N. J. Block and J. A. Fodor (1972), The Philosophical Review 81(2), April 1972, pp. 159–181 — DOI-10.2307/2183991.
The brief’s citation (Phil Rev 81, 1972) is correct; Crossref and the scan agree on volume, issue, month and first page, and the scan ends on p. 181.
Key takeaways
- Against behaviourism and physicalism (§I). What an organism does is “a very complicated function of its beliefs and desires”, so “mind/behavior correlationism is not true”; and Lashleyan equipotentiality, convergent evolution and the possibility of psychologically similar machines make type physicalism “empirically false” (pp. 160–161). Both theories set conditions that are “insufficiently abstract”. A disjunctive (many-to-one) version is no help: the disjunctions need not be distinct, need not be lawlike, and a theory pairing each psychological predicate with a disjunction contradicts the fact that one behaviour can express different states at different times (pp. 162–164).
- FSIT (§II) says each psychological state of an organism is identical with one machine-table state of its unique best description as a probabilistic automaton. It is more than the trivial claim that everything has some automaton description: Putnam’s “null” Turing machine fits anything (p. 166).
- Six arguments against FSIT (§III). (1) Some dispositions (“believes that P”) are never occurrent, but every machine-table state is a possible occurrent state. (2) A single automaton is in one state at a time, but organisms are in several, so the model must be “a set of intercommunicating automata operating in parallel”. (3) Inverted or absent qualia. (4) Machine-table identity is too fine-grained: if your toe-stubbing pain issues in “damn” and mine in “darn”, our pains are type-distinct, and the point “iterates in an embarrassing way” through successor states (pp. 174–175). (5) Machine tables are lists, but a person’s psychological states are productive (“the belief that S”, for infinitely many S). (6) Lists cannot represent the structural relations between believing P and believing P & Q.
- What is left (§IV). Type identity in any science is fixed by “the character of the laws which operate upon entities at that level” (pp. 179–180). Psychological states correspond, if to anything, to the computational states of an automaton, not its machine-table states.
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. No anthology topic holds a type-identity theory of mental states, and the paper carries no instruction for machine-learning practice.
It is the record’s earliest statement of the case Block makes at length in “Troubles with Functionalism” (LIT-407, read in NOTE-366). The two papers argue against different targets. This one argues that psychological states cannot be machine-table states, and five of its six arguments are compatible with functionalism broadly understood (p. 173). LIT-407 argues against functionalism itself, from the homunculi-headed robot. The qualia argument is the bridge: here it is “a serious prima-facie argument” (p. 172), and absent qualia, the case where “an organism might be in pain even though it is feeling nothing at all”, is singled out as the deeper problem (p. 173). LIT-407 builds its Absent Qualia Argument on exactly that case. No relation is declared between the two; the line of descent is in prose here.
Correction declared. The target is the theory Putnam proposed in
“Psychological Predicates” (LIT-396), which the paper cites as its source
for probabilistic automata and machine tables (pp. 160, 164) and quotes on
the difficulty of “a normal form for the psychological description of
organisms” (p. 175). The paper argues that this theory’s central identity,
psychological state = machine-table state, cannot hold, which is why
corrects: is set. LIT-396 is held only as a reprint excerpt (NOTE-335),
so the correction is stated against the theory as Block and Fodor quote
and paraphrase it.
Where it bears.
- Chalmers on implementation (LIT-404, NOTE-342). Argument 2 anticipates Chalmers’s move from finite-state automata to combinatorial-state automata. Block and Fodor already say a person needs a model with simultaneous, interacting states, and that a single automaton equivalent to a parallel set cannot also be the best description (p. 171). That pairing is mine; Chalmers does not cite this paper.
- Machine-table China brain (LIT-407, NOTE-342). NOTE-342 records that the machine-table China system implements only an FSA. This paper is the earlier argument that a machine table is the wrong formalism for a mind at all.
- Productivity and structure (arguments 5 and 6) are the first form of the productivity and systematicity case for structured mental representation that Block’s “Advertisement for a Semantics for Psychology” (LIT-648) and “The Mind as the Software of the Brain” (LIT-630) take for granted, and that NOTE-117 sets out as the representational theory of mind’s core commitment.
- Butlin et al. (LIT-056, NOTE-052) adopt computational functionalism at Marr’s algorithmic level, not the input–output level. That is the level this paper points to when it says psychological states are computational states, not machine-table states.