Contribution
It gives, in Block’s words, the book’s main claim and its structure, and updates it in one direction: since 2022 the working-memory evidence has made Block doubt that “raw” perceptual representations are used in cognition at all. The border between seeing and thinking is then a border of format, policed by recoding when percepts enter working memory.
Key insight
Read the global workspace not as what makes a percept conscious but as what puts a percept to cognitive use. Percepts compete in sensory areas; winners trigger frontoparietal “ignition”; the reverberating front-and-back coalition is the “cognitive envelope”, in which perceptual content is held in an abstracted, partly conceptual form for reasoning (Figure 1, pp. 274–275).
Assumptions
- Perception is constitutively iconic etc.; cognition only paradigmatically discursive (p. 273).
- Active working memory is conscious and maintained by broadcasting; “activity-silent” retention of unconscious information for up to 4 s is set aside as a different phenomenon (pp. 275–276).
- Content-based reasoning requires working memory, because premises must be held together (p. 276).
Key results
- Infant reasoning (p. 276). Hamlin’s helper/hinderer results, including preference for “the hinderer of the hinderer”, give “some reason” to credit 6–8-month-olds with content-based reasoning in working memory, though leaner iconic accounts are not ruled out.
- Nine differences between working memory and perception (pp. 277–279), listed in the LIT; Block notes they “do not show a change in format”, and a change could be within iconic representation.
- Iconicity (pp. 279–280). Three conditions: representations respond to environmental differences by degree; changes in them change what they represent by degree; relations among properties are mirrored by analogous relations among representations. Burge’s notion adds a requirement of naturalness, which Block rejects. Gross’s upper/lower-case adaptation challenge is answered by phonological and lexical association.
- Iconic thought (pp. 280–281). Carey’s parallel-individuation system (capacity about 3 items in infants, 4 in adults) and entropy-based same/different matching are iconic, but enclosed in a cognitive envelope, so not “entirely non-conceptual or non-propositional”; Block concedes he “might have scanted the extent of iconic representation in thought”.
- Attribution and discrimination (p. 281). Still no fact of the matter which is more basic, but now for different reasons, and “perceptual discrimination” is “importantly ambiguous”.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Perception is constitutively iconic, nonconceptual and nonpropositional | stated, argued in the book | p. 273 |
| C2 | The global workspace is better read as conceptualisation than as consciousness | assertion here | pp. 274–275 |
| C3 | Working-memory representations of percepts are recoded abstractions, not held percepts | moderate | nine cited lines of evidence, pp. 277–279; Duan & Curtis 2024 |
| C4 | Iconic thought in infants and animals sits inside a cognitive envelope | weak–moderate | pp. 280–281; concession that iconic thought may be underrated |
Concepts
- Cognitive envelope — the links from perceptual representations to cognitive faculties in the global workspace (p. 274).
- Iconic representation — a family of representations meeting the three tracking and mirroring conditions (p. 279).
- Active working memory — “active maintenance of representations for use in reasoning” (p. 275).
Connections
- The book (LIT-636), Block’s BBS target article “Perception is iconic, perceptual working memory is discursive” (2023) and the Analysis symposium (2023) are cited; none is read here.
- Dehaene’s global neuronal workspace supplies Figure 1.
- Ivy & Mroczko-Wąsowicz (LIT-099, NOTE-150) cite the book for a border without penetration.
Bearing on the record
- THEORY-102’s finding of sustained ventral content and an onset-only prefrontal event is compatible with, and on Block’s reading predicted by, the cognitive-envelope view. A THEORY combining the two would need the book’s arguments, not only this précis.
- No instruction for machine-learning practice.
Limitations
- An update, not a précis: the arguments for the main thesis are not here.
- The working-memory evidence is cited, not assessed.
- The cognitive-envelope reading of the workspace is asserted, not argued against the consciousness reading.
Open questions
- Does recoding in working memory change format (iconic to discursive) or only grain within iconic format? Block leaves both open (p. 279).