LIT-642: Psychologism and Behaviorism

Ned Block (1981), The Philosophical Review 90(1), January 1981, pp. 5–43 — DOI-10.2307/2184371.

The brief’s citation (Phil Rev 90, 1981) is correct; the scan runs from p. 5 to p. 43, which matches the reference Block gives in LIT-630.

Key takeaways

• Psychologism is “the doctrine that whether behavior is intelligent behavior depends on the character of the internal information processing that produces it” (p. 5). Two systems could be exactly alike in actual and potential behaviour, dispositions, capacities and behavioural counterfactuals, and differ in processing so that “one is not at all intelligent while the other is fully intelligent”.
• The target is the strongest behaviourist conception Block can build. He grants that “sensible” can be defined behaviouristically (p. 11), then replaces Turing’s judge with the neo-Turing Test conception: intelligence “is the capacity to produce a sensible sequence of verbal responses to a sequence of verbal stimuli, whatever they may be” (p. 18). The move from disposition to capacity disarms the Chisholm–Geach, perfect-actor and paralytic objections, which only ever told against behaviourist necessary conditions (pp. 14–19).
• The machine (pp. 19–21). List every typable string of an hour’s conversation in which one party is sensible; the machine finds the strings that begin with the conversation so far and types the next sentence. A tree-searching variant stores one reply per branch. It has the capacity the conception requires, “But actually, the machine has the intelligence of a toaster. All the intelligence it exhibits is that of its programmers” (p. 21). It needs only logical possibility (Objection 6), though Block argues it may be nomologically possible too.
• Psychologism is not chauvinism (pp. 6, 21): it “requires only that intelligent behavior not be the product of a (at least one) certain kind of internal processing”, so Martians with very different but rich processing still count as intelligent.
• Patching the test concedes the point. Requiring that the system “avert exponential explosion of search” (Dennett’s amendment, n. 29) is itself a condition on processing, hence psychologistic, and it faces a dilemma: “postponing” lets the machine in, “avoiding” may exclude us (pp. 38–40).

Standing in the record

Filed on 2026-10-03 at the owner’s request, with Block’s other papers on functionalism, meaning, AI and perception. It is a philosophical argument about the concept of intelligence, and no anthology topic holds that question as such.

Flagged anthology-candidate (ADR-005). Its central claim, that a behavioural test of any length can be passed by stored responses and so cannot certify intelligence without a claim about how the responses are produced, is a claim about “what a measurement cannot tell you”, which is how the anthology’s analysis-and-evaluation topic describes itself. It carries no instruction for practice, and the reading here is about the nature of intelligence, so it stays here. Whether the anthology wants it as a source on memorisation and behavioural evaluation is the question the flag asks.

Relations declared.

• extends: LIT-407. Note 17 says “A version of this machine was sketched in my ‘Troubles with Functionalism’”, and NOTE-366 records that sketch: a list-searcher over all “smart speakable strings” that “clearly has no mental states at all” (pp. 294–295 of LIT-407). This paper develops it into a full argument against behaviourism about intelligence, and n. 30 defends LIT-407’s homunculi-head intuition against Lycan.
• corrects: LIT-394. Block quotes Turing (n. 13) settling for passing the imitation game as a sufficient condition of thinking (“we need not be troubled by this objection”), and the machine is a counterexample to exactly that condition, in its strongest capacity form. NOTE-355 reads Turing’s paper itself.

Where it bears.

• Searle (LIT-400, NOTE-349). Note 30 notes that Searle’s forthcoming Chinese room uses “an example of the same sort as mine”, a person following a manual in a language they do not understand, and rejects Searle’s wider conclusion: “some symbol-manipulating homunculi-heads are intelligent”, and what disqualifies the one described is that “the causal relations among their states do not mirror the causal relations among our mental states”.
• Dennett (LIT-440, LIT-398). Dennett is named as close to a behaviourist analysis of intelligence (n. 12) and as the advocate of the exponential-explosion amendment (n. 29); this is the record’s statement of the position his intentional-system view is pressed with.
• Lycan (LIT-419, Deferred). Note 30 is the record’s only first-hand report of what “Form, Function, and Feel” argues, in Block’s words: that the intuition that a homunculi-head lacks qualia “could be made to go away by imagining yourself reduced to the size of a molecule”, and is “an illusion produced by missing the forest for the trees”. Block’s reply is a single-homunculus variant where no such illusion can operate. This is Block’s report of Lycan, not a reading of Lycan.
• AI consciousness (THEORY-023; Butlin et al., LIT-056, NOTE-052). THEORY-023’s behavioural half says mimicry undercuts behavioural evidence, and NOTE-052 records Butlin et al. discounting behavioural tests because systems can be trained to mimic. This paper is the classic argument that mimicry by stored response defeats a behavioural sufficient condition for intelligence. It is about intelligence, not consciousness, and its machine is a lookup table, not a trained network, so it supports THEORY-023’s premise by analogy only.