A Defense of Dualism


By "dualism" I mean the thesis that the mind and its contents are radically nonphysical, that is, that they are neither themselves physical, nor the logical product of anything physical, nor, except causally or nomologically, dependent on anything physical.1 It is worth noting that dualism in this sense, is compatible with monism. For while a dualist cannot hold that reality (or as I would prefer to say, ultimate reality)2 is wholly physical, he can hold that it is wholly mental. He can even hold this view while accepting what for the purposes of this discussion I shall pretend to accept,3 the reality (ultimate reality) of the physical world. This possibility, however, will not concern us till the very end.

It is nowadays generally assumed that dualism has been discredited. Even those who see problems for the alternatives usually are unwilling to countenance dualism as a possible solution. I believe that dualism is correct, and the purpose of this chapter is to defend it in two ways. Thus in section I, I shall consider the alternatives to dualism and try to show why they are unacceptable (some of the arguments here are familiar, but they are worth rehearsing). And in section II, I shall consider some of the main objections to dualism and try to show why they are ineffective. In addition to this twofold defense, there is a short postscript (section III), in which I consider the possibility of mentalistic monism.

I. The Alternatives to Dualism

i. Conceptual Materialism. The most radical alternative to dualism is "materialism," the thesis that mental facts (or states of affairs) are nothing over and above physical facts (or states of affairs). This thesis can be interpreted in two ways. I shall start by considering the most straightforward interpretation: all mental truths (true propositions about the mind) are logically deducible from physical truths (true propositions about the physical world). I shall call this version of the thesis "conceptual materialism." The label "conceptual" is not intended to signify that the materialist thesis is offered as a conceptual (a priori) truth but only to signify the deductive (a priori) character of the logical relation between mental truths and the physical truths from which they allegedly flow.

As defined, conceptual materialism is a very general thesis: it leaves open a wide range of possibilities. But, typically, what the conceptual materialist has in mind is something like this. Take a human organism O at a certain time t. Suppose, for the sake of argument, we have a complete description of the physical structure of O at t and know all the physical laws of nature. From this information we can, it is claimed, deduce a complete description (or as close to complete as we care to make it) of how O is designed to function in different physical environments.4 Not all aspects of this total functional organization are relevant to the mental activities of O. So we next select those that are. For example, we select as one mind-relevant item the functional organization of O's visual system, noting how, through the reception and processing of different light inputs, this system equips O to behave in discriminatory ways with respect to the arrangement of colors in his environment. Likewise, we select as another mind-relevant item the functional organization of O's hunger system;, noting how, by the monitoring of nutrient levels and stomach condition, this system disposes O to behave in food-seeking ways (partly under the guidance of the visual system) when appropriate. Once we have selected all the relevant items and, integrating them, have a complete description of the total mind-relevant organization of O, we then consider the physical state of O at t and work out its mind-relevant significance with respect to this organization. From this, it is claimed, we can deduce a complete description of O's state of mind at t.

It is, typically, something like this which the conceptual materialist has in mind—the mental description of an organism being deducible from its functional description and its functional description being deducible from its physical description and physical laws. But my objection to conceptual materialism applies quite generally. The objection, which is a very familiar one, is that such materialism fails to do justice to the nature of conscious experience. There are many ways of putting this objection. Here is one way which I find compelling. Consider the situation of someone who is congenitally blind. Could such a person derive from physical information about the sighted a full knowledge of the character of visual experience? Could he deduce from the physiology and functional organization of the sighted human organism (and any other physical truths that might be thought relevant) a full knowledge of what it is like, experientially, for such an organism to see? It seems clear that he could not. But why not if conceptual materialism is true?5

Does the conceptual materialist have any replies? Well, in the first place, he might say that what prevents the blind man from deducing the experiential conclusion is that being congenitally blind, he is not conceptually equipped to understand it. But this is clearly hopeless. If conceptual materialism were correct, there would be nothing to prevent such a person from acquiring the requisite experiential concepts, assuming he was sufficiently intelligent and was conceptually equipped to understand the relevant physical truths. The presumption that his blindness prevents such acquisition is already a tacit acknowledgment that the physical truths do not implicitly specify, in full, the nature of the sighted organism's experience. In any case, it is not essential to the example that the blind person should lack the requisite concepts. We can suppose that his visual cortex is sufficiently operational to give him, from time to time, visual experiences of a rudimentary kind, though ones which are not, and which he does not take to be, perceptions of his physical environment. These experiences could equip him to understand the experiential conclusion. But it still seems clear that he could not deduce it from the physical premises. For the physical information about the sighted, however comprehensive, would not logically establish that their experience was of the sort to which his experiential concepts, furnished by his own visual experiences, applied.

Alternatively, the conceptual materialist might claim that the blind man is not conceptually equipped to understand all the relevant physical premises. Thus he might argue that one of the physical truths which would be required as a premise is that the sighted organism is equipped by its visual system to behave in ways which are discriminatory with respect to physical color, e.g., in the case of a motorist, to stop when the traffic lights are red and continue when they are green. This is a truth which the blind man cannot fully understand, not having an adequate conception of physical color. But this reply too is quite hopeless. The blind man can acquire an adequate conception of physical color insofar as it is described scientifically, i.e., in terms of wavelengths of light and the reflective properties of pigment. What he lacks is merely a conception of how physical colors look to the sighted. But this deficiency is just one aspect of his not knowing what it is like, experientially, for the sighted organism to see. The only respect in which he fails to grasp what is meant by "the traffic lights are red" is that he does not know what sort of experience the sighted person has when he receives light of the relevant wavelength. But this knowledge is precisely what, if conceptual materialism were true, he should be able to deduce from the physical information available. So the blind man's inability to achieve an adequate conception of color merely reflects the falsity of the materialist's thesis.

The conceptual materialist still has one final line of defense. He might claim that the blind man can, from the physical truths, derive a full knowledge of the character of visual experience, and that what misleads us into supposing otherwise is that the physical truths do not afford that kind of knowledge which is available, through introspection, to the sighted. On this view, there is no more to the character of visual experience than the blind man can deduce; it is simply that, being blind and lacking even the capacity to visualize, he cannot conceptually focus on this character in the way the sighted can. He knows exactly what such experience is psychologically like but cannot achieve, in his imagination, the viewpoint of those who have it. This reply looks more promising than the others. At least the distinction between merely knowing the character of a certain type of experience and knowing it in introspective perspective is sound. To take a different example, there are some pitches which are too high for me to hear or frame a mental image of. Suppose P is the highest pitch I can hear or image. There is a perfectly good sense in which I know what it is like, experientially, to hear the pitch which is an octave higher than P: I can exactly fix the position of such an experience in the auditory spectrum. All I lack is the kind of knowledge available to someone who has the experience or can achieve, in his imagination, the viewpoint of one who has it. However, I do not think that this distinction can be of use to the conceptual materialist. For while it is possible for someone to know the character of a certain type of experience without knowing it in the introspective manner, it is surely only possible in those cases where the person can define that character by reference to other types of experience whose character he does know in that manner. I can grasp the experiental character of hearing the relevant pitch because 1 can introspectively grasp both the hearing of P and the octave relation. The problem for the blind man is that the physical truths provide no introspective reference points at all. And for this reason, I cannot see how he could derive from them, in any sense, a full knowledge of what it is like, experientially, for the sighted organism to see.

2. Metaphysical Materialism. The failure of conceptual materialism does not entail the failure of materialism as such. For the thesis that mental facts are nothing over and above physical facts can be interpreted in another way. It can be interpreted as meaning that all mental facts are the necessary consequences of physical facts by some mode of necessity which, while in a broad sense logical (i.e., stronger than mere natural necessity), is not that of deductive entailment. Since this mode of necessity is often called "metaphysical," I shall label this position "metaphysical materialism." It might be objected that this so-called interpretation of the materialist thesis does not do justice to the force of the phrase "nothing over and above." But this is a purely verbal issue. The fact remains that metaphysical materialism as defined is incompatible with dualism.

How could mental facts be the logically (stronger than naturally) necessary consequences of physical facts except by some relation of deductive entailment? Well, the metaphysical materialist invites us to accept a certain analogy. Consider the relationship between the two propositions (A) that this spoon is hot, and (B) that the molecules of this spoon are agitated. Let us assume that our ordinary (i.e., prescientific) concept of physical heat is as that property, whatever it is, which induces heat sensations in us. From this it immediately follows that (A) is not deducible from (B), since it is not an a priori truth that what induces such sensations is molecular agitation. At the same time, it is a truth, and one which science has established. And from this, together with the assumption, it follows that heat is (i.e., is identical with) molecular agitation. But, for reasons which Kripke has made familiar, this obliges us to say that there is no logically possible world in which (B) is true and (A) is false.6 In this sense, if the spoon is hot, its being so is a logically necessary consequence of the fact that its molecules are agitated. In the same way, it is claimed, mental facts about a person are the logically necessary consequences of physical facts about his body. Thus our ordinary (i.e., prescientific) concept of pain, it is said, is as that state, whatever it is, which reveals itself to the subject under a certain introspective appearance and is expressed overtly through certain forms of behavior. We cannot deduce that a person is in pain from a physical description of his body (not even from one which includes a formulation of physical law), since such a description does not tell us how things appear introspectively: it does not address itself to the character of pain as a subjective experience. But we should, it is claimed, anticipate the scientific discovery that the state which meets our conceptual requirements—the state which, in the actual world, has the required introspective appearance and behavioral influence—is, in its intrinsic nature, physical. And if pain is (i.e., is identical with) a physical state, then it is a necessary truth (true in all logically possible worlds), though not a priori, that a person is in pain if and only if his body is in that state. Admittedly, it may well turn out that pain in one species of organism is a different physical state from pain in another, and it may even turn out that, within a single species, pain in one member is a different physical state from pain in another. But while this would prevent an exact analogy with heat, it does not affect the fundamentals of the account.

However, as Kripke has shown,7 the suggestion of any analogy at all is totally misconceived. What allows us to identify physical heat with molecular agitation is that we can detach the sensible appearance of physical heat from its essential nature: we can regard the sensible appearance as just the contingent effect of physical heat on human experience. If we are to identify pain with a physical state (whether absolutely or in a way which is species/organism-relative) we must similarly detach the introspective appearance of pain from its essential nature. But how could this be done? Only by separating what pain is in itself from how pain feels to the person who has it. And such a separation is surely impossible. Moreover, even if it were possible, it would be, from the standpoint of the materialist, self-defeating. For, if the experiential character of pain is not part of its essential nature, the identification of pain with a physical state does nothing to make the experiential facts necessary consequences of physical facts and hence contributes nothing to the materialist's program. Nor can I see any other way in which the materialist could attempt to justify his consequentialist thesis without reverting to some form of conceptual materialism (e.g., by analyzing experiential concepts in functional terms), and this we have already refuted. Of course, the materialist could simply assert that experiential facts are metaphysically necessitated by physical facts and challenge us to refute him. But without some explanation of how such necessitation might obtain, we can hardly take this assertion seriously.

3. The Token-Identity Thesis. The fact that experiential states, as mental types, cannot be identified with physical states does not, as such, entail that experiences (experiential events), as mental particulars, are not identical with physical events. Thus a position which is becoming increasingly fashionable is to hold that mental events are identical with certain neural events, but that the mental properties of these events are not identical with any of their physical properties or indeed with anything, such as their functional properties, whose instantiation could be deduced from purely physical truths. This position is not a form of materialism as defined; it does not claim that mental facts are, either deductively or by metaphysical necessity, nothing over and above physical facts, though in some versions it does claim that mental facts are supervenient on physical facts, i.e., that there cannot be mental differences without physical differences. Rather, the position combines a kind of ontological monism with an attributive dualism. When someone is in pain, the pain event is part of the material world (e.g., the firing of his C fibers at a particular time); but its being a pain (its mental character) is something genuinely additional to, and not logically necessitated by, its physical properties or anything which physical science could describe. This position, or, strictly, the monist part of it, is known as the "token-identity thesis." It is thus contrasted with the stronger "type-identity thesis," which claims that the mental types of which mental events are the tokens are identical with certain physical types, i.e., that mental properties are identical with physical properties.

Once again it is Kripke who has exposed the flaw in this weaker position.8 Suppose Smith is in pain. Let us call the particular pain event (the particular experiential event in Smith's mind at that time) E. And let us call the particular neural event with which E is supposedly identical (e.g., some particular firing of Smith's C fibers) B. Now it is clear that the experiential property "being a pain" is an essential attribute of E (the experiential event); that is, there is no logically possible world in which E exists and is not a pain. The only remotely feasible way of denying this claim would be by adopting a functionalist account of experiential properties, as in certain versions of conceptual materialism. For this would allow one to say that E qualified as a pain only contingently, in virtue of the causal role which events of E's (presumably physical) intrinsic type played in the whole physical or psychophysical system. But, as we have already seen, the functionalist account is inadequate; the arguments that refute it as an adjunct to conceptual materialism refute it quite generally. But while "being a pain" is an essential attribute of E, it is surely not an essential attribute of B (the neural event). For surely there is a possible world (perhaps it has to be one with different psychophysical laws) in which B exists and has no experiential character at all. Of course, this would be rejected by someone who identified experiential properties with physical properties, but, as we have seen, this type-identity thesis is unacceptable. But if E is essentially a pain and B is not, then E and B cannot be numerically identical. For if they were, they would have all their properties, including their modal properties, in common.

Granted the rejection of functionalism and the type-identity thesis, the defender of token-identity has, as far as I can see, only one way of trying to resist this argument. He has to maintain that while the experiential character of B is not part of the essential nature of the type of physical event of which B is an instance, it is, nonetheless, essential to the identity of B itself. But even this reply fails, since the identity of B is wholly determined by physical factors. Thus, suppose we try to envisage a world W which results from the actual world by (a) removing B, (b) filling the gap with an event of exactly the same physical type, composed of exactly the same particle events, with exactly the same causal antecedents, but without B's experiential character, and (c) making whatever further adjustments this requires, e.g., changes in psychophysical law and similar (putative) replacements of other B-type events. It is surely obvious that the W event which is the counterpart of B is none other than B itself. Being of exactly the same physical type, having exactly the same spatiotemporal location, being composed of exactly the same particle events in the same brain, and being caused in exactly the same way, the putative replacement just is numerically the same physical event. It follows that if W is genuinely possible, "being a pain" is not essential to B. And consequently, the only way in which we could hold it to be essential would be by claiming that the experiential character of B was logically determined, either deductively or metaphysically, by its physical properties and other physical factors on which its identity depends. This claim would exclude the possibility of W, but it would oblige us to accept a full-blooded materialism of the kind we have already rejected and which the token-identity thesis was designed to avoid.

4. The Logical Dependence Thesis. The token-identity thesis was one way of trying to get a compromise between full dualism and full materialism. Another way would be to claim that while mental facts are not logically determined by physical facts, they are at least logically dependent on them. There are several positions of this sort. Among the most familiar are the claim that any subject of mental states has to be (or at least to have been) embodied and the claim that the connection between mental states and their behavioral manifestations is not purely contingent. I do not have time to examine such claims in detail, though I find none of them at all convincing. My main objection to all of them is that, in the last analysis, they rest on an indefensible form of the verification principle. They all, in one way or another, appeal to the alleged fact that dualism generates skepticism—for example (and this is the case most often cited), skepticism about other minds—and that such skepticism can only be avoided by moving, to some extent, in the direction of conceptual materialism—for example, by making certain concessions to behaviorism. But the fact that a philosophical position generates an epistemological problem is not as such a reason for rejecting it. Moreover, in the particular case of other minds, the problem that dualism allegedly generates has been greatly exaggerated. Even from a dualist standpoint, most of our commonsense beliefs about the mental states of others can be justified by an inference to the best explanation. If it is said that such justification falls short of conclusive verification, this is quite true. But it is surely very implausible to maintain that in the case of statements about other minds, conclusive verification is available. The only area, in the topic of other minds, where dualism seems to generate a really acute epistemological problem is that concerned with sense qualia—e.g., how can I tell that your color spectrum is not inverted with respect to mine? But here it seems to me that the problem is a genuine one and that a theory which did not allow it to arise would be defective.

5. The Restriction of Dualism to Experience. Assuming we reject the logical dependence thesis, we cannot, I think, avoid a dualistic account of experience: experiential states and experiential events are radically non- physical, i.e., are neither physical, nor the logical product of anything physical, nor, except causally or nomologically, dependent on anything physical. But we could still reject dualism with respect to the residual contents of the mind. Thus we could still claim that propositional attitudes, such as belief and desire, are physical states of the brain, or functional states whose instantiation is deducible from a physical description of the brain, or whole organism, together with certain physical laws. And we could still claim that propositional acts, such as thoughts and judgments, are physical events in the brain, or functional events whose occurrence is deducible from a brain/organism-description and physical laws. My arguments against materialism and the token-identity thesis do not directly exclude such claims, since they have exclusively concerned the case of experience.

Nonetheless, it seems to me that once we have accepted experiential dualism, the extension of dualism to the rest of the mind becomes unavoidable. The first point to stress is that although, as formulated, my arguments against materialism and the token-identity thesis have only concerned the case of experience, they can be generalized to cover all those aspects of our mental life which are essentially conscious—all those aspects which, of their very nature, form part of how it is, subjectively, with the subject. For it was precisely the conscious character of experience that made it resistant to the materialist and identity accounts. This does not oblige us to adopt a dualist account of propositional attitudes, such as belief and desire, since these are not essentially conscious in the relevant sense: a person who is sound asleep can still be said to have beliefs and desires. But it does, I think, oblige us to adopt a dualist account of propositional and quasi-propositional acts, such as thoughts, judgments, and decisions. For these, it seems to me, are essentially conscious. It might be objected, by philosophers of a radically empiricist persuasion, that it is only the phenomenal, not the conceptual, aspects of these acts which are essentially conscious. For example, suppose that in the course of performing some larger arithmetical calculation, I make the judgment that 9*7 = 63. An empiricist might argue that the only essentially conscious element in this judgment is my mental voicing of the sequence of sounds "nine sevens are sixty-three" and that this phenomenal act only derives its conceptual significance, as a vehicle for my propositional act, from a surrounding mental framework (involving such items as my mastery of English and my knowledge of arithmetic) which is, or could be, subjectively silent. If this were so, then the conceptual aspects of the judgment— in effect, the propositional act itself—could be construed in a nondualistic way. However, I find this empiricist account very implausible. I think we can best see this by beginning with the case of perception. It is quite clear that in most cases of perception, at least after early infancy, there is a large element of conceptual interpretation within the conscious experience itself. Thus in visual perception, the experience consists of more than just the presentation of a color array; it includes seeing the color array as a physical scene of a certain sort. (Seeing Wittgenstein's ambiguous picture as a duck or as a rabbit is just a dramatic example of this.) Now consider the particular case of hearing the sequence of sounds "nine sevens are sixty-three." There is clearly an experiential difference between hearing it just as a sequence of sounds (this might be the experience of someone who did not understand English) and hearing it as a sentence of English expressing the proposition that 9 * 7 = 63 (this would be the normal experience of a numerate English-speaker). But given that a propositional interpretation of the sounds can be part of the total perceptual experience when the sounds are heard, it can also be part of the total conscious state when they are mentally voiced. And surely this will always be so when the mental voicing is the vehicle for a propositional act.

We are still left with the case of propositional attitudes, which, as I have conceded, are not essentially conscious. Here, the case of dualism rests on the fact that such states, while not in themselves states of consciousness, are essentially linked with states of consciousness. Let us focus on the example of belief. There is no denying that a person can hold a certain belief at times when he is not performing the corresponding act of judgment. But it is surely part of the essential nature of belief that it should dispose the subject to make such judgments when the need arises. Thus it is surely logically impossible for me to believe that 9*7 = 63 unless I am thereby disposed to judge that 9*7 = 63 when I address myself to the question. If so, then beliefs cannot be identified with physical states, such as neural networks, since such states would at best sustain such dispositions only contingently, i.e., through the contingent obtaining of certain psychophysical laws. Nor, given that judgments themselves are nonphysical, can beliefs be construed as functional states of a kind whose instantiation could be deduced from purely physical truths. It might be objected that in rejecting the identity thesis, I am overlooking the distinction between types and tokens. Thus it may be that as a mental type, a belief state cannot be physical. But why shouldn't the tokens of this type—the particular instances of this belief state in particular minds at particular times—be physical? Why not say that my current token-belief that 9*7

63 is some current neural network in my brain, while allowing that this network only qualifies as a token-belief in virtue of its contingently disposing me to make the appropriate conscious judgment? But the trouble with this is that while we are free to recognize an ontology of token- beliefs, we cannot do so in a way which leaves their belief character (their character as token-beliefs) only contingent. For our only conception of a token-belief is as something which is, by its essential nature, the particular occurrence of some type of believing, e.g., my believing now that 9*7 = 63. And if the token-beliefs are essentially token-beliefs, then their dispositional connection with conscious judgments must be essential too.

Given all this, the only way of avoiding a wholly dualistic account of belief would be to adopt a hybrid form of functionalism, in which belief is construed as a functional state defined partly in terms of its output in nonphysical conscious judgment and partly in terms of its output in physical behavior—in effect, the combination of certain mentalistic and certain behavioral dispositions. All I can say against this is that I find the behavioristic element implausible. It is inconceivable that someone should believe that 9*7 = 63 without being disposed to make the corresponding judgment. But it is surely quite conceivable that someone should hold this belief without being equipped with the right kind of neuromuscular system to express it overtly. Indeed, in the case of stroke victims, this frequently happens. It may be objected that such victims only retain their beliefs because they once possessed the neuromuscular capacity to express them. But we can surely also envisage cases in which someone acquires beliefs without ever having such a capacity.

As I see it, then, we are obliged to give a wholly dualistic account of belief. (Exactly what account we should give I shall not pursue; the fact that belief essentially involves a judgmental disposition may seem to support some kind of dualistic functionalism, but for reasons I shall not go into, I doubt if this is so.) Obviously, if we are dualistic about belief, we must be similarly dualistic about other propositional attitudes. Since we have already established dualism for experience and other essentially conscious states, this means, in effect, that we must accept dualism as such—accept the thesis that the mind and its contents are radically non- physical.

II. Four Objections to Dualism

I have tried to show why the alternatives to dualism are unacceptable. If my arguments have been correct, there is a sense in which I could afford to rest my case there, since if the alternatives are wrong, dualism must be right. The reason I continue is that dualism itself is thought to be vulnerable to certain objections, and these objections (or at least some of them) merit answers. Of course, I do not have space to consider all the objections. Some are nothing more than rhetoric (I would put into this category such objections as "nonphysical entities are queer" and "postulating nonphysical entities is ontologically extravagant"); others would require a separate chapter to deal with them effectively (in particular, those objections which concern the problem of other minds and the issue over private languages, though, in the case of other minds, I have given some indication in section 1.4 of the line I would take). Apart from these, however, I have tried to cover those objections which I regard as the most important, either because of their intrinsic merits or because of the influence they have exerted.

As it turns out, the four objections I consider all concern, in one way or another, the dualist's account of psychophysical causation—his account of the causal relations between body and mind. Dualism itself, of course, does not entail that there are such relations. There is no contradiction in maintaining that the mind, as well as being nonphysical, is causally isolated from the physical world; and with the help of a suitable theism, it may even be possible to explain why things are empirically organized as if there were psychophysical causation, when there is not. But since I do not want my defense of dualism to force me into such an eccentric position, I shall work on the assumption that parallelism, in this extreme form, is untenable and that mind and body are indeed causally related. This still leaves the dualist with a choice between interactionism, which takes the causal relations to run in both directions, and epiphenomenalism, which takes them to run only from body to mind. Here my sympathies, again in line with common sense, are with the interactionist. And consequently, I would feel at least very uncomfortable if there were something which obliged me to choose between abandoning dualism altogether and adopting its epiphenomenalist version. That I do face such a choice is the substance of the fourth objection.

1. The Traditional Objection. Perhaps the oldest objection to dualism is that if the mind is nonphysical, the very idea of psychophysical causation—of the body causally affecting the mind or the mind causally affecting the body—is deeply puzzling, if not incoherent. How can such different kinds of thing—the physical and the nonphysical—come into causal contact? How can the material body gain purchase on the immaterial mind, or vice versa? However, put like this, I cannot see what the problem is supposed to be. Obviously, dualistic causation does not operate through physical contact, as when one billiard ball displaces another.

But why should it not just be in the nature of things that in certain psychophysical conditions, certain types of neural event cause certain types of mental event, or vice versa? According to Thomas Nagel, we cannot understand how such causation would work.9 Nagel assumes, 1 think correctly, though contrary to Hume's account, that causation involves some kind of objective necessitation. He then argues, in effect, that so long as we think of mental and physical events as radically different in their intrinsic nature, such necessitation is incomprehensible: "what we cannot understand is how . . . the brain process necessitates the sensation."10 But what I cannot understand is how this "how"-question arises. Since the necessitation is causal, rather than logical, there cannot be any question of construing it as some kind of a priori entailment. Moreover, since the causation is direct (the brain event directly causing the mental event, or vice versa), there is no question of an intervening mechanism. Perhaps it is this very directness which Nagel finds puzzling. For in the physical world causal processes are, in general, spatiotemporally continuous, thus providing an intervening mechanism between a cause and any subsequent effect. But there is surely nothing incoherent or problematic about the notion of direct causation. And where causation is direct, i do not see in what sense there could be a question of how it operates, except as a request to specify the causally relevant properties and covering laws.

2. The Problem of Causal Pairings. While the traditional objection to dualistic psychophysical causation is totally misconceived, there is a related objection which is more troublesome and for which, ironically, I have to take the credit.11 It is normally assumed that where two events are causally related, they are so wholly by virtue of the way in which, via their noncausal properties and relations, they fall under some natural law. Thus if, on a particular occasion, my heating of a lump of metal caused it to melt, it is assumed that what makes this true is that the metal was of a certain type and reached a certain temperature and that it is a law of nature (or a consequence of a law of nature) that whenever a lump of metal of that type reaches that temperature, it melts. However, when we apply this model to the case of psychophysical causation, dualistically conceived, we encounter a problem. Suppose that B is the event of Smith's brain being in state cj) at time t and that £ is a mental event of type which occurs in Smith's mind a tenth of a second after t and as the direct causal result of B. What psychophysical law could we postulate to account for this causal episode? We might begin by postulating the law (L,) that whenever a brain is in a state (J) a mental event of type v|j occurs a tenth of a second later. But this would be inadequate. For suppose that Jones's brain is also in state 4> at f, giving us the event B' which is a simultaneous duplicate of B, and that this causes in Jones's mind, a tenth of a second later, an event £' of type i|(, which is a simultaneous duplicate of E. Ex hypothesi, B is the cause of E and B' is the cause of £'. But the law L^ does not account for these causal pairings. Because it only specifies the temporal relation between cause and effect, it is neutral between these pairings and the alternative, but false, hypothesis that B is the cause of £' and B' is the cause of £. The obvious remedy is to replace L, by the stronger law (L2) that whenever a brain x is in state cf> a mental event of type >J) occurs a tenth of a second later in that mind which x, or the x-containing organism, embodies—in other words, in the mind of that subject whose brain x is. L2 would then yield the unique causal pairings of B with £ and of B' with £'. But the problem with this is that, for a dualist, the relation of embodiment itself must be analyzed, wholly or partly, in causal terms: at least part of what makes a particular brain x the brain of a particular subject y is that things are psychophysical^ arranged in a way which gives x (and x alone) the capacity to have a direct causal influence on y (and y alone) and, for the interactionist, vice versa. It would be circular to account for this arrangement by laws like L2 and, because of the original problem, impossible to account for it by laws like L,. In view of this, it might seem that the right solution is to abandon dualism. For if we identified mental events with neural events, we could envisage laws that guaranteed unique causal pairings by specifying the precise spatiotemporal relation between cause and effect. Indeed, we could hope to account for psychophysical causation wholly in terms of the ordinary laws of physics and chemistry.

When I first wrote about this problem back in 1968, I argued that the right response, for the dualist, was to postulate psychophysical laws restricted to particular pairs of brains and minds (or brains and subjects). Thus in the case of Smith and Jones, we can secure the correct causal pairings by postulating a separate law for each person—the law (relevant to Smith) that whenever brain ^ is in state c{), a i|j event occurs a tenth of a second later in mind M\ (or to subject S) and the law (relevant to Jones) that whenver brain x2 is in state 4>, a event occurs a tenth of a second later in mind Mi (or to subject J), where x, and M| are the brain and mind of Smith and x2 and M2 are the brain and mind of Jones. More recently, I have tried to meet the problem in a quite different way.12 I have argued that even in the physical realm we can envisage cases in which the fundamental laws do not account for the causal pairings. Thus suppose that, for a certain kind K of metal, it is a law that when any spherical K lump reaches a specified temperature, a flash occurs a tenth of a second later somewhere (unspecified) in the region of points no farther from the center of the sphere than twice its diameter. Suppose further that there is no stronger law which fixes the position of the flash more precisely. Now imagine the case in which two adjacent K spheres simultaneously reach the critical temperature and, a tenth of a second later, two flashes occur, both within the specified region for each sphere. It is coherent and indeed plausible to suppose that each flash is the effect of just one of the sphere- temperatures and that each sphere-temperature is the cause of just one of the flashes. But the causal pairings are not determined by the law and the noncausal conditions, since each flash falls within the specified region for each sphere. From this I concluded that we should abandon the assumption that causal relations can be wholly accounted for in terms of non- causal properties and laws. And once this assumption is abandoned, the problem of psychophysical causal pairings no longer arises; it no longer matters if the pairing of B with E and B' and E' is not determined by the noncausal conditions and the covering law.

Reviewing these alternative responses, I now think that in a sense, both were correct. On the one hand, the hypothetical case of the K spheres does indeed show that causal pairings can be law-transcendent; and, in that respect, the original objection to the dualist's position fails. On the other hand, it does seem quite plausible to envisage laws of the restricted kind as underlying the relation of embodiment. As we have said, at least part of what makes a particular brain x the brain of a particular mental subject y is that things are psychophysical^ arranged in a way which gives x (and x alone) the capacity to have a direct causal influence on y (and y alone)— and, for the interactionist, vice versa. It is hard to see in what this arrangement could consist except in certain psychophysical laws which specifically link x and y, and such laws would guarantee unique causal pairings. At all events, it is clear that the dualist has sufficient resources to meet the objection.

3. Davidson's Objection. Even if causal pairings can be law-transcendent, it is still plausible to insist that if an event x causes an event y, there is some covering law which ensures that in relevantly similar conditions, any event of the (relevant) x-type stands in a relevantly similar relation to some event of the (relevant) y-type. According to Donald Davidson, even this relatively weak claim, combined with the acceptance of psychophysical causation, commits us to the token-identity thesis.13 The reason, argues Davidson, is that the "disparate commitments of the mental and physical schemes" preclude the existence of strict psychophysical laws.14 "It is a feature of physical reality that physical change can be explained by laws that connect it with other changes and conditions physically described. It is a feature of the mental that the attribution of mental phenomena must be responsible to the background of reasons, beliefs, and intentions of the individual. There cannot be tight connections between the realms if each is to retain allegiance to its proper source of evidence."1 s As he explains the point in a subsequent essay, "my general strategy for trying to show that there are no strict psychophysical laws depends ... on emphasizing the holistic character of the cognitive field. Any effort at increasing the accuracy and power of a theory of behaviour forces us to bring more and more of the whole system of the agent's beliefs and motives directly into account. But in inferring this system from the evidence, we necessarily impose conditions of coherence, rationality and consistency. These conditions have no echo in physical theory, which is why we can look for no more than rough correlations between psychological and physical phenomena."16 Davidson concludes that where there are causal relations between mental and physical events, the covering law must be physical. This in turn requires identifying mental events with physical events and claiming that it is in respect of their physical, not their mental, properties that mental events have physical causes and physical effects. He calls this position "anomalous monism": "monism, because it holds that psychological events are physical events; anomalous, because it insists that events do not fall under strict laws when described in psychological terms."17

When Davidson denies the possibility of strict psychophysical laws, he only intends this, I assume, to apply to those types of mental phenomena, such as beliefs and desires, which have propositional, or in some other way conceptual, content. He is not, I assume, excluding the possibility of a psychophysical law ensuring that a certain type of brain process always produces a certain type of nonconceptual sensation. However, even when we confine our attention to the relevant types of phenomena, Davidson's argument seems very strange. No doubt he is right to stress the holistic character of the mental: we cannot, I suspect, even make sense of the claim that someone has just one propositional attitude, not forming part of some coherent system of attitudes; and certainly, as Davidson sees, it is only by finding evidence of some larger system in a subject that we can be justified in ascribing a particular attitude to him. No doubt too Davidson is right in claiming that the considerations of coherence, rationality, and consistency which form an essential part of the evidence (or of the way the evidence is interpreted) "have no echo in physical theory." My difficulty is in not being able to see why these points should rule out strict psychophysical laws. Davidson seems to be arguing that because the epistemo-logical methods of commonsense psychology and physical science are quite different, psychological and physical facts cannot stand in any tight lawlike connection. But this is just a non sequitur.18 The only point which even seems to pose a threat to strict psychophysical laws—and it concerns the logical, not the epistemological, holism of the mental—is that the in smallest viable unit of propositional mentality is a coherent system. But the most that follows from this is that some of the psychophysical laws would have to be very complicated, e.g., ones which determine how complex neural networks causally sustain complex systems of belief and desire. Maybe such laws would be too complicated for us to discover; maybe even some of them could not be exhaustively specified; but that line not mean that they do not exist. Nor should we conclude that all psychophysical laws are, in respect of their psychological content, of this complexity. If there are laws determining how complex neural networks sustain complex systems of propositional attitudes, then a fortiori there are laws determining how complex networks sustain individual attitudes. Moreover, in the case of causal links from mind to body, there is nothing in prevent there being laws which determine how, given certain brain c0nditions, certain types of volition (such as the attempt to move one's ii in in a certain way) cause certain types of neural response. The fact that such volitions are possible only in the framework of a complex cognitive field makes no difference.

All in all, it seems to me that while we can accept much of what Davidson says about the holistic character of the mind and the difference between the epistemological methods of psychology and physical science, there is nothing here which even looks like an argument against the existence of strict psychophysical laws.

4. An Objection to Dualistic Interactionism. Some dualists, while accepting that there are psychophysical causal relations, hold that they only i mi in one direction—-from body to mind. They accept, for example, that if  I am stung by a wasp, the subsequent pain is caused by the neural response which the sting induces, but they deny that if I decide to smoke my pipe, the subsequent bodily movements involving pipe and matches ire caused by my decision. And, quite generally, they accept that a person's mental states are causally affected by the states of his body but deny that his mental states have any causal influence on the states of his body or on anything else in the physical world. This version of dualism is known as Òepiphenomenalism," and the version it contrasts with, which accepts psychophysical causation in both directions, is known as "interactionism." I should add that most epiphenomenalists, as well as denying that the mind has any causal influence on the body, also deny that it has a causal influence on anything at all. In particular, they deny that mental states can cause, or contribute to the causation of, other mental states.

Like most other current philosophers, I regard epiphenomenalism as unnatural and implausible. In the first place, it is in radical conflict with our conception of ourselves as agents. If mental states have no causal influence on behavior, then behavior cannot be thought of as intentional in any decent sense, even if the subject happens to have certain intentions which it fulfills. And if behavior is not intentional, it does not qualify as action in a sense which distinguishes it from mere bodily movement. The epiphenomenalist might reply that the general conformity of our behavior to our intentions is not merely accidental; it is ensured by the very structure of our brains and their muscular extensions in the framework of physical and psychophysical law. But even so, the behavior would not be intentional in the requisite sense, since the intentions and the psychophysical laws that control their occurrence would be irrelevant to its production. Second, human behavior exhibits certain complex regularities that call for explanation and that, at present, we explain (at least partly) in psychological terms. These psychological explanations, though typically of a rational rather than a mechanistic kind, attribute a causal efficacy to the mental; they represent behavior as falling under the control of the subject's beliefs and desires, or under the control of the subject's decisions, which are responsive to (if not determined by) his beliefs and desires. And these explanations gain credence from the fact that, as well as being, in their own terms, successful, they cannot, at present, be replaced by non- psychological explanations that cover the same ground. Third, it is difficult to see how, if epiphenomenalism were true, the mind could form a subject matter for overt discussion. Certainly, if mental states have no causal access to our speech centers, the notion of an introspective report collapses: even if the subject retains an introspective knowledge of his mental states, his utterances would not count as expressing that knowledge if it contributes nothing to their production. But it is not even clear how, on the epiphenomenalistic view, our language, as a medium for our utterances, makes semantic contact with the mind at all. In what sense, for example, could the word "pain," as overtly used, be said to signify a certain type of sensation, if neither the occurrence of the sensations nor our introspective conception of their type affects the overt use? Quite generally, it seems that if the mental contributes nothing to the way in which the linguistic practices involving "mental" terms are developed and sustained in the speech community and in no other way affects the production of utterances employing these terms, then, in respect of their overt use, the terms should be analyzed in a purely behaviorist or functionalist fashion—which would deprive the epiphenomenalist of the linguistic resources to enunciate his thesis. It is true, of course, that each language user may mentally interpret each term as signifying a certain kind of (dualistically conceived) mental state. But how could such interpretations have any bearing on the objective meaning of the terms, as employed in speech and writing, if they are causally idle?

None of these points shows that epiphenomenalism is logically untenable, in the sense of being incoherent or self-contradictory. Even the third point, if correct, only shows that when overtly expressed, epiphenomenalism is self-refuting—that the very attempt to provide an audible or visible formulation of the thesis presupposes its falsity. Nonetheless, we have, I think, very strong, and perhaps even conclusive, reasons for rejecting it. And, because of this, I should not want my defense of dualism to involve my acceptance of anything but an interactionist position.

It is at this point that the fourth objection arises. For there is an argument which purports to show that dualism of any but an epiphenomenalistic kind is scientifically unacceptable. The argument runs like this:

1. The body is a physical system.

2. As such, the body must be subject to ordinary physical laws.

3.  Our theories as to what physical laws obtain are subject to revision. But our current scientific evidence strongly supports the view that at any level of description relevant to a theory of human behavior, these laws are, for all practical purposes, deterministic.

4.  So we can reasonably conclude that any bodily event, of a sort which might be cited in a description or explanation of behavior, is causally determined by prior physical events and conditions.

5.  But such a conclusion leaves no room for a nonphysical mind to have any causal influence on behavior.

6.   Hence, on our present scientific evidence, we face a choice between epiphenomenalism and some kind of identity thesis (if only of the token-token variety).

If it is correct, this argument constitutes an objection to dualistic interactionism. It also, on the face of it, constitutes an objection to dualism as such, if epiphenomenalism is, as I would concede, unacceptable. And, nowadays, it is normally as an objection to dualism, and in support of the identity thesis, that the argument is offered.19 But here we must be careful. It is true that epiphenomenalism is very implausible and, overtly expressed, may even be self-refuting. And this means that, other things being equal, we should reject it. But it does not mean that we should reject it at all costs. If the only alternative were the identity thesis, then, in my view, epiphenomenalism, despite its implausibility, would be the preferable option. For I think that my earlier arguments show that the identity thesis is incoherent. The choice between epiphenomenalism and the identity thesis only counts against dualism on the assumption, which 1 hold to be false, that epiphenomenalism is not only very implausible but less plausible than its rival. There is also a further point to bear in mind. If the identity thesis is only of a token-token kind, it is not clear how, even if it were coherent, it would avoid the implausibilities of epiphenomenalism. For if mental events have physical effects only by virtue of their physical properties, and if mental properties are not identical with physical properties, or with functional properties of a purely physicalistic kind, then mental properties are causally irrelevant. And if mental properties are causally irrelevant, the identity thesis does not accord the mind the kind of causal efficacy which common sense demands. It does not allow the mental an influence on behavior in any interesting sense. So to provide a genuine alternative to epiphenomenalism, the identity thesis would have to be of a full-blooded materialist form, involving, in addition to token-identity, some physicalistic construal of mental properties. This is something which defenders of token-identity do not always recognize. It is not, for example, recognized by Davidson.20

However, whether it is an objection to dualism as such or only to interactionism, I have to find some answer to the argument; for I am not willing to concede that epiphenomenalism and the identity thesis are the only empirically acceptable options. In fact, the error in the argument is not hard to identify. We must begin by distinguishing two ways in which science might provide evidence that the functioning of the body is wholly explicable in terms of deterministic physical law. The first way would be by direct research on the body itself—in particular, on the brain, since it is on brain activity, if on anything, that we might expect the mind to exert a direct causal influence. Thus, by monitoring neural activity in various parts of the brain (without disturbing normal functioning), scientists might build up a strong inductive case for the conclusion that the electrochemical state of any neuron at any time is determined by its immediately prior electrochemical state and the states of other neurons directly connected to it. The second way would be by discovering, without reference to the functioning of the human body, that the rest of the physical world seems to be subject to certain physical laws of a sort which, if they applied universally, would make the body a deterministic system. This is the evidence provided by the investigations of physics and chemistry into the properties of matter and energy in general. Now it is predominantly evidence of the second kind to which the argument appeals. The claim is not that a thorough sampling of brain activity reveals a wholly deterministic system, for no such sampling has been conducted. Rather, the claim is that the human body, including the brain, must be physically deterministic if it is to conform to those theories which apply to physical systems in general. But this evidence, just because it bears on the question of the human system only indirectly, is not decisive. It has to be weighed against what we know or have reason to believe, independently, about the relation between body and mind. In particular, it must be weighed against all that makes epiphenomenalism an implausible theory, and this, in turn, must be set against the background of the a priori objections to materialism and token-identity. When everything is taken into account, the most reasonable conclusion to draw is, surely, that through its attachment to a nonphysical mind, the brain is subject to certain influences which do not affect the other physical systems that science investigates and on whose behavior its nomological theories are based. It is conceivable that this conclusion will be called into question by future brain research (if it were, I should have to reconsider my rejection of epiphenomenalism). But as things stand, we are entitled to assume that it will not.

III. Turning the Tables

Given the current fashion for materialism or positions that come close to it, there is a final irony. For while it is certain that our minds are radically nonphysical, it is far from certain that the physical world is not wholly mental. And here the possibility I have in mind is not the obvious one of subjective idealism—the thesis that the physical world is, in some way, a construct of the human mind or the logical product of the nomological constraints on human sense experience. As it happens, this is a thesis which I accept and have defended elsewhere;21 but, for present purposes, I am willing to set it aside and adopt the commonsense position of physical realism. Rather, the possibility I have in mind is less extreme but more startling; that the physical world, while ultimately real and independent of human consciousness, is wholly mental in its intrinsic nature. I have called this position "mentalistic physical realism."22

At first sight the claim that the physical world might be intrinsically mental seems absurd. But if we find it so, it is because we are failing to appreciate the severe limitations on the extent to which the nature of the physical world can be specified in physical terms. As I have shown elsewhere,23 any correct physical description of the physical world, or any aspect of it, is of a highly "topic-neutral" character: it reveals formal structure and nomological organization but conceals what the physical items thus structured and organized are like in themselves. Thus whatever occupants of physical space we take as fundamental, it is impossible to provide, in physical terms, a "transparent" (genuinely revealing) specification of their intrinsic properties beyond a specification of their shape and size and other aspects of their spatial or spatiotemporal arrangement; beyond this, we can, in physical terms, only specify their intrinsic properties "opaquely" (in a non-genuinely-revealing way), as those properties, whatever they are, which, in conjunction with the laws of nature, generate certain causal powers and sensitivities. Moreover, it is impossible to provide, in physical terms, a transparent specification of the intrinsic nature of physical space beyond a formal specification of its geometrical structure—a specification which does not reveal what physical points and physical distance are like in themselves. If we are normally unaware of these limitations on the scope of physical description, it is largely because, prior to philosophical reflection, we think of the physical world as intrinsically characterized by the qualities of its sensible appearance. Once we have recognized, as we must, that these sensible qualities are, in their physical realization, nothing but powers to affect human experience, the topic-neutrality of any correct physical description of space and its occupants becomes clear. And it is this topic-neutrality which leaves room for mentalistic physical realism—leaves room for the possibility that the physical world is, in its intrinsic nature, wholly mental.

This is not the right place to elaborate these points in detail—to present the full argument for the topic-neutrality thesis and develop concrete versions of the mentalistic hypothesis. All this, in any case, I have done in my book.24 My reason for mentioning these points at all is to stress the extent to which current philosophy has got the real issues of mind and matter out of focus. And even here I have somewhat underplayed my hand. For what the topic-neutrality thesis shows is not just that the physical world might be intrinsically mental, but that unless it is, its intrinsic nature is, to us at least, incomprehensible—composed of properties of which we can have no conception.25 Whether this itself counts as an argument for mentalism is a moot point. But it certainly shows the extent to which the tables are now turned on the materialist.


1. I am indebted to Howard Robinson for his helpful comments on an earlier draft of this paper.

2. Thus see my The Case for Idealism (London: Routledge and Kegan Paul, 1982), chap. 1.

3. I, in fact, reject it. See ibid., part III.

4. On my interactionist view, which I defend in section II, this claim is, in fact, false.

5. Cf. H. Robinson, Matter and Sense (Cambridge: Cambridge University Press, 198Z), pp. 4-5.

6. Thus  see S. Kripke, Naming and Necessity (Oxford: Blackwell, 1980), pp. 131-34.

7. Ibid., pp. 148—54.

8. Ibid., pp. 146—47.

9. T. Nagel, Mortal Questions (Cambridge: Cambridge University Press, 1979), pp. 185-87.

10. Ibid., p. 187.

11. See my "Psychophysical Causal Relations," American Philosophical Quarterly 5,  no. 1 (1968).

12.  See my "In Self-Defense," in G. F. Macdonald, ed., Perception and Identity (London: Macmillan, 1979), pp. 168—70.

13. D. Davidson, "Mental Events," in L. Foster and J. Swanson, eds., Experience and Theory (London: Duckworth, 1970), pp. 79—101.

14. Ibid., p. 97.

15 Ibid., pp. 97-98.

16. "Psychology as Philosophy," in Davidson, Action and Events (Oxford: Oxford University Press, 1980), p. 2.31.

17. Ibid., p. 231.

18. This was pointed out by T. Honderich in "Psychophysical Lawlike Connections and Their Problem," Inquiry 24:191-93.

19. This, in effect, is how C. Peacocke argues for token-identity in his Holistic Explanation (Oxford: Oxford University Press, 1979), pp. 134-43.

20. For a fuller exposition of this point, see Robinson, Matter and Sense, pp. 8—13.

21.  In my The Case for Idealism, part IV.

22. Ibid., p. 13.

23. Ibid., part II.

24.  Ibid. For examples of mentalistic physical realism, see chaps. 7 and 11.

25. Thus see ibid., pp. 122—23.