*[[This is a commentary on John Perry's book Knowledge, Possibility, and Consciousness. It is part of a symposium on Perry's book in Philosophy and Phenomenological Research, vol. 68 no. 1, January 2004. The symposium includes a reply by Perry to this commentary. I have included a response to Perry's reply at the end of the commentary.]]
John Perry's book Knowledge, Possibility, and Consciousness is a lucid and engaging defense of a physicalist view of consciousness against various anti-physicalist arguments. In what follows, I will address Perry's responses to the three main anti-physicalist arguments he discusses: the zombie argument (focusing on imagination), the knowledge argument (focusing on indexicals), and the modal argument (focusing on intensions).
Zombies are unconscious creatures that are microphysically identical to conscious beings such as ourselves. If zombies are metaphysically possible, then materialism is false. The zombie argument first aims to establish that zombies are conceivable, infers that they are metaphysically possible, and concludes that materialism is false.
Perry's response is straightforward. He says (pp. 77-80) that the zombie argument is really an argument for epiphenomenalism: zombies will be possible if and only if epiphenomenalism is true. And he says that he cannot see any reason why one would think that zombies (and zombie worlds) are possible unless one were already an epiphenomenalist. So the zombie argument begs the question, by tacitly assuming epiphenomenalism as a premise.
I think the zombie argument is not just an argument for epiphenomenalism, and that the possibility of zombies is compatible with non-epiphenomenalist dualism (see the end of this section). But I will set that point aside here. Whether the zombie argument is an argument for epiphenomenalism or for dualism: if it rests on prior acceptance of epiphenomenalism, then the argument begs the question. (Here I work with the common understanding of epiphenomenalism as a form of dualism. Perry mentions a materialist version of epiphenomenalism, but it is incompatible with the possibility of true zombies, so it is irrelevant here.)
Whatever the merits of the zombie argument, however, it does not beg the question. To see this, note that the zombie argument is not based on a single premise — the possibility of zombies — but rather on two premises. The first premise is that zombies are conceivable, roughly in the sense that there is no a priori contradiction in the idea of a zombie. The second premise is that if zombies are conceivable in this sense, then they are possible. (Or they are "primarily possible", in that there is a possible world satisfying the relevant primary intension).
Each of these premises is accepted by many philosophers who clearly reject epiphenomenalism. For example, the first premise is accepted by many "type-B" materialists (those who accept an epistemic gap between the physical and the phenomenal but deny an ontological gap), al of whom deny epiphenomenalism: e.g., Ned Block, Chris Hill, Joe Levine, Brian Loar, and many others. The second premise is accepted by many "type-A" materialists (those who deny an epistemic gap between physical and phenomenal) and many interactionists, all of whom deny epiphenomenalism: e.g., Frank Jackson (current incarnation), David Lewis, Richard Swinburne, and others. Of course few or none of these people accept both premises. But their mere existence shows that the premises have support that is independent of a prior acceptance of epiphenomenalism.
To go beyond a roll-call, it can be seen straightforwardly that each premise has substantive support that goes beyond epiphenomenalism. The second premise rests on general considerations linking conceivability and possibility. The first premise rests partly on prima facie conceivability intuitions that many share, and partly on deeper considerations concerning the absence of any conceptual linkage between microphysical concepts (which are structural-functional in nature) and phenomenal concepts (which are not). In both cases, whether or not these premises are correct, their support presupposes nothing about epiphenomenalism.
I suppose that Perry might try to argue that the first premise tacitly builds in epiphenomenalism. Given all the non-epiphenomenalists who accept the premise, however, this would seem unlikely: it would require that they be deeply irrational, or have deeply divided minds. Further, Perry says a number of things elsewhere in the book that seem to come close to embracing the first premise (in the relevant sense of conceivability). For example, although he embraces an identity between phenomenal and physical properties, he denies (pp. 87-88) that consciousness supervenes logically on the microphysical (i.e., he denies a conceptual entailment from microphysical truths to phenomenal truths), which suggests that there will be no a priori contradiction in the notion of a zombie. And even in raising questions about zombies, he points to nothing like an incoherence in the notion; at best he points to a tension with certain a posteriori beliefs about the causal role of our own conscious states.
In any case, it seems clear that to reject the argument, Perry has to go beyond considerations about begging the question, and instead give substantive reasons to reject one of the premises. From Perry's overall position, I would expect him to be more likely to reject the second premise than the first. Like other type-B materialists, he can accept that zombies are conceivable in the relevant sense, but deny the principle linking conceivability (in this sense) with possibility. I will return to this matter in the last section.
One final point: it is worth noting that even the possibility of zombies does not obviously entail epiphenomenalism. To see this, note that an interactionist dualist can accept the possibility of zombies, by accepting the possibility of physically identical worlds in which physical causal gaps (those filled in the actual world by mental processes) go unfilled, or are filled by something other than mental processes. The first possibility would have many unexplained physical events, but there is nothing metaphysically impossible about unexplained physical events. Also: a Russellian "panprotopsychist", who holds that consciousness is constituted by the unknown intrinsic categorical bases of microphysical dispositions, can accept the possibility of zombies by accepting the possibility of worlds in which the microphysical dispositions have a different categorical basis, or none at all. So even if the argument had the one-premise structure that Perry suggests, then as long as it was appropriately neutral on the physical character of the world, it would not obviously beg the question in favor of epiphenomenalism.
The knowledge argument holds that there is phenomenal knowledge (e.g. concerning what it is like to see red) that is not deducible from physical knowledge, and infers that phenomenal facts are non-physical facts. Perry's response, essentially, is to analyze phenomenal knowledge as a sort of indexical knowledge. We know that there is indexical knowledge (e.g. concerning my current location) that is not deducible from complete physical knowledge, but this indexical knowledge does nothing to falsify physicalism. If phenomenal knowledge is indexical knowledge, then we can straightforwardly explain the epistemic gap between the physical and phenomenal domains, without requiring any non-physical ontology.
Perry applies this strategy to the new knowledge that (physically omniscient) Mary gains when she sees red for the first time, roughly as follows. Mary gains new knowledge of the form "red things cause experiences of type P". This knowledge crucially involves a phenomenal concept P — a concept of what it is like to have a certain sort of experience. Perry's strategy (pp. 145-48) is to analyze Mary's phenomenal concept P as a demonstrative concept — which he labels this_i — that functions, like other demonstrative concepts, to pick out whatever sort of experience she is currently attending to. Because of an underlying indexical character, knowledge involving demonstrative concepts cannot usually be inferred from complete objective knowledge. So Mary's epistemic gap is reduced to the familiar epistemic gap for indexicals and demonstratives.
My response is straightforward. I think that there are clear reasons to hold that phenomenal knowledge (of the crucial sort) is not indexical knowledge, and that phenomenal concepts (of the crucial sort) are not demonstrative concepts of this sort; so this analysis of the epistemic gap fails. Mary does gain indexical knowledge involving a demonstrative concept, but this indexical knowledge is not the phenomenal knowledge that is central to the knowledge argument, and this demonstrative concept is not the concept that Mary's central new knowledge involves. I have argued for this based on an independently-grounded analysis of phenomenal concepts elsewhere (Chalmers 2002a), but I will give some reasons here.
It is useful to consider analogies with other demonstrative knowledge of types. Let this_S be a demonstrative concept of certain shapes. Jill might tell Jack that she is about to show him her favorite shape. When she shows him a circle, he might form the thought "Jill's favorite shape is this_S", where this_S refers to circles. This is a clearly demonstrative thought. He might also form the thought "Jill's favorite shape is a circle". This is a clearly nondemonstrative thought: the right hand side is not a demonstrative concept, but what me might call a qualitative concept. Finally, he might form the thought "this_S is a circle". This is a substantive, nontrivial thought, taking the form of an identity involving a demonstrative concept and a qualitative concept. This sort of thought is very common with demonstratives — one conceives the object of a demonstration as the object of a demonstration, and at the same time attributes it substantive qualitative properties, conceived non-demonstratively.
On the face of it, Mary's situation is precisely analogous. She knows that she is about to have an experience of the sort usually caused by red things. Upon having the experience, she might form the thought "The experience usually caused by red things is this_I". But at the same time, she might form the thought "The experience usually caused by red things is R", where R is a qualitative concept of the sort of experience she is now having. This is brought out by the fact that just as Jack might think the substantive thought "this_S is a circle", Mary might think the substantive thought "this_I is R". Like Jack's thought, Mary's thought involves attributing a certain substantive qualitative nature to an object that is identified demonstratively. The concept R — her qualitative concept of the sort of experience in question — is not a demonstrative concept at all, as witnessed by the nontriviality of the identity involving the demonstrative concept this_I and R. And Mary's crucial new knowledge in the thought-experiment, the new knowledge that the knowledge argument turns on, is the substantive qualitative knowledge involving R (e.g. that red things cause R experiences), not the relatively uninteresting indexical knowledge involving this_I.
One might be distracted by the fact that Mary's concept R is a new concept, acquired upon having the experience, while Jack's concept "circle" is an old concept. But this is inessential to Jack's case. We can imagine that Jack has never seen a circle before (if one is worried about plausibility, make it another shape), but that on seeing a circle for the first time, he acquires the qualitative concept of circularity. He will then be in the position to think the qualitative thought "Jill's favorite shape is a circle", and to think the substantive demonstrative-qualitative thought "This_S is a circle. Mary's situation is just the same. Upon seeing red for the first time, she is able to form a qualitative concept of that sort of experience (based on her exposure to it), to think qualitative thoughts involving that concept, and to think demonstrative-qualitative thoughts in which both a demonstrative and a qualitative concept are deployed.
Perry might respond by holding that the cases are deeply disanalogous, and that where Joe has two concepts ("this_S" and "circle"), Mary has only one (this_I). But this seems quite false to the phenomenology of the cases, both of which involve substantive knowledge of the form "this is such-and-such", where "such-and-such" is a qualitative predication based on exposure to a demonstrated entity. We can also note that just as it seems that Joe could have had different thoughts of the form "this_S is a square", if he had been exposed to a different shape, it seems that Mary could have had different thoughts of the form "this_I is G" had she been exposed to a different sort of experience, such as an experience of green. So it seems that the qualitative concept of the experiential type is quite different from the demonstrative concept.
I conclude that Perry gives an adequate analysis of certain subsidiary demonstrative knowledge that Mary gains — knowledge that is analogous to demonstrative knowledge in other domains. But Perry gives no adequate analysis of Mary's substantive new non-indexical, non-demonstrative phenomenal knowledge, and in particular gives no way of reconciling this new phenomenal knowledge with the truth of physicalism.
(In Chalmers (2002a), I distinguish three sorts of phenomenal concepts: pure phenomenal concepts (such as R), demonstrative phenomenal concepts (such as this_I) and relational phenomenal concepts (such as "the sort of experience typically caused by red things"). Perry's discussion seems to acknowledge only two sorts, the demonstrative and the relational; or at least, it seems to assimilate the pure phenomenal and the demonstrative with each other.)
One can also make a direct case against any analysis of phenomenal knowledge as indexical or demonstrative knowledge, as follows. In the indexical case, any epistemic gaps disappear from an objective perspective. Say that I am physically omniscient, but do not know whether I am in the USA or Australia (let's imagine that there are appropriate qualitative twins in both). Then I have a certain indexical ignorance, and discovering that I am in the USA will constitute new knowledge. But if someone else is watching from the third-person point of view and is also physically omniscient, they will have no corresponding ignorance: they will know that A is in Australia and that B is in the US, and that's that. There is no potential knowledge that they lack: from their perspective, they know everything there is to know about my situation. So my ignorance is essentially indexical, and evaporates from the objective viewpoint. The same goes for indexical ignorance concerning what time it is, for demonstrative ignorance concerning what this is, and so on. In all these cases, the ignorance disappears from the objective viewpoint: an objectively omniscient observer can know everything there is for them to know about my situation, and there will be no doubts for them to settle.
Now consider Mary's ignorance. From her black-and-white room, she is ignorant of all sorts of facts: what it will be like for her to see red for the first time, what it is like for others to see red, and so on. Only the first of these looks even apparently indexical, so let us focus on that. In this case, a physically omniscient observer may have precisely analogous ignorance: even given his complete physical knowledge, he may have no idea what it will be like for Mary to see red for the first time. So this ignorance does not evaporate from the objective viewpoint. The same goes even more strongly for knowledge of what it is like for others to see red. For any observer, regardless of their viewpoint, there will be an epistemic gap between complete physical knowledge and this sort of phenomenal knowledge. This suggests very strongly that phenomenal knowledge is not a variety of indexical or demonstrative knowledge at all. Rather, it is a sort of objective knowledge of the world, not essentially tied to any viewpoint.
If this is right, then any analysis of phenomenal concepts as indexical or demonstrative concepts fails, and any attempt to explain Mary's epistemic gap in terms of the epistemic gap for indexical or demonstrative concepts fails. If this is right, then Perry's substantive account of Mary's epistemic gap, in terms of the "reflexive content" of certain beliefs, will also fail. Reflexive content is essentially indexical content (tied to the relation of this token to the world), so that an analysis in these terms applies only to broadly indexical knowledge.
A similar diagnosis can be applied to analyses of phenomenal knowledge by many other materialists who seek to acknowledge Mary's epistemic gap and to explain it away: any analysis of phenomenal concepts as indexical, demonstrative, or recognitional concepts will fall prey to similar objections. This does not show by itself that no analysis of phenomenal knowledge can be given that saves both physicalism and the epistemic gap but I think that it renders the prospects dim.
The third anti-physicalist discussion that Perry discusses is the "modal argument". Of course the zombie argument is itself a sort of modal argument. Presumably Perry intends a more general target here, but I think some of the same issues arise. The chapter is divided into a discussion of "Kripke's argument" and of "Chalmers' argument". His discussion of the latter focuses on my use of the two-dimensional framework in arguing against materialism; so this can be seen as filling in the issues that went unaddressed in his earlier discussion, concerning the relationship between conceivability and possibility. I will focus on this issue here.
Perry's discussion in this chapter is harder to follow than in the other chapters of the book, and it is not easy to clearly identify his responses to the arguments in question. His basic strategy appears to be to agree with his opponents that where there is "apparent contingency" (Kripke) or "conceivability" (me) of some mental-physical dissociation, there is some sort of possibility in the vicinity, but to argue that once the possibility is appropriately understood (in terms of reflexive relations to tokens and the like), it does not threaten materialism. I was left unclear, however, on how this strategy bears on the specific argument I put forward.
To summarize the argument briefly (see Chalmers 2002b for a more detailed treatment): Let us say that S is conceivable when it is not a priori that ~S. Any given statement S can be associated with two intensions: a primary intension, which is a function from centered worlds (worlds plus individuals/times) to truth-values, and a secondary intension which is a function from uncentered worlds to truth-values. When S is necessary, it has a necessary secondary intension (and vice versa); when S is a priori, it has a necessary primary intension (and somewhat controversially, vice versa). Let us say that S is primarily possible when its primary intension is true at some centered world. Let P be the complete microphysical truth about the world. Let Q be a phenomenal truth. Then the anti-materialist argument can be put as follows
(1) P&~Q is conceivable
(2) If S is conceivable, S is primarily possible.
(3) If P&~Q is primarily possible, materialism is false.
—
(4) Materialism is false.
Here, the first premise is simply a statement of the epistemic gap, grounded in the a priori coherence of zombies, or in the a priori coherence of inverted spectra, or in Mary's inability to deduce phenomenal truths from physical truths. The second premise is a core principle of the two-dimensional framework, linking apriority and primary possibility. The third premise is not completely obvious (since materialism requires the secondary possibility of P&~Q), but some relatively straightforward argument involving the nature of physical and phenomenal concepts takes one from the relevant primary possibility to a relevant secondary possibility.
(On the third premise: If a world satisfies the primary intension of P, it is at least structurally identical to our world. So either (i) the world is completely physically identical to our world but different in mental respects, so materialism is false, or (ii) it differs at most in the intrinsic microphysical properties that underlie this structure. Exploiting (ii) leads directly to the Russellian "panprotopsychist" position, which I count for present purposes as a version of nonmaterialism. The argument also needs supplementing to handle the role of indexicals and centering, but this is not hard: one can argue that premise 1 is true even when one supplements P with full indexical "locating" knowledge (see below for more on this), and then a corresponding version of premise (3) still goes through.)
Despite Perry's extensive discussion of the two-dimensional framework, I was left unclear on which premise of this argument he would reject, and why. For reasons discussed earlier, he seems to be committed to a version of premise 1. Many of Perry's fellow type-B materialists reject premise 2 (often allowing that it applies in all standard cases, but holding that the case of consciousness is exceptional or unique). Perhaps Perry would reject this premise, but he does not discuss it explicitly, so it is hard to tell. Perhaps more likely, given Perry's sympathy for the idea that where there is apparent contingency, there is possibility, it is not out of the question that he accepts the second premise and rejects the third, holding that the relevant possibility does not trouble materialism. But again, it is hard to tell from Perry's discussion, as he does not address the argument directly, and he does not argue against the relevant premises.
Insofar as I understand Perry's response to the two-dimensional argument (on pp. 198-200), it focuses on the status of an identity such as "Q_R = B_47, where Q_R is the concept of whatever sort of experience is caused by red things, and B_47 is the concept of brain states of a certain sort. He allows that the identity is "apparently contingent" (conceivable in the relevant sense), allows that there is a primary possibility in the vicinity, but explains the relevant primary possibility in terms of the fact that (i) Q_R might have picked out a different phenomenal property, and (ii) B_47 might have picked out a different sensation. In two-dimensional terms (modulo a few complications in the translation that I pass over here), it seems that Perry is accepting: (i) the primary intension of Q_R picks out a different phenomenal property in some centered worlds, and (ii) the primary intension of B_47 picks out a different sensation in some centered world. In effect, this is to explain the primary possibility in terms of the fact that Q_R and B_47 have distinct primary and secondary intensions. As such, it appears that Perry is here allowing premise (2) above, but denying the equivalent of premise (3): there is no argument from the relevant primary possibility to the falsity of materialism.
There are a few things to say here. First, my own argument here is not formulated in terms of identities, precisely because it is hard to mount an argument from the primary possibility of the negation of a physical-phenomenal identity to the falsity of materialism. (Perry characterizes my argument in terms of identity on pp. 188-89, but this is a mischaracterization.) When we invoke the full claim P&~Q, rather than the mere negation of an identity, the argument from primary possibility to the falsity of materialism is much more straightforward. So Perry is addressing something of a straw figure here.
Second, Perry invocation of the relational phenomenal concept Q_R is very odd in this context. This is a deeply extrinsic characterization of the phenomenal property in question, with a very different primary and secondary intension. For the anti-materialist argument it is much better to invoke something closer to a pure phenomenal concept, to accommodate Kripke's insight that for phenomenal concepts, there is no gap between reference-fixers and reference (or between primary and secondary intensions). Perhaps Perry denies that there are any such phenomenal concepts; but if so, he needs to give substantive arguments against the Kripkean claim. In any case, it is clear that by appealing to the relational concept, Perry is making things easy for himself. (At the very least, I would have expected to see the demonstrative concept this_I.) If there are phenomenal concepts with the same primary and secondary intension, then Perry's strategy here will fail. (In fact, even if phenomenal concepts have different primary and secondary intensions, the anti-materialist argument still goes through, but things are more complicated.)
Third, Perry appeals to the possibility that there are centered worlds in which (the primary intension of) B_47 picks out a different sensation. It is hard to see how this is supposed to go, at least if B_47 gives a full microphysical description of the relevant state. As such, it seems that there is no centered world in which B_47 picks out a different physical property. And presumably Perry holds that there are no centered worlds in which that very physical property is a different sensation (that would deny the necessity of identity). One way out might be to endorse the Russellian claim that physical concepts are topic-neutral, fixing reference to different underlying intrinsic properties depending on the state of the world. But this leads only to the "panprotopsychist" loophole, which I don't think Perry is exploiting. So it is hard to see how this move works for him.
I would have expected Perry to embrace an alternative strategy, suggested by his response to the knowledge argument. If A is "I am in the USA", then P&~A is conceivable and primarily possible, but the the primary possibility simply corresponds to the actual world with a different center. Such a centered world yields no argument against materialism. By analogy, Perry might hold that P&~Q is conceivable because of a tacit indexicality in Q, and is primarily possible, but the primary possibility simply corresponds to the actual world with a different center, again yielding no argument against materialism. Here the center would presumably need to be expanded beyond an individual and a time, but that could be justified if phenomenal concepts were truly analogous indexical concepts, so that phenomenal information is a sort of "locating information". For example, one might supplement the "you are here" and "now is here" arrows at the center with "this is here" arrows pointing to certain states as the referents of phenomenal demonstratives. Then the relevant centered world will simply be a world with different locations for the "this is here" arrows.
I think that this move does not work, for more or less the reasons given above. First, phenomenal concepts are not indexical or demonstrative concepts, for reasons given there. Second, phenomenal information does not disappear from the third-person viewpoint, so it is not locating information (it is epistemically objective information), and so not the sort of thing that can be built into the center of a world. Third, while Mary is in the black and white room, one can supplement her physical information with all the locating information one likes — "you are here", "now is here", and "this is here" with arrows pointing to various of her brain states — and she will still be in complete ignorance of what it is like to see red. So unlike the case of my physical location, the epistemic gap here is not closed by adding any amount of locating information. Still, this is an interesting strategy that is worth exploring, especially in the context of Perry's response to the knowledge argument.
In any case: it is possible that I have misunderstood Perry's intended response to the modal argument here. So I would be interested to see how he responds to the straightforward three-premise argument above.
Perry's responses to the anti-materialist arguments are philosophically deep and sophisticated. Nevertheless, I think that upon close examination, these responses fail in straightforward ways. Given Perry's depth and sophistication, I take this to be indirect evidence that the arguments are sound.
Chalmers, D.J. 2002. Does conceivability entail possibility? In (T. Gendler and J. Hawthorne, eds) Conceivability and Possibility. Oxford University Press. [consc.net/papers/conceivability.html]
Chalmers, D.J. 2003. The content and epistemology of phenomenal belief. In (A. Jokic and Q. Smith, eds) Consciousness: New Philosophical Perspectives. Oxford University Press. [consc.net/papers/belief.html]
Perry, J. 2001. Knowledge, Possibility, and Consciousness. MIT Press.
In his "Replies", Perry replies to my comments on the zombie argument, the knowledge argument, and the modal argument. I will take these one at a time.
(i) The zombie argument: Here I think Perry concedes my point that the conceivability-possibility argument does not presuppose epiphenomenalism. Instead, he makes it clear that he will deny the relevant conceivability-possibility premise. This reduces the issue here to the issue discussed under the modal argument: see below.
(ii) The knowledge argument: Here Perry makes clear that he accepts the distinction betweem what I call a demonstrative phenomenal concept and a pure phenomenal concept. Perry uses "this experience" for the former, and "the experience of seeing wow" for the latter ("while using "the experience of seeing red" for something like what I call the community relational phenomenal concept). I think I can agree with the spirit of most of what Perry says until the last paragraph of his reply here, when he brings the distinctions to bear on what Mary learns.
Here, Perry says that what Mary learns about her experience is: "This experience is the experience of seeing red". But my point was: while Mary does learn this (that is, she does gain knowledge equating the referents of demonstrative and community relational concepts), this is not the crucial knowledge that Mary gains. The crucial knowledge that Mary gains involves a pure phenomenal concept: in Perry's terms, it involves concept associated with "the experience of seeing wow". So Mary's key new knowledge involves the knowledge that red things cause the experience of seeing wow, i.e. that the experience of seeing red is the experience of seeing wow, and so on.
Then, my original worry was: insofar as Perry's argument relies on analogous with indexical knowledge, those analogies will work for the new knowledge associated with the demonstrative phenomenal concept ("this experience"), which works in a broadly indexical fashion, but they won't work for the new knowledge associated with the pure phenomenal concept ("the experience of seeing wow"), which is not an indexical concept at all. (For the reasons for this, see "The Content and Epistemology of Phenomenal Belief".)
I don't know what Perry would say to this: maybe he'd say that the latter concept also works in an indexical way, or maybe he'd say that his reply to the knowledge argument doesn't really rest on the analogy with indexicals. The latter is suggested to some extent by his reply, when he says that the new knowledge Mary gains is really knowledge at the level of reflexive content. It may be that Perry thinks this goes through regardless of whether the relevant knowledge is indexical. If so, this raises a different set of issues. See "Phenomenal Concepts and the Knowledge Argument" for some discussion of issues in the vicinity.
(iii) The modal argument: Here Perry makes a number of points:
(A) In developing the concept of primary intension Chalmers does not follow his own explanation of what primary intensions are (functions that model actual-world mechanisms of extension/reference), but instead takes primary intensions to be non-type-reflexive modes of presentation.
In reply: Perry seems to take the claim that primary intensions are reference-fixers as definitional of primary intensions, but I don't take it that way: the notion of a reference-fixer, or a reference-fixing mechanism is too loose and ambiguous to be helpful in giving a definition. Rather, it's better to define primary intensions in other terms, and see what sort of link to reference-fixation (and of what variety) emerges. In particular, I now think that it's clearly best to define primary intensions in epistemic terms (see lots of recent work on this, including "The Foundations of Two-Dimensional Semantics", on which Perry commented at the APA in March 2002.).
In any case, the substantive issue is independent of this. Perry thinks primary intensions would need to be type-reflexive in order to do the work I need them to do (so the primary intension of 'water' would have to be defined in terms of some relation to the expression type 'water'), but he thinks I regard primary intensions as non-type-reflexive (so the primary intension of 'water' is something like: the watery stuff around here).
In fact, my view is that primary intensions are sometimes type-reflexive and sometimes not. They're certainly not type-reflexive by definition (unlike Perry's reflexive contents), and there are plenty of expressions whose primary intensions are not type-reflexive at all, but there are plenty of expressions whose primary intension involves some sort of metalinguistic reflexivity. To a first approximation, this occurs whenever a user uses a term with an element of deference to the community: roughly, they use 'T' with the dominant intention of intending to refer to whatever other people in the community call 'T'. In these cases, one expects a primary intension to have a metalinguistic element.
Note that this isn't built in to the definition of a primary intension, but emerges as a consequence in specific cases, including many of the cases that Perry points to as cases that need reflexive modes of presentation. In particular, Perry's key example — the case of 'Gödel' — is precisely such a case. See section 8 of "On Sense and Intension" for discussion of just this case, and of other relevant issues.
(B) If we follow Chalmers' practice rather than his explanation, and take primary intensions to be non-type-reflexive modes of presentation, it is totally implausible to equate conceivability with contingency of primary intension.
Reply: As above, I don't say that primary intensions are never type-reflexive, so this criticism doesn't apply to my view. To support his argument Perry uses the case of 'Jonah is King Tut'. I agree with Perry that the 'Jonah is King Tut' case is one (like the 'Gödel' case) where the relevant primary intensions will have a metalinguistic element. With this understood, the conceivability-possibility inference will go through in these cases.
Perry also says that even in cases (such as 'water') where primary intensions can be understood in a non-metalinguistic way, the conceivability-possibility inference is questionable. His reasons here are a bit unclear. He says it's conceivable that watery isn't the watery stuff: e.g. it's conceivable that water is found in lakes but not in rivers (he gives a specific case). I'm happy to agree that this is conceivable: at best it makes the familiar point that any descriptive chaacterization of a primary intension (e.g. as "the watery stuff") is imperfect, and what really does the work is evaluation of specific cases. But in any case, it's hard to see how this undermines the conceivability-possibility inference: in this case, when Perry finds "water is not watery stuff" conceivable, there seems to be a genuine possible world nearby (the one Perry describes, in which a stuff falls from the skies to the rivers and is transformed into another substance before it reaches the lakes and is consumed), satisfying the primary intension of "water is watery stuff".
(C) Even if we ignore (A) and (B), there is a problem with Chalmers' treatment of sensation terms like "pain". He doesn't choose, for the primary intension of "pain", the actual-world mechanism of reference. Nor does he find some non-type-reflexive mediating mode of presentation. He simply takes pain itself to be the primary intension of pain. This choice is unmotivated, and in fact quite baffling.
It's true that I think the primary intension of pure phenomenal concepts pick out the corresponding phenomenal qualities in all world, so that the primary intension of pure phenomenal concept of "pain" picks out the phenomenal quality of pain in all worlds. (I'd prefer to put it this way than to say that the primary intension of "pain" is pain, which is the way Perry puts it. I don't think pain is an intension, for a start.)
This thesis isn't needed for the argument, though. As I've said in a number of places, the anti-materialist argument goes through even if this thesis is false. In particular, the thesis played no role in the argument given above. There, all that mattered was the claim that a world that satisfies the primary intension of ~Q is different from our world. So Perry is wrong to say that this thesis plays a crucial role in my argument.
As for motivation for the thesis, see "The Content and Epistemology of Phenomenal Belief", where I discuss the motivation in depth. It comes down to the special way that pure phenomenal concepts behave across epistemic possibilities.
Perry thinks that the primary intension associated with a "pain" concept ought to be somthing like "the thing that people around here call 'pain'". I think that we have concepts with primary intensions like that, but they aren't pure phenomenal concepts. It may also be that there's a different way of understanding primary intensions (as akin to Perry's reflexive contents) on which even a pure phenomenal concept wlil have a primary intension of this sort. But if we stipulate that primary intensions are to be understood in the way I prefer, i.e. in terms of evaluation of epistemic possibilities, then I think it's clear that the primary intension of pure phenomenal concepts aren't metalinguistic in this way.
Perry raises the interesting case of someone with the phenomenal concepts of smelling cinnamon (C) and smelling nutmeg (N), but for whom the smells have faded. He thinks that it could be conceivable for them that C=N while it's still false that C=N, so that (if identities are necessary and if phenomenal concepts have the same primary and secondary intension) no world satisfies the primary intension of C=N.
In reply: it seems to me that in the case as Perry conceives it after the memory of the smells has faded, the relevant concepts are not pure phenomenal concepts any more. The sense in which it's conceivable that the smell of cinnamon is the smell of nutmeg, I take it, is that it's conceivable that the original smell of cinnamon was the same as the original smell of nutmeg. But these clearly aren't pure phenomenal concepts: they're past-directed concepts of a fine-grained property, whose primary intension nonrigidly picks out (in a given centered world) the smell that one used to smell as cinnamon in that world. After the memory of the smell has faded, the subject may have no pure phenomenal concepts at all. At best they'll have very coarse-grained concepts of a coarse-grained phenomenal quality. If N and C are supposed to work this way, then either they're concepts of the same coarse-grained range, in which case it will be a priori that if something has N it has C, or they pick out different coarse-grained ranges, in which case it will be both conceivable and possible that something has N without having C (or vice versa).
(D) Finally, even if we follow Chalmers' choice for the primary intension of sensation words like "pain", the rest of his argument doesn't work. It is based on ignoring the type of physicalism I advocate, that is, the identity theory.
Here Perry rejects premise 3: "If P&~Q is primarily possible, then materialism is false." His grounds are that an identity theorist could hold that even if the primary and secondary intensions of Q coincide, the primary and secondary intensions of P might not: the secondary intension of P could require certain qualities that the primary intension do not require, so that P&~Q could be primarily possible but not secondarily possible.
In response: I discuss this loophole at a number of places: it's mentioned briefly in my reply to Perry (just after the formal argument is set out), discussed a bit in the book (section 4.2), and discussed further in "Consciousness and its Place in Nature" (sections 6 and 11). The upshot of these discussions is while this position is available, it requires embracing a strong and distinctive claim about the categorical basis of microphysical dispositions, the claim I call "panprotopsychism" or "type-F monism". In effect, this is the variety of identity theory that Grover Maxwell advocated in his "Mind-Body Identity and Rigid Designators". I don't have any argument against this sort of identity theory, and in fact I'm fairly sympathetic to it. But nothing Perry says elsewhere suggests any sympathy with it.
General: In his general comments Perry says that my anti-materialist reasoning stands or falls with "ignoring identity in favor of supervenience" and my committing the "subject matter fallacy". I don't think the former point can be right: even Perry will agree that if I have a good argument against supervenience of the phenomenal on the physical, then I have a good argument against materialism. Perhaps he just means to make the point in (D) above: that my argument against supervenience fails because I overlook distinctive resources available to the identity theorist in grounding supervenience. If so, see my reply above.
As for the "subject matter fallacy": I passed over this in my commentary as I'm not sure what it's supposed to come to. The basic "fallacy" is the claim that new knowledge always should be understood in terms of the things in the world that the knowledge is of, as opposed to knowledge concerning the way that thought and language hooks on to the world. Now it's not clear to me that such a view is obviously a "fallacy", but in any case, I didn't think that I held it. After all, it's a central part of my own view that new knowledge may well involve new modes of presentation of the things that the knowledge is about (i.e. a new primary intension associated with an old intension). I'd have thought that giving this role to modes of presentation avoids the "fallacy" in question.
Perhaps Perry understands the "fallacy"
more broadly so that even an appeal to (broadly Fregean) modes of
presentation is an instance of the fallacy: if so, I'll simply deny
that there's any problem with such an appeal. See in particular "On Sense and Intension" (especially
sections 6-8) for my take on how a broadly Fregean view can handle the
central problems that worry Perry, including problems about names and
indexicals.