a range of factors that the agent has higher-order attitudes about those factors, or about his own attitudes to them. An agent may have secondorder attitudes which cause first-order attitudes, and thereby influence his actions. Such is the case of the Puritan who has a secondorder desire not to act on certain of his first-order desires. If our Puritan is not capable of preventing those second-order attitudes from influencing him, he is not free with respect to them, or with respect to the relevant first-order desires. To have, for any given factor that may influence one, the capacity to decide effectively whether to let it influence one is to be distinguished from the following capacity: the capacity to decide, for all the factors which may influence one, whether to let all of them influence one. To suppose that the first capacity requires the second is analogous to moving from the premise that one has, for each book in the British Library, time to read it before one dies, to the conclusion that one has time to read them all before one dies. Actually, it is worse than that transition, for the conclusion, in the case of factors that may influence one, is incoherent, and not merely, in Russell's phrase, a medical impossibility. Whatever process takes place of scrutinizing and weighting factors which may influence one, once a decision is made, there is always something left unscrutinized, on pain of the task being uncompletable. Such scrutiny is, in the phrase Ryle memorably used to describe the impossibility of a certain kind of self-prediction, 'logically condemned to eternal penultimacy'.5 The point applies also to the nonrational factors influencing the choice of which of several prima facie reasons is to be operative with one, if there is some explanation of why the agent makes the decision he does. Whatever the nonrational factors that affect that final choice, they will not themselves have been the subject of a decision on whether they should be influential. 6 G. Ryle, The Concept of Mind (London: Penguin, 1963), p. 186. 354

The Modality of Freedom A fully rational and free thinker does not need to aim at the incoherent goal of scrutinizing everything that may influence a decision. If a thinker is entitled to believe that further proper scrutiny of rational and nonrational factors will not affect the direction of his decision - is entitled to believe that his decision is stable as we may call it - then he can rationally cease the process of scrutiny. Ill 'Could have done otherwise': The Closeness Account When the intuitive description of being free with respect to a particular factor applies to a person, it will be true that the subject could have let the factor influence him, and could equally not have let it influence him. Whatever he decided, he could have decided and done otherwise. Accounts of the nature of the modality in 'he could have done otherwise' offered hitherto by compatibilists, and in particular the account offered by Moore, have been found very unconvincing by incompatibilists. I find them unconvincing too. Some have tended to move from this point in the discussion, together with the non-negotiability of some form of modal requirement in our ordinary conception of freedom, to an incompatabilist conclusion. That was certainly how I myself thought for many years. But I have come to think that there is a compatibilist option which has been overlooked, and this section is devoted to developing it. Suppose you travel on a train through the Channel Tunnel, and there is a fire in the engine. Suppose also that the only reason that the fire does not spread poisonous smoke through the ventilation system is that some luggage, which could easily have been placed in a different configuration, happens to set up a draught which diverts the smoke from the ventilation system. It is true to say of this situation that there could easily have been a fatal accident. This is the kind of 'could have' with which we are concerned when assessing safety. It also seems that this species of possibility is compatible with determinism. If it were determined, on this particular occasion, that the luggage be stored in that configuration, perhaps because of the particular practices of the individual baggage-handlers on duty that day, that is not enough to establish that that particular journey was safe - to establish that, in the particular sense with which we are concerned, there could not easily have been an accident. Intuitively, only small variations from the actual conditions, small variations which there is no occasion-independent mechanism preventing, would have resulted in an accident. Happily favourable initial conditions are not sufficient for safety, nor do they imply that there is no kind of possibility under which an accident was possible. 355

Christopher Peacocke The relevant kind of possibility is one under which something's being not possible means that in a certain way one can rely on its not obtaining. Another area in which we employ this kind of modality is epistemology. It would be widely agreed that if someone is in a region where there are, to use the time-worn example, unbeknownst to him, many convincing barn-facades scattered through the countryside, he cannot learn that something is a barn just by looking at it. This correct verdict on the case is unaltered if we lived in a deterministic, Newtonian universe, so that it is determined that it is a barn, and not a barn-facade, that he is now seeing. The method of taking such perceptual experiences at face value still cannot be relied upon in those circumstances. If conditions had been only slightly different — for instance, if our subject had turned left rather than right at the last junction - he would have been confronted with a barn-facade rather than a barn, and the method would have led him into error. We can hear some species of possibility in the statement 'The method could have led him into error' on which it is true, even in a deterministic world. Let us call the kind of possibility involved in the safety and in the knowledge examples 'closeness possibility'. Closeness evidently needs elucidation, but a great deal of what I have to suggest involves only the existence of such a kind of possibility, and is independent of particular analyses. So for the present, let us just specify that we are concerned with the kind of possibility involved in those examples, whatever its proper elucidation may be. A closeness elucidation of freedom could then be offered. It states that someone is free to F just in case a. he could (closeness possibility) try to F, and b. he would F if he tried to. Clause (b) may be negotiable down to 'he might F, if he were to try to'. There are some arguments for the stronger versions: it may be said that when only the weaker conditions are met, it is true only that the agent is free to try. There might also be some indeterminacy in the ordinary meaning; in any case, the difference will not be crucial to the issues I will be discussing. The position I will propose could equally be developed using the weaker version. So, according to the closeness elucidation, I am free to act on (free with respect to) a prima facie reason on which I do not in fact act, if (a) it could easily have been the case that I try to act on it, and (b) if I had tried to act on it, I would have succeeded in bringing about that for which it is a prima facie reason. The friend of the closeness elucidation will say that it is because we take it that this condition is fulfilled with respect to a particular factor that we count a particu356

The Modality of Freedom lar person as free with respect to it. By contrast, it is not true of the kleptomaniac wandering in the department store that it could easily be the case that she does not steal some object within her grasp. Our conception of kleptomania is that of a state which, when someone is in it, it is a prima facie law that the person will steal when the opportunity arises, and this state is not one which, when in the department store at least, she could easily rid herself of. The closeness account is a compatibilist account. It is, though, noteworthy that there are three points on which it actually agrees with the criticisms which incompatibilists have levelled against other compatibilist attempts at elucidation. 1. Successive generations of thinkers have complained against early Moore-style compatibilist accounts of freedom which say that 'He was free to F' means only that 'If had tried to F, he would have' (or perhaps with 'choose' instead of 'try'). The entirely compelling objection, voiced by A. Campbell Garnett,7 accepted by Moore,8 and also emphasized by Chisholm9 and Wiggins,10 is that one is not free to F if one is not free to try F, or is not free to choose to F. Berlin gives some further history of the point." Far from being an objection to the closeness elucidation, however, the point which all these thinkers rightly insisted upon is entailed by the closeness elucidation. The closeness elucidation entails that if there is no close world in which the subject tries to F, then he is not free to F. The point that one must be free to try to F if one is to be free to F is, then, not the exclusive property of incompatibilist positions. Its soundness can equally be explained on a properly marshalled compatibilist position which appeals to closeness. 2. A second point on which the closeness theorist will agree with a classical incompatibilist criticism of other compatibilist positions is this. Excusing conditions, the closeness theorist will insist, do not have to reduce merely to an unexplained list, with no underlying principle of unification. One class of excusing conditions will be unified by the condition that each is one whose obtaining implies that there is no close world in which the agent tries to act in the rel7 A. C. Garnett, 'Moore's Theory of Moral Freedom and Responsibility', in The Philosophy of G. E. Moore, Volume I, ed. P. A. Schilpp (La Salle, IL: Open Court, 1968). 8 G. E. Moore, A Reply to My Critics', in The Philosophy of G. E. Moore, ed. P. A. Schilpp, Volume II, (La Salle, IL: Open Court, 1968). 9 R. Chisholm, 'Human Freedom and the Self, in Free Will, ed. G. Watson (Oxford: Oxford University Press, 1989), esp. Chapter 4. 10 D. Wiggins, 'Towards a Reasonable Libertarianism', in Needs, Values, and Truth (Oxford: Blackwell, 1987). 11 I. Berlin, 'From Hope and Fear Set Free', in Concepts and Categories: Philosophical Essays, ed. H. Hardy (Oxford: Oxford University Press, 1980). 357

Christopher Peacocke evant way. On this approach, one will distinguish between those cases in which the excuse really does involve a lack of freedom, from those in which the subject is free to act, but the costs of acting are too high. Here I am in agreement with Williams.12 3. The modality involved in freedom, on the closeness account, is not merely epistemic. It is metaphysical. Closeness possibilities do not have to align in any straightforward way with epistemic possibilities. It may seem, from the best available information, that something could easily have been the case, when in fact because of some hitherto undiscovered scientific principles and other conditions which could not easily have been different, it could not easily have been the case. The converse is possible too: it may be that, unbeknownst to us, the earth could easily have been destroyed by a passing asteroid a century ago. Similarly, chaos theory in effect shows that many conditions which one might have thought could not easily have obtained in fact could have come about with only tiny changes from the way the actual world is. So, when incompatibilists object to those compatibilists who offer merely epistemic elucidations of 'could have done otherwise', our closeness compatibilist will agree. What is the relation between the notion of a close possibility which I have been using, and the closer than relation which is used in some possible-worlds semantics for counterfactuals? It should not be assumed that the notion of closeness I have been using is simply the positive form of some concept whose comparative is the closer than relation used in some possible-worlds semantics, not even under the supposition that some such semantics is correct. Suppose for present purposes that Lewis's semantics for counterfactuals is correct13. (We could equally make the corresponding points for Stalnaker's treatment14.) So that we have a notation which does not encourage any begging of the question, let us indicate the three-place relation used in Lewis's semantics with the expression 'world u is L-closer to world w than is world v\ Lewis's semantics states that an arbitrary counterfactual 'If A were to be the case, then C would be the case' is true at w iff some (accessible) world in which A and C are true is L-closer to w than any world in which A and ~"C are true. (A person could agree to this without accepting Lewis's own philosophical theory of the nature of the three-place relation 12 B. Williams, 'How Free Does the Will Need to Be?' in Making Sense of Humanity and Other Philosophical Papers (Cambridge: Cambridge University Press, 1995), p. 4. 13 D. Lewis, Counterfactuals (Oxford: Blackwell, 1973). 14 R. Stalnaker, 'A Theory of Conditionals', in Studies in Logical Theory, ed. N. Rescher (Oxford: Blackwell, 1968). 358

The Modality of Freedom L-closer(u,v,w).) Now let us return to consider the unsafe, but causally determined, train trip. Suppose that we also hold the following principles, which may be found tempting: (a) that backtracking counterfactuals ('If I were to strike the match now, something in the past would be different') are false; (b) that close worlds contain no violations of laws of nature. If we agree that there is a close possibility that there is a fire on the train, then any world-history in which this close possibility is embedded must have at some point a different past from that of our actual world, given the supposition of determinism (and no violations of the laws). If this world were an Inclose world, there is a danger that some backtracking counterfactuals will then be counted as true in cases in which they are not true. One way (though not the only way) out of this is to distinguish sharply between the closeness property I have been discussing, and any positive form derived from the relation of L-closeness. The matter could be given extensive separate discussion. At any rate, the important point for present purposes is that the notion of closeness I am using is introduced by way of the examples of our apparent use of it. Any connections between the notion so introduced and the relation L-closer(u,v,w) and others needed in modal semantics would need to be established by further argument. IV A Puzzling Inference We can clarify the closeness conception by considering a puzzle. The puzzle concerns a certain form of inference. Suppose someone is not free to be not-F, and it's causally (nomologically) necessary that if he is F then he is also G. Does it follow that he is not free to be not-G? We can abbreviate the inference in question thus: (1) ifree-'F (2) causally necessary (if F, then G) (L) (3) -ifree-'G, The puzzle emerges if we raise the question: is this inference, which I label '(L)', valid on the closeness conception of freedom? It may strike one as valid. It may also appear that there is a sound argument from the semantics of 'is free to' on the closeness conception to the validity of (L). The argument would run thus. It seems reasonable to suppose that laws of the actual world are also laws of close worlds. If that is so, then premise (2) of argument (L) implies that in any close world in which our agent is F, he's also G. Now suppose, contrary to the conclusion (3), that our agent were free to be 359

Christopher Peacocke ~"G. By the closeness account, this implies that there is a close world in which he tries to be ~<G. It would certainly be puzzling if under this approach to freedom, none of the close worlds in which he tries to be ~>G is one in which he succeeds. By (2), any close world in which he so tries and succeeds in being ~"G will also be one in which he is ""F. Won't he then be free to be ~"F, simply by trying to be ^ F by in turn trying to be "'G? This then contradicts (1). So, it may seem from the semantics, the argument-schema (L) must be valid. So far, no puzzle. The puzzle emerges only when we add that (L) is very close to, indeed something which has an instance, the form of argument classically used by libertarians, and by incompatibilists, in attempts to establish that freedom conflicts with determinism (van Inwagen;15 Wiggins16). Yet the closeness conception was put forward as a compatibilist elucidation of freedom. So somewhere on this short journey, a mistake must have been made, maybe more than one. What is it, or what are they? An analogue of (L) is indeed valid for some intuitive notion of determination. If we substitute 'determined' for '"'free"1', then I have no quarrel with the validity of the resulting schema. The same applies if we substitute a specifically determinism-related 'open' for 'free'. However, in offering a compatibilist elucidation of freedom, we will be developing an approach on which those substituted notions are distinguished from what they are replacing. We will be in agreement with David Lewis's point that not all ways of being determined not to do something are ways which amount to inability to do it.17 And indeed, (L) in its original form, a schema involving the notion of freedom, is in fact invalid. Consider an instance of (L), as incompatibilists commonly do, in which 'F' is replaced by some predication about some time t, and G about some later time t+n, and the operator 'is free to' is in both (1) and (3) understood as concerning the later time t+n. That is, we are considering the case: (la) He is not free at t+n to be "'F at t (2a) Causally necessary (if he's F at t, then he's G at t+n) (3a) He is not free at t+n to be ~"G at t+n. 15 P. van Inwagen, 'A Formal Approach to the Problem of Free Will and Determinism', Theoria 40 (1974), 9-22. 'The Incompatibility of Free Will and Determinism', Philosophical Studies 27 (1975), 185-99. 16 Wiggins, 'Towards a Reasonable Libertarianism'. 17 D. Lewis, 'Are We Free to Break the Laws?' in Philosophical Papers, Volume II (New York: Oxford University Press, 1986), p. 292. 360

The Modality of Freedom An agent is not free to change the past. That is uncontroversial on the intuitive understanding of freedom. It is also the verdict of the closeness elucidation, since an agent's being free to change the past would require its being true that if the agent tried to, he would - and of course he wouldn't. (No doubt the agent is free in some sense to have had a different past, but this just means that in the past, he was free to have acted differently.) So (la) is true. We can suppose G chosen so that (2a) is also true. Nonetheless (3a) may still be false. It is false if there is a close world in which he tries to be ~"G at t+n, and if it is true that were he to try to be ^G at t+n, he would succeed. Nothing in the premises (la) and (2a) rules out the holding of those two conditions. A close world in which he is G at t+n, and which has the same laws as the actual world, must of course have a different past from the actual world. But nothing in the premises (la) and (2a) rules out the existence of such a world, and nor do the other requirements on close worlds. The fallacy in the semantic argument occurred when it said 'Won't [the agent] then be free to be ~>F, simply by trying to be ""F by in turn trying to be ""G?" The answer to the quoted question is negative, in the case in which trying to be ~<G involves the agent's trying to bring about the truth of some proposition about a time earlier than that of his attempt, where the proposition is false of that earlier time. I am developing a treatment of the case on which it matters that some close worlds are worlds in which the past is different from the way it actually is. The intuitive examples by which we introduced closeness possibilities should make us recognise that. We said that the following combination is coherent: the train trip is not safe, even when it is determined from fortunate initial conditions that there will not be an accident. If lack of safety consists in the closeness of a world in which there is an accident, then that close world must be one whose initial conditions are also different from those of the actual world, if it has the same laws as the actual world. We could make a corresponding point about lack of reliability in the knowledge example. So, on this treatment, the holding of (4) the agent is not free at t+n to be "'F at t. does not imply that (5) there is at t+n no world close at t+n at which he is F at t. There may be such a close world. The closeness account provides for at least two different ways in which a statement of inability (or more strictly, unfreedom) may be true. First, it may be true because 361

Christopher Peacocke there simply is at t+n no close world in which he tries to do the thing and succeeds (or perhaps he can't even try). This is the case which includes my current, and no doubt permanent, inability to jump 8 feet high. The second way a statement of inability may be false is this: although there are close worlds in which the agent has the property in question, the counterfactual 'if he were to try to have the property, he would have it' is false. This applies to any case in which the property concerns a time prior to t+n, and is one the agent did not in fact have at t+n. (I return at the end of section V to address those incompatibilists who think this sort of defence weakens the compatibilist's position.) A less general form of inference (L+) is valid on the closeness conception, the form in which the time indices in F and G are the same. This instance of (L+) is a valid argument: (lb) He is not free at t+n not to keep the air pressure in the cabin constant. (2b) It's causally necessary that if he keeps the air pressure constant, the temperature stays constant. (L+) (3b) He's not free to vary the temperature in the cabin. Suppose, once again, that (3b) were false — that our agent is free to vary the temperature. Then we ask the question 'Won't he then be free not to keep the air pressure constant simply by trying to make it vary by varying the temperature?' Here the answer to the quoted question is affirmative. A close world in which he tries to vary the temperature and succeeds will also be one in which he varies the pressure, if the world has the same laws as the actual world. ((2b) assumes that we are outside the range in which the cabin blows up with increasing pressure.) The form (L+), unlike the more general (L), is valid. Its validity may help to explain the attraction of (L) to some incompatibilists. On the present theory, though, it cannot justify that attraction. V Elaboration of the Closeness Conception Kant claims that 'we must necessarily attribute to every rational being who has a will also the idea of freedom, under which only can such a being act'.18 Does a rational being also act under the idea of freedom as the closeness conception elucidates freedom? It seems that we do act under the idea that in engaging in ordinary practical 18 I. Kant, Ethical Philosophy, trans. J. Ellington (Inianapolis: Hackett, 1994), Grounding 488 on Akademie pagination. 362

The Modality of Freedom deliberation, the options we are considering are ones that we could try or decide to act on, and be effective in so acting or deciding. The 'could' in this claim seems to me to be the 'could' with a closeness elucidation. Suppose there is no close possibility in which we realize options other than the one actually chosen. I noted earlier that for the closeness 'could have', there is a correlative 'reliably', of such a kind that 'there is a close possibility that p' is equivalent to 'not reliably ~"p'. If there is no close possibility in which we realize options other than the one actually chosen, that is equivalent to its reliably being the case that we don't realize those other options. (This of course doesn't mean that the actual deliberation isn't causally effective.) It seems to me that ordinary rational deliberation about a range of options presupposes that those options are ones the deliberator could realize, where this is the 'could' of closeness possibility. A rational deliberator who became convinced that a certain subset of his apparent options are ones which he could not, in any close possible world, realize cannot rationally continue to include them in the range of options about which he is deliberating. It also seems to me that we want it to be the case that, over a certain range, we could try to act on desires or values other than those which were in fact operative with us on a particular occasion, and be effective in doing so. Freedom with respect to a factor can be something worth having. The qualification 'over a certain range' also matters. I would rather be of a psychological make-up of such a sort that there is no close world in which I can even bring myself to try to act cruelly, for instance. What is involved in properly assessing which worlds are close to a given world? What is the right way to assess whether there is a close possibility that a person, object or system of things be other than it actually is? Three kinds of factor should enter the assessment. i. The ceteris paribus laws of a given world w are preserved at worlds which are close to w. That is, they are preserved as ceteris paribus laws. In worlds close to the actual world, it is also true that ceteris paribus a rise in interest rates in a country produces an inflow of capital to that country, true that ceteris paribus meandering rivers erode their outer banks, and so forth. There is no close possibility of something which involves violation of a ceteris paribus law, that is, involves failure of a law in conditions which there is not some independent reason, of a sort which would apply equally in the actual world, for declaring that other things are not equal. ii. We have a conception of some properties and relations of a given system of objects at a given time as being much more robust than others, and these, again ceteris paribus, are also preserved as properties and relations of that system, at that time, in close worlds. 363

Christopher Peacocke Precisely what we aim to put in place in making the train safe are devices which, for instance, detect smoke, insulate from heat or will not shatter dangerously, in a wide variety of conditions. (This is closely related to Nancy Cartwright's notion of an object's having a certain capacity).19 iii. Assessment of which possibilities are close depends not only on the factors (i) and (ii), but also on what, contingently, is the case outside the system for which close possibilities are in question. Consider again, for instance, whether the earth could easily have been destroyed in the last century by some collision with some other massive object in space. This question cannot be answered just by considering the ceteris paribus laws describing the stability of orbits in the solar system and the robustness of the arrangement of most the solar system's into planets and a sun. Whether a collision is a close possibility depends also on whether comets or asteroids far away in time and space from the earth in the nineteenth century could easily have traced a somewhat different course, and eventually have collided with the earth in that century. If there were no such heavenly bodies anywhere near the regions they would have occupied for a collision to occur, and there could not easily have been, then our intuitive verdict would be that there could not easily have been a collision. One proper way, then, to make it plausible that there is a close possibility in which a given object F at t is to make it plausible that (a) a little before t, some small difference from the actual world could obtain, a variation which some ceteris paribus law implies is sufficient for the object's being F, and (b) that there would in those circumstances be no changes in the robust conditions which undermine the applicability of the ceteris paribus law. There are negative existentials in these conditions, so a certain open-endedness is present. In general such claims of close possibilities will be potentially open to undermining by the discovery that some unobvious, robust conditions were preventing the earlier condition from holding. Equally they could be undermined by the discovery that it is much easier than was previously thought for the conditions under which things are no longer equal to obtain. Whether an object which is not actually F could easily have been F can vary with time. Such time-dependent variation is precisely what we are aiming to achieve when we try to make our trains safe, try to cure kleptomaniacs or try to make our belief-forming methods more reliable means of reaching the truth. Correspondingly, a notion of 'could have chosen otherwise' explained in terms of close19 N. Cartwright, Nature's Capacities and Their Measurement (Oxford: Oxford University Press, 1989), esp. Chapter 4. 364

The Modality of Freedom ness will be significantly time-relative. What could easily happen to a train may vary over time; what a person is capable of choosing may change over time. The notion of freedom identified by the closeness conception is genuinely historical, a function of the agent's situation at the time. There is also a relativity to circumstances: the full form is the four-place x is free to F in circumstances C at time t. Since closeness is evidently a matter of degree, the closeness elucidation makes freedom a matter of degree. Is this a problem? No doubt freedom is commonly thought to be a matter of degree, and we use the comparative and superlative forms. Abraham Pais, in his celebrated biography of Einstein, writes: 'Were I asked for a onesentence biography of Einstein, I would say, "He was the freest man I have ever known".'20 However, the common conception of freedom as a matter of degree, to which comparatives and superlatives can legitimately be applied, is arguably concerned just with the range of factors with respect to which a person is free. The freer person is free with respect to a wider range of factors. It is not so clear that the ordinary conception allows that freedom with respect to a given factor can be a matter of degree, or may have borderline cases, or can admit of an intelligible comparative. Yet the closeness account is committed to all of these. The closeness theorist should reply that it can sometimes come as a surprise which concepts admit of degree, and correspondingly exhibit a certain sort of borderline case. Let us take one of our parallel cases again. The factors which underlie knowledge of a given proposition, as opposed to how knowledgeable someone is, may seem initially not to be a matter of degree. Yet one can be forced to soften such a position by examples. There is a spectrum of cases, perhaps many spectra, which show that the factors which underlie knowledge are matters of degree. Consider our subject who believes 'That's a barn', when taking his visual experience at face value. He does not know it to be a barn, even when he is seeing a barn, if there are barn-facades scattered around nearby. But of course the barnfacades might not have been installed, and he still not know, if there had merely been a delay in installing them, so that he could easily at that time not have been in a facade-free environment. Or we can consider the case in which they were installed somewhere else altogether, but the film director had still been considering our believer's location as a filming site; or the case in which he was still choosing between installing barn-facades and real barns, and had not yet selected a site; or the case in which he was deliberating between 20 A. Pais, 'Subtle is the Lord...' The science and the Life of Albert Einstein (Oxford: Oxford University Press, 1982), p. vii. 365

Christopher Peacocke filming only in a studio, and filming in some countryside somewhere or other... and so forth. As we know from much reflection on vagueness, we cannot avoid this element of degree by moving from a requirement 'no close worlds of such-and-such a kind' to 'no definitely close worlds of such-and-such a kind'. 'Definitely close' seems to display the same problematic phenomena, the borderline cases and the like. I think that, contrary to first appearances, the factors which underlie knowledge of a given proposition is to be classified with other predicates which are, apparently, matters of degree. The closeness theorist will say the same of being free with respect to a given factor. Further on, I will offer some independent evidence for this position. Alternatively it may be objected that the closeness possibilities I have been identifying may be real enough, but that I have not shown that they are under the agent's control. What might be meant by this, and what does the objector want? The complaint should not be that on the closeness conception, which of various close possibilities is realized does not depend on the agent's choice or decision. That would be false: on the closeness conception, what the agent does causally depends upon his decision (a standard compatibilist remark). The closeness possibilities which the present position identifies as crucial to freedom have the feature that the agent could try to realize them, and would succeed if he were to try. What more could be required for the possibilities to be under the agent's control? The objector may rather be concerned that there is nothing in the closeness conception which entails that it is indeterminate, prior to the agent's making a choice, which course of action he will pursue. The closeness conception is at least in the target area of that remark, for indeed it has been put forward as a compatibilist view. This last construal of the objection is just a classic statement of the incompatibilist intuition, and I will consider it in section VII. How does the closeness conception compare with some other recent compatibilist elaborations of freedom? By addressing this question, we can reach a better articulation of what distinguishes the closeness conception. David Lewis presents a classic compatibilist position in his 'Are We Free to Break the Laws?'21 Suppose I am free to go to the meeting, but do not in fact go. Lewis highlights the distinction between (a) my doing something such that, were I to do it, either it or one of its effects would be a breaking of an actual law, (b) my doing something such that, were I to do it, some or other earlier small miracle would have occurred. Lewis's points are that it is (a) which would have to be involved in a freedom to break the 21 Lewis, 'Are We Free?', as in note 17 above. 366

The Modality of Freedom laws, and that nothing in his position commits him to (a). On Lewis's own theory of counterfactuals, it is only (b) which is entailed by my freedom to go to the meeting. On that theory, if I were to go to the meeting, some or other small 'divergence' miracle would have occurred earlier. I think that Lewis's response to van Inwagen-style incompatibilism is the right one for someone who accepts all of the Lewisian approach to counterfactuals — the full Lewisian, as we can call him. There is indeed a position in logical space which combines the full Lewisian view with the closeness analysis of freedom which I have been presenting. Under this combination, the role of the closeness analysis is to state the conditions which some event, in a nonactual possible world, of my going to the meeting must satisfy in order for it to be true in the actual world that I am free to go to the meeting. It is, however, very important that the closeness analysis is quite independent of any commitment to the full Lewisian treatment of counterfactuals, with its reliance on his particular theory of the counterfactually significant similarity relations between worlds. In particular, the closeness elucidation can also be accepted in combination with Jackson's theory of counterfactuals.22 On Jackson's approach, a sequential counterfactual 'if it were to be the case that p at t, then it would be the case that q at t+n' is true iff q is true at t+n in all />-worlds meeting the following three conditions. First, their causal laws at and after t are the same as those of the actual world. Second, their time-slices at t are most similar in matters of particular fact to ours. Third, they are identical in matters of particular fact to the actual world prior to t. Jackson emphasizes that on his theory, nonvacuous and empirical sequential counterfactuals are true only if there are appropriately sustaining laws. Similarity not based on laws cannot sustain such counterfactuals. Still, even the combination of the closeness conception of freedom and the Jacksonian treatment of counterfactuals may be thought to have unacceptable consequences. Must it not still be involved with miracles in the worlds that verify the counterfactuals? In particular, if I do not in fact leave for the meeting at 8.00 p.m., must not this combination count the following counterfactual as true. 'If I were to leave for the meeting at 8.00 p.m., then some miracle would have occurred in the period of time up to 8.00 p.m.'? Must not something have gone wrong in a compatibilist account of freedom which, if freedom exists, is committed to the nearby possibility of small miracles? I argue that there is no commitment, provided the case for the 22 F. Jackson, 'A Causal Theory of Counterfactuals', Australasian Journal of Philosophy 55 (1977), 3-21. 367

Christopher Peacocke closeness elucidation of freedom is properly marshalled. We should distinguish very sharply between counterfactuals of the form If it were to be the case that p at t+n, then it would have been the case that q at the earlier time t and conditionals of the form If it were to be the case that p at t+n, then that could only be because q was the case at the earlier time t or of the form If it had been the case that p at t+n, then it would have had to be the case that q at the earlier time t. Though counterfactuals of the first of these three forms - the backtrackers - are sometimes regarded as mere terminological variants of the latter 'could only' and 'would have to have' conditionals, as by Bennett,23 it seems to me that they have a quite different meaning. Let us call a nonvacuous, empirical backtracking counterfactual with a consequent which is false with respect to the actual world a threatening backtracker. It seems to me that the defender of the combination I am advocating should insist that threatening backtrackers are true only if there is backwards causal influence. That is what assertion of the threatening backtrackers would properly be used to express. The defender of the present combination can and should agree with the truth of such conditionals as 'If I were to leave for the meeting at 8.00 p.m. (which I am not), then that could only be because [or: then it would have to have been the case that] some conditions earlier were different from those which actually obtain.' The worlds which verify these conditionals do not involve the occurrence of any miracles (however small). This combination differs also from the positions offered to the compatibilist in other recent discussions. In his valuable book The Metaphysics of Free Will,2'' John Martin Fischer considers how the compatibilist might respond to the challenge that an agent cannot do something of which it is true that were he to do it, the past would have been different (his principle '(Fpnc)', p. 79). Fischer offers the compatibilist a position on which in certain examples both a 'can' claim is true, and so too are certain backtrackers. He considers a seadog who would never go sailing at noon unless the weather fore23 J. Bennett, 'Counterfactuals and Temporal Direction', Philosophical Review 93 (1984), 57-91. 24 J. M. Fischer, The Metaphysics of Free Will: An Essay on Control (Oxford: Blackwell, 1994). 368

The Modality of Freedom cast earlier in the morning had been favourable. Fischer (p. 91) says the compatibilist can say, even after an unfavourable weather forecast, both that the seadog can go sailing, and that 'if the seadog were to go sailing at noon, the past would have been different from what it actually is' (p. 91). On the combination I am proposing, though, since there is no real possibility of backward causal influence, such backtrackers are never true. No correct position can require their truth. It is rather the 'could only have been because...' and 'would have to have been because...' conditionals which are true in the Fischerian examples. Their truth does not involve any denial of the fixity of the past. (In effect, this is to suggest that Fischer's (Fpnc) principle, in which the 'nc' indicates the claim that it is a non-causal principle, is in fact causal after all.) How then does the combination I am favouring respond to Fischer's 'Basic Version' of incompatibilism (pp. 87-94), the principle that 'an agent can do X only if his doing X can be an extension of the actual past, holding the laws fixed'? (Fischer, p. 88, also cites Ginet.)25 On one intuitive understanding of the phrase, the closeness account does give an account of what is involved in an agent, with her actual past, being free to do X. Her actual past is highly relevant, since it determines whether she is in a state which prevents there from being any close world in which she tries to do X (or any close world in which she tries and succeeds). Fischer would protest, though, that that intuitive construal is not what he means. What he means by the Basic Version is that an agent can at t do X only if there is a possible world coinciding at all times prior to t with the actual world, with the same laws as the actual world, and in which the agent at t does X. This, though, seems to be too strong as a necessary condition of possibility at t, even outside cases of agency. In our example of the deterministic world in which at t our subject sees a barn, we emphasized that there is a sense in which that subject could at t have been seeing a barn-facade. In that example, there is no possible world coinciding at all times prior to t with the world there envisaged as actual, with the same laws envisaged as actual there, and in which the agent is not seeing a barn. Nonetheless, it seems that there is some sense in which our subject could have been seeing a barn-facade; and that is why he does not count as knowing that there is a barn in front of him. On the way I have been developing the closeness conception, corresponding points hold for the 'could have' involved in freedom. 1 C. Ginet, On Action (Cambridge: Cambridge University Press, 1990). 369

Christopher Peacocke VI Non-Theoretical Construals of Freedom I now turn to consider some options I've implicitly rejected in putting forward the closeness conception. I will consider in a little more detail views that construe ascriptions of freedom or responsibility as a manifestation, perhaps an expression, of a practical attitude, rather than of a theoretical belief which might be false. This is the kind of position famously associated with Peter Strawson, but a position of this general character is developed in detail, with great resourcefulness, by Christine Korsgaard, in the title essay of her collection Creating the Kingdom of Ends.16 She writes 'Responsibility is construed practically by those who think that holding someone responsible is adopting an attitude towards her, or, much better, placing yourself in a relationship with her' (p. 198). She contrasts the practical construal she favours with one on which 'Deciding whether to hold someone responsible is a matter of assessing the facts; it is a matter of arriving at a belief about her' (p. 197). Korsgaard's view is that the practical rather than the theoretical construal of responsibility 'is implicit in our actual practices' (p. 197). I want to consider briefly the phenomena she cites in support of that, and to argue that those phenomena can be explained on a theoretical conception of freedom and responsibility. 1. On Korsgaard's practical conception, there is, she notes, some distance between the practical issue of whether to hold someone responsible and the question of whether he acted voluntarily (p. 198). She writes 'there is neither need nor reason to... say that people under severe emotional stress cannot control themselves. We do not need to understand a form of debilitation as a form of impossibility in order to make allowances for it; we need only to know what it is like.' (p. 198) This seems to me intuitively correct. But it is also to be counted as correct on the closeness view. It will, on that view, be a matter of degree how easy it is to overcome some factor which, unless one makes some effort, will cause one's actions. It may be easy, somewhat hard, hard, ... through to extremely difficult to overcome it. These distinctions in degree correspond to differences in how close are the worlds in which one does overcome it, the more difficult cases being less close (though still close enough for the agent to be free to overcome it). A theoretical construal of freedom with respect to a factor does not need to, and should not, take it as an all-or-nothing matter. In sum, what we earlier raised as an objection to the closeness account, its recognition of degrees of freedom with respect to a factor, should properly be counted as a virtue of the approach. 2. Korsgaard also very acutely notices that 'it may be perfectly 26 Korsgaard, Creating the Kingdom of Ends, pp. 188-221. 370

The Modality of Freedom reasonable for me to hold someone responsible for an attitude or an action, while at the same time acknowledging that it is just as reasonable for someone else not to hold the same person responsible for the very same attitude or action. Perhaps it is reasonable for you to forgive or overlook our friend's distrustful behavior on the grounds that he has suffered so much heartbreak, but not for me, not because I fail to appreciate how hurt he has been, but because I am the woman whose loving conduct is always met with distrust' (p. 199). Korsgaard gives this as an example of a possibility 'that would not make sense if responsibility were a fact about the person' (p. 199). But it seems to me that the theoretical view can accommodate this phenomenon too, provided that it recognizes differences in degree of difficulty in carrying through an action the agent is free to perform. The different relations in which two people may stand to the agent - in Korsgaard's example, one being an outside observer, the other engaging in loving conduct towards the agent - may generate different entitlements or legitimate expectations in the degree of effort they can properly demand of the agent in overcoming difficulties in pursuing a course of action, before they forgive or overlook some of the agent's behaviour. It seems to me that Korsgaard is right to say that the loving woman has a right to require more than the disinterested observer of the situation may reasonably expect. What I cannot see is that the phenomenon is inexplicable on the theoretical conception of freedom. It is the different relations to the agent which are generating the differences in legitimate expectations. The example does not force us to say inconsistent things about the factual issue of the freedom of the agent. 3. I am much less sure than Korsgaard that there is so tight a connection between one's placing oneself in a relationship with someone, and attributions of responsibility and freedom. She discusses the reciprocity and openness involved in close friendship, and ways in which this may be abused. As she notes, we may eventually 'write someone off, and in extreme cases cease to have reactive attitudes to them altogether (p. 200). Yet it is not at all clear to me that this must involve thinking of this other person as unfree or as not responsible, as opposed to just being awful, manipulative or utterly egocentric. These latter characteristics seem to me, unfortunately, to be compatible with freedom. VII Libertarianism Russell at one point in Human Knowledge: Its Scope and Limits wrote: 371

Perhaps in the brain the unstable equilibrium is so delicate that the difference between two possible occurrences in one atom suffices to produce macroscopic differences in the movements of muscles. And since, according to quantum physics, there are no physical laws to determine which of several possible transitions an atom will undergo, we may imagine that, in a brain the choice between possible transitions is determined by a psychological cause called 'volition'. All this is possible, but no more than possible.27 I want to introduce an issue by the apposite comment made on this by a distinguished libertarian, David Wiggins, who wonders 'Could not the incidence of human acts of "volition" upon quantum phenomena upset the probability distributions postulated by the quantum theory?'28 Wiggins also observes that if the volitions were postulated to have some immaterially realized source (perhaps to connect them with the agent's character) the theory would be unacceptably Cartesian; he writes that 'We need not trace free actions back to volitions construed as little pushes aimed from outside the physical world' (p. 292). It seems to me, however, there is a problem lurking here for the libertarian conception of freedom which can be formulated in a way quite independent of any commitment to Cartesian mythology. The assumption required for formulation of the problem is rather the non-Cartesian principle that mental events and states supervene on physical states and events (not necessarily restricted, in fact necessarily not restricted, to physical states of the brain given externalist theories of content). Suppose an agent has a choice between an action-type A and an action-type B, and suppose that he cannot do both. Let SA be the set of physical states on which his choosing A would supervene, and let SB be the set of physical states on which his choosing B would supervene. Just before the moment of choice, his being in some or other state in SA and his being in some or other state in SB each have certain probabilities. (The probabilities cannot be calculated in practice.) If the libertarian theory is that the subject's complete freedom to choose either A or B implies that it is completely indeterminate which he will choose, then won't the libertarian theory imply that an agent with this freedom can make either choice with whatever frequency he pleases? Cannot a collection of agents, similarly situated, make either choice with whatever frequency they please? That would then eventually conflict with the 27 B. Russell, Human Knowledge: Its Scope and Limits (London: Routledge, 1992), Part I, chapter 5, pp. 55-6. 28 Wiggins, 'Towards a Reasonable Libertarianism', p. 292. 372

The Modality of Freedom probabilities and frequencies implied by quantum mechanics. This means that the original point, rightly raised by Wiggins against the conception Russell described, applies equally against a range of libertarian views. It is not only volitional theories and Cartesian variants thereof which are vulnerable to the objection. It is worth noting that only supervenience has been employed in this argument. There has been no commitment to token-identity theories, nor even to the existence of realization or constitution relations between mental events and physical events. I am inclined to draw from these reflections three points. The first is that a probabilistic micro-theory is just as threatening as classical determinism to the existence of freedom as conceived by libertarians. Second, these reflections also make me wonder whether a libertarian theory will not have to be non-naturalistic to a degree that its proponents may not have envisaged. Third, if neither deterministic nor indeterministic physical theories of the sort we currently have can be squared with the libertarian elucidation of freedom, it must follow that either we are not free, or the libertarian elucidation is incorrect. This is a kind of modern fork against the libertarian, though it does not involve quite the same conclusions and commitments as Hume's fork. What I have said does suggest that any form of physical theory will be problematic for the libertarian, at least given supervenience. But, unlike Hume, I do not conclude that freedom requires determinism (all to the good, since there is evidence that determinism is false). My own view is that, even on a nondeterministic conception of causal explanation, when it comes to free choices, there are cases and cases in respect of the classification of the choice as rationally caused and explained. Sometimes a free choice is rationally caused and fully rationally explained, of course. But sometimes it is part of the nature of the case that it is not. Consider for instance someone in the position of Buridan's ass, making a choice between two exactly equally attractive options, with no reason to choose one over the other. Unlike Buridan's ass, we do succeed in choosing; but the rational agent may in such a case have no reason for plumping for the one he actually selects. The choice between it and its equivalent alternative is not one which falls within the ambit of rational psychological explanation. Any explanation there is will be, at best, subpersonal. But the choice can nevertheless be a free one. My view, in short, is that freedom does not require determinism, and does not require indeterminism either. What then produces the illusion, if it is an illusion, that freedom requires indeterminism? One source is the genuine recognition that 373

Christopher Peacocke there is a 'could have done otherwise' requirement, when that recognition is combined with the belief that only an indeterministic construal of the modality can be given. I have tried to give an alternative construal of the modality. I think, though, that part of the explanation of the illusion may also be the fact that when an agent is presented with various prima facie reasons for different courses of action, there is apparently no ceteris paribus psychological law of commonsense psychology about which he will choose, not even when he has deliberated and found one of the courses best all things considered. All kinds of impulsive, and/or weak-willed, but nevertheless free, choices may be made by the agent. Since there is no such psychological law, and we also consider the alternatives between which the agent is deliberating to be genuinely open to her, it may be tempting to conclude that there is no law at all explaining her choice. Put like that, though, any such conclusion would be a nonsequitur. It would be entirely consistent with those premises that there is some other kind of ceteris paribus, and perhaps probabilistic, law which, in the circumstances in which the agent is placed, explains the choice under its intentional description. It may be anything ranging from the subpersonal-computational, or the psychoanalytic, to the sociological or the economic. All our premises said is that there is no ceteris paribus psychological law of commonsense psychology. VIII Concluding Remarks As you might expect, I am inclined to present the closeness account as a form of the very first of the six theoretical possibilities I distinguished when there is trouble squaring the metaphysics and the epistemology in a given domain. The closeness account of freedom seems to be an account of the metaphysics of freedom which makes freedom an intelligible possibility, and makes the distinction between the cases in which an agent is free to F, and those in which he is not, an empirical matter. To this, it may be objected that some intrinsically problematic conception of freedom is inextricably entwined in our normal thought and practices, and that because this is so, the closeness account must rather be revisionary. The objection is that the closeness conception is rather an example of the fourth of the six options I identified, the option in which we offer a surrogate rather than a full-width truth-condition for statements of the problematic kind. My immediate reaction to this objection is one which, despite our disagreement on the correct positive account, I share with Peter 374

The Modality of Freedom Strawson, when he writes that 'the idea that an entire range of emotions which pervade our personal and social lives as thoroughly as those in question should be thus linked to a condition which cannot be coherently described has a degree of implausibility which it would be difficult to rival'.29 It may help to consider a parallel. Suppose we had a benighted community which believed in an extreme form of absolute space, and with it a distinction between absolute rest and absolute motion. This would certainly be a case in which the integration of the metaphysics and the epistemology would present severe (I would say insuperable) problems. Suppose we point out to members of this community that cases in which they speak of absolute rest really involve rest relative to some specified object or array of objects, and that is how the truth conditions of their statements about absolute rest and motion should be given. Would it be right for members of this community to object that the truth-conditions we propose would really be revisionary, would be an example of case (iv) rather than case (i), and thus not really capturing what they meant? Well, our proposal would certainly be revisionary of a misconception of the nature of what they call rest, or motion, sans phrase. Maybe the misconception is indeed so inextricably involved in their thought that they would not know what to say in their old vocabulary once the situation is pointed out to them. What is clear, however, is that it would be futile in the spatial case to hope for a better solution than that we have offered them. Now no one could honestly say that the case of freedom is as clear as that of absolute location. Whatever one's favoured solution, philosophical humility is the only appropriate mode in which to present it, in the face of the tremendous difficulty and recalcitrance of the problem of freedom. I do, though, want to suggest one similarity between the case of freedom and this spatial example. I suspect that we cannot coherently have more than the closeness account offers; while to settle for less seems to leave us in the rather queasy position of engaging in practices without entitlement. To establish that that suspicion is correct, much further argument would be needed. But if it can be established, then the closeness account would be something with which it would be reasonable to rest content. 29 P. Strawson, 'Replies', in Philosophical Subjects, Z. Van Straaten (Oxford: Oxford University Press, 1980), p. 265. 375

Dualism in Action JENNIFER HORNSBY We know what one dualist account of human action looks like, because Descartes gave us one. I want to explore the extent to which present-day accounts of physical action are vulnerable to the charges that may be made against Descartes's dualist account. I once put forward an account of human action, and I have always maintained that my view about the basic shape of a correct 'theory of action' can be combined with a thoroughgoing opposition to dualism. But the possibility of the combination has been doubted1, and it will remain doubtful until we have a better understanding of what makes an account objectionably dualistic. In this paper, I hope to deflect some of the criticisms aimed against what I shall call my account,2 and to show that when they are turned onto their proper path their actual target is some physicalist accounts. I shall have to rely on one intuitive understanding of physical action here. According to this, where there is a physical action, a person moves, and there is a psychological explanation of a certain sort of something that she thereby does. This takes it for granted that human agency is evinced when someone does something intentionally,3 and that when people do things by moving their bodies, they are involved in events.4 Using this conception, and assuming a 1 The account I gave in Actions (London: Routledge and Kegan Paul, 1980) has often been accused of Cartesianism: there are more details in section V, and see note 30 infra. 2 I speak of 'my account' for the sake of having an easy label for what I defend. Despite the label, I do not mean to suggest either that it originates with me, or that it is the whole of an account of anything. Brian O'Shaughnessy defended something similar in The Will (2 vols., Cambridge: Cambridge University Press, 1980). The details of what I call 'my account' are to be understood as those of Actions. (I think that avoiding Cartesianism requires rejection of some of what O'Shaughnessy said in support of his account: see note 23 infra.) 3 The general idea that there is human agency when a person intentionally does something is relatively uncontroversial. It can be sustained by seeing what sort of trio the concepts of belief, desire and intention form, and thus what kind of psychological explanation an action explanation is. 4 Some resist the assumption that an action is ever an event. Resistance may stem from the thought that actions should not be reckoned among 'mere happenings'. I hope that it will become clear why, on my own view, there could be no reason to treat actions as 'mere happenings'. 377

Jennifer Hornsby certain account of events' individuation,5 one can say that any action is some person's moving her body (usually her moving of a bit of it). This understanding will serve in the present context, because the debates about action which are of concern here take place in the domain that it carves out.6 I Dualism vs. Physicalism Before I come to allegations of dualism made against accounts of action, I should say something about what dualism itself amounts to. I think that in the present state of play, many philosophers have an inadequate conception of this. Naturally enough, dualism is contrasted with physicalism.7 We know that there are various versions of physicalism advocated in the philosophy of mind. For a start, there is the mild sort - so called token-token identities physicalism - and the stronger sort - socalled type-type identities physicalism. And then there are versions of physicalism which hold that composition or constitution, rather than identity, is the relation holding between mental and physical states and events. Whatever the details, it can seem as if we might put physicalist doctrines onto some sort of scale — a scale on which dualism might be supposed to feature at the opposite pole, as it were, from the strongest physicalist doctrine. It seems, then, as 5 See e.g. D. Davidson, 'Agency', in Agent, Action and Reason, ed. R. Binkley et al. (Toronto: Toronto University Press, 1971); repr. In his Essays on Actions and Events (Oxford: Clarendon Press, 1980), pp. 43-61. 6 When actions are defined by reference to a class of physical events, the general idea of human agency is restricted in two different ways. (A) Left out from the account are things that fall into an intuitive category of the mental. Consider mental arithmetic; or consider the view that agency is evinced whenever there is an exercise of practical reason. (B) Left out from the account are things people intentionally do, the doings of which are not events. Consider an occasion on which a intentionally fails to greet b, and on which we might be apt to say that a did nothing, or say that nothing happened. Here a's intentionally not greeting b may be thought not to be an event; and, if it is not, then we have an example of agency - according to the intuitive conception of agency - but we do not have an action - not according to the restrictive conception of actions. For present purposes, it need not be a question whether restrictions (A) and (B) ought to be lifted by a correct conception of agency, because charges of dualism are faced by accounts of action which impose the restrictions and deal with 'physical actions'. 7 Throughout, I use 'physicalism' as the name of a kind of monism. I might have used the word 'materialism', or, introducing another bone of contention, 'naturalism', instead. 378

Dualism in Action though we could ask a person: 'How physicalist are you?' One possible answer would be 'Not at all'; and then, if this were the right way to think about things, we could place the person as a dualist. But this cannot be the right account of the matter. If dualists are to be contrasted with physicalists, then that is not because they reject rather a lot of the doctrine which we have come to associate with 'physicalism' at the end of the twentieth century. Dualists are distinguished from physicalists inasmuch as a dualist answers Two to a certain question, to which any monist - including a physicalist - answers One. The question to which Descartes's answer of Two earns him the title of dualist is the question 'How many sorts of substance inhabit the world?' Not only is he a dualist, but also (what matters here) Descartes's account of action is dualistic in a straightforward and obvious sense. It is true that when res cogitans first appears in the Second Meditation, 'it is in the strict sense only a thing that thinks': 'I am a mind or intelligence or intellect or reason', Descartes says. But Descartes widens 'thought' to include volitional, as well as intellectual, activity. 'What is a thing that thinks?', he asks; and answers 'A thing that doubts, understands, affirms, denies, is willing or is unwilling' .8 Volitions, in which a res cogitans participates insofar as it 'is willing', belong to substances which are entirely separate, and radically different in their nature, from any bodies. Let us call such substances 'souls'.9 Then one disavows dualism of Descartes's sort by saying that there are no such things as souls. In that case there cannot be any need to locate oneself on any scale of 'physicalism' in order to avoid dualism. But by the same token, there must be more to Descartes's way of thinking about persons than is elicited in contrasting it with contemporary physicalists' ways of thinking. We are often encouraged to think that 'Cartesianism' still rears its ugly head. From a variety of sources, we are familiar with attacks on the idea of mental states as inner, private states whose content can be specified without appeal to any8 My italics. Descartes actually adds imagining and having sensory perceptions onto this list of attributes characterizing a soul, but these come (by the Sixth Meditation) to be treated in a special category of their own, so that Descartes's account of perception is not straightforwardly dualist. See John Cottingham, 'Cartesian Trialism', Mind XCIV, No. 374 (April 1985), 218-230. The question how straightforwardly dualist Descartes's account of action is comes up in Appendix B. 9 I use 'souls' throughout to stand for what Descartes called sometimes 'esprit' (or 'mens'), at other times 'arae' (or 'anima'). We are familiar enough with 'minds' used as afacon de parler, supposed to make no commitment to non-physical substances, that 'souls' serves better to register such commitment. 379

Jennifer Hornsby thing outside the consciousness of the person whose states they are: these are attacks on Cartesianism.10 Cartesianism is arguably implicit in Descartes's method, and is usually supposed to be secured by substance dualism of Descartes's sort. But if charges of Cartesianism are still with us today, it seems that Cartesianism cannot actually require substance dualism of Descartes's sort. One can see that contemporary physicalist orthodoxy might not be proof against Cartesianism by noticing an ambiguity in 'substance dualism'. Substance dualism is often understood as the doctrine that mind and body are two different kinds of substance — so that, in the terminology being used here to register Descartes's view, there are souls as well as bodies, souls being of a different kind from physical things." But substance dualism might be understood more broadly — as the doctrine that a mind, whatever kind of thing it may be, is a substance different from any animal body. In this broader sense, substance dualism is compatible with versions of physicalism. Indeed any physicalist who tells us that minds are brains would seem now to be a substance dualist (no matter what he has to say about states and events).12 Underlying substance dualism in the broad sense is the idea that those persisting things which have mental properties are separable from all such things as lack mental properties (no matter whether the things have mental properties are actually physical things). This idea does not require souls to be present in the world. And it may be that some of the hostility to 10 By 'Cartesianism', I mean a conception of mind which, for instance, has been to be the butt of many of Wittgenstein's remarks. Assuming that a doctrine of substance dualism of Descartes's kind is to be avoided, I want to encourage the thought that some of its errors may actually attach to a Cartesianism which it brings with it, and which may attach also to other doctrines. 11 The matter is more complicated than this allows, because Descartes, though he thought that individual souls were substances, took individual bodies to be modifications of stuff, not substances proper. (Those who speak of Descartes simply as a mind/matter dualist ignore his different treatments of individuals in the realm of mind and individuals in the realm of matter. And I too ignore them pro hac vice.) 12 I make the assumption here that brains are substances. In the literature on personal identity, one sometimes encounters the claim that persons are brains; those who advance it do not intend to deny that persons are substances (in the relevant sense). Presumably those who say that minds are brains (who are rather more numerous than those who say that persons are brains) do not have their own special understanding of 'brain'. And we do not need Descartes's demanding notion of a substance to understand substance dualism in the broad sense (or even in the narrow one: see note 11 supra): 'persisting things' might serve for 'substances' here. 380

Dualism in Action Cartesianism is not hostility to souls as such, but is directed towards treatments of the mental as a self-standing, inner realm. One does not automatically escape such treatments by adopting the tenets of contemporary physicalist orthodoxy. Two possibilities have emerged here. First, it may be that one can be anti-Cartesian without endorsing any orthodox physicalist doctrine. Secondly, it may be that some of those who go in for orthodox physicalist doctrine are still Cartesian. I myself think that both of these possibilities are actual. At any rate, you will need to appreciate them both in order to understand how it can be that, in resisting the charge that ray own account of action lines up with Descartes's, I should avoid endorsing any of the going versions of physicalism. II A Very Short History of Action Theory We can look at a very truncated history of action theory in order to reveal where questions about Cartesianism impinge upon debates about action. This should start with Descartes. We have seen already that he thought that volition is a faculty of souls. Here is what he said about the soul's production of movements." The soul has its main seat in the little gland which is in the middle of the brain, from where it radiates throughout the rest of the body by means of the animal spirits, the nerves, and even the blood... [T]he machine of the body is made [so that]... this gland's being moved by the soul drives the surrounding spirits into the pores of the brain, which conduct them through the nerves into the muscles, by means of which it causes them to move the members [of the body]. Only the dualism here needs emphasizing now. The human body is one thing, a machine whose members are caused to move by muscles which {via the spirits) are driven by the gland wherein the soul resides. The soul itself is another thing: intellectual and volitional properties attach to it. The resulting picture of human action has been called volitionism. According to this, where there is an action: A soul's volition IS CAUSALLY RESPONSIBLE FOR a movement of a body In order to move to contemporary debate, we need to skip three 13 The Passions of the Soul, Article 34. 381

Jennifer Hornsby hundred years. This takes us to recent opposition to volitionism. The anti-volitionists of the 1950s and 1960s thought it an error to suppose that the question 'What makes a bodily movement voluntary?' should receive a causal answer.14 They wanted to avoid the soul and its modes of affecting things; and they thought that these could be avoided if causal connections were left out of an account of the explanatory relations involved in understanding what people do. Their opposition to volitionism, then, was anti-causalist. In the 1960s, the tide turned. Donald Davidson's paper 'Actions, Reasons and Causes' was largely responsible for that.15 Nowadays this paper is read as providing arguments for a particular causal thesis, rather than as reacting specifically against the anti-causalism of the anti-volitionists. But situating it by reference to the thinking which prevailed when it was written, we can be aware of the care which Davidson took to avoid any events that might have been supposed to play the causal role that volitions play in Descartes's picture. Davidson thought that there is no need for any volitional items in order for causality to have its rightful place in an account of action. (His view was, and is, that beliefs and desires cause actions.16) Although the anti-causalism which Davidson was reacting against was popular at one time, it is not very popular any longer:17 our powers as agents surely are power to change things; it can seem absurd to suppose that we might capture the idea of human agency without treating human beings as part of the causal world within which they operate. If Davidson showed that we can have causation on the scene without volitions there, then he might seem to have put an end to the debate about volitionism. 14 This is not a question that Descartes himself ever attempted to answer. But it is plausible that the attractions of a volitionism like Descartes's may have derived from thinking that having a mental cause could serve to distinguish the bodily movements that occur when there is voluntary (or intentional) action from all other bodily movements. 15 Journal of Philosphy 60 (1963); repr. In his Essays 3-19, and in The Philosophy of Action ed. A. Mele (Oxford: Oxford University Press, 1997), pp. 27-41. 16 I criticized this view in 'Agency and Causal Explanation', in Mental Causation, ed. J. Heil and A. Mele (Oxford: Clarendon Press, 1993); repr. in The Philosophy of Action, and in my Simple Mindedness: A Defense of Naive Naturalism in the Philosophy of Mind (Cambridge, MA: Harvard University Press, 1997). The criticisms leave intact a broadly causal picture of human action by allowing that one can provide a causal explanation of what an agent does by saying what her reasons for doing it were. (Although this leaves me opposed to the anti-causalists, I object not only to Davidson's version, but also to all of the usual versions, of causalism: see section VI infra.) 17 Although it is still defended: see Michael Morris, in this volume. 382

Dualism in Action There must be more to be said, however. We can pose questions which are simply not addressed in Davidson's account. In the first place, there are other concepts than 'belief and 'desire' which apparently have a peculiar relation to action; and we can ask how those concepts fit in. Secondly, Davidson spoke to the causation of actions, not the causation of the bodily movements of Descartes's picture; and we can ask about this — about such events as arms' going up, movements of lips, or whatever.18 Action theory of the 1970s and 1980s provided accounts which attempted to answer questions of these two sorts, by going into detail about the relations between the various events that happen when someone moves her body and thereby does something intentionally. The account I defended myself exploits a connection between what is done intentionally by an agent (i.e. what may be explained by allusion to what she wants and what she thinks) and what the agent tries to do: (T) She V-d intentionally —• she tried to V In the presence of the understanding of an action that we are working with (sc. an event of a person's doing something intentionally), (T) ensures that every action is a person's trying to do something. Allowing, then, that in the case of physical action, it is because she is trying to do something that a person's body moves, one reaches an account which can be summarized thus: A person's trying to do something is CAUSALLY RESPONSIBLE FOR a movement of her body... 18 At least it is natural to suppose (i) that the phrase 'bodily movements' subsumes events such as these, and (ii) that these events are not actions (where an action is an event of someone's intentionally doing something). I used to say that 'bodily movement' is ambiguous - so that it could mean either (say) the movement of a person's arm or a person's moving her arm (Actions, chapter 1). But I now think that I was over-generous to my opponents when I suggested that their claim that actions are movements relied upon an ambiguity. The verb 'move' is ambiguous, of course — between transitive and intransitive occurrences: 'move' plays different roles in 'She moved her arm' (where it is a transitive verb) and in 'her arm moved' (where it is intransitive). But this ambiguity appears not to carry over to the nominal 'movement'. When a trace of the transitive verb occurs in a description of an event, we have (say) 'a person's moving her leg', and it is not evident that the word 'movement' can serve for this. If it cannot serve, then it would take a serious argument to show that a hand's going up (which is a bodily movement) is the same event as a person's raising her hand (which is apparently not a movement). Someone equipped with such an argument will say — as Davidson and others do — that actions just are bodily movements. But the arguments seem to me ill-motivated: see section V infra. 383

Jennifer Hornsby Evidently this account is readily associated with Descartes's: at first blush it might seem simply to rename Descartes's volitions using 'trying to —'. It could appear, then, as if the result of filling in a causalist account were to return one to the very volitionism that the anti-causalists had reacted against. But I think that this is a false appearance. And I want to free myself from guilt by association. So I shall defend (T) against charges of Cartesianism (section III)19. And I shall show how superficial the similarity is between Descartes's account and the one I have just summarized ('my account': see note 2 supra). Ill The Import of Thesis (T) By introducing 'try to', (T) brings in antecedents of bodily movements which fall into an intuitively mental category. So my account's seeming similarity with Descartes's appears to come in through its endorsement of (T). I need to explain why (T) should not be thought responsible for any items' being conceived of in Descartes's way. Notice, first, that it is not only someone who accepts (T) who might have to guard against objections of Cartesianism. (T) makes a very general claim about what is required to do something intentionally.20 But even someone who rejects this general claim will surely accept that there are occasions when a person moves her body in trying to do something or other. On such occasions at least, a movement of a bit of the person's body arguably depends upon her try19 Since delivering the lecture on which the present paper is based, I have come across Descartes' Dualism, by Gordon Baker and Katherine J. Morris (London: Routledge, 1996), in which the authors argue that Descartes did not hold the doctrine (sc. 'Cartesian Dualism') which contemporary philosophers attribute to him. If they are right, then we may be less well placed than I suppose we are to base knowledge of a dualist account of action on our understanding of Descartes. I have responded by adding Appendix B. 20 (T) is to be read as a schema: in any instance 'V is to be replaced by a verb, and the tense of the verb at its left-hand-side occurrence is to match the tense introduced into the 'try to' that occurs before the verb's infinitive occurrence on the right-hand-side. 'My account' actually requires only that whenever there is an action, there is at least one thing that the agent intentionally does which is something she tries to do (at least one substitutend for 'V gives a truth). My 'quick and simple' argument {infra) suggests that agents try to do everything they intentionally do; but this fully general claim, which schema (T) catches, actually need not be at issue. 384

Dualism in Action ing to do whatever it is.21 Suppose, for example, that someone moves her fingers against the keyboard trying to type a '£' sign, but because the key has been reassigned she actually types an '@'. Her finger's movement then depends upon her trying to type a '£' sign. If there were objections to the very idea that the movement of a person's body might depend upon her trying to do something, then the objections would apply in this case. And this means that a charge of dualism (if such a charge can really be made) is likely to be made in particular cases even if it is denied that 'try to' has the pervasive application which someone who endorses (T) believes that it does. The crucial questions here do not turn upon the correctness of (T). Notice, secondly, that even though we are making a general assumption that actions require bodily movements, and (T)'s claim in respect of actions is very general, (T) does not make a claim about moving the body. You might hold that a person moves her body whenever there is a physical action of her doing something, and hold (as (T) says) that a person tries to do everything she intentionally does. That does not amount to your holding that people try to move their bodies whenever they do something intentionally. For it might be that people's intentions sometimes take off at points beyond their bodies. (T) can be acceptable, then, even where it is denied that something an agent always does is to try to move her body. Endorsing (T) does not force one to speculate about what it is to move the body. Notice, thirdly, that someone who accepts (T) will think that nearly all of the things that agents try to do are things that they actually succeed in doing; and that even where an agent fails to do something that she tries to do, she usually succeeds in doing something (there is something else she does - other than what she tries to do). Thus an ordinary case of someone's trying to do something, whether successful or not, is just an ordinary case of action. (T) should not lead anyone to believe in things called 'mere tryings'.22 21 An argument would require the distinctness of actions (e.g. her depressing the key marked '£') and bodily movements (e.g. her finger's moving against the key). Cf. note 18 supra. 22 When 'tryings' (simply) are spoken of, people conjure up a picture of 'mere tryings': they forget the adverbial characters of 'try to'. (See my 'Reasons for Trying', Journal of Philosophical Research 20 (1995), 525-39.) It is hard to find a natural terminological policy which enables one both to speak generally and to avoid the misleading impression that there might be 'mere tryings'. The policy I have adopted here where the context allows is to use 'try-to' (rather than just 'try') for shorthand, and to use 'try to —' as a sort of schematic verb: the intention is to keep it in mind that to try is always to try to do something. 385

Jennifer Hornsby By accepting (T) and a claim about the causal dependence of agents' effects in the world upon events in which agents participate, one arrives at a quite natural account of the difference between successful and unsuccessful attempts. Thus: someone who tries-tohave-an-effect-and-succeeds is someone who participates in an event which has some result she intended, whereas someone who tries-to-have-an-effect-and-fails is someone who participates in an event which doesn't have some result she intended. On this account, 'try to —' appears as a sort of common denominator, which is present both in intentional doings and in unsuccessful tryings. But those who have their doubts about (T) will wonder why 'try to' should be supposed to have any application at all when an agent actually does what she means to. The doubters may think that anyone committed to 'try to"s having such a pervasive application as (T) suggests must have been involved in a search for a common denominator — a sort of search which prescinds from the world surrounding the agent and considers only the agent herself and how things might have seemed to her. Well, it is certainly true that philosophers have given arguments for (T) in which such considerations are very much to the fore.23 But there is a quick and simple argument for it which requires no speculations about the phenomenology. All that this argument needs is that 'try to do something' can be glossed as 'do what one can to do the thing'. The agent who is influenced by having a reason to do something does what she can to do it. But what one does for a reason, one does intentionally. And in doing what one can to do something, one tries to do it. So agents try to do what they intentionally do. This argument will be too quick to satisfy.24 But my purpose is not 23 In volume II of The Will (Cambridge: Cambridge University Press, 1980), O'Shaughnessy announces that a Gricean argument supports the claim about 'trying to' which he and I accept. But he proceeds to give (among others) an argument from illusion, whose tenor is certainly Cartesian. Suffice it here to say that I do not think a defence of (T) (still less of the weaker claim which is really at issue: see note 20 supra) need advert to 'trying to do seeming 0', or take a view on the 'epistemological status of bodily tryings'. A properly Gricean argument can certainly be much simpler than O'Shaughnessy's argument from illusion: see note 24 infra. 24 A Gricean argument which I stated in my Actions, pp. 34—5 (which is an argument from ignorance, rather than an argument from illusion) also seeks to show that the background facts which conduce to an instance of 'She V-d intentionally' suffice for the relevant instance of 'She tried to V. There is a particular case where 'mere tryings' have seemed to be in question — the case of an agent who tries to do something actually does nothing. I discuss this in Appendix A: On Landry's Patient. 386

Dualism in Action to vindicate the account I outlined, but only to distance it from Descartes's. It is enough here to say that (T) will seem plausible only when it is understood that it can be true that someone tries to do something without the fact that she tries to do it being at all a usual or useful thing for anyone to say or to think. Usually, of course, people simply can do the things which they do-what-theycan to-do. Otherwise life would consist mainly of frustrated attempts. That is why there is usually no point in thinking of the person who has done what she set out to do as having done what she could. Certainly there is no need for the agent herself to think of herself as trying to do that which (in fact) she tries to do. So (T) need not be responsible for the musings of those philosophers who conceive of 'tryings' (as they call them - cf. note 22) exclusively from the standpoint of the agent. These points all help to show that (T) is not an accomplice in Cartesianism. But they do not speak directly to the similarity of my account and Descartes's. What I shall do next is to consider lines of objection which might be thought to have application equally to both accounts. I hope to show that their proper target is Descartes's account alone (sections IV and V). IV A Mysterious Gulf? In Gilbert Ryle's description of Descartes's account, 'mental thrusts, which are not movements of matter in space, can cause muscles to contract'; and mental thrusts work 'in some way, which must remain forever mysterious'.25 Ryle is one of the anti-causalist anti-volitionists. He wanted to know how something purely mental could have a causal influence in the material world where muscles contract. How is the gulf between mind and matter bridged? Descartes for his part saw no problem here. He once said 'if "corporeal" is taken to mean anything that can in anyway affect a body, then mind too must be called "corporeal" in this sense'.26 Of course, we are unlikely to be much impressed by this: a philosopher who tells us, as Descartes did, that the properties of thought and of extension are mutually incompatible can hardly be entitled to claim that there is any sense in which a thinking thing 'must be called "corporeal"'. But the possibility of using the claim in response to 25 Gilbert Ryle, The Concept of Mind (London: Hutchinson, 1949), pp. 62. 26 Letter to Hyperaspistes, August 1641, at 112 in Descartes' Philosophical Letters, tr. and ed. Anthony Kenny (Oxford: Clarendon Press, 1970). 387

Jennifer Hornsby Ryle shows that an objection of 'mental thrusts' on its own is only as powerful as the very familiar general objection to Descartes's sort of substance dualism - the objection which says that souls, being of a different kind from physical things, are alien to the world of causes and effects. Descartes hoped that his detractors might be persuaded to stop thinking of volitions as alien to movements by constructing a category, the 'corporeal', to which volitions and movements both belong. Evidently an analogous step could be taken in respect of my account if it seemed to need defence. In order to demonstrate that there is a category to which events of trying-to and movements both belong, one could say that an event of a person's trying to do something is, in some sense, physical. That would be enough to put any version of the familiar general objection to Descartes to rest. But Ryle's objection is actually more powerful than this allows. To see this, imagine someone who says that she can only conceive of electrochemical impulses as 'thrusts', and that she is puzzled about how their causing muscles to contract could illuminate human action. No doubt one sort of difficulty is alleviated if she is brought to see that there is a level of physical description which subsumes both the electrochemical and the mechanical. But even when the operation of neural transmission is made to seem unmysterious to her, she is not helped in understanding what a person's intentionally doing something consists in. If you hope to be better placed to understand those powers of persons which allow them to get things done by moving their bodies, then you would seem to be no better served by an account of neural transmission than by an account of a gland's being moved by a soul. Descartes's elaborate story (quoted above, invoking the animal spirits) is presumably meant to help us to understand the rational soul's active powers. The problem which Descartes faces and which could never be solved by calling the soul corporeal begins to emerge when we consider that story. In order for the soul's action to be found intelligible, the goings on around the pineal gland must be related to an understanding of human agency. The soul is a rational being, having intentional states. So we can ask Descartes: What does it will? The answer to this cannot be that the pineal gland moves. For a rational soul need not concern itself with the gland (just as ordinary active people need not concern themselves with neural transmission). It must be, then, that a soul is supposed to will (say) that a finger moves. But in that case the soul seems to have a magical power - the sort of power that we should attribute to a person if we could believe that she could directly move something remote from her. This is why Descartes's account of ordinary 388

Dualism in Action physical agency has been said to involve psychokinesis.27 The only thing that a soul can move directly is the pineal gland. But we can understand a rational being's capacity to move x mdirectly, only by thinking of it as having knowledge of how x can be affected by something that it can move directly.28 The trouble then is that souls do not have knowledge of how glands have to be affected for body parts to move, so that we lack any understanding of how something placed as the Cartesian soul is could be in a position to move (say) the little finger of a certain body. This remains, as Ryle said, 'forever mysterious'. The mystery here is created by the situation of the soul, and is independent of its non-physical nature. And my own account would introduce a mystery if 'try to —' were taken to apply to something that lacks capacities for movement. But there is no possible basis for supposing that 'try to —' could apply to something lacking such capacities.29 When accounts like mine are described, one often finds that 'try to —' is applied to nothing: philosophers often speak simply of 'tryings' - as if these might be unowned and (as it were) free-floating events. But of course what has to be meant by 'a trying', in any particular case, is someone's trying to do some particular thing. In my account, then, there is a place for things to which we actually predicate 'try to —'. Such things are human beings, whom we can readily conceive as having capacities to move their bodies. In order to conceive of them thus, we have only to think of ourselves, and to hold fast to the truth that there are no souls that our selves are. Whereas human agents are lost sight of in Descartes's picture of human action, they feature in mine. One can think that a person's action requires an event of her trying to do something without thinking of the person as composed from a proper part which tries. The claim that a person's trying to do something is distinct from " See Bernard Williams, Descartes (Harmondsworth: Penguin, 1978), pp. 288-92. 28 To use the terminology of basicness: everything an agent has it in her power to do is either something basic, or requires knowledge of how nonbasic things can be done by doing basic ones. The relevant notion of basicness here is a teleological one: see my Actions, Chapter 6. (I put the matter slightly differently from Williams Descartes, thinking as I do that a teleological notion of basicness is different from a causal one.) 29 In my Actions, I claimed that 'actions [and thus events of trying-to] occur inside the body'. The claim is misleading at best. But notice now that the idea was never that there is something inside the body to which predications of 'trying to —' attach. And see further the end of Appendix A infra. 389

Jennifer Hornsby her body's moving does not involve one in the idea that a person can be decomposed into a thing that tries and a body. V A Mysterious Inner Realm? A different sort of Cartesian malady has been thought to afflict my account. The allegation is not that I am involved in a distinction between mental and physical substances, but that I am involved nonetheless in a distinction which was bound to be present in Descartes, given his separating of souls from bodies. The distinction now is between a mental realm - wherein events of people trying to do things, or of souls' willing things may be supposed to occur - and a physical realm - wherein bodies move. In consequence of my holding that an agent's trying to do something results in a bit of her body's moving, it has been said that I am (i) Cartesian, and (ii) 'a mental action theorist'; and it has been said that, on my account of them, actions (a) are not 'overt', (b) are identified with 'purely mental acts of will', and (c) have their 'essence located in the will'.30 The critics who say these things recognize that even when persons are not problematically decomposed, the phenomenon of agency may still be. (Even where substance dualism is absent, Cartesian thinking may still be present - as we saw in section I.) Certainly, if these things were true, there would be more of an affinity between Descartes's account and mine than I have just allowed. In fact the allegations bring to light a difference between Descartes and me. I say that a person's trying to do something is an action (is her doing something that she does intentionally),31 whereas Descartes does not say that a soul's volition is an action. There is thus no question on my account, as there is on Descartes's, of an 30 See (i) R. A. Duff, Intention, Agency and Criminal Liability: Philosophy of Action and the Criminal Law (Oxford: Blackwell, 1990); (ii) Myles Brand, Intending and Acting: Towards a Naturalized Action Theory (Cambridge, MA: MIT Press, 1984), and Michael Moore, Act and Crime: The Philosophy of Action and its Implications for Criminal Law Intending (Oxford: Oxford University Press, 1993); (a) Brand; (b) Bill Brewer, 'The Integration of Spatial Vision and Action', in Spatial Representation, ed. Naomi Eilan, Rosaleen McCarthy and Bill Brewer, (Oxford and Cambridge, Ma: Blackwell, 1993), pp. 294-316; (c) Duff, Intention. 31 If there are cases in which a person tries in vain to move a part of her body, then the claim here is not a universal one. For present purposes, it makes no odds whether one accepts that there are such cases: the claim might be that where someone tries to do something and thereby intentionally does something, her trying to do the thing is her doing it. I discuss vain attempts to move the body in Appendix A. 390

Dualism in Action action's being 'partly in the mental realm', 'partly in the physical'.32 Still, this difference by itself will not impress the critics. For they think that a problem is exposed in my account as soon as a distinction between mental and physical is registered there. 'Even if actions themselves do not straddle the mental/physical divide,' they may say, 'it is objectionable that an account of action should straddle it. And as for actions themselves, these should be located firmly in the physical realm, not a mysterious inner one.' To get to the bottom of the objections envisaged here, we need to know why a claim of identities of actions with events of trying-to should be thought to make actions 'mental' and to place them beyond what is 'overt'. Suppose that you accept such identities - you accept, say, that her hitting the ball into the net was her trying to make a winning shot. Will you be led to say that her action of hitting the ball into the net must really be mental (seeing that is is describable using the word 'try')? Would you not rather say that her trying to make a winning shot must be physical (seeing that it is describable using the word 'hit')? You might equally well say either of these things. The claim of an identity of a putatively mental item (a person's trying to do something) with a putatively physical one (an action) might just as well be taken to reveal the physical character of the putatively mental item as the mental character of the putatively physical one. The objections can now be seen to rely upon the idea that a distinction between mental and physical corresponds to an actual division in the spatio-temporal world. If there were such a division, then no doubt one would be obliged to answer questions about which side of it actions, events of trying-to, and bodily movements fall on. But an event describable using both a piece of mental vocabulary ('try to —') and a piece of physical vocabulary ('hitting'), since it can equally well be said either to be mental or to be physical, might perfectly well be said to be both mental and physical. So the question 'Mental or physical?' has to be refused. The distinction between mental and physical does not partition the events that there are. And the assumption that there are boundaries in space between mental events and physical events must be rejected. Once the split between the soul and body has been renounced, there is no real divide for an account of action to straddle. 32 Descartes might have said that a volition is part of an action, the other part being a bodily movement. Not talking the explicit event language, Descartes did not in fact address questions about parthood. But some contemporary philosophers are explicit about actions having both mental and physical proper parts, taking this to be a sine qua non of action's psychophysical character. 391

Jennifer Hornsby If an assumption of a spatial mental/physical divide is made, then denying that actions are bodily movements appears to exclude them from the physical world. If an action is not a bodily movement but is someone's trying to do something, then, thinking of it with the putative divide in place, one conceives of it as hidden from view, as something which somehow initiates movements of a body. Actions then belong to a mysterious inner realm, separate from the outer realm inhabited by people's bodies. But when the assumption is rejected, there is no reason to think that actions belong anywhere in a picture containing the putative divide. Recognizing the identity of actions with events of trying-to helps to show (as we saw) that an intuitive distinction between mental and physical is inimical to such a divide. Refusing the identity of actions with bodily movements (we can now see) cannot create the mysterious realm which the divide introduces. For suppose that one really did have to say that actions, being causally anterior to bodily movements, must take up residence in a mysterious inner realm. Would it not then be in exactly such a realm that beliefs and desires were located by theorists who identify actions with bodily movements and who say that actions are caused by beliefs and desires? (Presumably beliefs and desires would be supposed to fall on the mental side of any mental/physical divide.) It is true that such theorists usually claim that the beliefs and desires which they take to cause actions are components of the same natural world as the physical things which they take all movements to be. But they are not entitled to such a claim if there is a problem with the idea that exercises of our powers as agents can be revealed in the items alongside which bodily movements are classified when the putative mental/physical divide does its work. There is a genuine difficulty about bringing events which are the doings of sentient beings who do things for reasons into relation with events conceived as on the farther side of a mental/physical divide. And bodily movements are often thought of as belonging on the farther side by philosophers: the claim that actions are bodily movements is often glossed as the claim that they are 'mere movements of the body', or that they are 'no more than bodily movements'.33 Bodily movements then come to be assimilated to items which might be there even if there were no persons whose bodies they were movements of. It is this assimilation, rather than anything in my account, which is the source of the genuine difficulty. If a difficulty sprang simply from denying that actions can be identified 33 See Davidson, 'Agency', at p. 59 in reprinted version; and Moore, Act and Crime, at p. 83, who announces that 'actions are no more than bodily movements' is a 'reductive' thesis. 392

Dualism in Action with bodily movements, then we should expect it to go away as soon as the identity was asserted. But in the presence of the difficulty, the step of identifying actions with bodily movements seems like subterfuge. The problem is to understand how a person's role in getting done the things that she does for reasons could be a matter of her operating on inanimate nature. The problem cannot be made to go away by declaring that an action (a person's doing whatever it is when she does something for a reason) is itself the operation of inanimate nature. A typical action theorist of today sees no problem at this point. Just as Descartes was content to call souls corporeal 'in a sense', in order to ensure that we should not have to think of causal transactions across alien kinds, so a typical action theorist of today is content with a homogeneous conception of those events which occupy the spatio-temporal world and participate there in causal relations. (Such a conception informs orthodox physicalism, as we shall see.) But perhaps even a typical action theorist has an inchoate sense of the problem. At least if he does, that would explain why it should be thought that distinguishing actions from bodily movements is tantamount to relegating actions to a mysterious inner realm (or to denying that they are 'overt', or to locating 'their essence in the will'). The problem, as I have said, arises from supposing that the bodily movements that there are when there are actions might be located in a world bereft of beings who do things for reasons - a world where so-called 'mere movements of bodies' belong. The supposition prevents one from treating movements in such a way that they can be rightly related to the agents who produce them. (And it makes no difference to this whether or not one says that bodily movements are the same as actions.) Those who make the supposition may see a point, as Descartes did, in calling a human body 'a machine'. They may forget that Descartes can be faulted for his assumption that corporeal substance excludes the features of thinking beings as much as for his more familiar assumption that the bearers of mental properties are not the sort of things to which physical properties attach.34 Ryle's objection to Descartes was that souls cannot be rightly related to what they are supposed to act upon. I said that the Rylean objection can be seen not to touch my account once it is allowed that 34 Descartes's belief in souls is normally thought of as arrived as through the introspective route he took in the Meditations. But part of his reason for attaching mental properties to a soul was a difficulty he thought he saw about attaching them to a substance whose principles of operation are purely mechanistic. See 'Descartes, Rorty and the Mind-Body Fiction', repr. In Simple Mindedness, pp. 24—41. 393

Jennifer Hornsby a human being is not detachable from an event which is her trying to do something. In allowing this, one rejects a Cartesian conception of people's possession of (intuitively) mental properties. What I say now is that the objection of a mysterious inner realm will present itself unless it is allowed that human beings are the bodily beings they are, and that the movements which they make are theirs. In allowing this, one rejects a Cartesian conception of people's possession of (intuitively) physical properties. VI Mental Causation: Dualism and Physicalism in Action I hope to have shown that there are no Cartesian assumptions in my account of action (section III), and that if there seem to be, that is because others read them in to it (section V). I want to suggest in conclusion that it is actually the orthodox physicalists' treatment of action, not mine, which is really aligned with Descartes's. The similarity of my account to Descartes's is partly to be blamed on their common focus of attention - on the agent's body. But notice that there are different reasons for this narrow focus. The reason for the apparent shortsightedness of my own account is simply a desire to generalize. If one hopes to say anything general about physical action, it is no good having one's sights on the world surrounding agents, because there are so many things of such various sorts that agents do. The thread running through them all is that the agent has to move to do them, and that is how the focus comes to be turned towards the agent's body. (The outlook of my own account is actually broader than the narrow focus suggests, because the things that agents can try to do are as many and various as the things they do.35 The point emerged in section III: my account does not deal only with people moving their bodies, but covers also all the more interesting things that they do.) In Descartes's account, attention to the agent's body has a different rationale. When the soul has been introduced, we are owed some account of its doings, and given the soul's situation, the close-up story of the production of movements is bound to be told. Descartes cannot simply acknowledge, as I do, a kind of being that has basic capacities of movement. But here the similarity, such as it is, shows up. Although I acknowledge that human beings have capacities of movement, I nevertheless discern a sort of causal complexity in exercises of those 35 There are plenty of substitutes for 'something' in 'the agent's trying to do something', plenty of verbs besides 'move the body' which can replace 'V in (T); and (T) introduces the agent's trying to do any of the things which she does intentionally. 394

Dualism in Action capacities. Human beings are complex beings; some of the events in which they participate depend causally upon others. The dependencies in the case of action, are dependencies of movements of parts of agents' bodies upon events of their trying to do things. So I think, as Descartes does, that when there is an action, a movement of the agent's body can be seen to depend causally upon something which is (intuitively) mental. This may be put in slightly different terms: both Descartes and I think that action involves 'mental causation'. Put in these terms, what I have been trying to establish, in order to show that the similarity does not go deep, is how very differently Descartes and I treat 'mental causation'. Since nearly everyone accepts that action involves 'mental causation', what distinguishes Descartes from me is something of which nearly everyone must take a view.36 'Mental causation' has been of great concern recently, especially among orthodox physicalists.37 None of the claims of orthodox physicalism was required to avoid Ryle's objection of a mysterious gulf. We saw that this objection is avoided by insisting on the sameness of that which tries to do something and that whose parts it can move. The movements which are caught up in the understanding of such a thing - of a human being — are then individuated as events in which someone's participation is crucial, and not as the subject matter of physics or of any other science. Bodily movements are physical of course. But the sense of 'physical' in which it is obvious that they are physical is not that which has informed the recent debate on 'mental causation'. 36 I say 'nearly everyone' to allow for the anti-causalists (see section II supra). The treatment of mental causation is a question for all causalists. It might seem that there is a special question for Descartes and me, because we accept (what many don't) that the agent's body is a locus of 'mental causation'. (Many think that one has to look to actions' antecedents — to what occurs before anything bodily - in order to find anything which is both psychologically describable and causally operative.) Still, we saw in section III that even someone who rejects my general claims about action may accept that there are occasions when a person moves her body therein trying to do something. So perhaps nearly everyone accepts that the agent's body is sometimes a locus of mental causation. That would ensure that there is in fact no special question for Descartes and me. But however this may be, nearly everyone allows that there is 'mental causation'. 37 And it has been the topic of a massive literature: see, e.g., the papers in Mental Causation, ed. Heil and Mele. I attach square quotes to 'mental causation', being reluctant to think of the causal dependencies which correspond to persons' causal complexity as marking out any kind of causation: see note 40 infra. 395

Jennifer Hornsby Most contemporary philosophers think that physicalism requires one to be able to see the mental's causal operation as an example of the world's working causally in such a way as to reflect its law-like workings. Their treatments of 'mental causation' encourage one to take the close-up view of the agent which Descartes took. They may say that events of trying-to, if they are causally responsible for movements of bodies, are, or are constituted by, 'brain events'. But an objection of Cartesianism arises however this is interpreted. If a brain event is something in which a brain participates, then the orthodox physicalist tells us that the causal transitions involved in human action are transitions between brain and body. In that case he accepts a version of substance dualism in the broad sense identified in section I. We saw in section IV that the principal objection to Descartes does not actually depend upon the nature of the soul; and this means that if one thinks of a person's trying to do something as the brain's doing something, one renders physical action mysterious as Descartes did. There is a kind of causal dependence encountered when effects are produced by a being with contentful states (a being that can will something, or try to do something); and this kind of dependence is not found intelligible when causal properties are attributed to something located inside a body. The other possibility is that calling something a brain event is a matter of locating it in the domain of neurophysiology (rather than of thinking of it specifically as the brain's doing something). Brain events in this case are among the flux of events in nature, unowned and free-floating, as it were; and the causal connections which are examples of 'mental causation' are discoverable without finding something to which 'try to —' can be predicated. But this is equally problematic. We have seen that the items of Descartes's story - the volitions which belong to souls, and the movements which belong to mind-excluding substances - are foreign to a proper account of physical action. Equally foreign must be the unowned and free-floating events. For the underlying difficulty is to think about the production of bodily movements as human action even when the causes of those movements are supposed to be identifiable without making reference to any bodily being. If one takes bodily movements to be robust presences 'in the physical world', then, in searching for their antecedents with the agent removed from the scene, one thinks of inner items, and then one may conjure up an inner realm for those items to inhabit. Here a Cartesian difficulty stems from attempting to find what are actually changes in a rational being inside a world which one had hoped to conceive of as physical in some exacting sense. The orthodox physicalist, in avoid396

Dualism in Action ing the mysterious gulf, puts herself under pressure to introduce the spurious divide between mental and physical.38 The problem here, for the orthodox physicalist, is the one we saw in section V - about understanding how causal transactions in inanimate nature could account for a person's role in getting done the things that she does for reasons. Causal dependencies which reflect the causal complexity of a human being are not examples of the world's working causally in such a way as to reflect its law-like workings. The phenomenon of 'mental causation', in which human beings show up as causally complex beings, cannot consist in pairs of particulars standing in a relation of causation as this is typically conceptualized by philosophers.39 So the dependencies encountered in human agency are not the 'physical causation' to which orthodox physicalists have wanted to assimilate 'mental causation'.40 We saw in section I that someone might be anti-Cartesian with38 Cf. John McDowell, Mind and World (Cambridge, MA: Harvard University Press, 1994), at p. 90. Put in the terms that McDowell takes from Sellars, what the present paper argues is that human physical action is situated in the space of reasons, where the space of reasons is to be contrasted not with the space of causes but with nomological space (and where the space of reasons, evidently, is not the space just of cognition). 39 For reasons to reject the typical conceptualization, see Helen Steward's arguments against what she calls the network model of causation in her Ontology of Mind: Events, States and Processes (Oxford: Clarendon Press, 1997). 40 In 'The Mental Causation Debate,' Aristotelian Society Supplementary Volume 69 (1995), 211-36, Tim Crane argues that the dominant contemporary versions of physicalism implicitly reject the assumption of 'homogeneity' — the assumption 'that mental and physical causation are the same kind of relation'. It seems then that I am on the side of contemporary physicalists in my treatment of mental causation. Well, I am more than happy to acknowledge Crane's point that there is a homogeneity assumption which provokes contemporary physicalist treatments of mental causation but which they find themselves forced to abandon. (I take their abandonment of the assumption to be symptomatic of a problem which is inherent in the orthodoxy and which I have tried to expose here.) But it would be an oversimplification of my own view to say that mental and physical causation are different kinds of relation. It is rather that we have to stop thinking that all causation can be understood by reference to the going model of 'physical causation' (cf. Steward, Ontology, and my 'Causation in Intuitive Physics and in Commonsense Psychology', in Simple Mindedness, pp. 185-94); and that we have to allow for the species of intelligibility that is peculiar to rational sentient beings. I should note that my arguments here - about treating events of tryingto as brain events - are directly addressed to a version of physicalism which does not flout the homogeneity assumption as Crane sets things up. But 397

Jennifer Hornsby out endorsing any orthodox physicalist doctrine, and that someone might be a Cartesian orthodox physicalist. What we see now is that endorsing physicalist doctrine is actually just a way of being Cartesian. Orthodox physicalisms's attitude to causation is a source of Cartesian thinking. APPENDIX A: On Landry's Patient In defending (T) against charges of Cartesianism, I pointed out that a person's trying to do something can usually be identified with an action. But if there are cases in which a person tries in vain to move a part of her body, then a person can try to do something without there being an action of hers. It has been thought objectionable that I should allow such cases. Bill Brewer puts an objection to my account of action, saying that 'the subject is distanced from movement in her body in such a way as to threaten her status as agent'.41 The case to which Brewer and others speak is the case of Landry's patient. The patient had lost all sensation in one arm. When his eyes were closed, he was told to raise his arm; unknown to him, his arm was held down — it was prevented from rising; and when he opened his eyes, he was surprised to find that it had not risen. It seems natural to say that, although he didn't raise his arm, he tried to. It may be that in contemplating Landry's patient, one starts to think in phenomenological terms about events describable using the word 'try': one thinks that it must have seemed to the patient just as it would have seemed if he had moved his arm (if the arm, in which he had no sensation, had not been prevented from rising); and then one may start to think of events of trying to do things under the aspect of seemings. But in fact one does not need to focus on the phenomenology to judge that the patient tried to move his arm. The judgment might be based on knowledge that he is obedient to instructions: obedient as he is, he does what he can to raise his arm when told to do so; and his belief that he has raised his arm, evinced in his surprise that it had not moved, is then a further piece of evidence that he tried to raise it. 41 Brewer, 'Integration', p. 306. when I say that it is part of the orthodoxy that one must be able to see the mental's causal operation as an example of the world's working causally in such a way as to reflect its law-like workings, I intend to speak to other versions of physicalism, including what Crane calls 'the constitution view'. I thank Paul Boghossian, David Papineau, Scott Sturgeon, and (especially) Tim Crane (who prompted me to re-read his 'Mental Causation Debate') for questions asked after the lecture on which this paper is based. 398

Dualism in Action It is not that we should necessarily go wrong if we considered how things seemed to the patient. But there is no reason to think that we must be working with some purely phenomenological notion when we think of Landry's patient as having tried to raise his arm. And of course it cannot be right generally to think about events of trying-to as seemings. (We could not have acquired the concept of trying to if we had had only phenomenology to work with.) We must not forget what an extraordinary epistemic and practical situation Landry's patient was in: he was not allowed to see; he was proprioceptively incapacitated; and his movements were obstructed. We can agree with Brewer that 'his status as agent is threatened'. Thoughts that we have about his case cannot be expected to generalize to other cases. Landry's patient's failed attempt is at least as unsuited to providing a model for action as the case of a false belief is unsuited to providing a model for knowledge. Still if we do accept that Landry's patient did try to move his arm, then the account I gave in section III, of the difference between successful and unsuccessful attempts, applies in this case. Thus: a person's trying to raise her arm is her raising it if is it causally responsible for her arm's rising, but is her unsuccessfully trying to raise it if no event of her arm's rising ensues. It is this to which Brewer really objects. Brewer wants to be able to say that someone's unsuccessfully trying to raise her arm is of a fundamentally different kind from her raising it. And he suggests that we need a 'disjunctive conception of tryings'. The trouble is that a disjunctive conception of tryings seems quite implausible in general. Remember the case of the typist who typed an '@' sign when the key had been reassigned. Is her moving her finger against the keyboard of a fundamentally different kind from the kind it is of when the key has not been reassigned and she actually types a '£' sign? Surely not. I believe that it is a disjunctive conception of bodily movements, not of events of trying-to, that we need if we want, as Brewer and I both do, to keep the subject in touch with movements of her body. (Such a conception is implicit in what I have said at section V and echoed in section VI supra, and it is explicit in my writing elsewhere.42) Landry's patient's case was one of the things that encouraged me to say (as I did once upon a time, cf. note 29 supra) that actions occur inside the body. That claim, though misleading at best, need not be Cartesian, because it could be that predictions belong properly to a whole substance, even where the events whose occurrence actually 42 'Postscript' to 'Bodily Movements, Actions and Epistemology', in Simple Mindedness. 399

Jennifer Hornsby makes those predictions true are locatable in a volume smaller than the whole substance. Consider: when I varnish the table, the event of its coming to be shiny is plausibly located at the surface of the table, even though being shiny is a property that the table comes to have. (This is only analogous in some respects of course. And one will not see the point in such an analogy until one has rid oneself of the orthodox physicalists' way of thinking about causation.) APPENDIX B: On Descartes' Dualism The distinction I made in section I, between dualism and Cartesianism, is different from the distinction Baker and Morris make between Descartes's dualism and Cartesian dualism.43 'Cartesian dualism' in the sense of Baker and Morris is certainly Cartesian; the argument of their book is that Descartes did not hold the doctrine known as, and criticized as, Cartesian dualism. Here I make some remarks about how Baker and Morris's challenge to the nowadays usual reading of Descartes might affect what I say. Baker and Morris's distinction shows that we can understand 'Cartesianism' if we know only the recent secondary literature. And for the argument of my paper, it would not matter if no-one actually held the views I attribute to Descartes (so long as they are wrong, and they line up more with orthodox physicalists' views than with mine). But I am inclined to think that the historical Descartes actually has slightly more in common with the Descartes to whom I attribute views than Baker and Morris would allow. Part of Baker and Morris's attack on the idea that Descartes was a Cartesian dualist is their claim that he was neither a volitionist nor an interactionist (in the usually meant senses). They suggest that the view 'that voluntary action is to be analysed in terms of volitions that are efficient causes of bodily movements' is ascribed to Descartes without any grounds. There are three things here of which Descartes's present-day expositor might be guilty: (a) crediting Descartes with analytical ambitions, (b) assuming that the volitions of which Descartes spoke are (to put it in my terms) unowned and free floating events, (c) assuming that Descartes held that 'cause' relates volitions and movements. On (a): I have not said that Descartes attempted to analyse voluntary action (see note 14 supra). On (b): I have been careful (as Ryle perhaps was not) to see the volitions of which Descartes speaks as some soul's volitions. On (c): I have deliberately used 'is causally responsible for' (which I take to be more open-ended than 'causes') in my statement of Descartes's 43 In Descartes' Dualism. 400

Dualism in Action account. I acknowledge that there is much more to be said on the subject of Descartes and causation, and that most expositors say very little. But I do not think that there can be any doubt that Descartes took his dualism to be the upshot of a correct understanding of causal transactions in the physical world (as I contended at the end of section V). That does not establish that Descartes was an interactionist, rather than an occasionalist. (And Baker and Morris endorse Russell's claim [which certainly seems sustainable] that occasionalism is derivable from premises in Descartes.) But the question for Descartes - about how to accommodate human action in a world of bodies as he conceived them — remains, whether or not, in giving his own answer to it, Descartes resisted a crude interactionism and plumped for occasionalism. Baker and Morris say that Descartes simply conceded that 'there could be no intelligible connection between soul and body' (p. 56, their italics). Well, presumably if Descartes was an occasionalist, then he might have allowed that a connection between the two was intelligible to God. And even if, as Baker and Morris say, Descartes did not expect us to understand soul/body transitions, we find Descartes trying to make them less unintelligible to us, e.g. when he says that the soul might be called corporeal. If Descartes had not been at all inclined to make any attempt to find the connection intelligible, then we should not expect to find the Passions passage (quoted in section II supra), which contains the close-up story. In any case, the stock criticism of Descartes, on which I rely, says only that the connections is not intelligible to a follower of Descartes's. (Here again there is much more to be said - now under the head of 'the substantial union of soul and body'.) What I call volitionism can probably be attributed to Descartes simply on the basis of the passage I quoted. To the extent to which it can be shown to be doubtful that he held that account, Descartes was a less consistent philosopher of mind - albeit it a more interesting one — than is commonly supposed. 401

Index of Names Agre, P., 35 Alexander, S., 3, 10 Allison, H., 326, 329, 332-4 Antony, L., 230 Armstrong, D. M., 5, 16, 237 Austen, J., 103-4 Ayer, A. J., 144 Bacandall, M., 168-70 Bacon, F., 78 Baker, G., 377, 400-1 Ballard, D., 39, 40 Baylis, G., 198 Bell, D., 232 Bennett, J., 368 Berkeley, G., Ill Berlin, I., 84, 357 Blackburn, S., 61, 79, 317 Block, N., 6, 17, 20, 83, 157, 182-4, 195 Bogdon, D., 44 Boghossian, P., 167 Boring, E., 3 Brand, M., 390 Bretano, F., 229-234, 238, 243, 246-7, 251-2, 255, 257, 263, 267 Brewer, B., 184-6, 220, 390, 398-9 Broad, C. D., 3, 5, 330 Brueckner, A., 279 Burge, T., 61, 213, 274-6, 280-1,283-9,296,300-1 Campbell, J., 190, 192-3, 197, 220,222 Carruthers, P., 53 Cartwright, N., 364 Caston, V., 229, 246 Chalmers, D., 16, 19-21, 159, 181,183-4 Cherniak, C, 96 Chisholm, R., 165, 247-8, 357 Churchland, P. S., 36, 38 Collingwood, R. G., 77-8, 84, 90 Cottingham, J., 379 Craig, E., 229, 344 Crane, T., 397 Crick, F., 23, 25-34 Curley, E., 332 Davidson, D., 138, 146, 254-71, 309-12, 316, 378, 382-3, 392 Davies, M., 83, 181,232, 249-50, 277, 279, 282-3 Decety, J., 46 Dennett, D., 44, 48, 160-2, 166-7, 232, 24, 246, 247-8 Descartes, R., 5, 145, 314, 332, 337, 379-84, 387-96, 400-1 Dilthey, W., 84, 90 Dretske, F., 162-3, 166, 176, 230-1,279,280 Dreyfus, H., 196 Driver, J., 198 Ducasse, C. J., 164-5 Duff, 390 Dummett, M., 244, 258, 261 Eccles, J., 140 Eilan, N., 190, 206, 222 Einstein, A., 365 Elman, J., 38 Evans, G., 190, 206-8, 217, 220, 222,281 Feigl, H., 4, 21 Fetz, D., 46 Field, H., 247, 253 Fischer, J., 368-9 Fodor,J.,5, 13, 37,49, 232, 257, 276, 279, 280, 283, 285-9, 305,312 Foster, J., 204 Freud, S., 231 Gardiner, P., 78-9 Garner, A., 198 403

Index of Names Garnett, A., 357 Geach.P., 113,207 Gibson, J., 35, 191, 195,220 Ginet, C, 369 Goldman, A., 57, 60, 63, 87, 323 Goodale, M., 26 Gordon, R., 63, 68, 77, 80, 87, 206 Grayling, A. C, 232 Grush, R., 45 Guyer, P., 330 Hamilton, W., 233 Hare, R. M., 8, 9 Harman, G., 177-8, 234 Harris, P., 68 Harrison, R., 331 Hartman, J., 26 Haugeland 37, 44 Heal, J., 57, 60, 71-3, 76, 87, 313 Heidegger, M., 196-7 Hempel, C, 74 Henik, A., 192,197, 198 Honderich, T., 138, 151 Horgan, T., 9, 278 Hume, D., 132, 320-4, 328, 373 Husserl, E., 229, 231,244 Hutchins, E., 35, 48 Ito, M., 46 Jackendoff, R., 48 Jackson, F., 16, 20, 178, 236-7, 367 James, W., 192 Kahneman, D., 72, 191-2, 197, 198,200 Kanizsa, G., 192 Kant, I., 77, 84, 90, 149, 186-7, 306, 324-46, 362 Kinsbourne, M., 133 Kirsh, D., 35 Kitcher, P., 327 Knierim, J., 38, 49 Knudsen, C, 229, 232 Koch, C, 23, 25-6, 28-34, 36 Korsgaard, C, 350, 370-1 Kripke, S., 303-4 Kuhberger, A., 70 Kwato, M., 46 Lear, J., 337 Leslie, A., 69 Levin, J., 61 Levinas, E., 244 Levine, J., 16, 20 Lewis, D., 5, 16, 358, 360, 366-7 Locke, J., 321-22, 324, 328, 348 Mackie, J. L., 236 Malcolm, N., Ill Malpas, J., 249 Marr, D., 11,35,39 Martin, M. G. F., 234, 236-7, 251 Matraric, M., 40 McCulloch, G., 242, 249, 251, 275,282 McDowell, J., 76, 107, 119, 177, 186-7,192,203,205,273, 281,304,308,314,397 McGinn, C, 230, 232-5, 249-51,276-7,280-1,317, 323 McGinn, C, 230, 232-5, 249-51,276-7,280-1,317, 323 McKinsey, M., 279 McLaughlin, B., 277 Melnyk, A., 12 Merleau-Ponty, M., 35, 42 Millikan, R., 41, 280, 287, 295, 297,300 Milner, A. 26 Moore, G. E., 8, 9, 164-5, 351, 355,357 Moore, M., 206 Moran, R., 77, 231 Morgan, C, 11 Morris, M., 382, 400-1 Nagel, T., 54, 141-2, 173, 338 Nagel, E., 13-5 Naughton, B., 43 Neisser, U., 200 404

Index of Names Nichols, S., 54, 58, 64, 67, 69-70, 80, 92 Nidditch, P., 321 Nietzsche, F., 352 Nisbet, R., 66, 67 O'Shaughnessy, B., 384, 386 Pais, A., 365 Papineau, D., 279, 280 Peacocke, C, 177-8, 210-11, 213,309 Penfield, R., 28 Pettit, P., 216, 273 Place, U. T., 3-4, 248 Plato, 149 Platts, M., 306, 307 Pollen, D., 25 Popper, K., 140 Price, H., 163-6, 172-5 Putnam, H., 5-6, 11, 17, 19, 118, 177, 216, 273-4, 277, 281 Pylyshyn, Z., 5, 37, 39 Quine, W. V. O., 247, 254-61, 264-71 Quinton, A., 137 Ravenscroft, I., 58 Reid, T., 165 Roessler, J., 206 Rorty, R., 250-1 Rubin, E., 198 Rumelhart, D., 38, 48 Russell, B., 354, 371-3, 401 Ryle, G., 3, 21, 354, 387-9, 393, 395, 400 Sartre, J.-P., 229, 231,242-3, 245 Schopenhauer, A., 264 Searle, J., 7, 19, 23-4, 27-8, 141, 143, 230,238-44, 246, 250, 260, 263-4 Sellars, W., 165 Shoemaker, S., 177, 234, 240, 249 Smart, J. J. C, 3-4, 21 Snowdon, P., 177 Sorenson, R., 57 Sperber, D., 72 Sprigge, T., 141 Stalnaker, R., 358 Stein, E., 96 Steward, H., 397 Stich, S., 54, 58, 64, 67, 69, 70 Stone, T., 83 Stout, R., 190 Strawson, P. F., Ill, 186, 206, 219,250,325,328,340-3, 350,375 Strawson, G., 264, 319, 350 Thelen, E., 36-37, 45 Tomasello, M., 123 Turing, A., 49 Tversky, A., 72 Tye, M., 165, 232, 234, 238, 243-4, 277, 279 Ullman, S., 200 van Ingwagen, P., 360 van Gelder, T., 36-37, 41-42, 44,50 van der Meulen, J., 45 Varela, F., 35 Vico, G., 84 Walsh, W., 323, 327 Walton, K., 58 Weiskrantz, L., 208 Wiggins, D., 219, 357, 360, 372-3 Wilkes, K., 250-1 Williams, B., 59, 358, 389 Wimmer, H., 58 Wittgenstein, L., 3. 101, 108-22, 133-3,258,323,337,339,380 Woodfield, A., 216 Woods, M., 220 405

Philosophy Editor ANTHONY O-HEAR M.A., Ph.D., 14 Gordon Square, London WC1H OAG. Published for the Royal Institute of Philosophy. The journal is concerned with the study of philosophy in all its branches: logic, metaphysics, epistemology, ethics, aesthetics, social and political philosophy and the philosophies of religion, science, history, language, mind and education. The journal is not committed to any particular school or method and contributors are expected to avoid needless technicality. There is a section on new books which includes reviews, booknotes and a list of books received. Subscription (until 31 December 1998) for four issues and two supplements is £130.00 (US$222.00 in the USA, Canada and Mexico). Single parts: £25.00 or $40.00. Members of the Royal Institute of Philosophy receive the journal free and may purchase the supplements separately. Membership costs £25.00 or $45.00 (Students £10.00 or $22.00). ISSN 0031-8191 Royal Institute of Philosophy Supplements Volume 26 1989 27 1990 28 1990 29 1991 30 1991 31 1992 32 1992 33 1992 34 1993 35 1993 36 1994 37 1994 38 1995 39 1995 40 1996 41 1996 42 1997 Philosophy and Politics Explanation and its Limits Wittgenstein Centenary Essays Human Beings A. J. Ayer: Memorial Essays Religion and Philosophy Philosophy, Religion and the Spiritual Life The Impulse to Philosophise Philosophy and Cognitive Science Ethics Philosophy and the Natural Environment Philosophy, Psychology and Psychiatry Philosophy and Technology Karl Popper: Philosophy and Problems Philosophy and Pluralism Verstehen and Humane Understanding Thought and Language From Volume 13 onwards the Series is published by Cambridge University Press and some earlier titles are also available. ©The Royal Institute of Philosophy 1998 CAMBRIDGE UNIVERSITY PRESS 1358-2461(1998)72+2;1-7