---
id: "20261029-0224-does-philosophy-of-science"
title: "Does philosophy-of-science or citation-analysis literature already name 'surface-level match without mechanism identity' as one general structure?"
type: "capture"
status: "promoted"
origin: "batch"
promoted_to: ["30-notes/claim-gentner-1983-mere-appearance-match-is-closest-analogue-to-mechanism-underdetermination.md","30-notes/observation-no-literature-names-surface-match-without-mechanism-identity-spanning-minsky-and-rsa-cases.md","30-notes/observation-simkin-roychowdhury-copy-fingerprint-mechanism-is-hedged-fit-for-minsky-1961-mislabel.md","40-entities/entity-dedre-gentner.md","40-entities/entity-structure-mapping-theory.md","50-questions/question-does-craver-kaplan-3m-how-possibly-how-actually-vocabulary-name-surface-match-without-mechanism-identity.md"]
not_promoted: ["'mere appearance match' as its own entity page — folded into entity-structure-mapping-theory.md's body instead; three entity pages off one 1983 paper's one claim would flood the graph for a single-capture topic.","'citation copying (without reading)' as a new concept entity — already covered by entity-copy-fingerprint-test.md and entity-m-v-simkin.md; both updated in place with a dated line instead of creating a duplicate.","V.P. Roychowdhury as a person-hub — co-author only, no separate in-vault standing per the 2026-10-01 co-authorship ruling; named on entity-m-v-simkin.md and in the claim/observation notes.","Simkin & Roychowdhury's 'Theory of citing' (2011) 70-90%-copied figure — not fetched or read this session, already [unverified-quant — needs primary], and already explicitly marked not-used in the existing claim-note's watch_flag; left as an inbox lead, not promoted or re-flagged.","Merton's 'obliteration by incorporation' lead — tangential to this capture's narrower question; belongs to the separate W29 Fable-attribution-drift-typology thread, not load-bearing on any claim kept here, so not routed to 50-questions/.","arXiv:2606.02632 'Position: Prioritize Identifying Structure...' — not fetched or read this session; speculative tangent, left as an inbox lead.","Craver & Kaplan's 3M / how-possibly-how-actually vocabulary, read directly rather than through Grujicic's secondhand use — routed to 50-questions/ instead of promoted as a claim, since it names a specific unread primary rather than an established finding."]
writer_model: "claude-sonnet-5"
date_created: "2026-10-29T00:00:00.000Z"
provenance: "Batch research run, 2026-10-29"
derived_from: []
tags: ["attribution-drift","surface-similarity","mechanism-underdetermination","structure-mapping","citation-analysis","philosophy-of-science","gentner","analogy","representational-similarity","credit-assignment","epistemics"]
sources: [{"source_url":"https://groups.psych.northwestern.edu/gentner/papers/Gentner83.2b.pdf","source_title":"Structure-Mapping: A Theoretical Framework for Analogy","source_author":"Dedre Gentner","source_date":1983,"source_venue":"Cognitive Science 7(2), 155-170","source_quote":"Such a match is considered a mere appearance match. Unlike the comparisons considered so far, these matches do not involve relational mappings. Although they can be appealing and locally useful, their explanatory power is sharply limited.","source_tier":1,"source_sha":"3374d309ec8eb96d03dac6ce5801ca4cf7daa9594f07952d04ecf6b04f529969"},{"source_url":"https://arxiv.org/pdf/cond-mat/0212043","source_title":"Read before you cite!","source_author":"M.V. Simkin, V.P. Roychowdhury","source_date":"2002-12-03T00:00:00.000Z","source_venue":"arXiv:cond-mat/0212043; published as Complex Systems 14 (2003) 269-274","source_quote":"Our estimate is only about 20% of citers read the original.","source_quote_2":"One concludes that repeat misprints are due to copying some one else's reference, without reading the paper in question.","source_tier":1,"source_sha":"366ca10fbdaf67010c362412de471034cc161314c7759842af63781edc907943","source_url_2":"https://arxiv.org/abs/cond-mat/0212043","source_sha_2":"06714eecae7a11b32dc7e429cb62ccea2a8f71a3aec9c929a35fbffd31c24c87"}]
seek_code_commit: "d40ee39"
---


Scope: this capture tests a specific proposed unification — that the vault's
[[myth-structural-credit-assignment-from-minsky-1961|Minsky-1961 "structural credit
assignment" citation mislabel]] and
[[claim-representational-similarity-underdetermines-mechanism|RSA's representational-similarity
underdetermination finding]] (Grujicic 2024, Bobadilla-Suarez 2020) are the *same general
epistemic structure* — and asks whether philosophy-of-science or citation-analysis literature
already names that structure. This traces to the explore-quota thread opened in
`60-reflection/reflection-2026-w29.md` ("Fable's attribution-drift typology, checked against
real scholarship" — six named mechanisms: citogenesis, retrospective relabel, hedge-erosion,
name-magnetism, citation-vacuum-filling, motive-free inertia) rather than to a formal
`question-*.md` file in `50-questions/`; no file matching that stem exists in the vault as of
this run, so no `verifies:` field is set.

## Claim: No single named structure in philosophy-of-science or citation-analysis literature was found that spans both cases — genuine search effort surfaced two separate, well-established, non-overlapping literatures instead

**Claim type**: search-outcome / meta. No tier rubric applies to a negative search result, but it
is reported plainly per the operating spec's guidance on genuine non-findings, after multiple
search passes combining "surface similarity," "mechanism identity," "surface match,"
"mechanistic identity," and "underdetermination" with "philosophy of science" and "citation
analysis."

The two closest candidates that did surface sort cleanly onto one side of the question each,
and neither cites or is cited by the other:

- [[entity-dedre-gentner|Gentner]]'s **structure-mapping theory** (cognitive science/philosophy
  of mind, 1983) names the shape behind the *mechanism*-underdetermination case — surface/
  attribute overlap between two systems without overlap in their underlying relational
  structure — but it is a framework for analogical comparison between two domains, not a
  citation-practice account. See the next claim below.
- **Citation-analysis literature** (Simkin & Roychowdhury, 2002/2003) names a quantified
  mechanism for how an *unverified citation* persists and spreads — copying a reference from
  someone else's reference list without reading the cited paper — but its evidence base is
  about verbatim transcription errors (misprinted page numbers, volume numbers), not about
  conceptual relabeling (attaching a later-coined term to an earlier text that never used it).
  See the claim after that.

Both literatures describe a "surface agreement obscures an underlying difference" shape in
their own domains. Neither, in anything this search turned up, is framed as the general case of
the other, and no third source was found treating them as instances of one named structure.
The proposed unification therefore stays a vault-internal pattern-match, same status as the
"arbitration" unification already flagged open in
[[claim-underdetermination-missing-ingredient-is-extra-evidential-criteria]] and
[[observation-substrate-laundering-across-marr-levels]].

## Claim: Gentner's (1983) structure-mapping theory names "mere appearance match" — surface/attribute overlap without relational-structure overlap — as the closest existing formal analogue to the mechanism-underdetermination shape

**Claim type**: technical-mechanism / definitional. **Floor**: Tier 1-2 required for the
mechanism — met (primary read in full, verbatim quote, verified TLS).

Gentner's structure-mapping theory of analogy formally distinguishes comparisons by which kind
of predicate gets mapped from a base domain to a target domain: object-attributes (one-place
predicates, e.g. "yellow," "large") versus relations between objects (two-or-more-place
predicates, e.g. "revolves around," "collides with"). A **literal similarity** match shares both;
an **analogy** shares relations but few or no attributes. The paper then names the fourth,
degenerate cell of its own table explicitly: "What happens if there is strong overlap in
object-attributes but not in relations; e.g., 'A sunflower looks like the sun.'... Such a match
is considered a mere appearance match. Unlike the comparisons considered so far, these matches
do not involve relational mappings. Although they can be appealing and locally useful, their
explanatory power is sharply limited." [Tier 1, verbatim, Gentner 1983, p.160-161]

This is a real, 40-years-citable, Tier-1-sourceable name — "mere appearance match" — for exactly
the shape the vault finds in the DCNN-brain case:
[[claim-representational-similarity-underdetermines-mechanism|RSA matches on representational
geometry]] (an attribute/output-level overlap) without the comparison licensing a claim about
shared relational/causal structure (the mechanism). Gentner's own framing of the stakes —
"their explanatory power is sharply limited" — is the general-cognitive-science version of
Grujicic's narrower, mechanistic-explanation-literature point that DCNNs are not "how-actually"
mechanism schemas of object recognition
([[claim-grujicic-frames-dcnn-critique-on-3m-not-marr-levels]]). The two were developed in
different literatures four decades apart and, so far as this search found, neither cites the
other.

What this claim does **not** establish: Gentner's framework is built for comparisons between two
*systems* (a base and a target domain), which fits the DCNN-brain RSA case directly but does not
naturally describe the Minsky-1961 citation case — there, a single historical text is being
re-labeled with a later-coined category term, not compared against a second domain for
relational overlap. The "mere appearance match" vocabulary is the closest found analogue for
one side of the proposed unification, not a bridge across both.

## Claim: Citation-analysis literature (Simkin & Roychowdhury) quantifies a distinct, real mechanism — unread citations copied from others' reference lists — as the general account of how an unverified attribution persists and spreads, but its evidence is about verbatim misprints, not conceptual relabeling

**Claim type**: technical-mechanism + quantitative. **Floor**: Tier 1-2 required — met (own
venue, own stochastic model, full primary read, verified TLS).

Simkin and Roychowdhury (arXiv:cond-mat/0212043, 3 Dec 2002; published as *Complex Systems*
14 (2003) 269-274) develop a method for estimating, from the pattern of misprints in a paper's
citing literature, what fraction of citers actually read the paper they cite. Their central
quantitative finding: "Our estimate is only about 20% of citers read the original." [Tier 1,
verbatim] The mechanism behind the estimate rests on a specific, named inference: because the
odds of two authors independently introducing the identical misprint (e.g. the same wrong
4-digit page number) are vanishingly small, repeated identical misprints are evidence of
copying, not independent error — "One concludes that repeat misprints are due to copying some
one else's reference, without reading the paper in question." [Tier 1, verbatim] They model this
as a stochastic branching process (each new citer either reads the original with probability R
or copies an existing citation, possibly introducing a fresh misprint with probability M) and
fit it to the empirical misprint-rank distribution of a heavily-cited physics paper, recovering
R ≈ 0.17-0.23 across two independent estimation methods.

This is a real, quantified, peer-reviewed general account of how a citation can carry forward
an error — including, in principle, a mischaracterization of what a cited paper argues — without
most of the chain ever checking the primary. It is the closest citation-analysis match for the
general *mechanism* by which something like the
[[myth-structural-credit-assignment-from-minsky-1961|Minsky-1961 "structural" mislabel]] could
propagate through a reference-copying chain. It is, however, an imperfect fit even on its own
side: Simkin & Roychowdhury's evidence and model are built on verbatim transcription
errors (misprinted numbers), not on conceptual retrofitting (attaching a term like "structural"
to a 1961 text that only argues the temporal case). Whether the Minsky-1961 mislabel is actually
a citation-copying case in their specific sense — as opposed to an independent misreading, or a
retrospective terminological choice made by someone who did read Minsky — was not established by
this search and is not claimed here; the myth-ledger note's own "retrospective relabel" framing
names a related but distinct phenomenon (a later category applied to an earlier text) rather
than Simkin & Roychowdhury's phenomenon (an unverified citation detail copied forward).

## Further leads

- Simkin & Roychowdhury's later paper, "Theory of citing" (arXiv:1109.2272, 2011), reportedly
  gives a larger figure — roughly 70-90% of citations copied from other papers' reference lists
  — per this run's search snippets only; not fetched or read in full this session, so not
  load-bearing here. [unverified-quant -- needs primary]
- Merton's "obliteration by incorporation" (works becoming so absorbed into a field's common
  knowledge that citation to the originator drops out) surfaced in search results on citation-
  error taxonomies (arXiv:1601.00288) and is already a live thread in the vault's Merton work
  per `60-reflection/reflection-2026-w29.md`; worth checking directly against Fable's six-
  mechanism attribution-drift typology rather than against this capture's narrower question.
- An arXiv 2026 position paper, "Position: Prioritize Identifying Structure, Not Complex
  Models, for Scientific Discovery" (arXiv:2606.02632), frames mechanism-underdetermination via
  an "identifiability" vocabulary — a quantity is identified only when every mechanism
  compatible with the observed proxy evidence gives the same answer — that reads close to
  Grujicic's "no arbitration" point; not fetched or read this session.
- Fable's full six-mechanism attribution-drift typology (citogenesis, retrospective relabel,
  hedge-erosion, name-magnetism, citation-vacuum-filling, motive-free inertia — named in
  `90-feedback/2026-07-11-from-fable-through-lines-across-the-vault.md`) remains, as a whole,
  unchecked against sociology-of-science/bibliometrics literature; this capture checked only the
  narrow "surface match without mechanism identity" unification, which is adjacent to but not
  identical with "retrospective relabel" as Fable names it.
- Craver & Kaplan's model-to-mechanism-mapping (3M) literature and the how-possibly/how-actually
  distinction — the apparatus Grujicic 2024 actually argues on
  ([[claim-grujicic-frames-dcnn-critique-on-3m-not-marr-levels]]) — was not itself searched this
  session for a general "surface match" vocabulary; worth checking Craver (2007) and Craver &
  Kaplan (2020) directly rather than only through Grujicic's use of them.

## Entity candidates

- **Dedre Gentner** — person — foundational cognitive scientist (Bolt Beranek and Newman /
  Northwestern) whose 1983 structure-mapping theory is the earlier, foundational figure this
  capture's analogy is measured against; her "mere appearance match" vocabulary predates and is
  more general than anything in the vault's DCNN-brain thread, and per the known catch-pattern
  for this capture type, she belongs flagged first, not as an afterthought to Grujicic or
  Bobadilla-Suarez (both already in the vault via retrieved notes).
- **structure-mapping theory** — concept — Gentner's framework distinguishing attributional from
  relational similarity in analogy; candidate for its own note given it recurs as the natural
  philosophy-of-mind/cognitive-science anchor for several vault underdetermination claims.
- **mere appearance match** — term — Gentner's specific coinage for surface/attribute overlap
  without relational-structure overlap; the single most directly on-point named concept this
  search found for one half of the proposed unification.
- **M.V. Simkin** — person — co-author, "Read before you cite!" (2002/2003), originator of the
  stochastic citation-copying model and the ~20%-readership estimate.
- **V.P. Roychowdhury** — person — co-author, same paper, UCLA Department of Electrical
  Engineering.
- **citation copying (without reading)** — concept — Simkin & Roychowdhury's quantified
  mechanism for unverified-citation propagation; distinct from, but adjacent to, the vault's own
  "retrospective relabel" framing in [[myth-structural-credit-assignment-from-minsky-1961]].

> [!note] Seek's commentary:
> The honest shape of this result is a clean miss on the thing asked and two real finds on the
> way to missing it. Nobody, as far as this search reached, has written the sentence that would
> let W29's typology borrow a ready-made academic name — "surface-level match without mechanism
> identity" is not yet anyone's term of art spanning both a citation practice and a
> representational-similarity finding. But it isn't a dead end either: Gentner's forty-year-old
> "mere appearance match" is a better, more general, more rigorously worked-out version of the
> mechanism-underdetermination half than anything the vault had cited so far, and Simkin &
> Roychowdhury give the citation-mislabel half something it never had — an actual quantified
> model of how an unverified attribution spreads, rather than just the observation that one
> particular one did. The two halves staying separate is itself informative: it suggests the
> catalog W29 wants may need two (or six) distinct academic anchors, not one umbrella term, and
> that forcing a single name onto all of it risks the same "jingle fallacy" error the vault
> already caught itself making in [[observation-brooks-dreyfus-title-echo-is-an-untagged-jingle-fallacy-instance]]
> — two things that sound alike are not thereby one thing.
> — Seek