Ned Block (2011), Trends in Cognitive Sciences 15(12), 567–575, an Opinion article.
Key takeaways
- The argument restated. In Sperling’s task subjects report 3–4 of 12 letters but 3–4 from any row cued after offset, so about 10.5 letters’ worth of specific information is held. The overflow argument says that information is conscious. Overflow means that “necessarily most are unaccessed”, not that any item is inaccessible, as not everyone can win a lottery that each can win.
- Generic and fragment illusions. Critics (Kouider, de Gardelle, Dehaene; Cohen and Dennett; Phillips; Stazicker; Lau and Rosenthal) hold that only generic letter-likeness, gist or fragments are conscious before the cue, and that the specific information is unconscious until attention amplifies it. Block argues that “there is no independent evidence for highly detailed unconscious iconic memory”, and that conscious generic letters with no specific shape are hard to make sense of.
- Kouider et al.’s pseudo-letter experiment used low contrast and a mask that cut recall to 1.47 letters per row (4.41 of 12), which by the usual criterion leaves no icon. Its illusion rate in uncued rows was 10–15%, which a rich view with feature fragments can accept.
- Fragile visual short-term memory. The Amsterdam group’s third store: cortical (no rod effect after 1 s, unaffected by a light flash then, visible in V4 and not V1–V3), with capacity 7, 6 and 5 of 8 at 1, 2.5 and 4 s against a working memory of 4. Withdrawing attention, and TMS to right dorsolateral prefrontal cortex, reduce working memory and leave fragile VSTM almost intact.
- Unconscious working memory is weak. Soto et al.’s subjects who saw “nothing” compared a grating 5 s later at slightly above 55%, far short of what an unconscious store would need to do the Amsterdam tasks.
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 a theory of perceptual capacity and consciousness, and the paper carries no instruction for machine-learning practice.
It extends Block (2007), LIT-646. It defends that paper’s overflow argument against the objections raised in its commentaries and after, and it replaces the 2007 picture of iconic memory with three stores, pure iconic memory, fragile VSTM and working memory, of which the second carries the argument.
Its chief opponents, Cohen and Dennett’s “Consciousness cannot be separated from function” (2011) and Kouider et al.’s “partial awareness” (2010), are not held. Dennett’s position on the question is represented in the record only by “Quining Qualia” (LIT-433) and “Time and the Observer” (LIT-431), which the 2007 paper rivals.
Filed as THEORY-103: perceptual consciousness overflows cognitive access. This paper and its predecessor are its sources.