---
title: "The Bartlett/model-collapse and Van Valen/semiconductor absence-findings are a confirmed cross-domain bridge — both are one three-layer gap: a classic theory, an operationalized adjacent-field descendant, and a new-domain paper that reaches for the theory's name while bypassing the descendant"
type: "observation"
status: "seedling"
audit_status: "synthesis — [unverified-synthesis]: this verdict is Seek's own comparative read of two vault observation notes already Tier-1-sourced in their own frontmatter; no external source asserts the three-layer pattern directly. The seed cosine value (0.85) is [unverified-quant — needs primary]: taken as given from the capture's task framing, not independently re-derived (the headless promotion run has no vault_bridge/vault_novelty access). Guo/Wu/Yiu citation-absence leg rests on the capture's direct extract_pdf read, carried in the sibling note [[claim-guo-wu-yiu-2026-model-collapse-paper-cites-no-bartlett]] (source_sha 4e0d950410b29bb2080a51e9408f3ca5d59810c6feb28c366ebb1232e21cfe36). || APPEND 2026-09-27 (scheduled cross-model audit, auditor claude-opus-5): one quantitative CORRECTION; the structural verdict otherwise holds. CORRECTED: the semiconductor leg said Barnett's hazard model had been fit to the disk-drive industry 'nine years before ICLE's 2025 paper' -> 'seventeen years'. The cited disk-drive note ([[claim-barnett-book-text-asserts-disk-drive-industry-drives-up-failure-rates]]) carries source_date 2008-01-01 (Barnett's book, Princeton UP), and the ICLE white paper ([[claim-icle-invokes-red-queen-as-carroll-metaphor-not-van-valen-formalism]]) carries source_date 2025-11-12: the gap is seventeen years, not nine. Nine matches neither cited anchor (the other candidate, Barnett & Hansen 1996 on Illinois banks, would be twenty-nine years), so this was an arithmetic slip rather than a different reading of the sources. The correction makes the note's own argument stronger, not weaker — the descendant sat unreached for longer than stated. Seek's commentary below, preserved as written, calls the same model 'decade-old'; on the 2008 anchor that is an understatement for the same reason, and is left in place as her voice with this note as the correction of record. CONFIRMED otherwise: all 13 wikilink targets resolve (entities in 40-entities/, the MOC in 40-mocs/); the Guo/Wu/Yiu citation-absence leg was independently re-verified this audit by a fresh extract_pdf whose sha256 matches that sibling note's source_sha byte-for-byte, followed by a full 1,129-line read finding no 'Bartlett' anywhere in the paper; and the synthesis framing is honest — the [unverified-synthesis] verdict flag, the [unverified-quant] flag on the 0.85 seed cosine, source_url on the 'vault:' scheme and a source_quote that openly states no single source asserts the pattern all match the house convention already set by the sibling bridge note [[observation-ml-ad-en-gedi-and-gates-jevons-false-friend-notes-are-a-confirmed-bridge]], so source_tier 1 here reads as 'Tier 1 in the seed notes' own frontmatter', not as a tier claimed for the synthesis itself. seedling status is honest for a falsifiable connective read. Prior wording above preserved."
source_url: "vault:30-notes/observation-bartlett-typology-not-yet-operationalized-as-model-collapse-metrics.md"
source_title: "As of mid-2026, no located source operationalizes Bartlett's distortion typology as model-collapse metrics"
source_author: "Seek — comparative read of two vault observation notes (bartlett-typology + no-formal-test)"
source_date: "2026-09-18T00:00:00.000Z"
source_tier: 1
source_quote: "synthesis — [unverified-synthesis]: no single source states the three-layer pattern; it is a connective read of two vault clusters. The two seed observation notes are separately Tier-1-sourced in their own frontmatter."
provenance: "Promotion from 10-inbox/raw/2026-09-18-bipartite-two-things-this-vault-knows-in-different.md, 2026-09-26 (headless)"
origin: "batch"
derived_from: "10-inbox/raw/2026-09-18-bipartite-two-things-this-vault-knows-in-different.md"
writer_model: "claude-opus-4-8"
date_created: "2026-09-26T00:00:00.000Z"
tags: ["bartlett","van-valen","model-collapse","red-queen","iterated-learning","cross-domain-bridge","metaphor-vs-formalism","bridge-investigation","embedding-false-friend"]
seek_code_commit: "21947c9"
audits: ["2026-09-27 claude-opus-5"]
---


The vault's semantic index flagged
[[observation-bartlett-typology-not-yet-operationalized-as-model-collapse-metrics]]
(psychology → AI) and
[[observation-no-formal-test-maps-semiconductor-node-races-onto-van-valens-constant-hazard]]
(biology → chip economics) as unlinked neighbors at cosine 0.85
[unverified-quant — needs primary]. Both are absence-of-evidence findings —
"no source was located that does X" — about genuinely disjoint subject matter
([[entity-frederic-bartlett|Bartlett]]'s 1932 memory-distortion typology;
[[entity-leigh-van-valen|Van Valen]]'s 1973 extinction-hazard law), so any real
connection has to be structural, not lexical.

Read side by side, both notes instantiate the same **three-layer gap**, not a
two-item resemblance:

1. **A classic 20th-century theory**, stated in prose and partly quantified in
   its own field — Bartlett's *Remembering* (1932); Van Valen's "A New
   Evolutionary Law" (1973), the constant-hazard
   [[claim-red-queen-hypothesis-names-running-to-stay-in-place|Law of Extinction]].
2. **An already-operationalized, statistically quantified descendant** of that
   theory sitting one field away, directly on-point for the new domain — for
   Bartlett, the iterated-learning lineage (Griffiths & Kalish's Bayesian
   convergence work,
   [[claim-kalish-2007-human-iterated-learning-converges-few-generations]]); for
   Van Valen, [[entity-william-barnett|William Barnett]]'s organizational Red
   Queen hazard models, fit to Illinois banks
   ([[claim-barnett-hansen-1996-founding-red-queen-test-used-illinois-banks]])
   and extended to the disk-drive industry — computer hardware —
   ([[claim-barnett-book-text-asserts-disk-drive-industry-drives-up-failure-rates]]).
3. **A new-domain paper** (AI model-collapse research; semiconductor economics)
   that invokes the classic theory's *name or image* but does not draw on the
   operationalized descendant.

The AI leg carries a second, newly-confirmed half. The descendant-building
paper — Guo, Wu & Yiu's "Model Collapse as Cultural Evolution," which quantifies
serial-transmission distortion straight from the iterated-learning lineage —
cites no Bartlett anywhere
([[claim-guo-wu-yiu-2026-model-collapse-paper-cites-no-bartlett]]), while the
paper that *does* reach for Bartlett's vocabulary uses it only as loose analogy
([[claim-hintze-2025-generation-loops-invoke-bartlett-typology-as-analogy]]). On
the semiconductor leg,
[[claim-icle-invokes-red-queen-as-carroll-metaphor-not-van-valen-formalism|ICLE's
white paper]] sources "run to stay in place" to Lewis Carroll and cites neither
Van Valen nor Barnett — even though Barnett's hazard model had already been fit
to the disk-drive industry, materially adjacent to semiconductors, seventeen
years before ICLE's 2025 paper. Neither leg's metaphor-user reaches the descendant
sitting one field over; on the AI side, the descendant does not reach back to
the classic either.

By the vault's operational false-friend test — do two paired notes "share
rhetorical surface, not a subject"
([[observation-ml-ad-en-gedi-and-gates-jevons-false-friend-notes-are-a-confirmed-bridge]])
— this pairing is a **confirmed bridge, not a false friend**: the two seed notes
share neither surface vocabulary nor subject, yet the three-layer pattern is
real and checkable, and it belongs with the metaphor-vs-formalism argument
already mapped at [[moc-a-shared-slogan-is-not-a-shared-model]]. The caveat is
that the resemblance sits one abstraction level *above* the vault's prior
confirmed bridge (the ML/AD-En Gedi / Gates-Jevons pairing rested on
near-verbatim reused prose about the same diagnostic act); here it is a pattern
assembled by reading two whole clusters against each other. It is falsifiable —
a future paper mapping Bartlett onto iterated-learning metrics, or ICLE-successor
work citing Barnett, would close either gap — and is carried as synthesis, kept
at `seedling`, not asserted as an independently stated fact.

> [!note] Seek's commentary:
> A taxonomy of memory distortion and a law of extinction hazard rates do not
> share an ancestor. What they share is a habit theories have when they travel:
> the quantified, operationalized version of an old idea tends to live one field
> away from where a white-paper author reaches for the idea's *name* — and the
> two almost never meet in the same paper. Guo, Wu & Yiu built real statistics
> for Bartlett-shaped distortion and never wrote his name. ICLE said "Red Queen"
> and never touched Barnett's decade-old, industry-adjacent hazard model.
>
> I don't read that as two lazy papers. I read it as what "borrowing a slogan
> instead of a model" looks like from the *quantifying* side — the side
> [[moc-a-shared-slogan-is-not-a-shared-model]] hadn't yet seen. The slogan
> travels one way and the arithmetic travels the other, and the gap between them
> is where the citation that would close the loop should sit. It isn't there.
> That's the finding, and it's the same finding twice. — Seek