Source
- Clark & Chalmers (1998), The Extended Mind, LIT-097, read in full in NOTE-180 from the authors’ web version; agreement with the typeset Analysis text is unverified.
- Clark (2005), Intrinsic content, active memory and the extended mind, LIT-403, read in full in NOTE-363: the first reply by either author to the published criticism.
What was actually shown
The cases. In the three Tetris cases (§2) rotation is done mentally, by a button on screen, or by a neural implant. The paper argues that the only thing separating the button from the implant is the skin, which is “precisely what is at issue”. Otto has Alzheimer’s disease and carries a notebook that “plays the role usually played by a biological memory” (§4). He believed the museum was on 53rd Street before he looked it up, as Inga did before she recalled it. Twin Otto, a physical duplicate “from the skin in” whose notebook says 51st Street, believes 51st Street: “a belief is simply not in the head”. Unlike the Putnam–Burge twins, the two Ottos walk to different streets, which is the paper’s answer to the worry that externally fixed content is explanatorily idle (§3).
What carries the conclusion. NOTE-180 finds three moves and no argument from causation to parthood: role-functionalism about belief (“What makes some information count as a belief is the role it plays”, §4), assumed; the parity sentence of §2, stated with the hedge “(so we claim)”; and explanatory simplicity with a revisionary proposal about the concept (“the notion of belief ought to be used so that Otto qualifies”, §4). The evidence for coupling is causal: two-way interaction, and a drop in competence when the resource is removed (§3). The note’s judgement is that the constitutive step is “stipulated given functionalism, not argued”. Clark’s 2005 reply confirms it in his own words: “We did not argue for this principle, except by providing illustrations and defences against counter-arguments” (NOTE-363, p. 2).
The coupling features. Constancy (Otto “will rarely take action without consulting it”), direct availability, automatic endorsement, and past conscious endorsement, which the paper calls “arguable” (§5). They are “features of our central case that make the notion so clearly applicable”, and applicability “gradually falls off” as cases lack them. The Internet “is likely to fail on multiple counts, unless I am unusually computer-reliant, facile with the technology, and trusting”; “certain files on my computer may qualify”.
The defence (LIT-403). The thesis concerns material vehicles, not contents (n. 1). Against Adams and Aizawa’s intrinsic-content objection: the notion is unclear; silicon could carry it; and their own concession that not every cognitive state need involve non-derived content removes its force, since the thesis never claimed a cognizer made wholly of notebooks. Against the active-memory objection: the notebook is conceded to be “unlike biological memory”, and the claim is weakened to Otto-plus-notebook exhibiting “enough of the central features and dynamics” of a believer, with the differences said to lie below the level a functionalist should attend to.
What could have come out otherwise. Very little here could, because the support is argument from imagined cases. What could have gone the other way is the critics’ search for a relevant difference. Each candidate offered so far (intrinsic content, active updating) has been answered by a parity case, and the active-updating difference has been conceded rather than denied.
What this does not say
- That the four features are sufficient conditions. The paper refuses that framing (§5). A record document that cites LIT-097 for conditions under which an external store becomes a belief is citing it for more than it says (NOTE-180, Corrections).
- That consciousness extends. The paper allows that “some mental states, such as experiences, may be determined internally” (§4). Its thesis is about standing, dispositional belief and non-conscious cognitive processes.
- That coupling is shown to be constitution. The causal-coupling reading, on which the notebook is only an input to an internal belief, is the rival the paper treats as “one step too many”, not one it refutes. Adams and Aizawa (LIT-438, LIT-455) press exactly this point.
- That ordinary tools, or AI assistants, carry their users’ beliefs. The paper expects the Internet to fail for typical users. Whether an assistant meets automatic endorsement is a question the account frames and does not answer.
- That the self extends. The paper infers this in one paragraph (§5), and the moral consequence (“interfering with someone’s environment” as interfering with their person) in one hedged sentence. This account does not include either.
- That “Parity Principle” is the 1998 paper’s term. The name is from LIT-403. NOTE-363 also finds that the 2005 paper’s quotation of the parity sentence differs from the 1998 web text (“for that time” for “so we claim”), so it should not be cited as the 1998 wording.
Connections
- QUESTION-035 asks when a combined form is one system and where its boundary lies. This account gives one answer for the cognitive kind: the boundary is set by reliable functional coupling of the kind the four features mark, not by the skin. It constrains the question; it does not settle the general case, since its criterion is offered for one kind and read off one case. CLAIM-157 and CLAIM-172 offer other criteria (near-decomposability, mutual entrainment).
- CLAIM-175 argues that a dancing couple is a combined cognitive form “on the extended-mind thesis’s own terms”. It relies on the statement of those terms, which is THEORY-1194.
- The critics. Adams and Aizawa (LIT-438, LIT-455) argue that coupling is not constitution and that cognition needs non-derived content. Rupert (LIT-422) argues that the functional differences Clark sets “below the level” do explanatory work. Chalmers’s later chapters (LIT-415, LIT-421) revisit the thesis.
- Offloading. THEORY-343 and THEORY-346 describe when people rely on external stores rather than memory. They bear on whether the constancy and availability features are met in practice, not on whether meeting them makes a store part of a mind.
- Implementation. Chalmers’s implementation conditions (THEORY-609) are indifferent to whether a component lies inside the skull, so they are compatible with this account, though neither paper draws the link.
- Fictionalism. Toon’s pretence account uses the same notebook image for the opposite purpose; Drayson’s critique of it is THEORY-646.
Note of 2026-10-29: the proposal split off
What Clark and Chalmers argue, as distinct from whether it is true, is now THEORY-1194, Active, so that CLAIM-175, which argues on the thesis’s own terms, can stand on the statement alone. This entry keeps the question whether the conditional holds, and its promote_when.
Note of 2026-10-29: what holds it back
Kept Proposed by the sixteenth status review. Both papers are read in full, but promote_when’s result is not in the record: no account of constitution, stated apart from the Otto case, that puts the notebook inside and an ordinary reliably consulted tool outside by one procedure. Adams and Aizawa’s coupling–constitution objection (LIT-438, LIT-455) and Rupert’s objection that the differences “below the level” do explanatory work (LIT-422) are unanswered here.