Review of John Perry, Knowledge, Possibility and Consciousness, The MIT Press, Cambridge, MA, 2001. 231 + xvi pp. Philosophical Quarterly Forthcoming

John Perry’s new book is a defense of physicalism against three arguments, the zombie argument, the knowledge argument, and the modal argument. In chapters 1-3, Perry introduces the issues, explaining that his goal is the relatively modest one of showing that physicalism—‘antecedent physicalism’, he calls it (p. 26)—is not overthrown by the arguments. In chapter 4, Perry addresses the zombie argument, suggesting it is sound only if epiphenomenalism is true. He argues that, since epiphenomenalism is false, and is false whether one is a physicalist or not, the zombie argument is no threat to physicalism. In chapter 5-7, Perry addresses the knowledge argument. He argues that what is really at issue here is an assumption about meaning and knowledge he calls the subject matter assumption. Perry suggests that the subject matter assumption is false—again: whether or not one is a physicalist—and draws the conclusion that the knowledge argument is unsound. Finally, in chapter 8, Perry addresses the modal argument. He argues that, once one has rejected the subject matter assumption, the modal argument is revealed as no more of a threat than the knowledge argument. As Perry notes (p. 16), the zombie argument is a version of the modal argument—though they are certainly different enough to be treated separately—and so presumably his view is that the former is doubly wrong: insofar as it is a modal argument, it succumbs to the objections concerning the subject matter assumption, and insofar as it is a zombie argument, it succumbs to the objections concerning epiphenomenalism.

The overall position defended in the book is a version of the ‘two ways of thinking’ strategy familiar to readers of the philosophy of mind literature from Loar (1997) and others. But Perry’s discussion and development of the strategy is sophisticated and rewarding—not to mention clear and entertaining. What is particularly striking is that Perry embeds the strategy within a rich framework of ideas concerning the content of beliefs and statements, a framework that is at least in outline familiar from Perry’s earlier important work on indexicals and demonstratives. But how compelling is Perry’s two ways view? I will make three comments, corresponding to the three arguments with which he deals.

Perry discusses the zombie argument by considering the possibility of a zombie world, a world physically identical to the actual world but which lacks all conscious experience. Perry thinks that so defined a zombie world is impossible. A world without conscious experience would be a world in which events which in fact are effects of experiences either do not occur at all or else occur as effects of some other cause. But—he argues—on neither possibility is a world without experience physically identical to the actual world: in the first case, various physical events would take place in the actual world which do not take place in the zombie world; in the second case, "the physical principles of the zombie world are different from the actual world" (p. 75).

Now, there is a response to this argument provided the version of physicalism in play is the supervenience version. Given this version, the response is to say that the zombie world is not physically different from the actual world, but is different only in the principles that connect physical goings-on with phenomenal goings-on. Perry considers this and responds by rejecting the supervenience version of physicalism. A supervenience version, he says, implausibly requires "the definition, analysis or explication of a phenomenal concept of experience in terms of the way physical states function" (p. 88), but the identity version requires nothing parallel. So in effect Perry’s suggestion is that the physicalist should respond to the zombie argument by adopting an identity-based version of the doctrine rather than the supervenience-based version.

However, while there are certainly differences between the two versions of physicalism, it is difficult to defend this account of the difference. If it is true that identity statements do not require anything by way of explication of their constituent phenomenal concepts, that is presumably because identity statements are necessarily true. But the same goes for supervenience. Similarly, if it is true that supervenience statements do require something by way of explication of their constituent phenomenal concepts, that is presumably because we need to justify belief in supervenience. But the same goes for identity. In sum, it seems plausible to say that, whatever else divides them, supervenience and identity versions of physicalism are on all fours as far as demands of explication are concerned. If so, it is difficult to see that Perry has answered the zombie argument.

Turning next to the knowledge argument, Perry here articulates and rejects the subject matter assumption. In many ways, rejecting this is the heart of the book, and what Perry says about it is interesting and instructive. As I understand it, the basic point is as follows. Consider an identity statement ‘A=B’ and imagine that it is both true and cognitively significant. We may assume in such a case that the two expressions flanking the identity sign are associated with two distinct "ways of thinking" (p. 19) of the referent—indeed, this is just a way of recapitulating the fact that the statement is cognitively significant. According to the subject matter assumption, however, we should assume in addition that the referent of the two terms has two distinct properties, and that the ways of thinking are to be explained in terms of these two properties. Perry says that the subject matter assumption is false: in at least some cases, we may have two ways but not two properties. In developing this idea, he presents a number of intriguing ideas about ways of thinking—e.g., a neo-Humean distinction between ideas and impression, the idea of file folders, information games, and (what he calls) reflexive content—but so far as the knowledge argument is concerned, the key point is not what ways of thinking involve so much as what they don’t: what they don’t involve are distinct properties of the referent.

But how does rejecting the subject matter assumption answer the knowledge argument? As Perry sees it, rejecting the assumption allows one to explain what Mary learns when she comes out of her room consistently with the hypothesis of physicalism. What he suggests is that when Mary comes out of her room and sees a red thing for the first time, she gains a particular piece of knowledge, which in the scheme of the book (p. 100) is called (5):

5. Qr is this subjective character.

(Here ‘Qr’ expresses the concept of the phenomenal character associated with seeing red things, a concept presumed to be available to Mary while in her room; the demonstrative is supposed to pick out a quality that is made available to Mary when she first sees a red thing.) The interesting feature of (5) is that, if one is attending only to subject matter—that is, only to the properties and relations required for the truth of (5)—then one could not distinguish (5) from anything that Mary knew prior to her release. However, argues Perry, one should not attend only to subject matter—the subject matter assumption is false—and thus it is reasonable to suppose that (5) is distinct from anything Mary knew. So, on the one hand, (5) represents a genuinely new piece of knowledge; but on the other the fact that Mary learnt (5) doesn’t put pressure on physicalism, for physicalism is concerned only with subject matter.

Now, in order to assess this line of thought, we need to concentrate not on whether Mary learns (5)—it seems clear that she does—nor on whether the fact that she learns it is a threat to physicalism—it seems clear that it is not. What we need to concentrate on is whether (5) is all she learns. But at this point a serious problem for Perry emerges. According to him, when Mary learns (5) her way of thinking associated with ‘Qr’ becomes linked in the right way with her way of thinking associated with ‘this subjective character’—certain file folders become configurated in a way that previously they were not, and it is this different ‘reflexive content’ rather than subject matter which distinguishes the knowledge expressed in (5) with any previous knowledge. However, this new configuration of file folders is something Mary could presumably have predicted would happen once she is released. Moreover she completely understands what this configuration of file folders is; for example, in considering others looking a red things for the first time, she will know the particular configuration of their file folders. So when she comes out of the room and sees a red thing, it is natural to suppose, if Perry is right, that things are exactly as she predicted. But the whole point of the example is that things are not exactly as she predicted. Mary is supposed to be surprised by what she learns. But if this is right, it is hard to believe that learning (5), while certainly part of what happens, is the whole of what happens when she comes out. But if learning (5) is not the whole of what happens, Perry has no answer to the knowledge argument.

The last argument Perry examines is the modal argument. This argument focuses on the fact that the necessarily true statements advanced by the physicalist—such as ‘pain = c-fiber stimulation’—have (as Kripke famously put it) an appearance of contingency which needs to be explained away. Kripke himself thought this could not be done, and drew the conclusion that physicalism is false. But Perry thinks that it can be done—by appealing again to the notion of the reflexive content of the statement. The reflexive content of the belief expressed by ‘pain = c-fiber stimulation’ is, roughly, "that there is some state that is referred to by both the concept of pain and the concept of stimulated c-fibers" (p. 181). Perry does not suppose that this is an alternative content of my belief, nor that it is a different belief that I have in addition (p 132-3). Rather it is a fact about the belief—that its constituent concepts are not "linked" (p. 180)—which contributes to its role in the explanation of thought and behavior, and in particular—he thinks—will explain the appearance of contingency. Since this explanation is available to physicalists who suppose that ‘pain = c-fiber stimulation’ is necessarily true, the modal argument is answered.

A number of issues are raised by this suggestion—why is reflexive content a kind of content, for example, rather than simply a fact about the belief state?—but the one that seems most pertinent is that Perry’s suggestion overgenerates. For there are many statements which, on the one hand, have the reflexive content of the sort he associates with ‘pain = c-fiber stimulation’ but which do not exhibit the appearance of contingency on the other. Earlier in the book, for example, Perry noted (p. 143-4) that mathematical statements have a similar sort of reflexive content. But mathematical statements don’t have the appearance of contingency associated with psychophysical identity statements—mathematical statements seem palpably a priori and necessary, even if they are neither trivial nor obvious. Perhaps the notion of reflexive content explains why certain necessary truths are unobvious, but it does not explain why they appear contingent, and it is this second thing that requires explanation.

At one point Perry summarizes his approach by saying that it "won’t be physiological or neurological, nor even…very phenomenological. [It] will be logical, semantical and philosophical." (p.118). Knowledge, Possibility and Consciousness provides as extensive and subtle a defense of the logical and semantical approach to the problem of conscious experience as one is likely to find. The suspicion that the approach is unsuccessful—a suspicion which I think the preceding points make plausible—strongly suggests that the problem is not at root logical or semantical but must have a different source.

Daniel Stoljar

Australian National University