Dan Dennett, in his seminal work, Consciousness Explained, says some confusing things about the ontological status of claims about consciousness. Sometimes, he seems to say there are facts about consciousness that we can get right or wrong; at other times he compares claims about consciousness to the claims of fiction writers about their fictional worlds -- claims that simply can't be wrong, any more than Doyle could be wrong about the color of Holmes's easy chair (CE, p. 81). The first strand tends to be emphasized by those who find Dennett appallingly (or appealingly) committed to the possibility of pervasive and radical mistakes about consciousness (Alva Noe calls Dennett the "eminence grise" of the new skepticism about consciousness), the second strand by people attracted (or repulsed) by the promise of an end to questions about what our "real" conscious experience is, underneath our reports.
Both strands of Consciousness Explained have their appeal, but I can't seem to reconcile them. One can't get it wrong in one's reports about one's consciousness, it seems to me, if there are no facts about consciousness underneath one's reports. Fiction writers can't make errors of fact about their fictional worlds.
I've written a paper outlining my confusion more fully, forthcoming in a special issue of Phenomenology and the Cognitive Sciences. Dennett has written a very gracious reply to my criticisms in sections 2-3 of "Heterophenomenology Reconsidered". Among other things, he suggests a helpful analogy. Imagine cavemen brought into the 21st century; to describe what they see, they'll be forced to use metaphorical language (a pencil might be a slender woody plant with a black center that marks square white leaves, etc.). I'm amenable to this way of thinking about our phenomenological reports: Despite the apparent nearness and familiarity of experience, our tools for thinking about it and our conceptualizations of it are very weak and primitive. But nothing Dennett says in his response seems to me to remove the tension between the we-often-get-it-wrong strand of his work with the we're-fiction-writers strand. The cavemen, of course, aren't fiction writers. They can't, like Doyle, make their claims true simply by uttering them. Of course they use metaphor, as Dennett emphasizes; but metaphor and fiction are two entirely different beasts.
When I pressed Dennett about this in an email, he responded with the interesting suggestion that I was thinking too narrowly about fiction. Not all fiction is novels; there are "theoretical fictions" like quarks (maybe) or functionalist homunculi. And of course the rules governing the use of theoretical fictions in science are quite different from those governing novels.
If Dennett really endorses this (and I don't want necessarily to hold him to a quick remark in an email), it seems to me represent a shift of position, given his earlier talk about Doyle and Holmes's easy chair. But I don't know if it entirely resolves the tension, or how appealing it is as a view. Claims involving theoretical fictions, for instance, probably should not be evaluated as true or false. Rather, they are helpful or unhelpful, provide an elegant model of the observable phenomena or don't. Is this really how we want to think about our phenomenological claims?
Rather than novelists or positers of theoretical fictions, I'd rather see the person reporting her phenomenology as like a witness on the stand. She aims (if sincere) to be speaking the literal truth; and her claims can come close to it or can miss the mark entirely. Perhaps, in some ways, she will be like a caveman asked to report a drive-by shooting, stuck with inadequate conceptions and vocabulary, forced to (witting or unwitting) metaphor; but there's still a realm of facts that render her claims, independently of her or our judgment, true or false or somewhere in between.
Addendum, August 10: Pete Mandik reminds me in a comment that Dennett has spoken at length about "theoretical fictions" in earlier writings on belief and desire attribution -- for example in his magnificent essay "Real Patterns" (1991) and in the Intentional Stance (1987). There, Dennett's examples of "theoretical fictions" are things like centers of gravity and equators, not quarks and homunculi. (Dennett cited no particular examples in his email to me.)
Now, I'm more inclined to think that claims about centers of gravity are literally true than claims about homunculi. So if that's the kind of thing Dennett has in mind, my last remark above may be off target. On the other hand, the rules governing "theoretical fictions" of that sort match very nearly those governing literal language. This brings us even farther from the Doyle, saying-it-makes-it-true model in Consciousness Explained.