Ned Block (2007), Behavioral and Brain Sciences 30(5–6), 481–499, a target article with open peer commentary and the author’s response (the section runs to p. 548). The issue is dated 2007 and was published online in March 2008.
Key takeaways
- The methodological puzzle (§§1–3). To learn whether unreportable states are phenomenal we need the neural basis of phenomenology in clear cases. But to fix that basis we must already decide whether it includes the machinery of cognitive access, and any evidence seems to come through access. Water does not raise the puzzle, because no one thinks our access to water molecules is part of what water is.
- Correlationism (§7). Metaphysical correlationism holds that the access machinery is part of phenomenology (Papineau; Crick and Koch on V1; Prinz’s AIR). Epistemic correlationism holds that the question is scientifically intractable (Dehaene and Changeux; Chalmers 1998; Kouider et al.). Block’s alternative is inference to the best explanation (§8).
- Overflow (§§9–11). In Sperling’s arrays subjects say they see all or almost all of 9–12 letters but can report about 4; cued, they can report most of any row. In Landman et al. the capacity for 8 rectangles is about 4 without a cue but 6–7 with a cue up to 1.5 s after offset, and Sligte et al. extend this to 32 items and 4 s. The persistence is cortical, not retinal. The capacity of phenomenology exceeds that of a working memory of about four items (§12).
- Against the refrigerator light illusion (§11). Dehaene and colleagues say subjects mistake potential access for actual phenomenology. Block replies that the impression does not depend on being asked about items, is not curable like cognitive illusions, is borne out by partial report and by Loftus’s grid task, and would be a “hyper-illusion” of a kind with no clear instance.
- The mesh (§§13–14). The control of working memory is prefrontal. The core basis of seen motion is a recurrent V1–V5 loop in the back of the head. Sensory coalitions that lose the competition for the workspace can be nearly as strong as winners (Kouider et al. 2007; extinction patient G.K.). If phenomenology does not include workspace activation, those losers explain overflow; if it does, nothing does.
Standing in the record
Filed on 2026-10-03 at the owner’s request, with six other papers by Block on consciousness. No anthology topic holds the methodology of consciousness science, and the paper carries no instruction for machine-learning practice.
It extends Block (2005), LIT-629. The phenomenal and access NCCs of that paper become the “core neural basis” of phenomenology and the workspace, and its MT/V5 evidence is restated in §13. This paper adds the overflow argument and the mesh argument for the separation that the 2005 paper could only call “a glimmer”. Block notes that it argues for less than his 1995 paper (LIT-643): not that Sperling directly shows inaccessible phenomenal states, only that phenomenology overflows access (§9).
It rivals Dennett and Kinsbourne (LIT-431). That paper argues (§2.2) that whether a content was experienced and then revised (Orwellian) or revised before it was experienced (Stalinesque) has no answer, because every datum, the subject’s included, fits both: “only a verbal difference”. This paper argues that the parallel question for Sperling, whether the uncued letters were seen before the cue or only seem to have been, has an answer and that inference to the best explanation gives it (§§8, 11). It names the general move it rejects: Putnam and Dennett “argue that two views of the facts about consciousness are ‘empirically indistinguishable’ – and then they in effect conclude that it is better to say that there are no such facts” (§7). Each account excludes the other, and neither tests the other.
It rivals “Quining Qualia” (LIT-433, read in NOTE-357). Block cites it (Dennett 1988) for exactly that move. “Quining Qualia” argues that the two surgical routes to an inverted spectrum are indistinguishable to the subject. Its dilemma is that if reactions constitute how things seem, qualia are relational, and if not, they are inaccessible to their owner; either way the traditional notion fails. This paper takes the second horn as a live empirical possibility. Phenomenology need not include the machinery of access, and whether it does is decided by evidence, not by its undetectability. It also rejects Dennett’s “first person operationism” by name (§8) as lingering verificationism.
Bearing on the new theories.
- THEORY-100 states that a report paradigm cannot separate access from phenomenal consciousness. This paper is where that problem is framed as “the methodological puzzle” and where the no-report paradigm’s motivation comes from. Block (2019), LIT-637, cites it as such and then argues the no-report paradigm does not solve it.
- THEORY-102. The mesh argument predicts that representations of unaccessed items involve recurrent activity in sensory cortex and need no sustained frontal activation (§14). A stable ventral temporal code that lasts as long as the image, with prefrontal content confined to onset, is the shape of evidence the argument asks for. Vishne et al. (LIT-644) test seen images only, so they do not reach unaccessed ones. The connection is mine.
Its interlocutors in the record: Chalmers is quoted, from a 1998 paper the record does not hold, as an epistemic correlationist. The coherence principle of his “Facing Up” (LIT-406) ties consciousness to availability for global control. Block answers Chalmers’s 1997 version of that claim: accessibility in the broad sense, which includes what could be accessed, “encompasses too much” (§11). That this answer reaches “Facing Up” is my inference. Butlin et al. (LIT-056) take indicators from the global workspace, which this paper argues is not part of the basis of phenomenology.