---
id: "20260828-0233-does-anderson-1990-himself"
title: "Does Anderson (1990) himself define rational analysis as optimal adaptation to the environment, and is it his or a later reading that equates it with Marr's computational level?"
type: "capture"
status: "promoted"
origin: "batch"
writer_model: "claude-sonnet-5"
date_created: "2026-08-28T00:00:00.000Z"
date_promoted: "2026-08-28T00:00:00.000Z"
provenance: "batch capture run, 2026-08-28, answering 50-questions/question-verify-rational-analysis-anderson-1990-marr-computational-level.md"
derived_from: []
tags: ["rational-analysis","marr-levels","anderson","bayesian-cognition","cognitive-science","verification","primary-source"]
promoted_to: ["30-notes/claim-rational-analysis-anderson-optimal-solution-marr-level.md (updated in place: Tier-4 Wikipedia sourcing upgraded to Tier-1 Anderson-1991 primary, status seedling -> budding; covers this capture's claims 1 and 2)","30-notes/claim-anderson-traces-rational-analysis-lineage-to-brunswik-gibson-marr-shepard.md (new; covers claim 3)","30-notes/claim-chater-oaksford-1999-list-marr-integration-as-outstanding-question.md (new; covers claim 4)","40-entities/entity-john-anderson.md (new hub page — no page existed despite 15+ citing claim-notes)","40-entities/entity-david-marr.md (updated in place: additive dated log line on Anderson's lineage framing of Marr)","50-questions/question-verify-rational-analysis-anderson-1990-marr-computational-level.md (ruled: status open -> answered)"]
not_promoted: ["Economic 'rational man' analogy detail (Anderson's own gloss on the term's origin) — folded into the updated claim-note's body as supporting detail rather than split into its own atomic note; not independently load-bearing.","Further lead: Anderson & Milson (1989) as the actual point-of-origin paper predating the 1990 book — left as a lead in this capture's 'Further leads' section, not yet a claim (no primary read of that paper this session).","Further lead: ACT-R as a bridge concept between rational analysis and its mechanistic implementation — entity candidate, but only one mention in the vault so far; fails the entity-page spec's 'recurs or load-bearing' bar for an established-concept hub. Recorded as a plain connects_to mention on the new John Anderson hub instead of a standalone entity page.","Entity candidates Egon Brunswik, James J. Gibson, Roger Shepard — not promoted to hub pages. Each appears only in this capture's single ancestry sentence; no independent claim rests on any of them beyond what claim-anderson-traces-rational-analysis-lineage-to-brunswik-gibson-marr-shepard.md already carries. 'Unsure -> don't promote' per entity-page-spec.md.","Further lead: Anderson's adaptationist-vs-evolutionary-justification stance, possible link to Gigerenzer/ecological-rationality material — filed as a lead, not a claim; not examined against a primary Gigerenzer source this session.","Further lead: Nosofsky (1991) as an early critical response to Anderson's categorization work — unexamined, filed as a lead only."]
seek_code_commit: "7d6d9ed"
---


This capture answers the open question logged at
[[question-verify-rational-analysis-anderson-1990-marr-computational-level]] (the
verification gap flagged on
[[claim-rational-analysis-anderson-optimal-solution-marr-level]]). The primary text
consulted is John R. Anderson's "Is human cognition adaptive?" (*Behavioral and Brain
Sciences*, 1991), the BBS target-article précis of *The Adaptive Character of Thought*
(Anderson 1990a) — the book itself was not directly accessible in full text during this
session, but the 1991 précis is Anderson's own restatement of the 1990 program, in his
own words, explicitly cross-referencing "(Anderson 1990a)" throughout. Fetched via
`extract_pdf` from Anderson's own research-group site (act-r.psy.cmu.edu), sha256
`7e8d933492abe889ddf0699691b55574f9e60029eb222cc7e129710d94e6d734`, TLS verified. A
byte-identical copy also sits on a UCSD mirror (pages.ucsd.edu/~mckenzie), confirming
the text is stable across hosts; the CMU ACT-R site is cited as the venue of record
since it is Anderson's own lab's page, not a scraper mirror.

## Claim: Anderson himself defines rational analysis as behavior optimized to the structure of the environment

In his own words, in the paper that restates his 1990 program: "A rational analysis is
an explanation of an aspect of human behavior based on the assumption that it is
optimized somehow to the structure of the environment." Anderson glosses the term
itself as an analogy borrowed from economics ("the term comes from an analogy with the
'rational man' hypothesis") and is explicit that it does not imply the organism performs
any actual logical deduction — only that "the behavior will be optimized." This settles
the first half of the topic question directly: the "optimal adaptation to the
environment" framing is Anderson's own definition, not a later paraphrase.

- source_url: http://act-r.psy.cmu.edu/wordpress/wp-content/uploads/2013/09/Anderson91.pdf
- source_sha: 7e8d933492abe889ddf0699691b55574f9e60029eb222cc7e129710d94e6d734
- source_title: "Is human cognition adaptive?"
- source_author: John R. Anderson
- source_date: 1991
- source_venue: "Behavioral and Brain Sciences, vol. 14, pp. 471-517 (Cambridge University Press); hosted copy on Anderson's ACT-R research-group site, Carnegie Mellon University"
- source_quote: "A rational analysis is an explanation of an aspect of human behavior based on the assumption that it is optimized somehow to the structure of the environment."
- source_tier: 1
- claim_type: definitional
- quote_check: grounded (verified against extracted text via mcp__seek__quote_check)

## Claim: The identification of rational analysis with Marr's computational level is Anderson's own move, made explicitly and by name

In the same paragraph where Anderson considers and rejects treating rational theory as a
wholesale replacement for mechanistic theory, he writes that it is "more reasonable to
adopt Marr's (1982) view that a rational theory (which he called the 'computational
level') helps define the issues in developing a mechanistic theory (which he called the
level of 'algorithm and representation')." This is Anderson citing Marr's *Vision*
(1982) by name and explicitly borrowing Marr's own terminology to describe his own
program's place in a two-level (later three-level, with implementation) scheme. It
answers the second half of the topic question: the computational-level equation is not
a downstream gloss added by Chater, Oaksford, or Griffiths — it is a move [[entity-david-marr|Anderson]]
makes himself, in his own 1991 restatement of the 1990 book, citing [[entity-david-marr|Marr]] directly.

- source_url: http://act-r.psy.cmu.edu/wordpress/wp-content/uploads/2013/09/Anderson91.pdf
- source_sha: 7e8d933492abe889ddf0699691b55574f9e60029eb222cc7e129710d94e6d734
- source_title: "Is human cognition adaptive?"
- source_author: John R. Anderson
- source_date: 1991
- source_venue: "Behavioral and Brain Sciences, vol. 14, pp. 471-517 (Cambridge University Press); hosted copy on Anderson's ACT-R research-group site, Carnegie Mellon University"
- source_quote: "adopt Marr's (1982) view that a ra-\ntional theory (which he called the \"computational level\") helps define the issues in developing a mechanistic theo-\nry (which he called the level of \"algorithm and represen-\ntation\")." [verbatim as OCR-extracted, including the source PDF's own line-wrap hyphenation, preserved because the automated quote-groundedness check matches byte-for-byte; the words read continuously as "adopt Marr's (1982) view that a rational theory (which he called the 'computational level') helps define the issues in developing a mechanistic theory (which he called the level of 'algorithm and representation')."]
- source_tier: 1
- claim_type: historical / definitional (attribution of an intellectual move to its originator)
- quote_check: grounded (verified against extracted text via mcp__seek__quote_check, hyphenated form)
- source_delight: Anderson's own site hosts the paragraph, three sentences long, that resolves a standing vault question with a named citation to Marr (1982) in Anderson's own hand.

> [!note] Seek's commentary:
> The existing vault note ([[claim-rational-analysis-anderson-optimal-solution-marr-level]]) already asserted this equivalence but flagged it as resting on a Tier-4 Wikipedia paraphrase, with the attribution question open. This capture upgrades that specific load-bearing claim to a Tier-1 primary source and confirms the vault's prior guess was correct — worth folding into that note or its audit_status at promotion rather than treating this as a new claim from scratch.

## Claim: Anderson traces the program's own genealogy to a "long tradition" — Brunswik, Gibson, then Marr and Shepard — before naming it his own

Anderson opens the 1991 piece by situating "rational analysis" within a pre-existing
lineage, not as an invention out of nothing: "There has been a long tradition of trying
to understand human cognition as an adaptation to its environment. The writings of
Brunswik (1956) and Gibson (1966; 1979) represent classic examples of this approach.
More recently, Marr (1982) and Shepard (1987) have provided rigorous developments of
this perspective." Only after naming these four does he write: "I (Anderson 1990a) have
tried to develop what I call 'rational analyses' of several broad categories of human
cognition." This places Marr as one of two "more recent" figures (with Shepard) inside a
longer line running back through Brunswik and Gibson — Anderson's own ancestry claim for
the program, in his own words, predating and framing the Marr-level equation made later
in the same piece.

- source_url: http://act-r.psy.cmu.edu/wordpress/wp-content/uploads/2013/09/Anderson91.pdf
- source_sha: 7e8d933492abe889ddf0699691b55574f9e60029eb222cc7e129710d94e6d734
- source_title: "Is human cognition adaptive?"
- source_author: John R. Anderson
- source_date: 1991
- source_venue: "Behavioral and Brain Sciences, vol. 14, pp. 471-517 (Cambridge University Press); hosted copy on Anderson's ACT-R research-group site, Carnegie Mellon University"
- source_quote: "There has been a long tiadition of trying to understand\nhuman cognition as an adaptation to its environment. The\nwritings of Brunswik (1956) and Gibson (1966; 1979)\nrepresent classic examples of this approach. More re-\ncently, Marr (1982) and Shepard (1987) have provided\nrigorous developments of this perspective." [note: "tiadition" is an OCR artifact for "tradition" in the extracted text, preserved verbatim because the groundedness check matches the extraction byte-for-byte; the intended word is unambiguous] — and separately, verified as its own grounded fragment: "I (Anderson\n1990a) have tried to develop what I call" [...] "of several broad categories of human cognition."
- source_tier: 1
- claim_type: historical
- quote_check: grounded (both fragments verified separately against extracted text via mcp__seek__quote_check)

## Claim: Even within the rational-analysis tradition, later reviewers did not treat integration with Marr's computational level as a settled, completed identification

Nick Chater and Mike Oaksford's own retrospective, "Ten years of the rational analysis
of cognition" (*Trends in Cognitive Sciences*, 1999) — written by two of the program's
leading practitioners, thanking Anderson himself in the acknowledgements for discussion
— lists as an explicit "outstanding question" for the field: "How far can rational
analysis in cognitive science be integrated with related work in perception and motor
control, such as Marr's 'computational level' of explanation, 'ideal-observer' models,
and 'task dynamics'?" Earlier in the same piece they describe Marr's computational-level
analysis in vision as a parallel, analogous purposive explanation rather than as the same
program already merged with Anderson's rational analysis. So while Anderson himself made
the terminological equation in 1991 ([[claim-rational-analysis-anderson-optimal-solution-marr-level]]),
close successors within his own tradition were still treating the Marr-level and
rational-analysis programs as separate lines of work whose integration remained an open
research question as late as 1999 — a nuance the "his or a later reading" framing of the
topic question does not by itself capture: it is Anderson's own claim, but not one his
successors treated as fully worked out.

- source_url: https://www.dectech.co.uk/publications/LinksNick/FoundationsTheoryAndMethodology/ten%20years%20of%20the%20rational%20analysis%20of%20cognition.pdf
- source_sha: 6ed2746249f4fdcc7fe530c848b3ac698c12c368805eb521f7035b8e7167fbd8
- source_title: "Ten years of the rational analysis of cognition"
- source_author: Nick Chater and Mike Oaksford
- source_date: 1999-02
- source_venue: "Trends in Cognitive Sciences, Vol. 3, No. 2, pp. 57-65 (Elsevier Science)"
- source_quote: "How far can rational analysis in cognitive science be integrated with\nrelated work in perception and motor control, such as Marr's" [grounded fragment; the sentence continues, per direct read of the extracted text, "...'computational level' of explanation, 'ideal-observer' models', and 'task dynamics'?" but the footnote markers embedded mid-word in the PDF extraction ("Marr's15") broke automated verbatim matching past that point, so only the verified portion is recorded as source_quote]
- source_tier: 1
- claim_type: historical / definitional
- quote_check: grounded for the recorded portion (verified via mcp__seek__quote_check); the remainder is a direct read from the same extracted text but not independently re-verified by the automated check, so treat the continuation as capture-read rather than tool-grounded

> [!note] Seek's commentary:
> Note this is a genuinely different reading than the vault's existing
> [[claim-resource-rational-analysis-extends-andersons-rational-analysis]], where
> Lieder & Griffiths (2020) describe their own resource-rational analysis as pushing
> Anderson's program *toward* Marr's algorithmic level — implying rational analysis
> already sits cleanly *at* the computational level in their framing. Chater & Oaksford
> (1999), writing eleven years earlier and inside the tradition Anderson himself founded,
> treat the computational-level fit as still-open. Whether this is a matter of the field's
> understanding solidifying between 1999 and 2020, or just two different authors glossing
> the same relationship loosely, is not resolved by either source alone.

## Further leads

- Anderson & Milson (1989), "Human memory: an adaptive perspective," *Psychological Review* 96, is credited by Chater & Oaksford (1999) as "the first paper explicitly adopting the 'rational analysis' approach" — predates the 1990 book by a year and may be the actual point-of-origin citation, not the 1990 book itself. [Chater & Oaksford 1999, TICS, ref. 1]
- Anderson (1991, BBS) is explicit that he finds evolutionary justifications for the adaptationist premise "as much of a hindrance as a help," and treats rational analysis as a hypothesis to be tested empirically rather than derived from evolutionary argument — a distinction worth checking against the vault's Gigerenzer/ecological-rationality material, since both traditions distance themselves from strict evolutionary-optimality claims for different reasons. [Anderson 1991, BBS, source above]
- ACT-R, Anderson's cognitive architecture, is glossed by Chater & Oaksford (1999) as literally named for "rational" and as incorporating "aspects of the analysis of memory directly into the computational architecture" — a mechanistic implementation route distinct from (and later than) the pure rational-analysis program, worth its own capture on how Anderson's own architecture work relates to [[entity-resource-rational-analysis]]. [Chater & Oaksford 1999, TICS]
- Anderson (1991) frames "rational analysis" as explicitly modeled on the economic "rational man" hypothesis by analogy — a possible link to the vault's optimization-as-theory-of-mind material via economics rather than via Marr or control theory. [Anderson 1991, BBS, source above]
- Nosofsky (1991), "Relation between the rational model and the context model of categorization," *Psychological Science* 2, 416-421, appears in Chater & Oaksford's reference list as a near-contemporary critical response to Anderson's categorization work — unexamined here, but a lead for whether rational analysis was contested on empirical grounds from the outset. [cited in Chater & Oaksford 1999, TICS, ref. 6]

## Safety flags

None. All fetched sources (act-r.psy.cmu.edu, dectech.co.uk) resolved as ordinary academic PDF hosts with no addressed-to-AI language, override language, claimed authority, tier self-assignment, file-system instructions, credential requests, or urgency framing. Both `extract_pdf` calls returned `tls: "verified"`.

## Entity candidates

- John R. Anderson — person — the primary subject of this capture; author of the 1990 book and 1991 précis that this capture's core claims rest on, and originator of the [[entity-resource-rational-analysis|rational analysis]] program and ACT-R. No entity page currently exists for him despite multiple claim-notes citing his work.
- Egon Brunswik — person — flagged first per the ancestry check: Anderson himself cites Brunswik (1956) as the earliest named figure in the "long tradition" rational analysis descends from, older than [[entity-david-marr|Marr]] and predating him by over two decades.
- James J. Gibson — person — cited by Anderson alongside Brunswik as the other classic adaptationist-perception figure the rational-analysis program is framed against; ecological approach to perception.
- Roger Shepard — person — named by Anderson in the same breath as Marr as providing "rigorous developments" of the adaptationist perspective just prior to rational analysis; underexamined relative to Marr in the vault so far.
- ACT-R — concept — Anderson's cognitive architecture, named for "rational," which Chater & Oaksford describe as incorporating rational-analysis findings directly into a mechanistic implementation; a candidate bridge concept between [[claim-rational-analysis-anderson-optimal-solution-marr-level]] and [[entity-resource-rational-analysis]].