Philosophy and Phenomenological Research Vol. LIX, No.2, June 1999 On David Chalmers's The Conscious Mind SYDNEY SHOEMAKER Cornell University One does not have to agree with the main conclusions of David Chalmers's book in order to find it stimulating, instructive, and frequently brilliant. If Chalmers's arguments succeed, his achievement will of course be enormous~ he will have overthrown the materialist orthodoxy that has reigned in philosophy of mind and cognitive science for the last half century. If, as I think, they fail, his achievement is nevertheless considerable. For his arguments draw on, and give forceful and eloquent expression to, widely held intuitions~ seeing how they go astray, if they do, cannot help but deepen our understanding of the issues he is addressing. I shall focus on three points: Chalmers's conceivability argument for the possibility of "zombies," which grounds his dualism about phenomenal consciousness ~ his "paradox of phenomenal judgment" ~ and the "dancing qualia argument" with which he supports his principle of organizational invariance. I Chalmers thinks that we can conceive of a world physically just like the actual world in which there are creatures, "zombies," which despite being physical and functional duplicates of conscious beings.in the actual world are themselves devoid of phenomenal consciousness. The states of these creatures lack phenomenal properties, or qualia. He takes the conceivability of such a world to establish its possibility. He takes this to show that phenomenal consciousness does not "logically supervene" on physical facts. And he takes this to show that it is not itself physical. Let "Q" name some quale, or phenomenal property. Chalmers is certainly right in holding that no matter what physical property "P" names, the statement (S) "If someone has a state with P, she has an experience with Q" will not be a priori true, and will be such that we can conceive of its being false. Since he thinks that in the case of concepts of qualia, the "prin1ary intension" and "secondary intension" are the same, and that this is also true of concepts of physical properties, and since he thinks that "primary propositions" are BOOK SYMPOSIUM 439
necessarily true if and only if they are a priori true, he thinks that statement S will not be necessarily true. That of course gives us the possibility of a world in which S is false; and generalizing this gives us the possibility of "Zombie World." But it is easy to see how S could be necessarily true despite not being a priori true. It is compatible with the claim that individual qualia are not functionally definable (because, e.g., functionally equivalent creatures could be qualia inverted relative to each other) that the relationships of qualitative similarity and difference are functionally definable. l The claim that these relationships are in fact functionally definable is of course controversial; but I do not think that it should be more controversial than functionalist claims Chalmers himself accepts, e.g., that judgment, memory, and various other "psychological" states or processes are functionally definable. If these similarity and difference relationships are functionally definable, there is no apparent reason why they should not be physically realizable. If they are physically realizable, then fixing all of the physical facts is at the same time fixing all of the qualitative similarity and difference relations between experiences. Fixing these relationships will involve fixing the qualitative character of experiences, i.e., what qualia they instantiate. And this in turn will involve fixing the truth-value of statements like S; there will be a vast number of such statements that will have the status of being necessary a posteriori. Chalmers claims that "If there were a functional analysis of the notion of experience or phenomenal quality, then the analysis in question would yield functional analyses of specific phenomenal properties..." (p. 23). Since having a functional analysis of qualitative similarity and difference would amount to having a functional analysis of "the notion of experience or phenomenal quality," Chalmers is here asserting in effect that the functional definability of qualitative similarity and difference would require the functional definability of individual qualia. And that seems plainly to be a mistake. In holding that qualitative similarity and difference are functionally definable I am denying Chalmers's claim that Zombie World (a world physically like our~ in which there is no consciousness) is conceivable. Other critics of Chalmers have conceded this conceivability claim but denied the legitilnacy of inferring from it to the metaphysical possibility of Zombie World. 2 I of course agree that the move from conceivability to metaphysical possibility can be questioned, even when the intensions involved are primary-for I think that the falsity of any statement like S can be conceived (i.e., no such statement is a priori), yet I think that some such statements are necessarily I have defended this claim in "Functionalism and Qualia" (Philosophical Studies, 27, 1975, reprinted in my Identity, Cause and Mind, Can1bridge, 1984) and in a number of other papers. E.g., Brian Loar, in his excellent "David Chalmers's The Conscious Mind," delivered at the meetings of the Central Division of the American Philosophical Association in 1997. 440 SYDNEY SHOEMAKER
true. (And since I think that nomological necessity is a special case of metaphysical necessity, I think that there are many other cases in which the move from conceivability to metaphysical possibility is illegitimate.) But while I think, in agreement with other critics, that the claim that Zombie World is conceivable does not get Chalmers what he wants, I also think that it is more than he is entitled to-and it is worth pointing out that (contrary to what he supposes) it is not implied by the conceivability claim he is entitled to, namely that given any "supervenience conditional" such as S, one can conceive of its being false. II I turn to the "paradox of phenomenal judgment." Because he accepts the causal closure of the physical realm, Chalmers concedes that his view makes phenomenal states "explanatorily irrelevant" to the phenomenal judgments that are supposedly about them. He therefore faces the objection that his view cannot account for either our knowledge of such states or our ability to refer to them. His presentation of this objection seemed to me brilliant. His replies struck me as unconvincing. Although Chalmers holds that the explanatory irrelevance of phenomenal states does not strictly imply that they are epiphenomenal, he concedes that his view precludes him from holding that a causal theory of knowledge explains our knowledge of phenomenal states, or that a causal theory of reference explains our ability to refer to them. He thinks, however, that while such theories may have application elsewhere, there is no compelling reason why they should have application here. But one does not have to maintain that such theories have universal application in order to hold that where states do stand in appropriate causal relations to judgments, the judgments are about the states and express knowledge of them. The holding of certain sorts of causal relations could be sufficient for reference and knowledge even if it is not necessary for this. So consider the hypothetical zombie who is an exact physical and functional duplicate of Chalmers. Nothing that Chalmers has argued excludes the possibility that this (supposed) zombie has states, physical or functional ones, whose causal relations to its phenomenal judgments are such as to make it true that those judgments are about those states and express knowledge of them. 3 But what- Chalmers does say that his zombie twin means to refer, in his "clain1s of consciousness," to something other than a functional property, and gives that as a reason for saying that this zombie's claims of consciousness are false, unless understood in a "deflationary" way (p. 204). But this of course cannot provide a good reason for saying that, e.g., the claims of consciousness of my zombie twin are false! And, putting aside the zombie twins, it is not just on a "deflationary" interpretation of claims of consciousness that it would be absurd to say that because Chalmers and I differ in our theoretical views about consciousness, one or the other of us must be wrong in our everyday claims of consciousness. BOOK SYMPOSIUM 441
ever case there is for holding that the phenomenal judgments of the zombie are about physical or functional states of it will also apply to Chalmers and his phenomenal judgments-for Chalmers and the zombie are physical and functional duplicates. And in that case (a) the "zombie" will not be a zombie, and (b) Chalmers's nonphysical phenomenal states, supposing he has such, will be irrelevant to the truth of the phenomenal judgments he makes about himself-either that or, weirdly, some of Chalmers's assertive utterances express two different judgments, one self-ascribing the sort of phenomenal state his zombie counterpart self-ascribes, and the other (somehow) selfascribing a phenomenal state that is explanatorily irrelevant to it. Since Chalmers thinks that the zombie's phenomenal judgments are false (or perhaps in some cases without truth value, owing to reference failure), and since he thinks that what is required to justify a phenomenal judgment is the subject's having the phenomenal property it ascribes, this being the evidence required for such a judgment, he thinks that the zombie's phenomenal judgments are without evidence and unjustified. But this cuts the notions of justification and evidence loose from their moorings. There is obviously no way in which the zombie is epistemically culpable in making the phenomenal judgments it makes. His making them does not display any failure of rationality. There is nothing that anyone could point out to him that should convince him that he should repress these judgments-nor is it in his power to do so. Insofar as the notions of justification and evidence are normative ones, notions having to do with what it is reasonable or rational for a creature to do or to judge, it seems that there could be no difference between their application to the zombie and their application to one of us. III While Chalmers thinks that it is logically possible for there to be creatures physically indistinguishable from us who are qualia inverted (e.g., spectrum inverted) relative to us, he thinks that it is "naturally" impossible for there to be creatures who are functionally like us (at a fine grained level) and qualia inverted relative to us. What he thinks is logically possible I think is logically impossible, for the same reasons I think that Zombie World is a logical impossibility. And what he thinks is naturally impossible I think is naturally possible-or, more guardedly, I think we have no reason, and in particular no a priori reason, to think that it is not. His case for the natural impossibility of qualia inversion is his "Dancing Qualia Argument." One assumes for purposes of reductio that one has a functional isomorph who is qualia inverted relative to one. One constructs a series of cases intermediate between oneself and one's isomorph. In the first case one of one's neurons is replaced with silicon, in the next two are replaced, and so on. The changes preserve fine-grained functional organization. If one's isomorph is qualia inverted relative to one, the series must include "two sys- 442 SYDNEY SHOEMAKER
terns that differ in at most 10 percent of their internal makeup, but that have significantly different experiences" p. 267). Chalmers lets one of these be himself, and the other someone he calls "Bill"-Bill is the one in which more neurons have been replaced with silicon. He then imagines that we install in him a silicon "backup circuit" just like one in Bill that is functionally identical to a neural circuit that is already in his head. And we install a switch that can switch between these circuits. When we flip the switch, the silicon circuit takes over from the neural circuit, and he becomes in essential respects like Bill. So (on our assumption) his experience will be qualitatively different from what it was. As we flip the switch back and forth, his experience will change, or "dance," before his eyes. But, since functional organization is unchanged, he will not notice any change. Chalmers says that "on any functional construal of belief, it is clear that I cannot acquire any new beliefs as the flip takes place" p. 269), and that it is extremely implausible that a single replacement of a neural circuit by a silicon circuit could be responsible for the addition of significant new beliefs such as "My qualia just flipped." He sums up the reductio ad absurdum as follows: "It seems entirely implausible to suppose that my experiences could change in such a significant way, with my paying full attention to them, without my being able to notice the change. It would suggest...a radical dissociation between consciousness and cognition" (p. 269). Let me first observe that Chalmers's denial that there can be change of belief "as the flip takes place" is question-begging. It is apparent that anyone who thinks that qualia inversion is possible must hold that not only are individual qualia not functionally definable but also that individual qualitative beliefs (beliefs ascribing specific qualia to experiences) are not functionally definable. 4 Presumably it is part of the functional role of qualia that the instantiation of a quale produces or tends to produce the qualitative belief that it is instantiated. So if the flip changes the qualia, it will also change the qualitative beliefs. But Chalmers is right, though for the wrong reason, in thinking that in this case there will be an unnoticed dancing of qualia. If functional organization is preserved, beliefs about the similarity and difference relations of current qualia to past qualia must be preserved. So in changing the qualia and beliefs about current qualia, the flip must also, in order to preserve functional organization, induce false memory beliefs about past qualia. So it is true that the subject will not notice the flip. Is this consequence absurd in a way that refutes the claim that qualia inversion is possible? No. If it seems so at first, this may come from the unthinking assumption that since functional organi- This point was made in my "Functionalism and Qualia," op. cit.; also in n1y "Absent Qualia are Impossible-A Reply to Block" (The Philosophical Review, 90, 1981, reprinted in my Identity, Cause and Mind). BOOK SYMPOSIUM 443
zation is preserved in the envisaged changes, we must have here a case of normal cognitive dynamics-together with the correct assumption that where the cognitive dynamics is normal, a marked change in qualitative character cannot go unnoticed. But the unthinking assumption is incorrect because the normal cognitive dynamics requires the functional organization to change when there is a change in qualia. Keeping the functional organization unchanged while changing the qualia requires a highly unnatural tampering with the cognitive dynamics, one which preserves beliefs (about similarity of present qualia to past qualia) in circumstances in which preserving them makes them false, and does so in part by inserting false beliefs about past qualia. Of course, such tampering would not be possible at all if individual qualitative beliefs were functionally definable. But as already noted, anyone who thinks that qualia inversion between functional isomorphs is possible will also think that qualitative belief inversion between functional isomorphs is possible; the latter implies that qualitative beliefs are not functionally definable, and Chalmers cannot assume otherwise without begging the question. 5 Thanks to David Barnett and Susanna Siegel for comments on an earlier draft. 444 SYDNEY SHOEMAKER