David Papineau has put online his paper "Phenomenal and Perceptual Concepts", forthcoming next year in the Alter/Walter collection Phenomenal Concepts and Phenomenal Knowledge: New Essays on Consciousness and Physicalism. In the paper he extends and elaborates the "quotational concepts" view he put forward in his 2002 book Thinking about Consciousness, and uses it to answer various anti-materialist challenges. In the last section of the paper, he tries to use the account to answer the challenge I put forward in "Phenomenal Concepts and the Explanatory Gap", posted here earlier this year, and forthcoming in the same volume.
In that paper, I argued that accounts of phenomenal concepts that attempt to explain away the explanatory gap face a dilemma: either there is an epistemic gap between P (physical processes) and C (the relevant features of phenomenal concepts), or there's not. If the former, then the relevant features of phenomenal concepts can't be physically explained. If the latter, then there's an epistemic gap between C and E (the epistemic feature we face with regard to consciousness), so that C can't be used to explain our epistemic situation. In his paper, Papineau claims to embrace both horns (!) of the dilemma. A response (adapted from recent correspondence with Papineau) is after the fold.
Although Papineau says he takes both horns of the dilemma, I think it's clear that he really takes the second horn. His embracing of the first horn requires that C employs the concept "phenomenal concept", but this is explicitly excluded (in my section on the first horn), as employing that concept would trivially yield an epistemic gap between P and C. Rather, C should be cast in more general architectural terms. It's not stipulated that it has to be a physical/functional account: it may make reference to various general conceptual and epistemic features, as long as it doesn't explicitly employ the notion of the phenomenal.
Some of Papineau's discussion of the second horn is somewhat tangential. E.g. he appeals to the metaphysical impossibility of zombies, but this isn't relevant here, since all occurrences of zombies in the argument are within the scope of a conceivability operator, and Papineau accepts that metaphysical impossibility is compatible with conceivability. To take the second horn successfully, he needs to make the case that P&~E is not conceivable, where E is cast in general epistemic terms, and he doesn't quite try to do that here.
It looks like Papineau thinks that upon examination, any conceivable zombie will share the relevant epistemic features with us, though directed at "schmonsciousness" rather than consciousness. Again, I discuss this move in the paper (under option 3), and it's not really clear to me what his reply is to the discussion there (though I think he takes what I call the "inflationary" option). I presume that he'd say that there's some sort of transparent explanation of the zombie's (and our) epistemic features in physical and functional terms, so that we can't really conceive of zombies that lack them. But he doesn't really explain how this would go, or explain away the intuition to the contrary. Perhaps he might take the option of conceding conceivability while rejecting the link between conceivability and transparent explanation, but even so the same sort of issue arises.
To put things independently of conceivability: what's required is a clear architectural explanation of just why we should expect that the quotational (or use/mention) structure he invokes should ground direct, semantically stable reference to the quoted state (whether in conscious beings, silicon zombies, or whatever), without presupposing that the state is conscious, or simply assuming some sort of special acquaintance with that state, or anything like that. Prima facie, as I say in the last section of the paper, all one would expect this structure to yield is some sort of quasi-indexical or demonstrative reference that does not ground any substantive knowledge of the state. Certainly, one wouldn't expect semantically stable reference to such a state, any more than one would expect it for indexicals or demonstratives. Papineau needs to meet this burden both to address my argument and for his explanatory project to go through more generally. Of course I think that the burden can't be met, for the reasons I give in the paper. But it would be interesting to see an attempt.