For more than a decade Ned Block has been defending the thesis that phenomenal experience can occur without cognitive access. In his target article here, he compellingly argues for the more restricted view that phenomenal experience outstrips accessibility in a narrow sense. In Block's terminology, a perceptual state is narrowly accessible if it has been encoded in working memory (or encoded in a “global workspace” in frontal cortex). It would be more perspicuous to describe such states as “accessed,” rather than “accessible.” In contrast to these, there are states that are accessible but not yet accessed, and states that are totally inaccessible (Dehaene & Nacchache Reference Dehaene and Naccache2001). Therefore, we may ask whether consciousness outstrips mere accessibility in addition to access. Block is largely agnostic about this question here, but he implies that phenomenology outstrips accessibility; his methodological claims seem to entail this conclusion, and his phenomenal/access distinction may depend on it (Block Reference Block1995b). In this commentary, I argue that Block can grant the existence of inaccessible phenomenal states only on pain of multiplying the number of phenomenal states beyond plausibility.

Elsewhere I argue that phenomenal consciousness requires attention (Prinz Reference Prinz, Brook and Akins2005; Reference Prinz2007; forthcoming). I define attention as the process that allows representations in perceptual centers to gain access to working memory. Attended representations are accessible, but not necessarily accessed, as confirmed by studies of divided attention and passive movie watching (Goldberg et al. Reference Goldberg, Harel and Malach2006; Kouider et al. Reference Kouider, Dehaene, Jobert and Le Bihan2007). If this account is right, the neural correlate of consciousness is the neurocomputational process that underwrites accessibility. That process might be a distinctive pattern of neural activation, or perhaps activation levels above a critical threshold. When the process occurs, the affected neurons can send afferent signals to working memory structures in frontal cortex, but task demands, lateral competition, and other factors determine which neurons in the window of attention propagate beyond perceptual centers and get encoded in working memory. Thus, there is a three-way distinction between unattended, attended, and encoded, or inaccessible, accessible, and accessed.

Block correctly concludes that phenomenology outstrips working memory encoding, but he underestimates the behavioral evidence for the view that accessibility is necessary for phenomenology. Research on inattentional blindness, attentional blink, extinction, and visual neglect strongly suggest that phenomenal experience disappears when attention is withdrawn. Block seems to disagree. He implies that phenomenology can occur without attention (consider his discussion of extinction and his enthusiasm for the view that reentry is the neural correlate of consciousness). Block implies that cases of inattention are comparable to Sperling cases (and standard change blindness experiments), in which participants report seeing a stimulus but cannot identify it. He may think that we have phenomenology in both cases. If so, this is a fatal flaw in his position. Sperling cases and cases of inattention are radically different. In cases of inattention, participants insist that they have seen nothing, and they are at chance in guessing whether there was an object presented in their unattended fields. There is no reason to postulate phenomenology under total inattention, and overwhelming reason to postulate phenomenology in Sperling cases (and cases of change blindness), where subjects attend to the stimuli and have vivid experiences, despite limitations of reportability. In my terminology, Sperling cases involve accessibility without (complete) encoding, and inattention cases render stimuli inaccessible. Therefore, consciousness depends on accessibility.

Block might counter with a methodological objection: The claim that phenomenology disappears under inattention is based on what participants report in the relevant studies, and taking reports as authoritative begs the question against those who believe that phenomenology outstrips accessibility. Against this, I have two replies. First, the evidence does not rely solely on reports. There are dramatic subjective differences between inattentional blindness and Sperling cases; anyone who has seen both kinds of stimuli can experience the contrast. There are presumably also psychometric and physiological differences that correlate with the subjective differences, and these can be used to test for phenomenology when reports are unavailable or unreliable (as with aphasia or anosagnosia). Under non-pathological conditions, every subject can subjectively confirm that when phenomenal experiences occur, we can at least report that something was experienced. That gives us reason to trust such reports in others. Reports can be subjectively confirmed.

Second, Block implicitly relies on reports in defending his own conclusions. For example, he dismisses the view that activations in low-level visual areas are conscious by appeal to the fact that there are low-level activations corresponding to unseen stimuli in binocular rivalry studies. But the assumption that participants in rivalry studies are not experiencing both concurrently presented stimuli is based on what they report. Block must either concede that reports are authoritative or deny that we can rule out the possibility of conscious states in V1 and earlier visual areas.

In response, Block should reject the conjecture that both stimuli are conscious in binocular rivalry on subjective grounds: When you participate in such a study, you feel like you are seeing just one stimulus. The subjective method of determining which states are conscious does not presuppose that consciousness depends on reportability; even if I don't trust the reports of other participants, I can participate in the studies and see for myself. If Block grants that subjective confirmation is authoritative, then he should concede that there is no experience of stimuli under conditions of inattention. If he denies that subjective confirmation is authoritative, then he will have no way of ruling out that we have conscious experiences of both stimuli in binocular rivalry – indeed, Block will have no way of ruling out the possibility that there are phenomenal states in the LGN and the retina. On the first horn, Block would have to concede that phenomenology requires accessibility, and, on the second horn, he would proliferate the number of phenomenal states beyond plausibility. In sum, I think Block can postulate inaccessible phenomenal states only on pain of rendering every visual response conscious. The dramatic subjective distinction between Sperling cases and cases of inattentional blindness undermines efforts to establish phenomenology without accessibility.