Contribution
The paper names and separates two things the word “consciousness” runs together: phenomenal consciousness (experience) and access consciousness (availability of a content for reasoning and the rational control of action and speech). It sets both apart from monitoring consciousness and self-consciousness. With the distinction in hand it diagnoses one argument form in the literature, from a deficit in “consciousness” to a function of experience, as a conflation. It also says what survives of that argument.
Key insight
When “consciousness” and some capacity go missing together, ask which consciousness is missing. In the standard cases both P and A are gone, and the capacity lost, using information to guide action, is just what access is. Crediting the loss to experience slides from an obvious function of A’s machinery to a non-obvious function of P. That P causes A is a coherent hypothesis. It is the slide that is the fallacy.
Assumptions
- Anti-functionalism about P (§3). “I take P-conscious properties to be distinct from any cognitive, intentional, or functional property.” Block says limited versions of his points survive for the functionalist (n. 16).
- Blindsight patients lack P in the blind field. Taken from the target reasoning, not argued. Block notes his argument does not depend on it: if it is false, the case reduces to the Searle–Penfield one (§6).
- Introspection as “the first word” on P (§4.2), though not the last.
- A is stipulated to be demanding enough that blindsight “guesses” do not count as access (§4). Block says there is “no right or wrong here”.
Key results
- A-consciousness (§4). A state is A-conscious if, in virtue of having it, a representation of its content is (1) inferentially promiscuous, poised for use as a premise in reasoning, (2) poised for rational control of action, and (3) poised for rational control of speech. The three are jointly sufficient but not all necessary. (3) has the least weight, so that non-linguistic animals can have A-conscious states.
- Three differences (§4). P content is phenomenal, A content representational (roughly; many phenomenal contents are also representational). A is functional and system-relative, P is not. There are P-conscious types (every pain feels like pain), whereas a thought accessible now may not be later.
- A without P (§4.1). The superblindsighter, trained to prompt himself, knows there is an X in the blind field but says he does not see it. His perceptual state is A-conscious and not P-conscious. Block knows of no actual case. The monkey Helen and patient GY are both doubtful.
- P without A (§4.2). The drill noticed at noon, heard all along: P-conscious throughout, A-conscious only at noon. Animals with intact P systems and damaged reasoning systems are conceptually possible, and the dorsal/ventral division (Goodale & Milner) suggests the two could be damaged separately.
- The fallacy (§6). Searle and van Gulick give no reason to think P is absent in petit mal automatism. Penfield’s own descriptions are of deficits in thinking, planning and deciding, “No mention is made of any sensory or phenomenal deficit”. For blindsight, the missing access explains why the thirsty patient does not reach. Searle contradicts himself, saying a “totally unconscious” epileptic can drive home and that a totally unconscious driver would crash: the second is A.
- Conflations named (§5). Baars, Shallice, Mandler, Edelman, Kosslyn and Koenig, Johnson-Laird, Kihlstrom et al., Andrade, Jacoby et al. and Schacter each say they are explaining experience and give an account of A, monitoring or self-consciousness. Baars’s global workspace is “a ‘global workspace’ model of A-consciousness”.
- Three models (§§6–7). Schacter’s (P is the gateway to the Executive System), the collapse hypothesis (P and the Executive System are one) and epiphenomenalism (the two transposed). Solutions that “pop into” P count against the collapse model, since the reasoning that produced them was neither P- nor A-conscious.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | P and A are conceptually distinct | moderate: argued from conceivable cases (superblindsight, the drill), not from data | §§4.1–4.2 |
| C2 | Inferring a function of P from blindsight or petit mal is fallacious | strong as a point about the argument form | §6; Penfield’s text quoted |
| C3 | The literature on consciousness routinely conflates P with A or self-consciousness | moderate: a dozen cited instances, read by the author | §5 |
| C4 | Dennett’s thesis that consciousness is a cultural construction is false of P and A and banal of sophisticated self-consciousness | moderate: textual argument against Dennett 1986, 1991 | §5 |
| C5 | There may be actual cases of P without A | weak by the author’s own account | §7: “don’t show anything on their own” |
| C6 | The collapse hypothesis is disconfirmed by high-level reasoning that is neither P- nor A-conscious | weak to moderate: introspective, and Block concedes variants of the model escape it | §7 |
Concepts
- P-consciousness: experience; “what it is like” to be in a state.
- A-consciousness: a content poised for use as a premise in reasoning and for rational control of action and speech.
- monitoring consciousness: inner perception, internal scanning, or higher-order thought about one’s own states (§4.2.2).
- self-consciousness: having and using the concept of oneself (§4.2.1).
- mongrel concept: a single word or mental representation that functions as one concept while covering several, inviting conflation (§4.2.2). Block contrasts it with a cluster concept.
- superblindsight: imagined blindsight in which the patient prompts himself and so has A without P (§4.1).
- Cartesian modularism: a model in which P is one module (§2), named after Dennett and Kinsbourne’s Cartesian materialism.
Connections
- Dennett and Kinsbourne (LIT-431, read in NOTE-367). Cited in note 2 for Cartesian materialism and for their “scorn” at Cartesian modularism. Block reads Multiple Drafts as a theory of A-consciousness (§5), which fits the record’s reading: NOTE-367 notes that the article accesses subjective sequence only through report.
- Chalmers, “Facing Up” (LIT-406, read in NOTE-336). Not cited. Chalmers’s verdict that Baars’s workspace “shows promise as a theory of awareness” and leaves experience unexplained is the same verdict as Block’s §5 on Baars, in different words: Chalmers’s “awareness” is close to Block’s A.
- Frankish (NOTE-143) cites this paper as defining phenomenal consciousness against function, and takes access consciousness to be “the whole story”. That is the view the paper’s §4 notes the functionalist can hold while still granting A without P.
- Butlin et al. (LIT-056, read in NOTE-052) take phenomenal consciousness as their sole target and distinguish it from access, in this paper’s terms.
Bearing on the record
- THEORY-100. The theory’s caveat, that report paradigms cannot
separate access from phenomenal consciousness, is this distinction. Its
promote_whenasks for a no-report paradigm in birds. Block’s later paper (LIT-637, NOTE-495) argues that removing report is not enough either. - THEORY-023. The A-without-P robot (§4.1) is the AI case. The paper holds it conceptually possible but takes no view on actual machines.
- No ML instruction. Nothing here belongs in the anthology.
Limitations
- The P-without-A cases are not evidence, as the author says. Sperling rests on Block’s own introspection and that of subjects reported by Baars. The later papers (LIT-646) drop the claim that Sperling directly shows P without A.
- A is defined to suit the argument (§4: “my choice here is mainly determined by the needs of the argument”).
- The commentaries were not read. Thirty-odd replies, including Dennett’s “The path not taken”, which Block (2005) cites as denying the conceptual distinction, are where the distinction was contested.
Open questions
- Is there any actual case of A without P? Block reports that each lead from training blindsight patients failed.
- Does P cause A, follow from it, or coincide with it? The paper leaves the three models open and offers introspective considerations only.
Corrections
- none to a seeded skim (there was no seed)