between having a feeling and making a judgement about it'. See M. G. F. Martin, 'Bodily Sensations', forthcoming in the Routledge Encyclopedia of Philosophy, ed. Craig. Nor are the inferences involving statements about objects of pain-states, by looking at the valid inferences which are made with statements concerning pain. For example: X has a pain in his foot; therefore there is something X has in his foot. The plausibility of these arguments is, in my view, relatively superficial, for the reasons given in note 21 below. 236
Intentionality as the Mark of the Mental But I do not need to dwell on the arguments for mental objects here, since the defence of intentionalism does not need to appeal to them. Intentionalism about bodily sensations can be defended instead by appealing to a perceptual account of bodily sensations, such as that of D. M. Armstrong, or the kind more recently defended by Michael Martin.22 On this account, bodily sensation is a form of perceptual awareness of one's body. It is by experiencing bodily sensations that we come to be aware of the state of our body, and of events happening within it. The qualities of which we are aware in bodily sensation - the sensory qualities of hurting, feeling cold or warm and so on - are predicated in these experiences of parts of the body. When one feels a pain, one normally feels it to be in a part of one's body; and even when a pain is felt where there is no body-part in which to feel it - as in the case of phantom limb pains - what subjects feel is that their body extends further than it actually does. They do not feel as if their pain exists in mid-air, a few inches from where they have lost their limb. The strongest considerations in favour of this view derive from this felt location of bodily sensation. An ache in my hand feels to be in my hand, not in my mind. Rather than being something which is contained within my body, it presents itself as something which my mind can concentrate on, attend to and try to ignore. In fact, this much is common ground between the believer in mental objects and the perceptual theory. But what tells in favour of the perceptual theory is the fact that to concentrate on the ache, I must necessarily concentrate on the part of my body which aches; the mental object theory cannot explain this necessity. Attending to bodily sensations is achieved by attending to a part of the body where these sensations feel to be. This is because bodily sensation is a form of awareness, the awareness of things going on in one's body.23 22 See D. M. Armstrong, A Materialist Theory of the Mind (London: Routledge and Kegan Paul, 1968); M. G. F. Martin, 'Bodily Awareness: a Sense of Ownership', in The Body and the Self, ed. J. L. Bermudez, N. Eilan and A. Marcel (Cambridge, MA: MIT Press, 1995); and his 'Sense Modalities and Spatial Properties', in Spatial Representation, ed. B. Brewer, N. Eilan and R. McCarthy (Oxford: Blackwell, 1993). 23 Note that an advantage of this view is that it can give a univocal account of both the bodily sensations which are naturally identified in sensations uncontroversial; for although we might be happy with the inference from 'X has a pain in his foot' to 'There is something which X has in his foot', the inference from 'X has a pain; Y has a pain; therefore there is something which X and Y both have' is clearly invalid if the something is supposed to be a particular object, and irrelevant to the present issue if it is supposed to be a property. 237
Tim Crane Why call this intentionality? What this perceptual theory says is that in bodily sensation, something is given to the mind, namely the body, or a body part. Calling this phenomenon 'intentionality' classifies it together with the case of outer perception, where the perceived portion of the world is 'given' to the mind; and with thought, where some object, property or state of affairs is 'given' to the mind. What is in common between these different states of mind is expressed in Brentano's formulation: 'in the idea something is conceived, in the wish something is wished'. And in the sensation something is sensed: the body. Ill The Intentionality of Emotion That is the basis of my case for an intentionalist view of bodily sensation. I now want to move on to the second kind of counterexample to Brentano's thesis: Searle's examples of 'nervousness, elation and undirected anxiety'. How should an intentionalist deal with these apparent examples of non-intentional mental phenomena? First we need to identify the phenomena in question. This is actually harder to do than it might initially seem. Everyone will agree that there is such a thing as being anxious and yet not being able to give an answer to the question 'what are you anxious about?' But this by itself does not show that anxiety can lack intentionality. For one thing, we have just seen that asking 'what is X about?' is not always the most uncontroversial way of deciding whether X is intentional. And more importantly, it should not be a condition of a state's being intentional that the subject of that state must be able to express what the state's content is, or even which kind of state it is. Every theory of intentionality must allow that subjects are not always the best authorities on all the contents of their minds. A possible intentionalist account of the state of mind in question would be to say that the intentional object of the state of mind is its cause. So on this view, when we describe ourselves as 'just anxious without being anxious about anything in particular', we mean that we do not know the cause of our anxiety. Now in some cases, it is terms of what they are of - warmth, cold, pressure etc. - and those which are not, like pains and so on. This version of the thesis that bodily sensations are intentional should be contrasted with Tye's view that pains give one non-conceptual representations of damage to one's body: see Tye, Ten Problems, chapter 4. Tye's view is, however, consistent with the view defended here. Pains may have many levels of representational content; my concern in this paper is with the uncontroversial phenomenological content they appear to have. 238
Intentionality as the Mark of the Mental certainly true that to identify the cause of an emotion is to identify its intentional object. But this cannot be true in general. For one thing, the object of the emotion might lie in the future. Or the cause of an emotion might be a past event which is too remote from the present manifestation of the emotion to be properly regarded as its object. (It may be true that the cause of someone's fear of dogs was a childhood encounter with a certain dog — but it would not always be right to say that that dog was the object of their current state of fear in the presence of a different dog.) Or the cause of the emotion might be something completely unrelated to its object. (A drug may cause you to hate some person or thing.) So the fact that an emotion has a cause does not by itself entail that it has an intentional object. The intentionalist cannot refute Searle merely by pointing to the fact that emotions have causes of which we are sometimes ignorant. But, as we have just seen, nor can Searle infer that there are nonintentional emotions merely from the fact that we sometimes say we are anxious without being able to say what we are anxious about. Searle presents the existence of non-intentional emotions as if it were something entirely obvious. An intentionalist, however, will deny that it is obvious. There can be no real debate about this matter if we are restricted to each participant stating what they think is obvious. So how can the debate proceed? In order to assess what is at issue between Searle and the intentionalist, we need to know more about how they would classify the various emotions into kinds. What is it that makes anxiety, for instance, the state it is? Whatever it is, it must be common to the cases where anxiety clearly does have an intentional object and the cases which Searle is calling 'undirected'. Remember that these are the cases where someone is anxious but it is not clear to them what they are anxious about. The issue between Searle and the intentionalist is whether the existence of these cases establishes that there are mental states which have no intentionality. If we learn more about what the intentionalist and non-intentionalist think emotions are, we can assess their competing claims over whether any of them are 'undirected' in Searle's sense. Let's start with non-intentionalism, Searle's position. Perhaps non-intentionalism could say that anxiety is distinguished from (say) an undirected state of contentment by the functional roles of the two states. The functional roles must be explicable in commonsense psychological vocabulary, since we are after a phenomenological classification of the emotions. And yet the functional roles must be relatively informative - 'behaving anxiously' will not do, in this context, as a characterization of part of the functional role of anxiety. So perhaps we can say that anxiety is characterized by the anx239
Tim Crane ious person's inability to concentrate, or by an obsessive concern with trivial details of life, or by a jumpy, nervous form of behaviour. Contentment, by contrast, might be characterized by a benign way of behaving towards the world, an enthusiasm for its daily tasks and so on. However, this style of identifying the functional roles of anxiety and contentment does so in terms of forms of behaviour which are manifestly intentional. So while it might suffice for an account of intentional ('directed') anxiety, it will not do for undirected anxiety. For the non-intentionalist, there must be something directed and undirected anxiety have in common, which licenses them both being called 'anxiety'. And this 'something' must be detectable from the subject's point of view, if Searle's claim is going to have any force — remember that Searle was appealing to what is obvious to us. Yet this something must also be non-intentional: it cannot be directed on anything. So the non-intentionalist must say that an emotion like anxiety (directed or undirected) has properties which are phenomenologically detectable to the subject, but are non-intentional, involving nothing beyond themselves. These properties must therefore be qualia: non-intentional, subjective properties. Just as there are (according to many philosophers) pain-qualia and seeing-redqualia, there are also emotion-qualia which give the emotions the characteristic phenomenal 'feel' which they have. Let us suppose, then, that anxiety is partly characterized by its distinctive qualia. Now it is a plausible general thesis about qualia that there is no intrinsic connection between any particular quale and being in any particular objectively-identifiable mental state. For instance, there is nothing intrinsic to the qualia involved in seeing red that links these qualia with the state which plays the functional role of seeing red in normal observers. The coherence of inverted qualia thought-experiments depends on there being no such links, and most defenders of qualia, like Shoemaker, believe that qualia inversion is possible.24 In fact, it seems part of the very idea of qualia that there be this possibility: for qualia 'point to' nothing beyond themselves, which would make them associated with one kind of objectively-identifiable state rather than another. So on the non-intentionalist view of emotion, it must be true that there is nothing about the qualia associated with anxiety themselves which make them anxiety-qualia: that is, associated with a state with the particular functional role of anxiety. Just as seeing-red-qualia could, in some other possible world, be associated with the state 24 See Shoemaker, 'Qualities and Qualia', pp. 108-13, where he discusses the inverted qualia speculation. 240
Intentionality as the Mark of the Mental which in the actual world is seeing green, so anxiety-qualia could be associated with some other emotion-state, say contentment. This is because there is nothing in the qualia themselves which connects them with particular kinds of emotion, objectively identified (for instance, in terms of functional role). So now it appears that a non-intentionalist has to accept the possibility that there is a world in which contentment feels to someone as anxiety feels to me. And while the inverted qualia story seems plausible when applied to simple colour-qualia - after all, why shouldn't green things look to you the way red things do to me? - the story is very hard to believe when applied to the putative emotion-qualia. For here we are supposing that the same emotion might feel in opposite ways to two subjects in different possible worlds — emotions have their distinctive 'feel' only contingently. But does this possibility really make sense? One might respond to this: so much for the plausibility of the view that there are emotion-qualia. And I agree: even if there are qualia, to assimilate anxiety to the experience of seeing red is a distortion of ordinary experience. But how else is the non-intentionalist going to describe the characteristic phenomenology of anxiety, undirected and directed? The nature of these states cannot be described in terms of how things seem to the subject, however vaguely stated. For descriptions of how things seem are patently intentional, and so they will not capture the phenomenology of undirected anxiety. Non-intentionalism is committed to emotionqualia because it is committed to emotions having properties which are non-intentional yet phenomenologically salient - and nonintentional, phenomenologically salient properties of mental states just are qualia, by definition. But what is the alternative to this non-intentionalist view? How should an intentionalist give an answer to the question about how to distinguish the different emotions? One answer has already been suggested. Someone experiencing anxiety might not be able to put into words what it is they are anxious about; but they may still be able to say how things seem to them in their state of anxiety. And even if they can't express it, there is still nonetheless such a thing as how things seem to them. To begin with, the intentionalist will start by distinguishing being anxious for oneself, and being anxious for another. This is clearly an intentional distinction: in the one case, one's mind is directed on oneself, in the other case, it is directed on another. The cases Searle mentions are not cases where one is anxious for another: otherwise it would be directed anxiety. So the intentionalist will say that these are cases where one is anxious for oneself — so in these cases, one's anxiety is directed upon oneself. 241
Tim Crane Being anxious in this way is a matter of having a certain attitude to oneself and one's position in the world: it is to regard the world, for example, as a potentially disturbing place for oneself. This is one way in which anxiety exhibits directedness. And it is an alternative to seeing Searle's cases as examples of mental states which are directed on nothing, as Searle does. It might be helpful to contrast, in these very general terms, anxiety with depression. In depression, the world seems to the subject to be a pointless, colourless place: nothing seems worth doing. The change involved in coming out of a depression is partly a change in the subject's apprehension of the world. Things seem to have a significance, a purpose which they previously lacked. And this can be true of a subject even when they cannot say what they are depressed about. In this way, the phenomenon Searle would call 'undirected depression' can be seen as having a certain directedness or intentionality. These brief remarks suggest that the difference between anxiety and depression resides in the different manners in which the world, and the subject's place in the world, are apprehended in the emotion. This was Sartre's view: My melancholy is a method of suppressing the obligation to look for... new ways [to realize the potentialities of the world] by transforming the present structure of the world, replacing it with a totally undifferentiated structure... In other words, lacking both the ability and the will to carry out the projects I formerly entertained, I behave in such a manner that the universe requires nothing more from me. This one can only do by acting upon oneself, by 'lowering the flame of life to a pin-point' — and the noetic correlate of this attitude is what we call Bleakness: the universe is bleak; that is, of undifferentiated structure.25 Sartre's view of emotions, in general, is that they are characterized by their intentionality. 'Emotion is a specific manner of apprehending the world,'26 he writes, and 'all the emotions have this in common, that they evoke the appearance of a world, cruel, terrible, bleak, joyful etc.'27 Sartre's view provides one general framework in which to defend the intentionality of all emotions - even those which Searle describes as 'undirected'. Let me summarize this line of thought. Searle says that there are 25 Sketch for a Theory of the Emotions (London: Methuen 1971; originally published 1939), pp. 68—9. For an illuminating introduction, see Gregory McCulloch, Using Sartre (London: Routledge, 1994), chapter 2. 26 Sartre, Sketch, p. 57. "Ibid., p. 81. 242
Intentionality as the Mark of the Mental emotions which have no intentionality. But this does not follow from the fact that people cannot say what it is that their emotions are about. Nor does its denial follow from the fact that the objects of emotions are sometimes their causes, of which we are sometimes ignorant. To decide the issue about whether there are non-intentional emotions, we should first ask what distinguishes, from the phenomenological point of view, the different emotions. The nonintentionalist answer to this question is committed to the existence of emotion-qualia, and the implausible possibility of inverted emotion-qualia. But the intentionalist who accepts (for example) Sartre's view of emotion as a mode of apprehending the world is not committed to this possibility. The differences between the different emotions would not be explained in terms of qualia but in terms of the different ways the emotions present the world and the subject's place in it. This is one way an intentionalist can characterize the emotions Searle is talking about, like anxiety and depression, where the subject is not able to say what they are anxious or depressed about. The phenomenology of emotion is a very complex area, and I have only touched the surface of the issues. What I have tried to do is to suggest a way in which an intentionalist can argue that these apparent counterexamples to Brentano's thesis are not really counterexamples.28 But what does this treatment of the counterexamples show about the nature of intentionality in general? My original question was: what would you have to believe about intentionality in order to believe that it is the mark of the mental? The way I have approached this question is to try and specify the sense in which something is 'given' to the mind in sensation and emotion, just as something is given to the mind in thought and experience. The heart of the view is inspired by Brentano's remark that in the idea, something is conceived; I say that in the sensation, something is felt, in the emotion, something is apprehended - and so on. The issue is in danger of collapsing into an uninteresting question of terminology if the notion I am identifying as intentionality had nothing in common with what others call intentionality. But this is not the case. It is possible to isolate two main elements of the concept of intentionality as discussed by recent philosophers.29 The first is the apparently relational structure of intentionality, the structure Sartre and other phenomenologists express by saying that 28 Of course, this is not the only way for an intentionalist to account for emotion. Compare Tye's views: Tye, Ten Problems, chapter 4. 29 For a representative of recent discussions, see Tye, Ten Problems, pp. 94-6. 243
Tim Crane consciousness is always the consciousness of something.30 While intentional states appear to be relations between thinkers and the objects of their thoughts, this cannot be true in general, since intentional states can be directed on things which do not exist, and relations entail the existence of their relata.31 (This point holds independently of the truth or falsity of the doctrine of externalism, since even the most extreme externalist must allow that intentional states can concern the non-existent.) The second element is what some call the perspectival or fine-grained nature of intentionality, what Searle calls 'aspectual shape'.32 This is just the familiar idea 30 For instance: 'all consciousness, as Husserl has shown, is consciousness of something' Being and Nothingness (London: Methuen, 1958; first published 1943, p. xxvii). Compare Searle: 'It is characteristic of Intentional states, as I use the notion, that there is a distinction between the state and what the state is directed at or about or of.' (Searle, Intentionality, p. 2); for a different way of formulating the same kind of point, see E. Levinas, 'Beyond Intentionality', in Philosophy in France Today, ed. A. Montefiore (Cambridge: Cambridge University Press, 1983), p. 106. 31 This fact gives rise to one of the main problems of intentionality. For an excellent presentation of this problem, see Michael Dummett, Origins of Analytical Philosophy (London: Duckworth, 1993), pp. 35—6. See also Caston, 'Aristotle'. Brentano came to appreciate the importance of this point when he wrote the appendix to his Psychology from an Empirical Standpoint. There he says that 'If someone thinks of something, the one who is thinking must certainly exist, but the object of his thinking need not exist at all.' He goes on to observe that 'we might doubt whether we are dealing with something relational here, and not, rather, with something somewhat similar to something relational in a certain respect, which might therefore better be called "quasi-relational"' (p. 272). Sometimes it is supposed (see Dennett, Content, and Tye, Ten Problems) that Brentano was concerned with the question of non-existence even before he wrote the Appendix to his Psychology. It is true that in a famous passage, Brentano says that the object of thought 'should not be understood as a reality' (p. 88); but by this he is just reminding his readers that he is talking about 'phenomena' or 'appearances', not about the 'underlying reality'. In this sense, the physical phenomena with which he contrasts mental phenomena 'should not be considered a reality' either. Compare, for example, the following passage: 'the phenomena of light, sound, heat, spatial location and locomotion which [the natural scientist] studies are not things which really and truly exist. They are signs of something real, which through its causal activity, produces presentations of them. They are not however, an adequate representation of this reality... We have no experience of that which truly exists, in and of itself, and that which we do experience is not true.' (Brentano, Psychology, p. 19). 32 John R. Searle, The Rediscovery of the Mind (Cambridge, MA: MIT Press, 1992), p. 155. 244
Intentionality as the Mark of the Mental that when something is apprehended as the object of an intentional state - whether a particular object, fact or property - it is always apprehended in a certain way. Both features of intentionality are present in my treatment of the counterexamples to intentionalism. I claimed that instead of seeing bodily sensations as instantiations of purely subjective, monadic properties, we should see these experiences as presenting something — a part of the body — as modified in a certain way. Bodily sensations, then, are primarily states of awareness, and therefore apparently relational states. They are only apparently relational since, according to the perceptual theory, phantom limb phenomena (e.g.) are cases of awareness of a felt quality in merely apparent body part. They are therefore analogous - in this respect only - to cases of perceptual hallucination, where one perceives a quality to be instantiated in an object which does not actually exist.33 If Sartre's account of the phenomenology of emotions is right, then there is a similar apparent relationality in emotional experience: there is the experiencing subject, the world experienced (or the thing in the world experienced) and the particular way of apprehending the world. The second element of intentionality — its fine-grained character - is also contained within my account of sensations and emotions. A pain in one's ankle is a state of awareness of one's ankle, presented as such, not as the organic organization of tendons, bone and muscle which one's ankle actually is. Similarly with the so-called 'undirected' emotions. In a particular undirected emotion, the same world appears under one aspect — bleak, terrible, threatening — rather than another. (Of course, there may be debate about whether the world could properly be said to have the aspects or properties attributed to it in an emotion - but this does not affect the present point.) So the core of the concept of intentionality, as discussed in much contemporary philosophy of mind, is present in the theses advanced by intentionalism. The dispute between the intentionalist and the non-intentionalist is substantial and not just terminological. Where this characterization does depart from some recent discussions is in not starting the discussion of intentionality with the notion of a propositional attitude. A propositional attitude is an intentional 33 Only in this respect, since it is not quite correct to say that a phantom limb pain is an illusory pain — the pain certainly exists, one just perceives it as having a location which it does not (indeed, in the circumstances, cannot) have. An analogy would be perhaps with some device which made it seem to you as if sounds were coming from one direction when they were in fact coming from the opposite direction (as when a ventriloquist 'throws' his voice). 245
Tim Crane state whose content - that which characterizes its directedness - is something evaluable as true or false. I do not question the applicability of the notion of a propositional attitude itself, but rather the tendency in some contemporary philosophers to see the propositional attitudes as the sole home of the concept of intentionality.34 Obviously, the form of intentionalism I am defending here cannot accept such a view, but even putting this to one side, the thesis that all intentional mental states are propositional attitudes lacks phenomenological plausibility. To take a nice example of Victor Caston's: when asked to think of a number between one and ten, what comes to mind is a number, not a proposition. And it is a familiar fact that certain emotions, notably love and hate, can be directed on objects rather than always on states of affairs. While the notion of a propositional attitude must play an important role in any theory of intentionality, it does not exhaust the application of the concept of intentionality. IV Intentionality, the Non-Mental and the Mark of the Mental I have been defending the claim that all mental phenomena exhibit intentionality. Now I want to return to the other part of Brentano's thesis, the claim that intentionality is exclusive to the mental domain. This will give me the opportunity to air some speculations about why we should be interested in the idea of a mark of the mental. Now, in the way I am suggesting we should think about intentionality, it is a concept which applies to all mental phenomena, including conscious, phenomenally salient mental states such as perception, sensation and conscious emotional episodes, but also unconscious beliefs, desires and other mental dispositions. The binding idea is captured by the Brentanian slogan that in the intentional state something is given. But can we find intentionality in the non-mental? It is sometimes said that Brentano's thesis is a threat to physical34 For a clear-headed (but in my view mistaken) statement of this policy, see Dennett, Content pp. 27—9. Even Searle, Intentionality, who admits that much intentionality cannot be expressed in terms of whole propositions (pp. 6-7), seems to commit himself implicitly to the opposite in his analysis of intentional states by analogy with his account of speech acts (p. 26). The tendency is still pervasive: see, for instance, the definition of 'intentionality' given in William Lyons, Approaches to Intentionality (Oxford: Oxford University Press, 1995), pp. 1-2. 246
Intentionality as the Mark of the Mental ism because it implies that intentionality can only be found in the mental and never in the physical. Dennett, for instance, says that 'the Intentionalist thesis... proclaims an unbridgeable gulf between the mental and the physical.'35 But we must distinguish between the view that intentionality is not present in the physical, and the view that intentionality is not present in the non-mental. For if physicalism is true, then the physical is not the same as the non-mental. Of course, Brentano himself — to whom the question of physicalism would have been of little interest - says that 'no physical phenomenon manifests anything like' intentionality. But if we want to remain neutral on the question of physicalism, we should prefer a weaker version of Brentano's thesis which only says that intentionality is characteristic of the mental alone. Whether the mental is reducible to the physical is a further question; if it is, then some physical things manifest intentionality. But no non-mental things do. However, some philosophers take a view of intentionality which makes it unproblematically a feature of many non-mental things. For instance, some follow Chisholm and Quine and take the nonextensionality of certain linguistic contexts as criterial for the intentionality of the phenomena described in those contexts.36 Chisholm's approach was to 'formulate a working criterion by means of which we can distinguish sentences that are Intentional... in a certain language from sentences that are not'.37 In essence Chisholm's criterion was that a sentence S is intentional iff: S contains a singular term yet does not entail the usual existential generalization; or S contains an embedded sentence in a non-truth-func35 Dennett, Content, p. 21. The point derives from Quine: see Word and Object (Cambridge, MA: MIT Press, 1960), p. 221. See also the opening pages of Hartry Field, 'Mental Representation', in Readings in the Philosophy of Psychology vol. II, ed. Ned Block (London: Methuen, 1980). Obviously, if one thinks of intentionality as a property of sentences (as Quine and Dennett do), Dennett's quoted remark makes more sense than it would do otherwise. I quote it here because the idea that Brentano's thesis presents a problem for physicalism has survived the waning of the popularity of the linguistic criterion of intentionality. 36 See Roderick Chisholm, Perceiving (Ithaca: Cornell University Press, 1957), chapter 12, and Quine, Word and Object, esp. the section on 'The Double Standard'. Note especially the following passage: 'the Scholastic word "intentional"' was revived by Brentano in connection with the verbs of propositional attitude and related verbs... e.g. 'hunt', 'want' etc. The division between such idioms and the normally tractable ones is notable. We saw how it divides referential from non-referential occurrences of terms.' 37 Chisholm, Perceiving, p. 170. 247
Tim Crane tional context; or the principle of the substitutivity of co-referring singular terms does not apply to S. This disjunctive criterion is supposed to establish the intentionality (in Brentano's sense) of the phenomena described by the sentence S. Dennett, for instance, says that 'Chisholm's three criteria come close to reproducing Brentano's distinction.'38 I will call this the 'linguistic criterion' of intentionality. Some of those who adopt the linguistic criterion take a deflationary approach to the distinctively mental characteristics of intentionality. They point out that intentionality, in their sense, is common to non-mental linguistic contexts, too — for instance: modal, causal, dispositional, probabilistic or functional contexts - and they draw various conclusions from this fact. They might draw the relatively weak conclusion that intentionality is not the mark of the mental; or they might draw the stronger conclusion that there is no special problem of intentionality, if intentionality is shared by so many different and (in some cases) unproblematic phenomena.39 The version of intentionalism defended here cannot accept this. This is not to say that it would have to reject the view that causal, probabilistic and the other contexts are non-extensional. Nor does intentionalism have to deny that the features of intentionality I have just mentioned receive expression in the linguistic structures which we use to describe it. So, for instance, the apparent relationality is evident in ascriptions of intentionality (in the failure of existential generalization in non-extensional contexts) as is the fact that intentionality is perspectival (in the failure of substitutivity of co-referring terms). What intentionalism must reject is rather the linguistic criterion of intentionality itself. These linguistic phenomena are guides to the presence of intentionality in ascriptions of intentionality, but they do not constitute its essence. And given the way I have been proceeding in this paper, this should not be surprising. Intentionality, like consciousness, is one of the concepts which we use in an elucidation of what it is to have a mind. On this conception of intentionality, to consider the question of whether intentionality is present in some creature is of a piece with considering 38 Dennett, Content, p. 23. Compare Searle, Intentionality, pp. 22-5, who takes the correct view of this matter, as I see it. See also William Kneale, 'Intentionality and Intensionality', Proceedings of the Aristotelian Society, Supplementary Volume 42 (1968). 39 A good example of this general approach is Enc, 'Intentional States'; see also C. B. Martin and Karl Pfeifer, 'Intentionality and the NonPsychological' Philosophy and Phenomenological Research 46 (1986), and U. T. Place, 'Intentionality as the Mark of the Dispositional; Dialectica 50 (1996). 248
lntentionality as the Mark of the Mental what it is like for that creature - that is, with a consideration of that creature's mental life as a whole. To say this is not to reject by stipulation the idea that there are primitive forms of intentionality which are only remotely connected with conscious mental life - say, the intentionality of the information-processing which goes on in our brains. It is rather to emphasize the priority of intentionality as a phenomenological notion.40 So intentionalists will reject the linguistic criterion of intentionality precisely because the criterion will count phenomena as intentional which are clearly not mental. This would be a perverse or circular way to proceed if we did not already have a grasp on the concept of a mind. But we do have such a grasp: it is that concept which we try to express when we say that to have a mind is to have a point of view or perspective on the world, or when we say that there is something it is like to be conscious, or when we talk about the world being manifest to a subject of experience, or when we talk about the world being a phenomenon for a subject. Some philosophers associate these ways of talking solely with the conscious or phenomenal side of the mind, where the conscious or the phenomenal is contrasted explicitly with the intentional.41 Consider, for instance, how McGinn formulates his pessimism about our inability to explain consciousness: We can, it is felt, explain what makes a mental state have the content it has; at least there is no huge barrier of principle in the way of our doing so. But, it is commonly conceded, we have no remotely plausible account of what makes a mental state have the phenomenological character it has.42 40 For the idea of intentionality as a phenomenological notion, see Gregory McCulloch, 'The Very Idea of the Phenomenological', Proceedings of the Aristotelian Society 93 (1992-93); and 'Intentionality and Interpretation', this volume; J. E. Malpas, in Donald Davidson and the Mirror of Meaning (Cambridge: Cambridge University Press, 1992), section 4.2, gives an interesting reading of intentionality as a phenomenological notion, drawing on the Heideggerian notion of a 'horizon'. For a survey of various ways in which the idea of intentionality can be applied beyond the central cases, see Martin Davies, 'Consciousness and the Varieties of Aboutness', in Philosophy of Psychology: Debates on Psychological Explanation, ed. Cynthia and Graham Macdonald (Oxford: Blackwell 1995). 41 For the contrast between the phenomenal and the intentional, see, for example, Sydney Shoemaker, The First-Person Perspective and Other Essays (Cambridge: Cambridge University Press, 1996), pp. 112, 138. 42 Colin McGinn, 'Consciousness and Content,' in The Problem of Consciousness and Other Essays (Oxford: Blackwell, 1991), p. 24. See also his later remark that 'subjective features lie quite outside the proper domain of the theory of content' (p. 33). 249
Tim Crane Here 'phenomenological character' is explicitly contrasted with 'content', as if the two categories were exclusive. Sometimes composite states are envisaged - as when perceptions are conceived of as having content and qualia. But in general, the picture of the mind which lies behind remarks such as McGinn's is one on which we have two kinds of mental state: intentional states which are not essentially conscious, and conscious states whose consciousness is intrinsically unrelated to any intentionality they may have. The trouble with this picture of the mind is that the classification of both kinds of phenomena as mental seems to lack a rationale. The most we can say is that mental is an accidental category, which presupposes no underlying nature to the phenomena it picks out. As Kathleen Wilkes puts it, 'it is improbable that something bunching together pains, and thoughts about mathematics, is going to be a reliable pointer towards a legitimate natural kind'.43 Wilkes here echoes Richard Rorty's complaint about the heterogeneity of the concept of mind: The attempt to hitch pains and beliefs together seems ad hoc - they don't seem to have anything in common except our refusal to call them 'physical'.44 There are two possible ways of reacting to these points. One is simply to accept that there is no more than a nominal unity to the concept of mind. The other is to object that there is something wrong with the whole picture - specifically, with the way of distinguishing between intentionality and consciousness that we find expressed, for instance, in the above passage from McGinn. If this is our reaction, then we need to find a way of characterizing mental phenomena which reflects the underlying unity of their classification as mental: that is, we need a mark of the mental. Some philosophers have argued recently that consciousness is the only true mark of the mental.45 But this view battles with the widely accepted and uncontroversial view that many mental states are unconscious, so its defence is an uphill struggle. The alternative, which I have been canvassing here, is that it is intentionality, the mind's directedness on the world, which should be thought of as the mark of the mental. If we take this view, then we must reject the 43 Quoted by Davies, 'Consciousness', p. 358. 44 Richard Rorty, Philosophy and the Mirror of Nature (Oxford: Blackwell, 1979), p. 22. 45 See Searle, Rediscovery; for his view of the unconscious, see pp. 155—56. A similar view is taken by Galen Strawson in Mental Reality (Cambridge, MA: MIT Press, 1995): 'the only distinctively mental phenomena are the phenomena of conscious experience' (p. xi). 250
Intentionality as the Mark of the Mental distinction implicit in McGinn's, Rorty's and Wilkes's remarks, that the phenomenal is one thing, the intentional another. Whatever the fate of qualia, we must accept that all mental states are permeated with intentionality, and characterizing their phenomenal character - giving a phenomenology - can be achieved by characterizing their intentionality.46 Brentano's view was that the science of psychology should be distinguished from both physiology and philosophy, not by its methods, but by its subject-matter. These days, it is less common for there to be serious dispute among psychologists about the subjectmatter of psychology. But there is perhaps more disagreement in today's philosophy of mind about what its subject-matter is, and in some cases there is even disagreement about whether it has one. Those who find this situation unacceptable may wish to reconsider the popular rejection of Brentano's thesis that intentionality is the mark of the mental, and therefore the subject-matter of the philosophy of mind. 46 For some different approaches to the same idea, compare M. G. F. Martin, 'Setting Things Before the Mind', this volume; and Gregory McCulloch, 'Intentionality and Interpretation', this volume. My remarks in this last section are highly speculative, and raise many issues which demand further elaboration. One question is whether the suggested 'unification' of the phenomena of mind by the concept of intentionality can be achieved within the weak intentionalist picture I defend here. For if one allows that the existence of non-intentional phenomenal properties (qualia) is compatible with the intentionality of all mental states, then it appears as if a question can be raised for weak intentionalism which recapitulates the question I am raising for the McGinn/Rorty picture. More needs to be said about non-intentional properties in order to assess the force of this question. Here I am indebted to participants in the discussion at the Royal Institute of Philosophy meeting, and especially to Paul Boghossian. 251
lntentionality and Interpretation GREGORY McCULLOCH According to Brentano in a much-quoted passage, Every psychological phenomenon is characterized by... intentional inherent existence of ... an object... In the idea something is conceived, in the judgement something is recognized or discovered, in loving loved, in hating hated, in desiring desired, and so This is a doctrine about the nature of thought or cognition, and some would say that the matters it raises have only fairly recently come to be directly engaged by mainstream philosophers of the analytical tradition. For analytical philosophy's approach to the question of intentionality has tended to be routed through concern with linguistic behaviour, its interpretation and its logical and semantic analysis. The depth of this concern with language is, if anything, emphasized when we consider more recent attempts by heirs of the analytical tradition to approach the question of intentionality more directly, missing out the trip through linguistic behaviour. According to one very influential version of this tendency, the problem of intentionality is still to do with the properties of linguistic structures, only now not public, spoken linguistic structures, but systems of mental representations in the brain.2 The tendency to miss out the trip through public language reflects, in part, a reaction against behaviourism and in favour of mentalism: just as one of the principal causes for the longevity of the result of the 'linguistic turn' was undoubtedly the influence of behaviourism on analytical philosophy of mind. I think the mentalists are right to be antibehaviouristic, but that their forerunners were also right to focus on the interpretation of linguistic behaviour when approaching cogniThanks to Tim Crane and Chris Hookway for valuable comments on an earlier draft. 1 F. Brentano, Psychology from an Empirical Standpoint, ed. L. McAlister, trans. A. Rancurello, D. B. Terrell, and L. McAlister (New York: Humanities Press, 1973), p. 88. 2 For example, see H. Field, 'Mental Representation', Erkenntnis 13 (1978), 9-61 J. Fodor, Psychosemantics (Cambridge, MA: MIT Press, 1987). 253
Gregory McCulloch tion in general and intentionality in particular. By the same token, then, there has usually been something wrong with analytical approaches in this area: too much behaviourism, or too little concern with linguistic behaviour. In this paper I want to recommend a more balanced approach, building on themes from Quine and Davidson. II A behaviouristic starting point is explicit in the Quine of Word and Object. one is ready to say of the domestic situation in all positivistic reasonableness that if two speakers match in all dispositions to verbal behavior there is no sense in imagining semantic differences between them.3 Consequently, he focuses on radical translation, on how a field linguist might penetrate an unknown tongue with no established links with her own. The reason is that 'all the objective data [the linguist] has to go on are the forces impinging on the natives' surfaces, and the observable behaviour, verbal and otherwise, of the native'.4 The behaviourism is notably radical: the idea is to make independent sense, in terms of these 'objective data', of mentalistic notions such as meaning, belief, intentional object. Quine's thesis of the indeterminacy of translation then amounts to the view that no such independent sense can be made: the relativity to non-unique systems of analytical hypotheses invests not only translational synonymy but intentional notions generally5 For example, the 'objective data' will support nothing better than stimulus meaning, and terms can be stimulus-synonymous without even being co-extensive, much less intuitively synonymous: thus, according to Quine, 'rabbit' and 'undetached rabbit-part'. He later dubbed this phenomenon the inscrutability of reference (later still, the indeterminacy of reference),6 and given his behaviourism it 3 W. V. Quine, Word and Object (Cambridge, MA: MIT Press, 1960), p. 79. 4 Ibid., p. 28. 5 Ibid., p. 221. 6 For 'inscrutability' see W. V. Quine, Ontological Relativity and Other Essays (New York: Columbia University Press, 1969), p. 37; and for 'indeterminacy' W. V. Quine, Pursuit of Truth (Cambridge, MA: Harvard University Press, revised edition 1992), p. 50. 254
Intentionality and Interpretation entails the impotence of intentionality. If conceiving is, at bottom, a matter of linguistic and other behaviour, then problems with reference just are problems with conceiving, and Brentano goes beyond the Quinean 'objective data' when saying 'in the idea something is conceived'. For Quine, the 'something' is neither rabbit, nor undetached rabbit-part, nor any other determined thing. One can no more direct attention at rabbits than use words to refer to them, and in his own words: the arbitrariness of reading our objectifications into the heathen speech reflects not so much the inscrutability of the heathen mind, as that there is nothing to scrute.7 Nor are things any better, from Brentano's point of view, if we draw the Quinean moral that inscrutability of reference delivers ontological relativity: the idea that ontological questions only make sense relative to this or that scheme. There is no mention of relativity in Brentano's thesis as stated, and we shall see there is no evident sense in applying the relativity point to the thesis correctly construed. Davidson, we know, self-consciously departs from the Quine of Word and Object over a number of central matters.8 In particular, he is explicitly anti-behaviourist, claiming that behaviour is no more than 'the main evidential basis of attributions of belief and desire'.9 He also denies that mentalistic notions, to be made respectable, need to be reduced to non-mentalistic ones. In this connection he drops Quine's notions of assent and dissent, and introduces the more mentalistic one of holding true. He finds no use for stimulus meanings, instead matching utterances with ordinary worldly conditions. He rejects theorizing in terms of translation, using a more demanding notion of structure-revealing interpretation: although he retains Quine's concern with the radical case, and I shall be much concerned with this important point. He doubts whether the scope for indeterminacy in semantics is as wide as Quine suggests. But despite all this, Davidson's approach to intentionality is very close indeed to Quine's. First, he argues for the inscrutability of reference: suppose every object has one and only one shadow... On a first theory, we take 'Wilt' to refer to Wilt and the predicate 'is tall' to refer to tall things; on the second theory, we take 'Wilt' to refer to 7 Quine, Ontological Relativity, p. 5. 8 Indeed, later Quine seems to move significantly towards Davidson: see, e.g. Quine, Pursuit of Truth, §29. 9 See D. Davidson, Inquiries into Truth and Interpretation (Oxford: Oxford University Press, 1984), p. 160. 255
Gregory McCulloch the shadow of Wilt and 'is tall' to refer to the shadows of tall things. The first theory tells us that the sentence 'Wilt is tall' is true if and only if Wilt is tall; the second theory tells us that 'Wilt is tall' is true if and only if the shadow of Wilt is the shadow of a tall thing. The truth conditions are clearly equivalent... What matters is that what causes the response or attitude of the speaker is an objective situation or event, and that the response or attitude is directed to a sentence or the utterance of a sentence. As long as we hold to this, there can be no relevant evidence on the basis of which to choose between theories and their permutations.10 Given Davidson's anti-behaviourism, commitment to the inscrutability of reference does not directly yield the impotence of intentionality: but armed with other Davidsonian themes, we soon get there. For he claims that there is no general grip to be had on the content of speakers' beliefs except through an account of what their words mean. On the Davidsonian approach, held-true sentences are matched with worldly conditions, and if the theory that supplies the matching satisfies certain constraints, these conditions are deemed the truth-conditions of the utterances. The utterances may then normally be taken as expressions of belief with those same truth-conditions, since it is assumed that speakers usually hold-true a sentence on the basis of a true belief (the principle of charity). Any potential for permutation in the apparatus that supplies the truth-conditions of utterances thus carries over to belief, at least in the case of language users and beliefs expressible in language. But then Davidson denies beliefs to non-linguistic creatures." And although his argument seems independent of the main body of his doctrines, overall we confront a position on which inscrutability of reference comports with impotence of intentionality. The problem over whether I am talking about Wilt yields a parallel problem over whether I am thinking about him, over whether I can direct my attention at him. In the idea Wilt is no more (and no less) conceived than Wilt's shadow. Davidson regards this as an acceptable result: Indeterminacy of meaning or translation does not represent a failure to capture significant distinctions; it marks the fact that certain apparent distinctions are not significant.12 10 Ibid., pp. 230-1. 11 Ibid., Essay 11; also D. Davidson, 'Rational Animals', in Actions and Events, ed. E. Lepore and B. McLaughlin (Oxford: Blackwell, 1985). 12 Davidson, Inquiries, p. 154. 256
Intentionality and Interpretation III One might suspect, in the light of the striking closeness to Quine, that Davidson's anti-behaviourism is only skin-deep. One could then go on to defend Brentano and the doctrine of intentionality by attacking the behaviourism, the message being that cognition can only be accommodated by mentalism. This could mean dropping the concern with linguistic behaviour, and focusing instead on a notion of mental representation presumed to underlie it, resulting in a version of behaviour-rejecting mentalism: a view which denies bodily behaviour any essential role in thought or cognition. According to a behaviour-rejecting mentalist, an appropriately stimulated and sustained ab initio vat-brain could have the same cognitive properties as I do as I sit here now. Thus Fodor: If you imagine a brain in a vat that's hooked up to this world, and hooked up in just the same way one's own brain is, then - of course — that brain shares one's thought-contents...13 As we shall see, there are signs of (unstable) behaviour-rejecting mentalism in Davidson. Nevertheless, moving from Davidson-^walover-of-linguistic-behaviour to behaviour-rejecting mentalism involves taking a lot of steps: more than I think justifiable, for reasons some of which will emerge. So I want to recommend a nonbehaviouristic defence of Brentano, against the impotence claim, which is much more accommodating to Davidson's over-arching concern with linguistic behaviour and interpretation, as well as to other broadly Davidsonian themes. This involves recommending behaviour-embracing mentalism, according to which embodiment and bodily behaviour are essential aspects - though not sufficient conditions - of thought and cognition.14 To make the argument, we need to begin with two doctrines common to Quine and Davidson. One is obviously not behaviouristic, the other not obviously so. These doctrines are (1) the primacy of sentences, and (2) the primacy of the radical case. 1. The primacy of sentences. On Quine's account, stimulus meaning approximates most nearly to meaning as intuitively conceived in the 13 Fodor, Psychosemantics, p. 52. Note that there is some absurdity in suggesting that disembodied brain could be hooked up to the world in just the way that an embodied brain is. This is not as trivial a point as it may look, given the possibility of behaviour-embracing mentalism (for which see immediately below). 14 See G. McCulloch, The Mind and Its World (London: Routledge, 1995), chapters 5-8. 257
Gregory McCulloch case of observation sentences, those 'occasion sentences whose stimulus meanings vary none under the influence of collateral information. .. These are the occasion sentences that wear their meanings on their sleeves.'15 Other sentences do less well, but indeterminacy attributable to the inscrutability of reference does not appear until sentences are broken into terms. Davidson also homes in on sentences: the evidence available [for a theory of interpretation] is just that speakers of the language to be interpreted hold various sentences to be true at certain times and under specified circumstances.16 As well as reflecting these philosophers' concern to give empirical substance to their accounts of language, these claims also highlight their view that sentences are in some sense primary. At its most innocent this derives from the centrality of inference and truth to any viable classical account of logical structure, and reflects also the observation, in Dummett's words, that 'we cannot ... do what Wittgenstein called "make a move in the language game" without, in effect, using a sentence'.17 Given this, sub-sentential semantic notions like reference and satisfaction come out as derivative, their utility exhausted by the role they play in helping systematize what Quine calls 'the interanimation of sentences'. As he puts it in a fairly recent paper on Davidson: On the one hand there is the set of theoretical sentences ... On the other hand, there is the observation sentence ... subject to a verdict by dint of sensory stimulation. Where complexity comes is in the relation of the set of theoretical sentences to the observation sentence. They are connected by a network of intervening sentences, variously linked in logical and psychological ways. It is only here that we have to pry into sentences and take notice of ... objective reference, as Davidson well argued...18 Once the focus shifts away from language as logician's abstraction to language as revealed in linguistic behaviour, the primacy of sentences quite smoothly transposes into the primacy of speech acts (more accurately, of assertion: or - given that it is thought we are 15 Quine, Word and Object, p. 42. Occasion sentences are those which 'command assent or dissent only if queried after an appropriate prompting stimulation' (pp. 35-6). 16 Davidson, Inquiries, p. 135. 17 M. Dummett, Frege: Philosophy of Language (London: Duckworth, 1973), p. 3. 18 W. V. Quine, 'Events and Reification', in Actions and Events, ed. Lepore and McLaughlin, p. 169; cf. Quine, Pursuit of Truth, chapter 2. 258
Intentionality and Interpretation ultimately supposed to be dealing with - judgement). And this much, so far, is not to be quibbled with. But problems threaten when it is combined with the doctrine of the primacy of the radical case. 2. The primacy of the radical case. As remarked, Quine's focus on radical translation is a device to make graphic behaviouristic assumptions. As also remarked, Davidson drops translation and moves to interpretation, but retains the focus on the radical case. Given that he cannot offer Quine's behaviouristic motivation, what does he offer instead? He writes: I propose to call a theory a theory of meaning for a natural language L if it is such that (a) knowledge of the theory suffices for understanding the utterances of speakers of L and (b) the theory can be given empirical application by appeal to evidence described without using linguistic concepts, or at least without using concepts specific to the sentences and words of L. The first condition indicates the nature of the question; the second requires that it not be begged.19 Given his aim to 'understand semantic concepts in the light of others',20 we need to exclude certain linguistic notions from the characterisation of the evidence for this or that interpretation. Clearly, focusing on the radical case is a way of doing this, and so it would not be easy to pin underlying or vestigial behaviourism on Davidson here. Any quasi-reductionist ambition, howsoever mild, might reasonably make use of the radical case. Whether even mild quasireductionism can be warranted in the philosophy of thought is another matter to which we shall return. First, though, we need to have before us a key Davidsonian argument that inscrutability of reference is inevitable once a focus on the radical case is added to the doctrine of the primacy of sentences. It turns on the fact that while the primacy of sentences merely entails that sub-sentential semantic notions are derivative, adding the primacy of the radical case converts this into the stronger idea that they are theoretical, 'non-observational' notions, whose role in saving the phenomena of sentence-production exhausts their empirical reality. Thus Davidson: I suggest that words, meanings of words, reference and satisfaction are posits we need to implement a theory of truth. They serve this purpose without needing independent confirmation or empirical basis.21 19 Davidson, Inquiries, p. 215. 20 Ibid., p. 219. 21 Ibid., p. 222. 259
Gregory McCulloch Given the focus on the radical case, all the evidence there is relates to the production of sentences in observable circumstances; and sub-sentential semantics is then, according to Davidson, in the same boat as theories about microscopic physical structure: 'we explain macroscopic phenomena by postulating an unobserved fine structure'.22 So: the idea is that primacy of sentences plus the radical case makes sub-sentential semantics theoretical, leaving room for permutations at the sub-sentential level which equally well save the observable phenomena. That's inscrutability of reference, and impotence of intentionality seems to follow on given the general plan of tying cognition essentially to linguistic behaviour in Davidson's manner. But think of the charge often levelled against Quine that he without good reason converts underdetermination into indeterminacy where translation is concerned. Now think of Davidson's analogy between physical and sub-sentential structure. The initial claim is that empirical or evidential reality is exhausted by a range of phenomena (macrophysical, linguistic) which can then be held fixed while the 'theoretical' story is permuted. But in the microphysical case, this certainly only reflects underdetermination, and there is no demonstration here that the whole reality of the microphysical is exhausted by its empirical reality. So why not take the same line in the linguistic case, and say that there is more to the reality of subsentential semantics than is given by what Davidson offers as its empirical reality? Such a move is exactly what one expects a behaviour-rejecting mentalist to make: beneath the overt production of sentences lies a cognitive mechanism that determines what is left underdetermined by correlation of sentences with observable states of affairs.23 Now in Quine's case, the reply to this is straightforward: to make the move is to abandon his axiomatic behaviourism, according to which the reality of 'theoretical' posits precisely is exhausted by their empirical reality: Language is a social art which we all acquire on the evidence solely of other people's overt behavior under publicly recognizable circumstances. Meanings, therefore, those very models of mental entities, end up as grist for the behaviorist's mill. 22 Ibid. 23 This appears to be the suggestion in J. Searle, 'Indeterminacy, Empiricism and the First Person', Journal of Philosophy 84 (1987), cf. Quine, Ontological Relativity, pp. 28-9. There are similarities between Searle's diagnosis of what drives Quine and Davidson towards indeterminacy, and the one offered in the present lecture. But for a very big difference in the countersuggestions offered, see section V below. Thanks here to Barry C. Smith. 260
Intentionality and Interpretation In psychology one may or may not be a behaviorist, but in linguistics one has no choice ... There is nothing in linguistic meaning beyond what is to be gleaned from overt behavior in observable circumstances.24 But the straight Quinean answer is not available to the anti-behaviourist in Davidson: so we need from him a different reason why empirical reality should be taken to exhaust the whole of reality where sub-sentential semantics is concerned. Having noted this absolutely crucial point, however, I shall delay further discussion of it for a short while. But we shall see that there is a major fault line in Davidson's position on account of this matter. IV I first want to counter a strong impression sometimes given by Davidson that his inscrutability and impotence result is entailed by the primacy of sentences doctrine alone. As Dummett has argued: Since it is only by means of a sentence that we may perform a linguistic act ... the possession of sense by a word cannot consist in anything else but its being governed by a rule which partially specifies the sense of sentences containing it. If this is so, then, on pain of circularity, the general notion of the sense possessed by a sentence must be capable of being explained without reference to the notion of the senses of constituent words...25 Sentence-meaning is primary when it comes to explaining what it is for words to have the sense they do, even though word-meaning is primary in another way: 'we derive our knowledge of the sense of any given sentence from our previous knowledge of the senses of the words that compose it...'26 There is no commitment here to the idea that sub-sentential semantics concerns the theoretical or nonobservational, much less the indeterminate. To see this, we need only consider a kind of approach which combines a Fregean truthconditional account of sentence-meaning with an ideational conception of word-understanding. On this approach, one would understand 'the cat is on the mat' in virtue of the ideas one associated with 'cat', 'mat' etc., even though the semantic complexity of the signifying words could only be fully explained in terms of the sentence's truth conditions. If the ideas here are construed in the 24 Quine, Ontological Relativity, p. 26 and Pursuit of Truth, pp. 37-8, respectively. 25 Dummett, Frege, pp. 4—5. 26 Ibid., p. 5; emphasis added. 261
Gregory McCulloch traditional style as objects of introspection or inner perception then there is no danger that they would be merely theoretical or nonobservational, at least from the first-person point of view: while any supposed problems from the third-person point of view would naturally be regarded, in the first instance anyway, as epistemological, and hence as indicating underdetermination rather than indeterminacy. Now such an approach is, of course, hopeless for a whole variety of reasons. But it is not true that the key problem is a questionbegging failure to theorize in terms of the radical case. So here we have the primacy of sentences without inscrutability of reference and impotence of intentionality. In the cat-idea cats are conceived. To take this admittedly hopeless line is not to revert to what Davidson criticizes as the Building Block theory, which tries to explain directly the referential properties of sub-sentential expressions and then to characterize truth and other sentential semantic notions on this basis. Davidson gives a lightning historical sketch of that approach, and concludes that: as the problems become clearer and the methods more sophisticated, behaviourists and others who would give a radical analysis of language and communication have given up the building block approach in favour of an approach that makes the sentence the focus of empirical interpretation. And surely this is what we should expect. Words have no function save as they play a role in sentences...27 But the problem with this passage is that it deals simultaneously with two issues: (1) the question whether sentential semantics is prior to sub-sentential; and (2) the proposal to give a 'radical analysis of language and communication'. Perhaps sentences are prior but there is something wrong with the idea of radical analysis. Dropping it would not then equate with reverting to the Building Block theory. In other words: one might join Davidson in rejecting the ambition of defining truth in terms of independently understood sub-sentential reference etc., without thinking that linguistic concepts can be understood in the light of non-linguistic ones, which is what 'radical analysis' means here. It may be that, historically, the two projects - defining truth, 'reducing' semantics - have gone together: but there is clear blue logical water between them. So I want to leave unquestioned the primacy of sentences doctrine and suggest instead that there is something wrong with the idea of 27 Davidson, Inquiries, p. 220. 262
Intentionality and Interpretation 'radical analysis'. The impotence of intentionality result looks as straightforward a reductio of an approach to thought as one could imagine.28 For present purposes, the key point here is that thinking, conceiving, doubting and so on can occur as conscious (and sharable) phenomena: there is such a thing as having direct conscious awareness as such of a piece of contentful thought, either one's own or someone else's. Content can be as much a constituent of the stream of consciousness - whatever that means - as itches or patches of red. Thus one may consciously feel an itch, 'see red', or suddenly think (or hear someone say) that the cat is on the mat. In the last kind of case, the propositional content figures in the same conscious domain as itches or flashes of red. Moreover, unsurprisingly, the content appears as structured appropriately: this is why one can intelligibly and consciously go on to infer that there are cats. In the context of all this - which was certainly Brentano's context - the claim that cat-thoughts have cats as intentional objects is not part of some underdetermined or relativized theoretical structure for delivering a theory of interpretation for 'cat'-utterances. Rather, it is an aspect of phenomenological analysis, part of the enterprise of saying what our conscious life is like in itself. That we can direct our thought and talk at cats rather than undetached cat-parts, not relative to this or that scheme imposed by someone else from outside but period, is a highly salient feature of this life. Any account of thought which denies this is simply, literally, failing to save the phenomena. This is not to say that focusing on interpretation and linguistic behaviour is a mistake. On the contrary: If content-bearing states can be conscious, then at least part of knowing the consciousness of another falls under the general heading of knowing their intentional states. And to know an intentional state as the intentional state it is involves knowing it as the state that thus and so, i.e. involves knowing its content. So the enterprise of gathering knowledge about the consciousness of subjects overlaps with the overall project of interpreting them: of making sense of their behaviour, linguistic and otherwise, by e.g. seeing what they are doing and hearing what they are saying, and ascribing beliefs, desires and other intentional states to them. Knowing what it is like to be someone involves being able to interpret or understand them for oneself. It follows that a complete - that is phenomenologically adequate - 'account' of their conscious life must be interpretational: any 'account' of their consciousness which did not enable us to interpret them would thereby be incomplete, since it would not tell 28 Compare Searle, 'Indeterminacy', 124—7, 136-7. 263
Gregory McCulloch us what it is like to be them. This aspect of their phenomenology would be missing.29 All of this should make it plain that to emphasize the phenomenology of content is not to revert to a version of the traditional ideationism mentioned above, or to accept what Quine calls 'the myth of a museum in which the exhibits are meanings and the words are labels',30 or to fall into the temptation warned against in this passage from Davidson: Perhaps someone ... will be tempted to say, 'But at least the speaker knows what he is referring to.' One should stand firm against this thought. The semantic features of language are public features. What no-one can, in the nature of the case, figure out from the totality of the relevant evidence cannot be part of meaning. And since every speaker must, in some dim sense at least, know this, he cannot even intend to use his words with a unique reference, for he knows that there is no way for his words to convey this reference to another.31 We do not need to make these mistakes because the point about the phenomenology of content does not concern the Cartesian theatre, but rather the interpersonal facts involved in consciously thinking, understanding and communicating. More generally, it is not even primarily a point about the first-person case,32 but is equally a point about the public practice of understanding each other. In a recent discussion of what he calls 'understanding experience', Galen Strawson cites an excellent statement of this part of the matter from Schopenhauer: While another person is speaking, do we at once translate his speech into pictures of the imagination that instantaneously flash upon us and are arranged, linked, formed, and coloured according to the words that stream forth, and to their grammatical inflexions? What a tumult there would be in our heads while we listened to a speech or read a book! This is not what happens at all. The meaning of the speech is immediately grasped, accurately and clearly apprehended, without as a rule any conceptions of fancy being mixed up with it.33 29 For much more on this, see G. McCulloch, 'Bipartism and the Phenomenology of Content' (forthcoming). 30 Quine, Ontological Relativity, p. 27. 31 Davidson, Inquiries, p. 235. 32 Pace Searle, 'Indeterminacy', 126, 126-7, 141, 145. 33 G. Strawson, Mental Reality (Cambridge, MA: MIT Press, 1994), p. 7, n4. 264
Intentionality and Interpretation Meaning figures in phenomenology not only in the first-person case, as when one is thinking consciously, but also in the third-person case, as when someone is speaking and one is aware of what they are saying (as when one interprets them successfully). It figures there directly, unmediated, but there is no need to claim that we are infallible detectors of meaning, our own or those of others: we can certainly misinterpret (hear the wrong meaning), and whatever has ultimately to be said about privileged first-person knowledge, there are well-known problems with making too much of it. But we should maintain that meanings, others' as well as our own, can figure as integral components of our conscious life. Moreover, these are indeed public features of the use of language, since the point is that public matters like speech, interpretation and communication are themselves fundamentally conscious phenomena. What I think is often what I consciously put into words; and suitable audiences frequently hear that same thing. To say all this is to stand on its head Davidson's reasoning in the last passage quoted from him. Speakers have a very strong conviction that they can intend to use their words with a unique reference or intentional object, and that there are situations in which their words do convey this reference to another. Given also the point that meanings can be phenomenologically available, speakers have, if anything, more than a dim knowledge that 'what no-one can, in the nature of the case, figure out from the totality of the relevant evidence cannot be part of meaning'. So the correct conclusion to draw is that it is Davidson's conception of the relevant evidence that is suspect. His focus on the radical case simply excludes some of the evidence that is available to speakers when they understand one another's utterances: it excludes the phenomenology of thought and communication. This now puts us in a position to see fairly easily why Davidson, despite his anti-behaviourism, moves directly from underdetermination to indeterminacy in the case of sub-sentential semantic reality. The answer lies in the passage under discussion, and specifically in this part of it: The semantic features of language are public features. What noone can, in the nature of the case, figure out from the totality of the relevant evidence cannot be part of meaning. This, I suggest, amounts to an acknowledgement, howsoever mishandled, that meanings are manifest: that the kind of content aimed at by a theory of interpretation is a phenomenological notion, something that can, for example, be directly seen and heard in the behaviour, linguistic and otherwise, of speakers. (Earlier, we encountered 265
Gregory McCulloch what amounts to a rather similar acknowledgement in Quine, when we quoted him as saying that 'There is nothing in linguistic meaning beyond what is to be gleaned from overt behaviour in observable circumstances.') Now if meaning is, indeed, a phenomenological notion, then of course there is a sense in which sub-sentential semantic reality is exhausted by its empirical reality, by what is available to our ordinary observations of the personal realm. I hear you say that the cat is on the mat, and thereby am (defeasibly) aware that you are directing your thinking at cats. However, if Davidson is to avoid Quine's radically behaviouristic conception of what it is for empirical reality to exhaust the whole of sub-sentential semantic reality, then he has to move to a richer conception of the empirical reality. In theorizing in terms of the radical case, he is trying to keep the empirical base thin in order to leave room for 'radical analysis'. But this procedure is in severe tension with the claim that meanings (rather than behaviouristic surrogates) are manifest. Yet only something like this claim - which he anyway rather appears to accept - could protect him from the charge of an unargued (or behaviouristic) slide from underdetermination to indeterminacy. So something has to give. If Davidson really is to occupy principled space between Quine's behaviourism, and a behaviour-rejecting mentalism that acquiesces in underdetermination but jibs at indeterminacy, albeit at the cost of trying to locate the essential determining facts behind linguistic behaviour, then the thin empirical base has to go. And it takes the notion of 'radical analysis', the primacy of the radical case, with it.34 It is very important to get this point right.35 The temptation is to retort that what I am calling the phenomenological availability of meaning and intentionality boils down to something like: speakers evince beliefs or so-called knowledge about what their words and sentences mean; so, of course, one is inclined to make utterances such as '"cat" as we use it refers to cats, not to undetached catparts'. But if that is all the point comes to, then there is a simple Davidsonian reply: namely, such utterances are as open to the permutation trick as any others, so that although on one theory 'refers' refers to reference, on another it refers to p-reference, where for 'cat' to p-refer to cats is for it to refer to undetached cat-parts. 34 This is not to say that reflection on the radical case has no merit at all. On the contrary, I think it can be used to make plausible the central insight of the Verstehen tradition that knowledge of minds as minds is fundamentally different from knowledge of body: see the brief mention of the epistemological Real Distinction in section VII below; also G. McCulloch, 'An Essentially Dramatic Idiom: Quine and the Attitudes' (forthcoming). 35 Thanks here to my colleagues Harold Noonan and Jose Zalabardo. 266
Intentionality and Interpretation In reply: It is question-begging to assert that the point about phenomenology 'boils down to' the fact that speakers evince beliefs or so-called knowledge and make utterances about the meanings of their words, all of which are open to the permutation point. That should be, at best, the result of an argument that starts with the need for 'radical analysis' and ends with the inscrutibility of reference and the impotence of intentionality (somehow one needs to have moved from points about the artificially constructed radical case to a generalization that embraces the home case). But the present argument concerns the initial propriety of aiming for 'radical analysis' and its concomitant thin empirical base. If we do not make Quine's behaviouristic assumptions, it is not a datum that 'radical analysis' should be undertaken. On the contrary: I am claiming as a datum that in the course of our conscious, communal mental life we direct our thoughts upon objects, unrelativized, in the way claimed by Brentano. If this fact cannot be coped with by attempts at 'radical analysis', then so much the worse for concern with the radical case. As for my claim about what is a datum: assessment of that has to be left as an exercise for the reader. Do you find yourself now and again directing your attention at cats? Do you sometimes hear other people talking about cats? Or what? VI Naturally, a lot can be said about Davidson's analogy between subsentential semantics and the microphysical. But here I shall make two comments. Suppose first that the analogy is strong. Then it is curious that Davidson should be so keen on 'radical analysis'. Even if, in the physical case, there is something relatively theory-neutral, a shadow of what traditional empiricism saw as the evidential basis of physical science, few would accept that this shadow exhausts the actual observational or phenomenological domain of scientific data. A lot of our ordinary observational classification of the passing show is not theoretically innocent: yet this is where the evidence for the underlying microphysical reality is gathered. This practically undeniable fact that theory taints the evidence gives rise to well-known problems: nevertheless, practically undeniable fact it is. Why should things be any different in the theory of meaning? Why shouldn't the way people appear to us be 'tainted' by semantic reality in just the same fashion? And if so, then oddly enough one might draw support here, from Davidson's own analogy with the scientific case, for the richer conception of the empirical reality of sub-sentential 267
Gregory McCulloch semantics canvassed above. The call for 'radical analysis' in the physical domain would be nothing less than a plea for a very unappealing empiricism. Why should this be any more appealing in the theory of meaning? All of that, of course, is a standard response for behaviour-rejecting mentalists, who hold that genuine semantic reality is 'underlying' in the same way that microphysical reality is. But I am interested in offering Davidson a different thought, based on the idea that the analogy between sub-sentential semantics and the microphysical is weak (to say the least): a thought that is more in keeping with his own mis-handled point that meanings are manifest, part of the phenomenological domain. It is not bold or interesting to complain that Quine's conception of 'the objective data' is tendentious: that is just another way of saying that it is behaviouristic, and therefore inadequate to its domain. The foregoing points about the phenomenology of content can then be seen as supporting the idea that adequate conceptions of thought require mentalism. But a lot here depends on what you mean by 'mentalism'. Behaviourism, at least when considered as an ideology as opposed to a methodology, centres on the determination to root out of our conception of the mind any suggestion of the occult. In practice this meant the elimination of mentalistic vocabulary or, at least, of any connotation such vocabulary may have of the inner, essentially private, or hidden. In so far as this ideology was part of a general positivism, it is not surprising that the mentalistic backlash should have involved the idea that mental reality is essentially a matter of the 'theoretical' underlying causes of behaviour - i.e. be behaviour-rejecting. But there is middle ground. I have claimed, in effect, that the 'behaviouristic' aspects of Davidson's account, evinced by its closeness to Quine's on many matters of substance, are a due if blurred reflection of the important point that meaning is a phenomenological notion. Equally, we might say that the incompatibility between this idea and Davidson's hope for 'radical analysis' reflects an aspect of mentalism: it reflects at least the thought that no adequate treatment of intentional matters can dispense with mentalistic, intentional vocabulary (something that Davidson himself accepts: for more on which, see below). All of this leaves in the air to what extent this vocabulary carries connotations of the inner causes of behaviour. But it is clear that one can make something of the thought that it does without going the whole behaviour-rejecting hog and regarding intentional reality as essentially 'theoretical'. Perhaps the distinction between genuine performance and mere simulation turns on the question of etiology. But it does not follow from this that such necessary conditions of cogni268
Intentionality and Interpretation tion are also sufficient. And although I don't have the space to argue the point here, I think it can be shown that behaviour-rejecting mentalists, who make this sufficiency claim, are as unable to accommodate the phenomenology of content as those who theorize in terms of the radical case.36 Quickly: think of the version of the ideational account of understanding sketched in section IV. I said there that the unavailability of ideas from the third-person point of view would naturally be regarded as epistemological. But, of course, things are much worse than that, as anyone knows who has considered the question of grafting a theory of communication on to such an account. The 'subjective' drops out as irrelevant: and nothing changes if we replace subjective items as traditionally conceived with expressions from the language of thought. The point is that since communication is both a public event and a sharing of thoughts, the bearers of thought-content themselves have to be public. That is one reason why Quine and Davidson are quite right to focus on linguistic behaviour when approaching the matter of intentionality. Where they go wrong is to focus on the radical case, since that simply washes away the public facts, the phenomenology, that they set out, insightfully, to capture. What is really required here, I am claiming, is behaviour-embracing mentalism: that is what the phenomenological facts dictate. And to say all this just is to say that the analogy between sub-sentential semantics and the microphysical is weak, or much worse. VII The foregoing has been critical of a salient aspect of Davidson's approach. So it is instructive and rather satisfying to note how easy it is to excise this aspect without warping the rest. One searches very hard to see why he follows Quine over the radical case. We noted a brief mention of not begging questions above, and the following passage actually contains an explicit argument: 'Theory of meaning' is not a technical term, but a gesture in the direction of a family of problems ... Central among the problems is the task of explaining language and communication by appeal to simpler, or at any rate different, concepts. It is natural to believe this is possible because linguistic phenomena are patently supervenient on non-linguistic phenomena.37 36 McCulloch, The Mind and Its World, chapters 5-8, and G. McCulloch, From Malicious Demon to Evil Scientist: How Much World Does a Mind Need} (Inaugural Lecture, University of Birmingham, 1997) 37 Davidson, Inquiries, p. 215. 269
Gregory McCulloch But the spirit of this passage is strangely counter to that of the following one on anomalous monism: Although the position I describe denies that there are psychophysical laws, it is consistent with the view that mental characteristics are in some sense ... supervenient on physical characteristics ... Dependence or supervenience of this kind does not entail reductibility through law or definition...38 The argument in the first passage is also weak, for reasons given in this continuation of the second: if [supervenience did entail reducibility], we could reduce moral properties to descriptive, and this there is good reason to believe cannot be done.39 Furthermore, Davidson is quite emphatic that the mental and the physical are fundamentally separate, answerable to their own 'disparate commitments'.40 What he appears to gesture at hereabouts is some version of what I have elsewhere called the epistemological Real Distinction,41 the central claim of the Verstehen tradition that knowledge of minds as minds is fundamentally different from knowledge of body. Here is what looks like a commitment: When we attribute a belief, a desire, a goal, an intention or a meaning to an agent, we necessarily operate within a system of concepts in part determined by the structure of the beliefs and desires of the agent himself ... this feature has no counterpart in the world of physics.42 Now all of this is entirely in keeping with Davidson's focus on interpretation, as well as the principle of charity and his truth-theoretic approach to meaning. More to the point, none of it is touched if the commitment to 'radical analysis' is abandoned. So why should Davidson have followed Quine so closely over the matter of the radical case, given that his arguments for so doing, such as they are, are weak, and the rest of his position does not require it? Summing up in 'Reply to Foster', he wrote that My way of trying to give an account of language and meaning makes essential use of such concepts as those of belief and intention, and I do not believe it is possible to reduce these notions to 38 D. Davidson, Essays on Actions and Events (Oxford: Oxford University Press, 1980), p. 214. 39 Ibid. 40 Ibid., p. 222. 41 McCulloch, From Malicious Demon to Evil Scientist. 42 Davidson, Essays, p. 230. 270
Intentionality and Interpretation anything more scientific or behaviouristic. What I have tried to do is give an account of meaning (interpretation) that makes no essential use of unexplained linguistic concepts.43 Given his views about the tight interdependency between thought and talk, it is strange that he should discriminate them so emphatically. We should, of course, remember that the notion of irreducible belief and intention he accepts, thanks to his focus on the radical case, is infected by the impotence of intentionality and so somewhat removed from the pre-theoretical. What he claims to be irreducible is not what we think we have before the arguments for inscrutability of reference swing in. But this makes it even more mysterious why he should see such a large gulf between the mental and the linguistic. Anyway, for what it is worth, I suspect the following. In rejecting Quine's behaviourism, Davidson more or less unthinkingly moves to behaviour-rejecting mentalism, which involves regarding cognition as essentially a matter of what happens 'behind' behaviour. Then even given his claim that thought requires talk, the fact still remains that linguistic behaviour is not, in itself, essentially mental, even though the capacity to exhibit it is held necessary to the having of beliefs and similar cognitive states. Talk is fundamentally distinct from though fundamentally involved with thought. And because this is behaviour-rejecting mentalism, talk must derive its 'intentional' or semantic properties from the underlying mental reality it purportedly reveals. Hence it is 'natural' (his word) to think that linguistic semantic concepts can be explained in terms of mental ones. If that is right, then it is Davidson's unthinking recoil to behaviour-rejecting mentalism that leads him to keep in place the aim for 'radical analysis' even after Quinean behaviourism has been rejected. In fact, given the general point raised above about how theory infects evidence, this kind of mentalism does not sit easily with the idea of 'radical analysis', so the resulting position is unstable. But much more importantly, neither 'radical analysis' (I have argued here) nor behaviour-rejecting mentalism (I have argued elsewhere) can accommodate the phenomenology of content: and this is a much deeper point than the observation that theory infects evidence. So a much more promising tack is to drop the aim for 'radical analysis', and to persevere with the remaining bulk of Davidson's position as an articulation of a kind of behaviour-embracing mentalism, on which semantic reality is located in the public, phenomenological domain, the impotence of intentionality is avoided - and proper sense can be made of the point that meaning is manifest. 43 Davidson, Inquiries, p. 176. 271
Externalism and Norms CYNTHIA MACDONALD We think that certain of our mental states represent the world around us, and represent it in determinate ways. My perception that there is salt in the pot before me, for example, represents my immediate environment as containing a certain object, a pot, with a certain kind of substance, salt, in it. My belief that salt dissolves in water represents something in the world around me, namely salt, as having a certain observational property, that of dissolving. But what exactly is the relation between such states and the world beyond the surfaces of our skins? Specifically, what exactly is the relation between the contents of those states, and the world beyond our bodies? I believe that the correct view of the relation between certain mental contents, the contents of at least some of our intentional states, and the world beyond our bodies is an externalist one. Crudely, externalism is the view that certain of our intentional states, states such as beliefs and desires, have contents that are world-involving.' Less crudely, it is the view that certain intentional states of persons, states such as beliefs and desires, have contentful natures that are individuation-dependent on factors beyond their bodies. My belief that salt dissolves in water, has a content, that salt dissolves in water, that is individuation-dependent on a certain substance in the world beyond my body, namely, salt. The roots of externalism lie in the work of Hilary Putnam, who was concerned to show something, not specifically about the nature of mental states, but about the nature of meaning.2 He argued that one's meaning what one does by a natural kind word, although intuitively a state of mind, is world-involving. It is world-involving because it is determined in part by the actual, empirically discoverable nature of something in the world external to one's body. So a person's meaning something by a natural kind word cannot be 1 would like to thank Graham Macdonald, Graham Bird, Anthony O'Hear, Michael Martin, Scott Sturgeon, Jan Bransen and Marc Slors for comments and discussion of issues in this paper. ' The term is Philip Pettit's and John McDowell's. See their 'Introduction' to Subject, Thought, and Context, ed. P. Pettit and J. McDowell (Oxford: Oxford University Press, 1986), pp. 1-15. 2 See 'The Meaning of "Meaning"', in Mind, Language, and Reality, vol. 2 Cambridge University Press, 1975), pp. 215—71. 273
Cynthia Macdonald determined independently of that person's relation to the physical world around them. Putnam reinforced this claim by invoking what is by now the familiar strategy of the twin earth thought experiments. These experiments invite us to suppose that the environments of two individuals might differ in certain ways while all the 'within-the-body' physical and phenomenological (or 'felt') facts about those individuals remain invariant. In that case, Putnam argued, the meanings of the words in those individuals' mouths would also vary: these within-the-body twins would then mean different things by their (indistinguishable) utterances. Tyler Burge took the moral of the twin earth thought experiments one step further.3 He argued that since, when a person is sincere, what she says is what she believes, the Putnam conclusion about meaning carries over to intentional states such as beliefs and desires. Burge argued that the twin earth thought experiments not only show that meaning is (partly) an external phenomenon, but that mental states like beliefs and desires, whose contents are typically specified by means of words whose meanings are determined by factors external to persons' bodies, are also partly external phenomena. Just as my twin and I might mean different things by our indistinguishable utterances of 'there is salt in the pot' because of differences in the chemical constitutions of superficially and phenomenologically indistinguishable substances to which we are related in our respective environments, so too might my twin and I think different thoughts when we think thoughts with those propositional contents. The twin earth thought experiments have been used by Burge and others to support the externalist view that certain intentional states have contentful natures that are individuation-dependent on factors external to the bodies of persons who undergo them. Put like this, it may make look as though there is one single, clear formulation of externalism and that there is agreement amongst externalists about what it entails with regard to the existence of objects beyond the bodies of persons who undergo intentional states with representational contents. But this is so far from being the case that part of my aim in this paper is to disentangle some of the different formulations and associated commitments of the view from others, in order to fix on what I take to be a central commitment common to all of them and to defend that commitment. 3 See, for example, 'Individualism and the Mental', in Midwest Studies in Philosophy 4 (Minneapolis: University of Minnesota Press, 1979), pp. 73—121; 'Other Bodies', in Thought and Object: Essays on Intentionality, ed. A. Woodfield (Oxford: Oxford University Press, 1982), pp. 97-120, and 'Individualism and Psychology', The Philosophical Review 95 (1986), 3-45. 274
Externalism and Norms Externalist theses can be strong or weak, and they can be strong or weak in different kinds of ways. However, most theses apply in the first instance to contentful intentional types or kinds, such as the kind, thinks that salt dissolves in water. Many thinkers can think thoughts with this content, and when they do they think thoughts that fall under a single contentful kind.4 It is this version of externalism that I wish to concentrate on in 4 See, for example, Burge, 'Individualism and Psychology', Jerry Fodor, 'Individualism and Supervenience', Proceedings of the Aristotelian Society, Supplementary Volume 60 (1986), 235—62, Psychosemantics (Cambridge, MA: MIT Press, 1986), and 'A Modal Argument for Narrow Content', Journal of Philosophy 88 (1991), and Gregory McCulloch, The Mind and Its World (London: Routledge, 1995). Many who are externalists with regard to contentful intentional kinds also endorse externalism with regard to individual states or events that fall under, or are of those kinds. Tyler Burge is one notable example; he is what might be called a token externalist as well as a type externalist (see 'Individualism and the Mental', and 'Individualism and Psychology', note 7). Token externalism is the view that the natures of individual intentional mental events or states are individuation-dependent on factors beyond persons' bodies. They are so because they are individuated by the contentful types or kinds under which they fall, which themselves are individuation-dependent on factors beyond persons' bodies. Since to be a mental event is to be an event of a contentful kind, and since contentful kinds are individuation-dependent on factors external to persons' bodies, mental events are themselves individuation-dependent on factors external to persons' bodies. Despite this natural association of type with token externalism, it is possible to be a type externalist without embracing token externalism, and vice versa. Both of these possibilities have been argued for, and in my opinion both positions are defensible. In particular, the combination of type externalism and token internalism is defensible. Whether one is a token as well as a type externalist depends on whether one thinks that it is of the essence of any mental event which is of a contentful type that it be of a contentful type. This is not a question about the truth of the claim that, necessarily, each event that has intentional content has intentional content. That claim is obviously and uncontroversially true. It is a question, rather, about the truth of the claim that necessarily, each event that has intentional content necessarily has intentional content. And this claim is not obviously and uncontroversially true. Whether it is true depends on the truth of other views. For instance, it depends on whether non-reductive physicalism is true and contingent. If it is, then token externalism is false, since nonreductive physicalism is committed to the view that the essences of mental events are physical, not mental. It may be true that mental events, qua mental, cannot be individuated independently of the contentful types or kinds under which they fall; but it does not follow that these events cannot be individuated independently of the contentful kinds under which they fall. For that depends on whether these events are essentially mental events. 275
Cynthia Macdonald the remaining sections of this paper. In section I below, I briefly outline a small number of type or content externalist theses, in order to fix on a core commitment that they share. I then formulate type externalism in these terms. Then, in section II, I focus on a debate between two very well-known adversaries, Tyler Burge and Jerry Fodor. This debate concerns the truth of anti-individualism, which differs from externalism in that it concerns how mental kinds are to be taxonomized for the purposes of a scientific psychology. However, the debate is instructive, since it helps to identify the source of individualism; of why both externalists and anti-individualists disagree with individualists with regard to the core commitment articulated in section I. Then, in section III, I defend Burge by anchoring the source of type externalism in a very general but distinctive argument, one that relies on the rationalistic normativity of the psychological domain. My defence trades on likenesses between psychological explanation and functional explanation in biology. If the defence succeeds, it succeeds equally for externalism and anti-individualism. Finally, in section IV, I conclude with some remarks about the consequences of this particular form of externalism. I Varieties of Type Externalism There is a central claim that almost all of the varieties of content externalism share, which concerns the dependency of contentful kinds on conditions or factors in the environment in which subjects are embedded.5 To see this, we need consider only a few of the ways in which type externalism is typically expressed. 5 McGinn {Mental Content (Oxford: Oxford University Press, 1989)) is an exception. He distinguishes between what he calls 'strong' and 'weak' externalism, and argues for the latter and against the former. By 'strong' externalism, McGinn means one which takes content-individuation to require the existence, in the environment in which a thinker is situated, of some object or objects external to the thinker's body. McGinn rejects this view, but endorses the weaker externalist view that content-individuation requires the existence, in the world of the thinker, of some object or objects beyond that thinker's body. This departure from most other externalism means that McGinn is not prepared to rest the truth of falsity of externalism on the existence of twin earth examples. Thus, he says: it understates the case to express the upshot of twin earth reflections as inconsistent with methodological solipsism, since those reflections imply strong externalism, not just weak. Such understatement can be misleading if it encourages the idea that the inapplicability of twin earth arguments to certain cases shows that internalism is true in those cases. 276
Externalism and Norms Externalism is often expressed in terms of some kind of supervenience claim regarding the contentful natures of certain intentional types.6 Broadly speaking, the claim is that such types weakly (in the case of the twin earth thought experiments) or strongly (for thought experiments involving other possible worlds) supervene on factors beyond the bodies of persons.7 However, this claim can itself be interpreted in a number of ways. The reason is that supervenience is a name for a very general co-variance relation, one which states that things cannot differ (or vary) in one respect without differing (or varying) in another, and this covers many different types of case.8 What is related by supervenience, and how it is related, can 6 See, for example, Martin Davies, 'Aims and Claims of Externalist Arguments', Philosophical Issues 4 (1993), 227-149, where externalist theses are explicitly formulated in these terms. Also, see Brian McLaughlin and Michael Tye, 'Externalism, Twin-Earth, and Self-knowledge', in Knowing Our Own Minds, ed. C. Wright, B. Smith, and C. Macdonald (Oxford: Oxford University Press, forthcoming), note 39, where externalism is formulated in terms of supervenience, and Burge, 'Individualism and Psychology', where individualism is formulated in terms of supervenience, externalism being the negation of that thesis. 1 The Putnam twin earth thought experiments concern weak supervenience, since Putnam envisaged twin earth as being a planet in our own universe, and so in the same possible world. Twins are members of different linguistic communities, but communities within the same possible world. 8 See Frank Jackson, 'Armchair Metaphysics', in Philosophy in Mind: The Place of Philosophy in the Study of Mind, ed. M. Michael and J. O'Leary-Hawthorne (Dordrecht: Kluwer Academic Publishers, 1994), pp. 23—42, who characterizes supervenience in similar terms, as lack of independent variation. You can be a weak externalist about a certain kind of content, and so reject methodological solipsism, and yet deny vehemently that a twin earth case can be given for the content at issue: that is in fact my position about certain kinds of content, as will become apparent. {Mental Content, p. 9, n. 13). If McGinn is right, then the truth or falsity of externalism is not decided by whether twin earth examples exist: although the existence of a twin earth example may be decisive for externalism with regard to certain contents, other contents may be externalistically individuated even when a twin earth example is not forthcoming. Although I do not subscribe to McGinn's brand of externalism, I agree with him that the truth or falsity of the thesis is not anchored in the twin earth examples. However, because his version of externalism departs markedly from most others, I set it aside for present purposes. 277
Cynthia Macdonald differ greatly from case to case; and the strength of that relation may vary also, in accordance with variation in the objects related and the nature of the relation. All of this will make a difference to how the claim of supervenience is to be understood, and whether it is likely to be true in any particular case. In short, supervenience itself is a name for a class of theses that may concern different objects, different kinds of relations between them, and different strengths of relations, each thesis itself requiring independent explanation and defence.9 Given this variety, one cannot expect there to be just one externalist thesis associated with any given claim of supervenience. And indeed there is not. Some have held that externalism commits one to the view that contentful intentional properties, properties associated with contentful kinds such as thinks that salt dissolves in water, actually entail the existence of objects or kinds of objects in the 9 For example, there are supervenience relations between logically or conceptually related properties, such as being coloured and being red, supervenience relations between what we might call 'metaphysically' related properties, such as moral or aesthetic properties and psychological ones, or psychological properties and physical ones, and supervenience relations between causally related properties, such as those that figure in causal laws. All of these conform to the formula that is thought to characterize supervenience relations generally, namely, no change in supervenient property without a change in subvenient property. So no psychological change without a physical change, no aesthetic change without a physical change, no change in effect property without a change in cause property. But the relations are really very different in these different types of cases. Although they all involve a relation between properties, they differ in the types of properties related, and they differ in the kind of relation that is thought to hold between them. Other theses differ from these in relating, not properties, but regions of worlds or worlds themselves, or events or states. Global supervenience claims typically concern worlds or regions of worlds. See, for example, Terence Horgan, 'Supervenience and Microphysics', Pacific Philosophical Quarterly 63 (1983), 29-43. Matters are more complicated still, since the strength of the dependency relations associated with these different kinds of supervenience relations also varies considerably. For example, the dependency relation associated with being coloured and being red is said to be logically or conceptually necessary. But this is not so for the relation that is thought to hold between moral and aesthetic properties and psychological properties, or between moral and aesthetic properties and physical properties. Here the relation seems to be weaker than one of logical-cum-conceptual necessity. It seems, rather, to be either metaphysically necessary, where this is understood not to require conceptual necessity, or physically necessary, a necessity that is weaker still, requiring only compatibility with the existing laws of nature in this world. 278
Externalism and Norms world beyond the skins of persons.10 This, it is said, is because externalism is committed to the claim that it is a conceptual truth that, for some propositional content C (such as that salt dissolves in water), and some proposition, P, not knowable a priori (such as salt exists), if a thinker knows that C, then P. Thus, for example, externalism is committed to the claim that it is a conceptual truth that if a thinker is thinking that water is transparent, then water exists.11 If 10 See, for example, Martin Davies, 'Externalism, Architecturalism, and Epistemic Warrant', in Knowing Our Own Minds, ed. Wright et al., and B. McLaughlin and M. Tye, 'Externalism, Twin-Earth, and SelfKnowledge', same volume. The twin earth thought experiments are standardly construed as supporting conceptually necessary externalist theses. This is what lies behind arguments of the kind that Michael McKinsey has advanced to show that externalism is incompatible with privileged access, or authoritative self-knowledge. His argument depends on externalism being committed to the claim that it is a conceptual truth that, for some thought content, C, which has externalistic individuation conditions (such as that water is transparent), it is a conceptual truth that if one is thinking that C, then P, where P is a proposition that cannot be known a priori (such as water exists). See Michael McKinsey, 'Anti-individualism and Privileged Access', Analysis 51 (1991), 9-16. For a reply which denies that externalism is committed to such a claim, see Anthony Brueckner, 'What an Anti-individualist Knows A priori', Analysis 52 (1992), 111-18. But many externalist theses do not purport to be conceptually necessary. See, for example, Fred Dretske, Knowledge and the Flow of Information, (Cambridge: MA: MIT Press, 1980) and Explaining Behaviour: Reasons in a World of Causes (Cambridge, MA: MIT Press, 1988), Ruth Millikan, Language, Thought and Other Biological Categories (Cambridge, MA: MIT Press, 1984), David Papineau, Reality and Representation (Oxford: Basil Blackwell, 1987), and Fodor, Psychosemantics. There are differences between the view known as anti-individualism and externalism. Fodor, for example, explicitly distinguishes the two, and claims that externalism is true, but anti-individualism is not (see 'A Modal Argument for Narrow Content'). Externalism is a view about how the contents of intentional states, states with propositional content, are correctly individuated. Anti-individualism, on the other hand, is a view about how the contents of intentional states are, or should be, individuated for the purposes of a scientific psychology, i.e. for the purposes of (causal) explanation in psychology. The distinction between externalism and anti-individualism raises important questions about the nature of psychological explanation and the nature of scientific explanation and taxonomy in general. However, these issues are largely irrelevant to the present discussion, and so the distinction will not play a role in the argument to be developed. 11 McLaughlin and Tye ('Externalism') have pointed out that no type externalist seems actually to have held a view this strong. Brueckner ('What an Anti-individualist Knows'), in his reply to McKinsey ('Antiindividualism') (whose argument is directed at Burge), points out that Burge (in 'Other Bodies') actually argues against this view. 279
Cynthia Macdonald this is so, then it must be conceptually necessary that contentful properties supervene on factors beyond the bodies of subjects that undergo states with those properties. Others deny that externalism is committed to anything as strong as this claim.12 Although it requires that the contents of certain intentional states be object-dependent, this is not a matter that can be known a priori, since one cannot know a priori that certain concepts, or propositional contents, are object-dependent. This seems especially plausible in the case of natural kind contents.13 Still others claim that externalism commits one to something stronger than a mere claim of object-dependency but weaker than a claim of conceptual entailment, since it requires dependency on objects with which persons causally interact in their environments.14 Teleological externalist theses, which require that content supervenes on the causal history of subjects and their interactions with objects in their environments, are theses of this kind. These are very different kinds of theses than either of the two just mentioned, and failure to distinguish them can only lead to confusion about the basic commitments of externalism and about whether externalism is itself a plausible or implausible doctrine. 12 Burge ('Other Bodies') is one. See also Fodor {Psychosemantics), Millikan {Language, Thought, and Other Biological Categories), Papineau {Reality and Representation), and Dretske {Knowledge and the Flow of Information and Explaining Behaviour). 13 The argument is this. Whether a concept is a natural kind concept cannot be known a priori, since it cannot be known a priori that there are natural kinds (and according to at least one version of externalism there can be no natural kind concepts without natural kinds). This can only be known a posteriori, if at all, since whether or not there are natural kinds is an empirical matter. But if it cannot be known a priori that the concept of salt is a natural kind concept because it is not knowable a priori that there are natural kinds, then it cannot be a conceptual truth that if one is thinking that water dissolves in water, then salt exists. See Brueckner, 'What an Anti-individualist Knows'. 14 I am thinking of Millikan, Language, Thought, and other Biological Categories, Dretske, Knowledge and the Flow of Information and Explaining Behaviour, Jerry Fodor, Psychosemantics, and Papineau, Reality and Representation. It is difficult to know where to place McGinn {Mental Content). On the one hand, he rejects the requirement of causal interaction with instances of the natural kind by individuals who possess concepts of that kind (and in this he commits himself to a thesis weaker than Millikan's and others), and on the other he seems to think that a thinker's thinking such contents conceptually entails that objects exist beyond the bodies of subjects who think them. For more on this, see McLaughlin and Tye, 'Externalism'. 280
Externalism and Norms These people disagree about the strength of the relation between the subvenient and the supervenient in externalist theses. Others disagree about the sorts of objects related. Externalists may take their commitment to externalism to entail the existence primarily of individual things, corresponding to the contents of singular thoughts such as the thought that Cicero was a Roman orator, or demonstrative thoughts such as the thought that this computer has a coloured monitor.1' Others may take the view to entail the existence of natural kinds of things, such as tigers, salt and water (corresponding to natural kind thoughts), but not necessarily to any individual instances of such kinds.1'1 Others still may take the view to entail the existence of both natural kinds and instances of such kinds with which persons who undergo thoughts with contents that are individuation-dependent on such kinds interact causally.17 Finally(!), still others may take the view to entail the existence of artefactual kinds, such as sofas and chairs (corresponding to thoughts concerning socially determined kinds).18 Since these views are compatible with one another, externalists may take the view to commit them to some combination of the above commitments. Despite all of these differences, however, type or content externalists are united in denying that the contentful nature of any intentional kind supervenes only on factors within the bodies of the subjects that undergo states of that kind. So all forms of externalism are committed to some kind of supervenience claim with regard to certain contentful intentional types. The claim is that certain intentional contents supervene on factors beyond person's bodies, in the sense that subjects' intentional states can vary or change with regard to their contents without varying with regard to all of their intrinsic physical properties.19 Given the variation 15 See John McDowell, 'On the Sense and Reference of a Proper Name', Mind 86 (1977), and Gareth Evans, 'Understanding Demonstratives', in Meaning and Understanding, ed. H. Parret and J. Bouveresse (Berlin: W. De Gruyter, 1981), pp. 280-303, and The Varieties of Reference (Oxford: Clarendon Press, 1982), chapters 4-8. "' See McGinn, Mental Content. 17 See Putnam, 'The Meaning of "Meaning"'. '* See Burge, 'Individualism and the Mental'. ''' I leave open the issue of whether such variation would entail variation in phenomenological, or 'felt' properties. It may be that variation in factors beyond the body of an individual would affect not only contentful states such as beliefs and thoughts, but also sensation states such as perceptual experiences. This is so, for example, for externalists who think that there is no non-conceptual content (see, for example, John McDowell, Mind and World (Oxford: Oxford University Press, 1995)). 281
Cynthia Macdonald amongst externalists in what factors these may be, this claim is best formulated in terms of the negation of an individualist supervenience thesis. And since variation in supervenient properties requires variation in subvenient ones, so that sameness with regard to subvenient properties prohibits the possibility of difference with regard to the supervenient ones, we can formulate the negation of that thesis as follows: 1. It is not (conceptually, metaphysically, physically) necessary that, for any two individuals x and y and any contentful property M, if x and y are indiscernible with regard to all of their intrinsic physical properties P, then x and y are indiscernible with regard to M. Or, 2. It is (conceptually, metaphysically, physically) possible that, for any two individuals x and y and any contentful property M, x and y are indiscernible with regard to all of their intrinsic physical properties P, but discernible with regard to M. What this says is that it is possible for two individuals to be the same with regard to their intrinsic physical properties but different with regard to their contentful mental properties. Short of a specific form of dualism, namely an internalist one, this possibility can only be because the natures of contentful kinds depend on factors or conditions external to the bodies of persons who undergo states of those kinds.20 Versions of externalism that are articulated in terms of this general supervenience claim are sometimes called modal externalist theses.™ These are concerned with the existence of twin earth examples. Since the twin earth examples make explicit the dependency of contentful kinds on factors or conditions external to subject's bodies, implicit in supervenience formulations of externalism is a claim which is sometimes called constitutive externalism. This is the claim that the correct philosophical account of the natures of certain contentful types takes them to have natures that depend on factors or 20 One might think that dualism alone is sufficient to account for the truth of this claim. However, dualism is silent on the internalism/externalism issue. It is consistent with dualism that mental contents should be individuation-dependent on factors external to the bodies of thinkers (and so external to the mind). See McCulloch, The Mind and Its World, p. 227, note 5. 21 This is Davies's terminology. See 'Aims and Claims of Externalist Arguments', p. 227-8. See also his 'Externalism, Architecturalism, and Epistemic Warrant'. 282
Externalism and Norms conditions that exist beyond the bodies of individual subjects that undergo states of those kinds.22 Constitutive externalism is the view I want to defend. Although it is a common strategy to employ the twin earth examples to establish it, I want to defend the view in a more direct way. The twin earth examples are best viewed as a kind of counterfactual test of the truth or otherwise of constitutive externalism. This test is meant to flesh out and validate intuitions about the object-dependence of contentful kinds. However, the test is only as persuasive as the intuitions that prompt it. If one is inclined to think that mental contents are object-dependent, then one will be inclined to accept that the twin earth examples are really possible and that they establish such object-dependence. If on the other hand, one is inclined to think that mental contents are not object-dependent, then one will be inclined to think either that the twin earth examples are not possible or that they do not show that mental contents are objectdependent.23 22 As Martin Davies puts it, constitutive externalism says that the most fundamental philosophical account of what it is for a person or animal to be in the mental states in question does advert to the individual's physical or social environment, and not only to what is going on within the spatial and temporal boundaries of the creature. ('Aims and Claims of Externalist Arguments', 230). Davies correctly points out that one can establish a constitutive externalist thesis by establishing that modal individualism is false, i.e. that the supervenience claim (1) stated above is true, but that one cannot establish modal externalism just by establishing that constitutive externalism is true. It may be, for example, that although constitutive externalism is true, modal externalism is false because there is a necessary connection between subjects' intrinsic physical properties and factors or conditions beyond that subjects' bodies, so that an environment in which the contents of subjects' intentional states varied would necessarily be an environment in which their intrinsic physical properties also varied. 23 This emerges in debates such as that between Burge and Fodor concerning the truth or falsity of anti-individualism. Burge argues that attention to actual descriptive and explanatory practices in psychology reveals that the taxonomy of both intentional and nonintentionally described behaviour and the taxonomy of intentional states to be non-individualistic. For the interpretation of these practices fails to respect local supervenience, and this is supported by the twin earth thought experiments. However, his arguments for anti-individualism, based on these arguments, have been charged with presuming the truth of anti-individualism. In a similar vein, Burge effectively accuses Fodor's arguments for individualism, which also make use of twin earth thought experiments, of presuming the truth of individualism. Fodor argues that since whether or not 283
Cynthia Macdonald I want, therefore, to ground the intuitions on which that test is based in certain features of actions and their explanation, where the relevant actions are ones based on perception. Like Burge, I see the source of externalism as lying in our actual descriptive and explanatory practices. And I believe that attention to these practices can help to explain certain of our intuitions in the twin earth cases. But the argument for externalism can be mounted independently of the twin earth cases. So our intuitions concerning externalism can be vindicated without appeal to them. The argument that I develop specifically concerns thoughts and other intentional states whose contents, widely construed, concern natural kinds, such as salt and water.241 believe that it can be generalized to other sorts of case, but I shall not attempt that here. 24 In fact, nothing in the argument to follow requires commitment to any doctrine about natural kinds, even though the examples concern what many would consider to be natural kinds. Natural kinds are typically employed in twin earth thought experiments in order to bolster the view that twin earth twins might have thoughts that are distinct despite the phenomenological indistinguishability of the objects or substances to which their thoughts relate in their respective environments. However, the thesis that is being defended here is constitutive externalism, not twin earth externalism. Further, the examples on which the argument is mounted make reference only to the observable effects on normal observers of objects in their environments. twins have type-identical states depends on whether they have the same causal powers, and since sameness and difference of causal powers must be assessed across contexts rather than within them (casual power being a counterfactual notion), whether twins have type-identical intentional states depends on whether their states have the same causal powers across contexts. Burge agrees, but argues that twin earth considerations cannot determine and distinguish causal powers of intentional kinds because one cannot decide which contexts are relevant for determining and distinguishing causal powers without making assumptions about the kinds in question. To suppose that the actual environment external to subjects' bodies is not relevant to determining causal powers, and so taxonomy of contentful kinds, is already to assume individualism. The moral for the twin earth thought experiments is that they play a more peripheral role in adjudicating between individualism and anti-individualism. The reason is that their employment is evidently not independent of individualistic/antiindividualistic assumptions. See the debate between Burge and Fodor in Philosophy of Psychology: Debates on Psychological Explanation, (ed.) C. Macdonald and G. Macdonald (Oxford: Basil Blackwell, 1995), containing Burge's 'Individualism and Psychology' and Fodor's 'A Modal Argument for Narrow Content', with a commissioned reply by Burge. 284
Externalism and Norms II The Source of Externalism It is common in debates between externalists and individualists for both parties to appeal to behavioural considerations in support of their claims about the individuation of contentful kinds. But it is important to see how this appeal is put to work in arguments for and against externalism, and how little it establishes in the way of externalist or individualist conclusions. Consider, for example, the debate between Tyler Burge and Jerry Fodor. Burge maintains that explanatory practices in psychology supports external ism/anti-individualism because the explananda in many cases, when they are behaviour, are commonly and clearly understood to be behaviour, relationally understood as involving relations between organisms and their environments.25 Thus, he appeals to the fact that one distinguishes a heart from a waste pump by its biological function in the organisms in which it performs its function, which cannot be determined to be what it is independently of the causal history of its ancestors in organisms of the same and similar kinds. Its function cannot be specified independently of relations it bears to its surrounding environment, and the way it is embedded in that environment. However, Fodor does not deny that many of the behaviours in which intentional creatures engage, intentionally described, are to be understood as involving relations between organisms and their environments. What he denies is that such relations are relevant to the taxonomy of intentional content, at least for the purposes of causal explanation employing such content. They are not relevant because they do not make a difference to the causal powers of contentful kinds. And the individuation of contentful kinds is sensitive only to their causal powers. Thus, he reasons that because twin earth twins are molecular duplicates and so their actual and counterfactual behaviours are identical in relevant respects, the causal powers of their mental states are identical in relevant respects. They therefore belong to the same natural kind of purposes of psychological explanation, and individualism is true.26 Fodor recognizes that this argument can be turned on its head simply by denying that the actual and counterfactual behaviours of me and my twin are identical in relevant respects. After all, when I 25 See 'Individualism and Psychology', and 'Intentional Properties and Causation', in Philosophy of Psychology, ed. Macdonald and Macdonald, pp. 226-35. 26 Fodor, 'A Modal Argument for Narrow Content'. Fodor has since given up his commitment to narrow content. See The Elm and The Expert (Cambridge, MA: MIT Press, 1994). 285
Cynthia Macdonald am thirsty, I reach for water, whereas when my twin is thirsty, she reaches for twater.27 Since the behaviours are not identical, neither are the causal powers of the mental states which explain them. Inasmuch as externalists and individualists are agreed that differences in behaviour, non-intentionally described, are not what is at issue, but rather differences in behaviour, intentionally described, it seems that this argument for individualism does not go through. Fodor, however, is not perturbed by this. He argues that the question of whether the relevant intentional kinds of twins are the same is a matter of their causal potentialities and that this is to be determined, not within contexts, but across them. So, for example, the fact that my beliefs on earth cause me to drink water whereas my twin's on twin earth cause her to drink twater does not show that these beliefs have distinct causal potentialities. What is relevant is whether my twin's beliefs would cause her to drink water on earth and whether my beliefs would cause me to drink twater on twin earth. By this (cross-context counterfactual) criterion, the causal potentialities of our beliefs are the same and the beliefs are typeidentical. This response doesn't quite work, since it is still vulnerable to the charge that when I utter the words 'Gimme water' on earth, I get what I ask for, but when I utter the words 'gimme water' on twin earth, I do not get what I ask for. Similarly for my twin. Our behaviours, intentionally described as water/twater requests, do not have the same causal powers, even across contexts. Fodor attempts to patch the criterion up by providing a general condition on when differences in properties of causes are differences in causal powers. His claim is that differences in properties of causes are differences in causal powers when those properties are not conceptually connected to the effect properties for which they are responsible. By these lights, such differences as there are in intentional behaviour between me and my twin cannot be attributable to differences in the contentful properties of the states which cause that behaviour, widely construed as beliefs about water and beliefs about twater. For those properties are conceptually connected to the properties of the behaviour which makes them intentional, i.e. actions, namely, water requests and twater requests. My water requests and your twater requests may differ, but this difference in behaviour does not mark a difference in content between my water beliefs and your twater beliefs, since the contentful properties of these beliefs, widely construed as beliefs about water and beliefs about twater, are conceptually connected to the behaviour those beliefs cause. 27 See Burge, 'Individualism and Psychology'. 286
Externalism and Norms What this debate between Fodor and Burge brings out clearly is that one can agree (1) that intentional content is to be taxonomized by its relation to behaviour, (2) that behaviour is decisive in determining the truth or falsity of externalism and (3) that behaviour is to be taxonomized for psychological purposes intentionally in ways that involve relations between organisms and their environment, and yet (4) still disagree about whether externalism is true or false. Burge and Fodor agree on all of these points, and even on the further two points that (5) mental kinds are to be taxonomized in terms of their causal powers for the purposes of psychological explanation and (6) psychological explanation is causal explanation. But despite all of this agreement, they disagree about whether externalism is true. What this shows, I think, is that the truth or falsity of externalism, inasmuch as it turns on the broad/narrow content distinction, depends on the issue of the explanatory efficacy of broad or wide content. That is to say, it depends on the issue of whether, in at least some cases of the explanation of action, the contentful kinds implicated in such explanations, to do their explanatory work, must be individuated widely, i.e. by relation to factors that exist beyond the surfaces of the bodies of organisms who undergo states of these contentful kinds. Burge thinks they must because individuation of contentful kinds is individuation by causal powers, but this is not independent of assumptions about the kinds in question. Contentful kinds, like biological-functional kinds, are not only causally but conceptually connected with their effect properties. So the taxonomy of the cause properties is not independent of conceptual connections with their effect properties.28 In Burge's view, this makes psychological explanation, like functional explanation in biology, explanation which is causal but which breaches the 'Humean' requirement of connecting effects with causes non-conceptually.29 Burge acknowledges that it breaches this 28 Thus he claims, One could plausibly claim that it is a conceptual truth that hearts differ from twin waste-pumps in that they pump blood. One could plausibly claim that it is conceptually necessary that if something is a heart, then when functioning normally, it pumps blood. ('Intentional Properties and Causation', p. 233). 29 According to Burge, that is. See 'Intentional Properties and Causation'. In this Burge concurs with Neander ('Functions as Selected Effects: The Conceptual Analyst's Defense', Philosophy of Science 58 (1991), 168-84.). But note that Millikan (Language, Thought, and Other Biological Categories) denies that there are such conceptual connections between functional properties and the effect properties to which their taxonomy is sensitive. Similarly for intentional properties. 287
Cynthia Macdonald requirement, but does not see that it presents any problem for the view that psychological explanation is causal explanation, since he rejects the 'Humean' requirement. Fodor, on the other hand, thinks that the contentful kinds implicated in the explanation of intentional behaviour or action need to be individuated narrowly, i.e. individualistically. This is because, although he agrees with Burge that individuation of contentful kinds is individuation by causal powers, he accepts the Humean requirement that causes and effects must be individuated in terms of properties that are conceptually independent of one another if the cause properties are to be genuinely causally potent with regard to their effect properties and psychological explanation is to be genuinely causal explanation. Since psychological explanation is genuine causal explanation, it too must meet this requirement. By that standard, widely individuated content gets ruled out from being genuinely causally potent, hence genuinely explanatorily potent. So the crucial issue that divides Burge and Fodor is whether the explanatory potency of intentional kinds requires that such kinds meet the Humean requirement on causal explanation of being conceptually independent of their effect properties. I think that Burge is correct in his claim that psychological explanation, explanation of actions by means of states with intentional content, works by way of broadly conceptual connections between explanans property and explanandum property. Such contentful properties do their explanatory work because they have causal powers which relate them conceptually to their effect properties. However, unlike Burge, I believe that attention to the ways in which psychological explanation is like explanation in functional biology shows it to be of a distinctive, normative noncausal type. Moreover, I think that by attending to the ways in which contentful properties are like biological-functional ones, and unlike physical ones, it is possible to mount an argument for externalism that does not lead to the kind of stalemate that seems to be the inevitable result of debates between externalists like Burge and individualists like Fodor. The dispute is between those who agree that taxonomy of contentful properties is taxonomy by causal powers but disagree about whether this supports externalism because they disagree about whether such taxonomy meets the Humean requirement that for a property to be a distinctive causal power, it must be contingently or non-conceptually related to its effect property. But this dispute seems to me to be unresolvable within the narrow confines of externalism. It simply relocates the disagreement in the issue of whether psychological explanation is like functional explanation in biology 288
Externalism and Norms or like causal explanation in such sciences as physics. However, one needs a principled reason for adjudicating between these two alternatives. I want to try to provide that principled reason by showing that and how the explanation of action by intentional content is like functional explanation in biology in a certain important respect. First, I shall mount the argument. Then I will locate the source of the externalist commitment, and indicate how like it is to the source of externalism in functional biology. Ill An Argument for Externalism I begin with the observation that the truth or otherwise of externalism does not depend on whether the explananda of psychological explanations are actions construed widely or actions construed narrowly (but intentionally). As the debate between Burge and Fodor illustrates, one can agree with an externalist that the explananda of psychological explanations are actions, widely construed, and disagree about whether this shows externalism to be true. However, I think it plausible that contentful states are employed as explanantia of both sorts of actions. Sometimes, for example, we may wish to explain why a subject washes her clothing with water (rather than with sand, or with Coca-Cola), where what seems to need explaining is why she engages in a particular type of action with regard to a particular object or type of object. But there are other cases where what we wish to explain is not why a subject engages in a certain type of action with respect to a particular object or type of object, but where we simply wish to explain why that subject engages in actions of a particular type at all. Sometimes, for example, we may want to explain why a subject eats every day, or goes to bed at night, where the actions that serve as explananda are actions, narrowly construed. It may be that a subject cannot eat without eating something, but what is eaten is not what one wants to explain. What one wants to explain is the activity of eating, or the activity of washing, itself. Phenomena like these are actions, narrowly construed. Narrow actions seem to be just the sort of phenomena whose explanation would only require narrow content, if any phenomena are. So let us concentrate for the moment on actions, narrowly construed. If the explanation of even these cannot be effected without appeal to wide content, then externalism will have been vindicated. Narrow-act explanation seems to require no mention of any particular object, or of any of a range of objects, on which such actions 289
Cynthia Macdonald depend. Because of this, the states which explain and make intelligible such behaviour also seem to be capable of doing so by means of narrow content. That such actions can be construed individualistically evidently supports the view that the contentful kinds that are required to explain them by making them intelligible can also be construed individualistically. For if their taxonomy does not depend on the existence of any particular object or range of objects, then they can evidently be made intelligible, or explained, by means of contents that also do not depend on the existence of any particular object or range of objects. This idea can be further supported by a twin earth thought experiment. Consider Sue, who washes with water, and her twin, who washes with twater. Although the activity of washing requires that there be something that one washes with, the activity itself, what Sue and her twin do with the respective stuff, is the same kind of thing. Since the activities are the same, it is plausible to hold that so too are the contentful kinds which explain them. I do not think, however, that this establishes individualism. The reason is that the individuation of actions, narrowly construed, can only take place against the background of wide-act individuation; and wide acts are only made intelligible by states with wide content. Actions are not only purposeful; at least sometimes they involve interaction with objects. Further, when these actions are successful, that they are object-involving is not an accidental feature of them. If such actions were not at least sometimes non-accidentally objectinvolving, they could not be purposeful. But if they could not be purposeful, they could not be actions at all. The point here is not that there must be successful actions if there are to be actions at all. It is true that actions, in being purposeful movements, aim at success. But this is consistent with the possibility that no action is actually successful; that creatures should regularly fail to succeed at what they aim to accomplish by moving their bodies in various ways. So it is not a necessary, but a contingent matter that there are successful actions in the world. It is a contingent matter that there are objects in the world with which human beings engage, and it is a contingent matter that by engaging with these objects they are both changed by and change the world. But that there are such objects with which human beings engage, and that they at least sometimes do so successfully, is, I take it, common ground between externalists and individualists. Given that there are successful actions in the world, intentions that engage with the world are required to make them intelligible. Without such intentions, one cannot make intelligible the non-acci290
Externalism and Norms dental connection between action and object when an action is successful. This is because without intentions that engage with the world, there is an explanatory gap which leaves it mysterious why that connection is non-accidental. One makes it intelligible by citing intentions concerning objects that match the objects with which subjects non-accidentally engage in their behaviour. The connection between purpose and the world beyond the body is required because successful object-involving action is non-accidentally successful. Wide content is needed to explain successful action, which happens to, but need not, occur. Think, now, of Sue, who washes every day. We, who want to make intelligible that activity, explain it in terms of her desire to make herself clean and her belief that by washing she will make herself clean. However, that belief and desire will only serve to explain her activity against the background of assumptions she has about the sorts of stuff that makes things clean. For not every substance is such that it can make things clean. Water can make things clean. Sand can make things clean. But Coca-Cola cannot make things clean. Nor can tar, mud or hydrochloric acid. In short, the intelligibility of what one is doing, narrowly construed as a successful activity, takes place against the background of assumptions about what it is appropriate to do it with. One does not make intelligible Sue's activity of washing every day just by mentioning her desire to make herself clean and her belief that by washing she will make herself clean. Her activity simply does not count as an activity of washing if she does it with mud, or with Coca-Cola, or with tar. And here I mean: successful activity. Her movements may be the movements of someone who takes herself to be washing. But movements are not actions; and their classification as actions, even narrowly construed, depends on what the appropriate objects are towards which they are directed. One's actions being the successful actions they are depends on the appropriateness of such objects to them. So narrow-act taxonomy is made intelligible against the background of wide-act taxonomy, taxonomy which is object-dependent at least to the extent that it requires appropriate objects in the environment in which agents are embedded towards which they can successfully act. And given that this is so, the explanation of even narrow acts by states with narrow content is made intelligible against the background of explanation of wide acts by states with wide or broad content. Sue's activity of washing herself is made intelligible by her desire to make herself clean and her belief that by washing she will make herself clean, only if she also has beliefs about what it is appropriate to wash with. Her success in washing depends on 291
Cynthia Macdonald this. That is to say, her washing depends on this, since the taxonomy of her behaviour, narrowly construed, is not independent of what objects are appropriate to wash with. Even her unsuccessful attempts at washing, using inappropriate substances such as tar or mud, require this. For her unsuccessful attempts are ones in which she mistakenly takes herself to be washing. But if she takes herself to be washing, then she takes herself to be doing what agents do when they wash. Sue can only be mistaken about whether she is washing if she has correct beliefs about the washing, which require beliefs about what it is appropriate to wash with. These observations about successful actions and appropriate objects may seem insufficient to establish externalism. For externalism requires that the contents of at least some intentional states be u\dW\du&X.ion-dependent on factors beyond their bodies. Sue's water content must be a water content, not just a content that depends on the existence of some substance or other. And it is not clear that appeal to successful actions towards appropriate objects establishes such dependence. Consider Sue's twin. She successfully does with twater what Sue does with water. And twater is, on twin earth, appropriate to wash with. On twin earth, twater gets things clean. Appeal to considerations about successful actions towards appropriate objects fails to discriminate between Sue's behaviour and her twin's behaviour, narrowly construed. Does this not show that the dependency of actions and intentional content on appropriate objects is insufficient to discriminate between water and twater contents, so that individualism is still viable? One cannot respond to this by saying that whether an object is appropriate for a given activity depends on its empirically discoverable nature. The problem that the twin earth examples pose is that although water and twater have different natures, both are appropriate to wash with. So it looks as though such differences as there are between their natures cannot make a difference to the determination of intentional content. The problem arises because the twin earth examples are designed to keep firmly in place the ordinary, day-to-day role that certain objects, substances or phenomena beyond persons' bodies play in their activities and, correspondingly, in their thoughts, while varying in their natures in ways that are hidden to the naked eye. Since our day-to-day activities can be and often are insensitive to such differences in the natures of things that do not manifest themselves to the naked eye, it is not surprising that water and twater should play the same (narrow) role in Sue's and her twin's day-to-day activities. 292
Externalism and Norms However, I think that externalism can be defended in the face of this. The twin earth examples are effectively tests of antecedently held intuitions, as I have said. How they are to be interpreted, and what they establish, depends on the intuitions that prompt them. The intuition that the argument for externalism just given set out to defend is that water contents are water contents because our day-today descriptive and explanatory practices can only intelligibly explain subjects' successful actions, even narrowly construed, against the background of successful actions which take water as an appropriate object, where these explanations require the employment of contents that are water contents. Water actions are made intelligible by states with water contents. Unsuccessful attempts at water actions that take objects other than water are made intelligible in part by states with water contents. It is true that on twin earth, it is stipulated that twater plays the role in twin-earthians' day to day life that water plays on earth. Why then do Sue and her twin think thoughts with different contents? Because whether Sue's thoughts are water thoughts depends on her actual behaviour, in the actual context in which she is situated. Her counterfactual behaviour - what she would do in other circumstances, and in other environments - depends on this. If she were to be transported to twin earth, what would she do? She would wash with twater, drink twater, and so on. But it would not be appropriate for her to do so. Why? Because appropriateness is context-dependent, and the organism is part of the context.30 Sue's behaviour is appropriate on 30 Two objections might arise here. One is that what is appropriate behaviour towards an object depends in part on how the type of object involved in that type of behaviour is specified, and that, specified more generally (say, as 'the thirst-quenching, odourless, transparent, colourless liquid'), twater is appropriate for Sue to wash with, because on twin earth, the stuff which satisfies that description gets things clean. The other objection is that appropriate behaviour must be, as the functional behaviour is, capable of allowing for novelty in the range of objects to which such behaviour can become adapted. Creatures move around and may, in new environments, encounter objects of kinds that are distinct from those of the kinds to which their behaviours were initially adapted. It may thus be accidental that these objects serve the needs for which the behaviours were initially selected. Still, engaging with them might prove to be beneficial for these creatures, and so it may be functional for them to behave in the same way towards these new items as they did towards the old ones. Consider the first objection. Suppose that we are concerned to specify the function of the frog's tongue-flicking behaviour. On one view, the function is to catch small dark moving things. On another, it is to catch frog-food. How we specify the object toward which the frog behaves mat293
Cynthia Macdonald Footnote 30 - continued ters because it makes a difference to whether the frog is functioning biologically normally rather than malfunctioning when it flicks its tongue at black spots in its visual field. Similarly, it might be argued, for Sue and water/twater. On one view, appropriate behaviour for Sue is behaviour towards water. On another, it is behaviour towards the thirst-quenching, odourless, transparent, colourless liquid. How we specify the object towards which Sue behaves matters here too because it makes a difference to whether Sue is behaving appropriately when she washes with twater. Indeterminacy of function-specification is a problem in an environment where both specifications apply precisely because of its consequences for what would count as malfunctioning behaviour. And it is a problem in Sue's environment, since both ways of specifying water are satisfied by water. However, the considerations that may lead one to think that the correct specification of the frog's behaviour is the more general one do not apply with equal force to Sue's behaviour. In the case of the frog, the inclination to specify its functional behaviour as frog-food catching behaviour seems poorly motivated in the light of the fact that the frog's perceptual system seems to be sensitive only to small dark moving things in its environment. To attribute more specificity in functional behaviour to the frog than this would require us to view the frog as capable of seeing small dark moving things as frog-food. But nothing in its behaviour gives us good reason to suppose that the frog has this capacity. The situation is different for Sue and water. The frog's environment contains many things which count both as frog-food and also as small, dark and moving. However, Sue's environment does not contain many stuffs that are phenomenologically indistinguishable from water. Whereas in its actual environment, the frog flicks its tongue at many small, dark, moving things which may not be frog-food, Sue does not in her actual environment wash with many thirst-quenching, odourless, transparent, colourless liquids which may not be water. Whereas, in the case of the frog, we see no reason to specify its behaviour in the more specific way, in the case of Sue, we have no motivation for specifying her behaviour in the more general way. It all depends on the organism and what is in its actual environment. So, in the case of Sue, unlike the case of the frog, we do have a reason to specify her behaviour as appropriate behaviour towards water. And so we have a reason for taking her behaviour on twin earth towards twater to be inappropriate, although intelligible. We can make sense of such inappropriate behaviour because, although the environment of twin earth happens to cooperate with Sue, that it does is an accident. Now consider the second objection. Suppose that the correct way to specify the function of the frog's tongue-flicking behaviour is in terms of its goal in catching frog-food. Still, it might be argued, different things in different environments might count as frog-food. Thus, suppose that the frog were to be placed in a new environment, one where creatures of a different type than those to which the frog's tongue-flicking behaviour was originally adapted nevertheless served to nourish the frog. Would it not 294
Externalism and Norms earth because of the beneficial effects washing with water has in that environment. But on twin earth those beneficial effects are accidental for Sue: on twin earth it is an accident that twater is appropriate for Sue to wash with. It is no accident that on twin earth twater is appropriate to wash with for twin earthians. And so it is no accident that Sue's twin acts in ways made intelligible to twin earthians by twater contents. But Sue's actions will not be made intelligible by beliefs and desires of hers with twater contents — not, at least, independently of the fact that Sue's actions are based on misthen be functional for the frog to flick its tongue at these different creatures? I want to say here that the behaviour in the new environment, however beneficial to the organism it may be, is not thereby functional for the frog. The reason is that whether a behaviour is functional depends on the types of objects to which the behaviour was initially adapted. It was frog-food, not small dark moving things, to which the frog's behaviour was initially adapted, and for which that behaviour was selected. It was creatures of a certain type to which that type of behaviour was initially adapted and for which that behaviour was selected. So it was frog-food of a certain kind to which the frog's behaviour was initially adapted and for which that behaviour was selected. And that is why it is functional for the frog to flick its tongue now in the presence of that kind of frog food. That different organisms in another environment nourish the frog, so that the its tongue-flicking behaviour in those circumstances is beneficial for the frog, is fortuitous. It is just good luck for the frog that its new environment obliges its need for nourishment by supplying different, but satisfying creatures for it to eat. (See Ruth Millikan, 'Compare and Contrast Dretske, Fodor, and Millikan on Teleosemantics', in White Queen Psychology and Other Essays for Alice (Cambridge, MA: MIT Press, 1993), pp. 123-33, especially 125-31.) Happy coincidence between producer and consumer does not thereby make for functional behaviour. Similarly for Sue's behaviour towards twater. It does not follow from this that the frog's functional behaviour cannot be adapted to new things, and that these things cannot come to figure in the process by which a type of behaviour or trait is selected. They can. And those that do will thereby figure in the specification of objects towards which that type of behaviour is functional. But that they are objects towards which a type of behaviour is functional depends on their role in the selection process, and not vice versa. Similarly for Sue and her appropriate behaviour towards water. In responding to these objections in this way I am presuming a particular view of biological function, namely a causal-historical view, such as that advocated by Millikan in Language, Thought, and Other Biological Categories. It contrasts with ahistorical accounts, such as propensity accounts (see John Bigelow and Robert Pargetter, 'Functions', Journal of Philosophy 84 (1987), 181-96). 295
Cynthia Macdonald perceptions of twater as water. Given that Sue thinks water thoughts and given that such differences as there are between water and twater do not manifest themselves in the day-to-day role that these substances play in the activities of agents on earth and on twin earth, it is not surprising that Sue should wash with twater. It is not surprising; but nor is it true that Sue's act of washing on twin earth is made intelligible independently of her water thoughts.31 Is this response question-begging against the individualist, whose criterion for the taxonomy of behaviour and intentional content is counterfactual? I do not think so, and the reason connects the taxonomy of psychological kinds firmly to the taxonomy of functional kinds in biology. In biology, the taxonomy of functional kinds is both teleological and what one might call 'effect-sensitive' in a normative sense of the term. The camouflaging behaviour of this chameleon is camouflaging behaviour, not because it has camouflaging effects in this chameleon, nor even because it tends to have camouflaging effects in the majority of chameleons. It is camouflaging behaviour in this chameleon because this type of behaviour had camouflaging effects in a sufficient number of its ancestors to lead to the proliferation of chameleons which displayed that behaviour. That type of behaviour exists in order to have camouflaging effects in this and other chameleons, whether it does so or not. And it exists in order to have such effects precisely because its presence in ancestors led to the survival and so to the proliferation of chameleons which displayed this behaviour. Similarly for the deer's flight behaviour, the bee's dancing behaviour, and so on. In biology, teleology arises from the working of natural selection 31 As Burge puts the point, Imagine that a heart and an organ that pumps digestive waste (from a completely different evolutionary scheme) were physically indistinguishable up to their boundaries. Clearly they would be of two different biological kinds, with different causal powers, on any conception of causal power that would be relevant to biological taxonomy. Judging the heart's causal powers presupposes that it is connected to a particular type of bodily environment, with a particular sort of function in that environment. One cannot count being connected to such a body to pump blood as just one of many contexts that the heart might be in, if one wants to understand the range of its biologically relevant causal powers. It would show a serious misconception of biological kinds to argue that the causal powers and taxonomically relevant effects of the heart and its physical twin are the same because if one hooked up the waste pump to the heart's body, it would pump blood and cause the blood vessels to dilate; and that if one hooked the heart to the waste pump's body, it would move waste. ('Intentional Causation and Psychology', p. 227) 296
Externalism and Norms on instances of physico-chemical properties of organisms.32 Turning green in this environment just in this chameleon's camouflaging itself given that it inhabits a green environment and that instances of this type of behaviour in this chameleon's ancestors helped them to avoid predators and so to aid survival and proliferation of descendants. It is the success of instances of certain types of behaviours in the past that gives rise to teleology in functions, and this teleology persists even when the effects which instances of such behaviour now have regularly fail to occur. Causes are designed to bring out certain effects, even when they do not. Still the functional pattern remains, and still the chameleon displays such behaviour, in order to have camouflaging effects. So in biology, certain types of behaviour too aim at success: their having teleology just in their aiming at success. To be a functional kind just is to aim at success, and this makes functional behaviour very like action in this respect. In biology, the fit between behaviour and environment is non-accidental when successful because the cause — say, the chameleon's turning green — is designed to have a certain (functional) effect, namely, to match the colour of the environment. So too in the domain of intentional psychology. In successful action, the fit between behaviour and environment is nonaccidental when successful because the cause - contentful intentional states - is designed to have a certain (purposeful) effect. The source of the design in the two cases may not be the same, since in biology it is brought about by the process of natural selection. But the design itself - that the cause exists in order to bring out a certain type of effect - is present in both. This is what makes for teleology in biology, and it is what makes for teleology in the domain in the psychological. If this is right, then the non-accidental fit between activity and object in actual cases of successful action is required for the taxonomy of action itself, just as the non-accidental fit between behaviour and environment in successful behaviours is required for the taxonomy of functional kinds. In the biological case, successful behaviour depends on the actual environment in which it was selected: a chameleon placed in a pink environment cannot camouflage itself.33 32 This is the view of biological-functional kinds advocated by Millikan in numerous works. See particularly Language, Thought, and Other Biological Categories, and White Queen Psychology and Other Essays for Alice. It is a causal-historical (in contrast to a propensity) account of biological function. 33 Millikan uses the term 'Normal' (with capital 'N') to distinguish the biological-normal from 'normal' in the sense of 'average' or 'usual' or 'typical'. See Language, Thought, and Other Biological Categories. 297
Cynthia Macdonald And in the case of action, the success depends on there being appropriate objects toward which agents can at least sometimes act. Sue can get it wrong occasionally when she takes inappropriate substances to wash with. She can even get it wrong much of the time. But she cannot get it wrong all of the time. Beyond a certain point, we are no longer prepared to say: she washes. And getting it right or wrong means getting it right or wrong in her environment. And so what is appropriate is what is appropriate in her environment. Similarly for Sue's twin. And this is why a counterfactual criterion is inappropriate for the taxonomy of action and content. In the case of biology, teleology is there because of functions, and functions arise from the workings of natural selection on instances of physico-chemical properties. Why is there teleology in the case of action? Here there is teleology because there is intentional content. Contentful kinds imbue behaviour with purpose. Behaviour counts as intentional, hence as action, only if it is caused by states with intentional content. But if so, then they aim at success precisely because their contentful causes themselves aim at success. Without teleology in the contentful cause, there is no teleology in the intended effect. This is not, of course, to say that without intentional states there would be no teleology in human behaviour. There would be teleology because human beings are biological creatures. But the teleology would not be the teleology of intentional behaviour. Movements would remain purposeful, but they would not thereby be intended, and so would not be actions. IV Conclusion This completes the argument for externalism. It remains to consider some of its consequences. The externalism argued for here is distinctive in being essentially normative. The objects (etc.) upon which narrow-act taxonomy depends are ones that are appropriate to those actions. As I remarked at the end of the last section, this normativity lies firmly in the mind. It lies in the essentially teleological nature of contentful kinds. However, for reasons which should be apparent from the comparison of psychological kinds with functional kinds in biology, I do not think that this teleology can exist independently of the actual environments in which agents are embedded. So what counts as an appropriate object towards which an agent can successfully act is not independent of the actual, empirically discoverable nature of that object. What makes Sue's washing with water a successful act is not independent of its being a washing with water; and water's 298
Externalism and Norms being an appropriate object to wash with is not independent of the fact that it is, unlike hydrochloric acid, H2O. Hydrochloric acid does not have a nature such that it is appropriate to wash with, nor does mud, nor tar. Given what has just been said, it would not be rational for Sue to attempt to wash with these substances. However, it would be intelligible for Sue to wash with twater on twin earth, since we can make intelligible why she might think it appropriate to do so, even though it isn't appropriate for her. For twater is phenomenologically indistinguishable from water and also gets things clean. Given this, it is understandable that Sue should do something that is inappropriate for her but appropriate for twin earthians. It is understandable, and perhaps also rational, in much the same way that a person who misperceives salt as sugar and pours salt in her tea can be seen to be behaving rationally because of this misperception. But is it not thereby appropriate, given that appropriateness is context-dependent and the individual is part of the context. There is slack between what is appropriate in the environment of twin earth and what is appropriate for Sue. This is not to say that whether an object or substance is appropriate to a certain activity reduces to, or can be determined only by, its empirically discoverable nature. It is true that whether a given object or substance is appropriate for a given activity depends in part on its empirically discoverable nature; but different natures can be equally appropriate for the same type of activity. On earth it is appropriate to wash with water; but it can also be appropriate to wash with sand. Apples are appropriate objects to eat, but so are walnuts and mushrooms. Because objects with different natures can be equally appropriate to a single activity, there is no telling in advance, there is no a priori limit on, what object or objects can be appropriate for a given activity. Appropriateness depends on actual effect - for example, in the case of washing, getting things clean - and this in turn depends in part on the nature of the objects with which one engages when one acts. But different objects can have the same effect. And so different objects can be appropriate for a given activity. It is appropriate to wash with water, and with sand, but not with tar, because water and sand gets things clean, but tar does not. However, the similarity between contentful intentional kinds and functional-biological kinds ends here. Specifically, the attribution of contentful kinds is, whereas the attribution of functional kinds is not, sensitive to both the perspective of others and the perspective of the subject. This makes the norms that govern the attribution of intentional content to subjects' states answerable both to the perspective of those subjects and to the perspective of others. There is 299
Cynthia Macdonald no analogue of this dual-perspective constraint on functional taxonomy in biology. Failure to appreciate this can lead to too close an assimilation of the psychological to the biological, with undesirable consequences. For example, Ruth Millikan has claimed that because externalism is true, a subject who is ignorant of the factors that determine intentional content might fail to recognize that two beliefs with the same content have the same content, and as a result hold contradictory beliefs, one the negation of the other. Many externalists would be prepared to concede this point.34 But she goes on to infer from this that the norms that govern rationality itself lie beyond the individual subject, in much the same way that the determinants of functional behaviour lie beyond the biological organisms that display it. She holds that externalism has the consequence that whether humans are rational is not determined by, or even partly answerable to, subject's perspectives.35 34 See, for example, Tyler Burge, 'Intellectual Norms and the Foundations of Mind', Journal of Philosophy 83 (1986), 697-720. 35 Thus she says, it is implicit in contemporary 'externalist' accounts of the contents of thought that what is consistent versus inconsistent, indeed, I will argue, what is rational versus irrational, is not epistemically given to the intact mind. ('White Queen Psychology', in White Queen Psychology and Other Essays for Alice, p. 281) And similarly, If the White Queen is right, then that Alice has a coherent system of thought, that she possesses, for example, only one thought of each semantic kind, and hence that she thinks in accordance with laws, say, of rational psychology, depends on a felicitious coordination between Alicethe-organism and Alice's environment. It depends, in fact, on much the same kind of felicitous coordination that constitutes Alice's thinkings of true thoughts; rationality fails to be in the head in the same sort of way as does truth. (Ibid., p. 285.) The illusion that modes of presentation will help save logical possibility also rests on a failure to see that rationality pivots essentially on referential content, or Bedeutung, and not at all on mode of presentation, that rationality cannot simply be lifted up and attached to mode of presentation. The capacity to reidentify content but only under a mode is a restriction on rationality, a lessening of rationality, not a removal of rationality into an inner and safer sphere. (Ibid., pp. 283-4.) Millikan apparently thinks that that rationality, like content itself, is world-involving, shows that it lies beyond the subject altogether ('rationality pivots... not at all on mode of presentation', 'rationality fails to be in the head in the same sort of way as does truth'). I deny that content exter300
Externalism and Norms But I deny that externalism has any such consequence. It is true that externalism implies that subjects might think thoughts with contents they imperfectly grasp. And since they might, they might mistakenly think that two thoughts have the same content when they do not, or that they have different contents when they do not. Fallibility in knowledge of one's own thoughts of this kind is indeed a consequence of externalism. But I do not see that the norms that govern rationality do not thereby operate 'from within' the individual subject. For these norms operate in epistemic ways, in ways that make the behaviour of agents intelligible, not only to others, but also to themselves. And it is a constraint on the attribution of content by others that such attribution respect the agent's perspective. In short, the factors that make a certain content the content it is do not thereby determine the acceptance or rejection of that content by a subject, or the patterns of reasoning in which that subject might engage with regard to that content. Externalism a metaphysical view about the factors that help to determine intentional content. But rationality, and the norms that govern it, is an epistemological matter. So it does not follow from the fact that externalism is true, hence that the determinants of intentional content lie beyond the individual, that the determinants of reasoning and behaviour lie beyond the individual also. On the contrary, it is plausible to maintain, in the face of externalism, that these norms are accessible to and operate within the subject, and guide the very formation, rejection, and assimilation, of contentful states.36 36 So I am recommending a combination of metaphysical externalism with regard to the determinants of intentional content and epistemological internalism with regard to the norms that govern rationality. Burge seems to pursue a similar strategy. See 'Our Entitlement to Self-Knowledge', Proceedings of the Aristotelian Society 96 (1996), 91-116. nalism has this consequence. That rationality is world-involving does not thereby show that it does not depend in any way on the perspective of the subject, and so does not show that the norms that govern rationality lie beyond the individual. That what is thought about when thinking a content lies beyond the individual does not show that how it is thought about is not also involved in thinking rationally. The rejection of narrow content does not force one to reject any role for the subject to play in rationality; nor is it a 'pernicious Cartesianism' to insist on the importance of that role. Without it, it is difficult to see why subjects should be critically reflective thinkers, and what role critical reflection might serve in an individual's psychology. 301
Mind, World and Value MICHAEL MORRIS Naturalism is the dominant philosophy of the age. It might be characterized as the view that the only real facts are facts of natural science, or that only statements of natural science are really true. But perhaps this scientistic formulation underestimates the depth and everydayness of the dominance of naturalism. More informally, we might say that naturalism is the view that the world is a world of natural objects and natural phenomena, that the only properties of these objects are natural properties, and the relations between them are all natural relations - in short, there are only natural facts, natural truths. There are obvious questions to be raised about the coherence of naturalism (for example: can the truth of naturalism really be supposed to be a natural truth?); but I shall not dwell on these here. I want to put naturalism into question in a different way: by suggesting an alternative to it in the philosophy of mind which is rich enough to stop naturalism seeming compulsory. It is often simply assumed that a good account of the mind must be naturalistic. What does this rule out? Sometimes it seems to be supposed that all that is ruled out is an account which presents the mind as something supernatural. (A particular kind of Cartesian bogeyman will be imagined.) But this is a shallow contrast which is in danger of blinding its opponents to the strength of naturalism's own commitments. For the concept of the supernatural is the concept of something which is of the right general kind to be given a naturalistic explanation - for which there ought to be a naturalistic account — but for which no naturalistic explanation is possible, for reasons other than the mere inadequacy of the minds of the explainers. A supernatural thing is a miraculous natural thing. The supernatural is an alternative to the natural conceived of from within naturalism. A more fundamental contrast is suggested by the thought that it is the business of natural sciences to describe how things are, and perhaps how they must inevitably be, but not how they should be. This ties in with Kripke's famous objection to dispositionalist accounts of meaning, that they explain what I will say rather than 303
Michael Morris what I should say;1 and with an older objection to psychologism in ethics. If this is right, what is distinctive of naturalism is that it finds no place for value in the world. The natural facts exclude facts about value; the world is value-free; natural science is dispassionate and value-neutral. This suggestion is not uncontentious, since it rules out both a traditional kind of ethical naturalism, and an alternative (perhaps Aristotelian) conception of nature as essentially value-rich.2 But I shall adopt it here without argument, because it seems to me to provide a fair characterization at least of an orthodox kind of naturalism, and because it suggests a way of developing a clear alternative to orthodox naturalistic theories of the mind. II We will get a clear alternative to orthodox naturalism about the mind, if we develop an account which makes explicit reference to value. Where should we begin? We might take a hint from that thought of Kripke's: he wanted an account of meaning to explain what we should say, rather than what we will say. Kripke was not concerned with etiquette, but with right answers: that is, with truth. So a first suggestion might be: truth is what we should say. This is uncompelling for two reasons: first, it might sometimes be right not to say what is true (for reasons of diplomacy, or just politeness); secondly, it cannot ever be required that we should say everything that is true, because there are absurdly many truths (start with simple addition and keep going). So here is a revised suggestion: it is distinctive of truth that one should only believe something if it is true. And conversely: it seems essential to belief (as the term is used in the analytical tradition), that it is a state which one should not be in unless what is believed is true. In this way we get the beginnings of an evaluative account of truth and of the mind at the same time. There will be an immediate objection: might it not be the case that it was psychologically necessary for someone to believe a falsehood? The answer to this is that is less than ideal to be in a condition in which it is psychologically necessary to believe a falsehood. So the revised suggestion could be revised again: one should be in a condition such that one only believes something if it is true. The 1 S. Kripke, Wittgenstein on Rules and Private Language (Oxford: Blackwell, 1982), p. 37. 2 See, e.g., J. McDowell, Mind and World (Cambridge, MA: Harvard University Press, 1994), pp. 78-84. 304
Mind, World and Value important point is that such counter-examples do not have any tendency to show that value can be dispensed with here. The attraction of locating the value that naturalism must miss in truth, specifically, is that it seems reasonable to hope that this will explain the importance of the normative in reasoning and thinking in general. Classical logic is based on the thought that truth preservation is the essence of validity. Almost every epistemology holds that what is distinctive of knowledge is that it provides some kind of guarantee of truth. The difficulty with the suggestion is that it is hard to see what kind of value truth might be: why should one only believe the truth? What is it about truth which matters so much? One thought might be that the value of truth lies in the evolutionary advantage which true belief confers. An obvious difficulty is that it is hard to see why truth should be especially advantageous: indeed, it is easy enough to imagine circumstances in which false belief might help. But there is a more fundamental problem with the suggestion, for our present purposes: it hardly provides an alternative to a naturalistic account. It is not merely that evolutionary accounts are commonly championed in the name of naturalism. The important point is that an evolutionary account makes no essential appeal to value. Indeed, the whole purpose of evolutionary explanation is to show how to dispense with the value-rich conception of the world which we so easily bring to our descriptions of nature.3 (It is tempting to suppose that, by a kind of reversal of the general tendency of evolutionary explanation, evolutionary theories of the mind have seemed attractive just because they have appeared to license an appeal to value; whereas in fact they undermine it.) Another suggestion might be this: the truth of one's beliefs is essential for successful action. There are a number of difficulties with this. First, this account cannot be deep enough. For successful action is presumably just action which achieves what one wants: we seem to need to appeal to an unexamined notion of desire. Secondly, it is not obviously true that the truth of one's beliefs is necessary for one to achieve what one wants: convenient accidents would do as well. Thirdly, the truth of our beliefs cannot have a merely instrumental value in the satisfaction of our desires: for without any true beliefs it is hard to see what content our desires could have at all. Fourthly, action is only a part of the range of behaviour in which belief can be manifest, and in which the truth of our beliefs seems to matter. Noticing the uncritical use of the notion of desire in this last suggestion might prompt us to wonder whether we could offer an 3 For support for this point, see J. Fodor, 'A Theory of Content, I' in his A Theory of Content (Cambridge, MA: MIT Press, 1990), p. 79. 305
Michael Morris explicitly evaluative account of desire to parallel the evaluative account of belief which was offered before. And we can. For it is surely true that one should only desire something if it is good, or at least not bad. There are two clear differences between this account of desire and the corresponding account of belief. First, this connection with value does not suffice to single out desire: for it is also true that one should only approve of something, or hope for something, or wish for something, or like something, or enjoy something, or be pleased at something, if the something is good, or at least not bad. Further distinctions between all these attitudes need to be drawn. The second difference between the account of desire and the account of belief lies in the nature of the value which is appealed to. In the case of desire, it seems clearly to be a kind of value which is at least continuous with the moral. It may not be appropriate to describe every issue about what one should want as a moral issue, particularly if the concept of the moral is reserved for a particular kind of social institution; but it seems clear that the question of what one should want will always fall within the range of that larger question, which has been taken to be the founding question of moral philosophy, of how one should live.4 This provides us with a contrast between true belief and good desire which should enable us to get clearer about the value of truth. But this link with the moral is likely to frighten some enough to drive them to an alternative account of the difference between beliefs and desires, in terms of 'direction of fit'. Here is Platts providing an orthodox characterization (even though he is not sympathetic to the view himself): Beliefs aim at the true, and their being true is their fitting the world; falsity is a decisive failing in a belief, and false beliefs should be discarded; beliefs should be changed to fit the world, not vice versa. Desires aim at realisation, and their realisation is the world fitting with them; the fact that the indicative content of a desire is not realised in the world is not yet a failing in the desire, and not yet a reason to discard the desire; the world, crudely, should be changed to fit with our desires, not vice versa.5 The first sentence provides a characterization of belief entirely in 4 In this respect I am distancing myself from Kant, who held that there was a fundamental distinction between moral motivation and desire. It is unclear how Kant could accommodate the notion of what one should want. 5 M. Platts, Ways of Meaning (London: Routledge and Kegan Paul, 1979), pp. 256-7. 306
Mind, World and Value accord with the suggestion I have offered. The account of desire which follows is just false, however. Desires do not aim at realization. Only the person whose desires they are aims at their realization. But this is just to say that she desires their realization. And since their realization is just the object of the desires, this amounts to no more than saying that she desires what she desires,6 which provides no account of desire. And it is false that the world should be changed to fit with our desires: if our desires are wicked, the world should not be changed to fit with them. Anyone tempted by the first sentence of Platts' characterization should realize that a truly comparable account of desire would be the one I have offered. And this will only suggest that there is a difference in 'fit' (though hardly direction of fit) if desiring what is good is not counted as fitting the world; which will only be if the world is not allowed to contain values; that is, if we accept an orthodox naturalism. Moral evaluation is largely concerned with behaviour in the widest possible sense: actions, certainly, but also certain kinds of non-intentional response, and certain tendencies of character. These are things that are characteristically explained by means of beliefs, desires, hopes, wishes and all the other propositional attitudes. (The concentration on action in moral theory matches the concentration on desire among 'motivational' attitudes in the philosophy of mind.) How does the moral (when it is clearly moral) evaluation of behaviour relate to the motivational attitudes of the person whose behaviour it is? The simple answer is that the moral evaluation of behaviour is just the evaluation of the motivational attitudes which explain the behaviour. An action is good insofar as it manifests a good desire. A glance is kind and considerate insofar as the thought 'behind' it is kind and considerate. A decision is morally good insofar as it is made for the right reasons. This conflicts, of course, with familiar consequentialist ways of evaluating behaviour; but the contrast with consequentialism is one which we will want to maintain in the construction of an alternative to naturalism. Evaluating behaviour simply in terms of its consequences, whether actual or merely likely, is no longer to think of behaviour as a special category, the object of a special kind of evaluation. Behaviour becomes just another causal factor, of no evaluative interest in itself. This attitude to behaviour is hard to square with the alternative to naturalism which I am suggesting. For the suggestion is precisely that what is distinctive of the 6 Someone may suspect that this argument involves a fallacious application of Leibniz's Law within an intensional context. I think there is no fallacy: the realization of a desire must be thought by the desirer to be the object of the desire. 307
Michael Morris mind, and hence of behaviour (as opposed to mere movement), is that it can only be understood in terms of value. Once we see that the evaluation of motivational attitudes is the same as the (moral) evaluation of the behaviour which they explain, we can give a characterization of an evaluative approach to the mind which makes it vivid. Behaviour simply strikes us as being appropriate for such evaluation.7 We simply see affection in a touch, cruelty in a sneer, aggression and despair in a slouch. Any sensitive description of behaviour must bring out more clearly what there is to be admired, what to be loved, what to be feared, what to be hated, what to be pitied in it: a sensitive characterization may not make unambiguous evaluation any easier, but it will make clearer what is at issue in the evaluation of it. If the evaluation of motivational attitudes is the same as the moral evaluation of the behaviour which they explain, it should also be true that the evaluation of beliefs, as true and false, is the same as a kind of evaluation of the behaviour which they explain. And we might hope that the evaluative approach to belief might be made vivid in the appropriate evaluation of behaviour. Here is a simple example. Someone walks upstairs reading a magazine. At the top she takes one step too many and stumbles. The stumble is comic and pathetic. This shows that the person thought there was another step. This belief was false. The falsity is manifest to us in the ludicrousness of the stumble: finding it ludicrous just is acknowledging the falsity of the belief which it manifests. This shows the nature of the value of truth in belief. If a belief is false, then the behaviour which depends on it is in a certain sense empty, or pathetic, or futile: it is, as it were, all for nothing. Our sense of the comic and the tragic often depends upon this. It may be, of course, that the person gets what she wants, luckily, anyway. But this does not detract from, even if it compensates for, the foolishness created by false belief. Here we can see the truth or falsity of belief in behaviour. But what is the belief itself? If someone acts on the belief that p, one will be able to recognize what she does as foolish or otherwise according to whether or not it is false that p. As a provisional suggestion, I propose the simplest possible extrapolation of this as an account of belief: someone believes that p at a certain time just in case it is legitimate, in virtue of how she is at that time, to count what she does as foolish or otherwise according to whether or not it 7 The idea of such rich perception is to be found in many places in McDowell's work: e.g. 'On "The Reality of the Past'", in Action and Interpretation, ed. C. Hookway and P. Pettit (Cambridge: Cambridge University Press, 1978), p. 140. 308
Mind, World and Value is false that p.* This does not mean that whatever she does at that time must be evaluated according to whether or not it is false that p; not even if her action is of a general type which might be explained by the belief that p. It means, simply, that if we were to assess what she does in these terms, we would be considering her on her own terms. Someone may complain that it is now very uncertain whether or not someone believes something if she is not acting on that belief. But this is an unreasonable complaint: it often is uncertain whether someone believes something if she is not acting on that belief. Did you believe that 'consequence' is a longer word than 'theory' before I mentioned it? An account in the same style can be provided for desire. If someone acts on a certain want, then what she does is bad or otherwise, according to whether or not what she wants is bad. And the simple proposal is: someone has a desire at a certain time just in case it is legitimate, in virtue of how she is at that time, to count what she does as bad or otherwise according to whether or not that is a bad thing to want. Here we have the skeleton of an evaluative conception of mind. We can now paste on some flesh, before concluding with some obvious worries. Ill The character of this explicitly evaluative conception of the mind is most clearly seen in the account it provides of the kind of explanation which is given when we explain behaviour in terms of states of mind. The dominant view of this kind of explanation is that it is a species of causal explanation. This dominance is not surprising: causal explanation is what natural sciences - in particular physics - are peculiarly suited to provide, and the conception of the world as the world of natural science is the dominant conception.9 But for all its dominance, the view has no compelling arguments to support it. There has been no significant advance on Davidson's two considerations in favour of the causal theory.10 One is that unless we sup8 This definition is provisional because it does not yet secure the right degree of intensionality for belief contexts. There are several possible revisions, one of which is suggested in my The Good and the True (Oxford: Oxford University Press, 1992), chapters 12-13. 9 Peacocke actually defines naturalism about explanation as the view that 'any explanation of an event or temporal state of affairs is a causal explanation': A Study of Concepts (Cambridge, MA: MIT Press, 1992), p. 127. 10 D. Davidson, 'Actions, Reasons, Causes', in his Essays on Actions and Events (Oxford: Oxford University Press, 1980). 309
Michael Morris pose that beliefs and desires (for example) cause behaviour, we cannot make sense of the difference between doing something because one has a certain belief, and merely doing it while having that belief. But although this is a reasonable ad hominem objection to some particular (rather vague) accounts of this kind of explanation - which was actually Davidson's original point — there is no ground for proposing it as a general claim: I shall suggest a different way of marking the distinction. (And in fact, as I shall argue, there is room for doubt about whether Davidson's version of the theory actually does provide an account of that 'because'.) The second of Davidson's considerations might be thought to support the first: we have no better model of the explanatory connection than the familiar causal one, so we should adopt the causal one. This argument (if it deserves the title) is doubly parasitic on the dominance of naturalism. First, in that naturalism makes the causal account seem the natural choice. Secondly, in the basic assumption that any model is better than none: this is legitimate only when the adoption of the model plays an instrumental role in the pursuit of some further project, as in the technological application of science; not when the goal is just the understanding provided by that model itself, unless we suppose - as some versions of naturalism indeed do - that all such understanding must have a merely instrumental value. In any case, this second consideration of Davidson's is vulnerable to the provision of an alternative model; which I shall supply. The causal theory typically brings with it a certain conception of the nature of the behaviour which states of mind are cited to explain, as well as a particular view of the nature of the explanation involved. It may be that not every causal theorist has felt committed to these things: I suspect that that is because the consequences of adopting a causal theory have not been thoroughly thought through. But my concern is to present what is distinctive about an explicitly evaluative approach to the mind; and for this it is enough to contrast it with the views typically associated with the causal theory. The causal theory is typically associated with a restrictive conception of the behaviour which states of mind are cited to explain: in the simplest case, the behaviour is thought of as just bodily movement. The restrictiveness follows from a certain (natural, and Humean) conception of causation: the nature of the cause is not intrinsic to the effect. It is possible to know everything about the intrinsic nature of an effect without being able to infer from that anything about the intrinsic nature of its cause." If we apply this principle to the case of explanation by means of states of mind, 11 This notion of intrinsicness is here left unexplained. It needs to be explained as part of a fuller account of causation. 310
Mind, World and Value then we have to suppose that the behaviour which states of mind are cited to explain does not in itself require that it be explicable in terms of states of mind. Putting it crudely, the behaviour is not intrinsically behaviour at all: it is movement, perhaps, or something else whose intrinsic nature is characterizable without recourse to any mental terms. We then count it as behaviour in virtue of having found something mental to be its cause. It is perhaps this commitment of the natural conception of causation which leads to the view that states of mind are hidden from us, and cannot be seen in behaviour. Any view which supposes that the mind lies somehow behind the behaviourism seems to be a distinctly causal view, and it certainly requires that states of mind are not unproblematically visible in behaviour. A similar result is reached from the thought that causes bring about events, or happenings (causes make things happen). To see something as a happening is to bring it under a general scientific view of the world. To say that actions are events is to say that there is nothing in their intrinsic character which sets them apart from the world of natural occurrences.12 Actions then become events with a particular distinctive kind of cause; but as events, they are intrinsically on a par with earthquakes, punctures and the gradual decay of buildings. To think of actions as events is to think of them as intrinsically (that is, in respect of their intrinsic nature) agentless. Any causal theory which accepts this sterilization of behaviour (and I suspect that all should, if there is to be any point in insisting that causes are involved) is led to an odd conclusion. The theory aims to provide an analysis of a certain explanatory relationship, between states of mind and behaviour. But what it suggests is a more fundamental description of that relationship — that it holds between states of mind and certain happenings or events — which would never normally be recognized as a description of that relationship at all. If anyone were to say that a particular state of mind made something happen, the something would not be the kind of thing we explain by means of belief in the basic examples of the kind of explanation we are concerned with. (Consider, for example: 'The managing director's uncritical confidence in the value of its product made the company collapse.') A similar forgetting of what the causal account was designed to explain is evident in another feature of many versions of the causal 12 Davidson's philosophical-logical argument for the claim that actions are events (which is questionable anyway) does not actually require more than that actions be entities of a much more general sort, over which one can quantify. See his 'The Logical Form of Action Sentences', in Essays on Actions and Events. 311
Michael Morris theory. In Davidson's account, for example, the status of states of mind as causes of the events which are described as actions is ensured only by the fact that both states of mind (or 'mental events', if you prefer) and actions are physical events, and as physical events fall under genuine 'homonomic' causal laws.13 The problem with this is that the causal account is supposed to be an account of the 'because' in such claims as 'She winced because she thought the gas was about to light'; and this 'because' is vindicated, if it is, by the fact that this thought, described as a thought with that content, is explanatory of the behaviour described as behaviour. If one were to appeal here to the familiar Davidsonian separation between causal relations and causal explanation,14 that would serve only to emphasize the fact that this version of the causal theory fails to explain what it was designed to explain. A similar difficulty infects certain traditional computational versions of functionalism. One argument for a representational theory of mind has been that it is not content which is causally efficacious in computational systems, but the non-intentional properties of the bearers of that content.15 But this, once again, is just to deny the genuine explanatoriness of those features whose presumed genuine explanatoriness led us to a causal theory in the first place. The theory undercuts its own motivation. An obvious response to this is to liberalize the conception of causal efficacy which creates the problems. If this goes far enough, it will be unclear what point there is in insisting that the explanatory connection between states of mind and behaviour is causal. And the central motivational difficulty remains: it is unclear how calling the connection causal helps us to understand what is explanatory about it. At this point it is hard to find any reason to adopt a causal theory of the mind apart from a blanket commitment to naturalism. A generally naturalistic conception of the kind of explanation involved here is evident when it is characterized as 'folk psychology'. The picture is of a primitive people's first attempt at science. The offensive condescension of this view is acknowledged in the amendment of the label to 'commonsense psychology', but the naturalistic commitment of the original description is left unrevised. To count the claim that a certain person maintained a cheerful demeanour because she thought it would cause most irritation as psychological at all is already to count terms like 'thought' and 13 See Davidson, 'Mental Events', in Essays on Actions and Events. 14 See Davidson, 'Causal Relations', in Essays on Actions and Events. 15 See, e.g., J. Fodor, 'Fodor's Guide to Mental Representation', in his A Theory of Content, 19-24. 312
Mind, World and Value 'desire' as theoretical terms of a commonsense science, and their subject matter as one in which science alone can provide authoritative advances in understanding. The idea will typically be that the concepts of thought and desire are integral to a project which is fundamentally oriented towards prediction, just as the physical sciences are. Insofar as the project is oriented towards prediction, its value will primarily be instrumental, although we can become interested in thoughts and feelings for their own sake - out of curiosity, for example. This psychological conception of our understanding of each other is required, I think, by the causal theory of the explanation of behaviour; but not conversely.16 For there can be instrumentalism in psychology, just as in any scientific field (even if its coherence in psychology is open to doubt); and an instrumentalist account might be seen as an alternative to, rather than as an analysis of, causal explanations in the field in question. Here is a quick sketch, then, of the kind of view of the explanation of behaviour which will be congenial to naturalism. When we say what someone thinks or wants, we are adopting an attitude to them which is fundamentally theoretical, and the terms we use are constrained fundamentally by their efficiency in the prediction of behaviour. On the dominant view, we are concerned with the diagnosis of the causes of events which are not in themselves intrinsically mental, but which we can come to see as mental in virtue of finding that they have mental causes, or which we initially presume to be mental in the expectation of finding that they have mental causes. Quite what role the content of thoughts and desires has in this explanation is left obscure. 16 I here blithely ignore the debate about 'simulation' accounts of our ascription of propositional attitudes. (See J. Heal, 'Replication and Functionalism', in Language, Mind and Logic, ed. J. Butterfield (Cambridge: Cambridge University Press, 1986).) There are two reasons for not considering it here. First, it is unclear to me that the 'simulation' or 'replication' theory offers a radically different account of the nature of states of mind from that provided by an explicitly theoretical conception, rather than simply a different account of the way we usually think about them. (What is it that is simulated, after all?) Secondly, Heal herself accepts (see, e.g., J. Heal, 'Simulation and Cognitive Penetrability', Mind and Language 11 (1996), 48) that a 'simulation' account of our thinking about other people's minds is not incompatible with our having a 'prototheory' of the mind; which is enough to bring the 'simulation' account within the scope of the characterization of typical naturalistic theories provided in the text. 313
Michael Morris IV The account of the explanation of behaviour suggested by an explicitly evaluative approach differs from this in every particular. In the first place, it is hard to see how it can be causal.17 Recall that I suggested that for someone to believe something at a particular time is simply for it to be legitimate at that time to count what she does as foolish if what she believes is false. There is surely no conception of causal relevance so relaxed as to permit its being legitimate to count what she does as foolish if something is false to qualify as causally relevant to someone's doing something. But this disqualification brings immediate advantages. To begin with, much of the motivation for the sterilization of behaviour is removed. Behaviour does not need to be thought of as in itself neutral with respect to whether it is the behaviour of a thinking agent. Nor need it be supposed to be just a matter of happenings. We should think of it as essentially and intrinsically valuerich. It is of the very nature of behaviour that something rides on it, that it is capable of being foolish or wicked. This is, of course, how it strikes us, and how we always describe it. We see gestures as tentative, gentle, bold and rough: these categorizations demand our evaluation. It is not that there need be any simple correspondence between one particular adjective and approval, or between another and disapproval: it is rather that they are categories for which it is essential that value is an issue. Within an evaluative conception of the mind, this is to say that the behaviour itself is intrinsically the behaviour of a thinking agent: the beliefs and desires which may be cited in explanation of the behaviour are internal to the behaviour itself. This seems to entitle us to say that what people think and feel, and that they think and feel at all, is in principle simply perceptible in their behaviour.18 There are two principal reasons for resisting this: acceptance of a causal theory of the mind; and a certain view of the kinds of things which are in principle available to perception. The first reason has been removed, and the second surely depends upon a restriction of what is strictly perceivable which is traceable through the empiricists to Descartes, and seems ultimately to depend upon naturalism. But the significance for our present purposes of recognizing that thoughts and feelings can be in principle simply perceptible lies not in the response it provides to the sup17 A more direct argument against the causal account, on the ground that it is incompatible with freedom, is provided in my The Good and the True,