Block's argument against correlationism depends in part on the assumption that subjects in the central experiments (i.e., Sperling 1960, Landman et al. 2003, and Sligte et al. 2008) are phenomenally aware of all or almost all items in the presented arrays. Block suggests that the main source of evidence for this assumption comes from introspection – that is, from subjects' introspective reports that they have seen all or almost all items during the initial brief presentation.

Someone who adopts the methodology of inference to the best explanation, as Block does, should be ready to defend a proposed range of data as relevant to his or her approach. And when it comes to the above-mentioned introspective evidence, at least, Block is sensitive to such an obligation. He defends his claim that we can take the reports in question at face value against the worry that subjects might confuse potential phenomenology with actual phenomenology.

But there is a more general worry about introspective judgements Block should consider. Introspective judgements might be reliable enough when it comes to off-the-cuff or approximate judgements about phenomenology, as for instance when one introspectively judges that it appears as if something is red. Visual phenomenology arguably does involve colour-related features and, arguably, introspection reveals this fact. For the purposes of theorizing about phenomenology, however, not much follows from this general fact. But introspective judgements are demonstrably unreliable about the subtle details of phenomenology, as the philosophical debate about the nature of perception makes perfectly clear. Participants in that debate disagree fiercely on how to characterize phenomenology on the basis of introspection. Views range from characterizations of phenomenology solely in world-directed terms to ones involving intrinsic properties of experience individuated independently of one's environment. The bone of contention here is not whether a given case involves actual phenomenology – on this most participants are agreed – but rather how to describe it accurately on the basis of introspection. I have argued elsewhere (Spener, under review) that the best explanation of such disagreements in the philosophical literature involve the idea that one's introspective grasp of phenomenology is thoroughly expectation-driven, and, for this reason, that introspection is thoroughly inadequate on its own to provide reliable knowledge of the finer details of phenomenology.

How does this worry affect Block's claim that we can believe subjects' reports that they saw all the items in the presented arrays? Note, first, that this particular bit of introspective evidence plays a crucial role in Block's argument. It provides the main reason to identify the experimentally elicited informational state with what is present to the subject phenomenally. This identification does not come for free, as demonstrated by the experimenters' explicit attempts to argue for it (see Sligte et al. 2008; Sperling 1960). Moreover, despite giving such arguments, Sperling is careful to say that “[s]hort-term information storage has been tentatively identified with the persistence of sensation that generally follows any brief intense stimulation” (Sperling 1960, p. 26, my italics). Without the introspective evidence, the claim that the detailed information about the array of items – that which is experimentally shown to be at the subjects' disposal – is indeed phenomenally present to the subject is on much weaker footing. In turn, this means that Block's argument against correlationism is on much weaker footing.

So, what sort of introspective judgements are we dealing with in Block's argument? Specifically, are they of a detailed sort and hence suspicious, or are they of an approximate sort and hence theoretically uninteresting? Given their role in Block's argument, it seems they have aspects of both. On the one hand, Block takes subjects to report introspective judgements about phenomenology which provide reason to identify the relevant phenomenal state with the short-term informational state demonstrated by the experiments. The latter concerns specific details about all the particular items in the presented array. This suggests that the judgements in some sense reveal detailed phenomenal information. On the other hand, though, subjects cannot exactly report their alleged phenomenal awareness of such plentiful detail, but can only “bring [these experiences] under a general concept” (sect. 9, para. 2). This suggests that the introspective judgements are of an approximate sort.

Recall that the introspective judgements in question are meant in some way to give good evidence for subjects' phenomenal awareness of all the items in the array, a kind of phenomenal awareness that can readily be identified with the informational state demonstrated by the experiments. I think there is trouble for that thought with respect to both the aspects just articulated. Consider the detailed-phenomenology aspect first. As the old debate about the problem of the speckled hen (for instance) makes vivid, the right way to characterize the phenomenology of an experience of a plurality of items based on introspection is a matter of great dispute (see, e.g., Chisholm 1942; Jackson & Pinkerton 1973). In this case, too, then there can be different introspective takes on the details of phenomenology. This makes room for my worry that introspection-based characterizations of the specifics of phenomenology are partly expectation-driven. They should not be regarded as straightforward, reliable evidence about phenomenology.

Consider now the approximate-judgement aspect of Block's introspective judgements. There clearly is a large gap between the general nature of the introspective judgement and the phenomenology of particular detail they allegedly reveal. That general introspective take on phenomenology could correctly capture several, non-equivalent ways that things look. It is hard to see how the general judgement could in turn provide evidence to single out one of those ways as captured by it.

By Block's lights, correlationism must be rejected because it places mistaken limits on the kind of data relevant and admissible to the enquiry. I suggest, more radically, that correlationism must be rejected because it limits itself to poor data. Introspective evidence should not be abandoned altogether, of course, but it also should not carry too much weight in an argument. I recommend spreading the weight equally, using introspective data alongside data from other sources, for instance, branches of science and philosophy in which perceptual experience figures prominently. The theory that fits such a motley bag of data best overall, should win our allegiance. I take this suggestion to be very much in the spirit of Block's methodological proposal.