Source
- Block (2007), Consciousness, accessibility, and the mesh between psychology and neuroscience, LIT-646, read in NOTE-497: §§8–12 (the method, Sperling, Landman et al., Sligte et al., the refrigerator-light reply and working-memory capacity). Target article only.
- Block (2011), Perceptual consciousness overflows cognitive access, LIT-640, read in NOTE-500: the restatement, the generic and fragment illusions, fragile VSTM, Soto et al., Box 2 and the two fallacies.
What was actually shown
The data both sides accept. Block (2007), LIT-646, §9, reports that in Sperling’s arrays of 9–12 letters subjects “said that they could see all or almost all of the characters” and could report only about 4, yet when a tone cued a row after the array went off they could report “all or almost all the characters in any given row”. In Landman et al.’s task with 8 rectangles, capacity (Cowan’s K) is 4 when the cue comes with the test array, almost 8 when it comes with the first array, and “between 6 and 7 for up to 1.5 seconds after the first stimulus has been turned off” when it comes in the blank. Sligte et al. extend this to 32 items and 4 s (§10). Working memory, across many paradigms, holds about 4 items, with “neither visual working memory system [having] a capacity higher than four” (§12). So subjects hold specific information about more items than they can access at once. That much is not in dispute.
The surplus is not retinal. Block (2011), LIT-640, splits the old “iconic memory” into a rod-driven retinal persistence of a few hundred milliseconds and a cortical fragile visual short-term memory. With cues up to 4 s after offset, its capacity is “7 of 8 items at 1 second, 6 of 8 at 2.5 seconds and 5 of 8 at 4 seconds – while working memory remains steady at 4 items”. A light flash before the cue affects only the earliest cue, and Sligte et al. image the persisting representation in V4, not V1–V3. Withdrawing attention, by three methods, and a TMS pulse to right dorsolateral prefrontal cortex reduce working memory and leave fragile VSTM almost intact. So the high-capacity store is a distinct system, neither the retina nor working memory.
The step that makes it overflow. The claim is that the surplus is experienced. Block’s case is cumulative (LIT-646, §11; LIT-640). Subjects’ introspective judgement that they see every item does not depend on being asked about any item. It predicted partial-report superiority, which was borne out. Subjects in Landman et al. recover size and orientation together, as from an image. The rival reading, that only a generic letter-likeness or fragments are conscious and the specific information sits in an unconscious store until the cue amplifies it, needs “highly detailed unconscious iconic memory”, for which Block finds “no independent evidence”. The one direct test of unconscious working memory he cites (Soto et al.) gives comparisons “slightly above 55%” at 5 s for a single grating rated unseen, far below what the Amsterdam tasks require. Hence “it is reasonable to tentatively accept the overflow interpretation” (LIT-640).
Accessible, not accessed. Block (2011) is explicit that overflow “does not claim that any of the items in the array are cognitively inaccessible, but rather that necessarily most are unaccessed”, as not everyone can win a lottery that each can win. The title keeps that reading. Block (2007), §9, already notes that this argues for less than his 1995 paper (LIT-643), which took Sperling to show inaccessible phenomenal states directly.
What could have come out otherwise: cued capacity could have fallen to working-memory level within the first few hundred milliseconds, as a retinal afterimage would, or could have been reduced by the same attentional and prefrontal manipulations that reduce working memory. It did neither.
The rival accounts
The sparse or illusion view, that subjects mistake potential access for
actual experience (Dehaene’s “refrigerator light”), or a generic image
for a specific one (Kouider et al.; Cohen and Dennett), is the account
this one excludes. The record holds no theory stating it as such. Parts
of Dennett’s position are now stated: “Quining Qualia” (LIT-433) as
THEORY-415, and Taylor’s real-pattern reading, with the
Orwell/Stalin indeterminacy, as THEORY-563; Frankish’s illusionism
is THEORY-633. None states the timing argument of Dennett and
Kinsbourne’s “Time and the Observer” (LIT-431), which is the one this
account answers, so no rival is declared. Block’s 2007 paper rivals
LIT-431 and LIT-433 at the level of the work. LIT-431 argues that whether a content
was experienced and then revised, or revised before it was experienced,
has no answer, because every datum fits both. Block argues that the
parallel question for Sperling, whether the uncued letters were seen
before the cue, has an answer that inference to the best explanation gives
(LIT-646, §§7–8). LIT-433 concludes from the subject’s inability to
tell two hypotheses apart that the notion they share is empty, the move
Block rejects by name. Neither is a theory, and Cohen and Dennett’s
“Consciousness cannot be separated from function” (2011), Block’s main
target in 2011, is not held, so no rivals is declared against Dennett.
The workspace side of the sparse view is now held as a theory: THEORY-096, from Mashour et al. (LIT-631), which takes global availability to be what is experienced and reads experience without access as untested. That is this account’s rival, and the relation is declared on that file. The two agree on the capacity data and disagree on whether the unaccessed surplus is experienced, which is the premise each adopts.
What this does not say
- It does not say the surplus is inaccessible. Any one item can be accessed when cued. It says that, necessarily, most are not.
- It does not rest on reports alone, but it does rest on them. The capacity data are not in dispute. That the surplus is experienced rests on reading subjects’ reports at face value, which is exactly what the illusion view denies. Block’s case against the illusion is cumulative, and he calls the issue “certainly still open” (LIT-640).
- It does not say fragile VSTM is shown to be conscious. Block concedes that the choice of the cued object “could” be unconscious in principle, and answers that “we will not get anywhere in consciousness research by accepting the less plausible account”. That is a plausibility argument (C5 in NOTE-500, weak to moderate).
- It does not say where experience is in the brain. The step from overflow to a claim about the neural basis is the mesh argument, filed separately as THEORY-098.
- The evidence is second-hand. Every experiment is as Block reports it, in an opinion article and a target article. The 32 BBS commentaries, where the objections were made, were not read.
Connections
- THEORY-098 presupposes this account: if the neural basis of experience excludes the workspace, losing sensory coalitions are the mechanism of overflow.
- THEORY-094 states the methodological problem this account answers by inference to the best explanation rather than by isolating a correlate.
- THEORY-102 reports a stable ventral temporal code for a seen image while it stays on. Fragile VSTM is placed in V4, the same territory, but after offset and for unaccessed items, which that study does not test.
Note of 2026-10-29: what holds it back
The step from held surplus to experienced surplus has no result behind it. The record holds no recording showing that the uncued items’ representations are recurrent like seen items’ rather than feedforward like masked ones, and no first-hand reading of Sperling, Landman et al., Sligte et al. or the critics Block answers. Block calls the issue “certainly still open” and rests the step on a plausibility argument (C5 in NOTE-500).