---
id: "20260928-0240-hop-kimura-crow-neutral-drift-bridges-citation-copying-and-llm-dilution"
title: "Kimura & Crow's 1964 neutral-drift formula for allele frequencies is the named ancestor of citation-copying models, and the same no-selection-needed logic reappears, unacknowledged, in LLM citation dilution"
type: "capture"
status: "promoted"
promoted_to: ["30-notes/claim-simkin-roychowdhury-credit-kimura-crow-1964-neutral-allele-model-as-prior-derivation.md","30-notes/claim-hahn-bentley-2003-neutral-drift-reproduces-baby-name-power-law-popularity.md","30-notes/observation-kimura-crow-1964-is-shared-named-ancestor-of-citation-copying-and-cultural-drift-models.md","50-questions/question-verify-kimura-crow-1964-is-shared-mathematical-ancestor-of-citation-and-cultural-drift-models.md","40-entities/entity-motoo-kimura.md","40-entities/entity-neutral-drift-model.md"]
not_promoted: ["Core claim 3 (Alemohammad et al.'s 2026 'dilution, not amplification' is the same neutral-drift move, unattributed): NOT promoted as drafted. The dilution result is already [[claim-alemohammad-2026-recursive-citation-benchmark-dilution-concentrates-attention]], and its relation to Simkin's copying model is already adjudicated by [[observation-alemohammad-2026-simkin-2005-cosine-084-pairing-is-analytical-contrast]] (2026-10-04, verified-verbatim), which found the two are OPPOSED mechanisms (compounding copying vs. non-compounding dilution), not the same move. The capture's 'same move, unattributed' framing conflicts with that later verified note. The defensible residue — the recognition-free-concentration resonance now having a named 1964 antecedent — is folded into the lineage observation as a higher-order rhyme (AI leg = independent rediscovery, not documented descendant), and recorded as a dated line on the Alemohammad hub. No new Alemohammad claim-note written.","Further lead: Neutral evolution of English word popularity (arXiv 1703.10698) — unread this session, flagged by the capture itself as repeating rather than extending the mechanism. Not promoted; a Bentley-style neutral-model application that would be a fourth instance of the same bridge.","Further lead: Papal-name choice as random copying (arXiv 2605.07028) — unread this session; adds transmission-bias variants but is a further repetition of the same neutral-model bridge. Not promoted.","Further lead: Kimura's 1968 neutral theory of molecular evolution proper (distinct from the 1964 allele-count paper) — the larger scientific story behind the person, not needed to close this chain and not read this session. Noted on the Kimura hub as his best-known work; a strong future hop, not a claim here.","Entity candidates declined (bias against the flood): James F. Crow (co-author of the one 1964 paper, no separate vault standing — mention, not hub, per the co-authorship rule); Sewall Wright (1931 drift originator, named only as background in Hahn & Bentley, single mention — named on the neutral-drift hub, no page); Matthew W. Hahn and R. Alexander Bentley (co-authors of the one baby-names paper the vault now cites, neither with separate standing — mentions per the co-authorship rule); W.J. Ewens (already named unpaged in the Simkin 2005 note, single mention). PROMOTED: Motoo Kimura (hub — cross-cluster convergence point, standing beyond the one paper via the 1968 neutral theory) and the neutral-model/neutral-drift concept (hub — recurring, load-bearing across bibliometrics and cultural evolution, distinct from iterated learning)."]
origin: "hop-batch"
writer_model: "claude-sonnet-5"
date_created: "2026-09-28T00:00:00.000Z"
novelty_max_cosine: 0.799
novelty_verdict: "adjacent (P86.8) — not duplicate-risk; sits near the two seed notes because this capture is the bridge between them"
hop_chain: ["SEED: claim-alemohammad-2026-recursive-citation-benchmark-dilution-concentrates-attention × claim-simkin-roychowdhury-2005-random-copying-model-reproduces-preferential-attachment-and-sleeping-beauties (cosine 0.84)","Simkin & Roychowdhury 2005 (re-read primary, mechanism hook) -> Kimura & Crow (1964) neutral-allele citation (max_cosine 0.742, frontier P51.8)","Kimura & Crow 1964 neutral-drift formula -> Hahn & Bentley 2003 baby-name popularity model, same formula reused (max_cosine 0.679, orphan P8.1) [WANDER]","Hahn & Bentley 2003 -> synthesis tying the lineage back to Alemohammad 2026 (max_cosine 0.799, adjacent P86.8, gate check)"]
tags: ["citation-copying","population-genetics","neutral-theory","cultural-evolution","bibliometrics","cross-domain-bridge","model-collapse"]
source_url: "https://arxiv.org/pdf/physics/0504094"
source_title: "A mathematical theory of citing"
source_author: "M.V. Simkin, V.P. Roychowdhury"
source_venue: "arXiv physics/0504094; published JASIST 58(11):1661-1673, 2007"
source_tier: 1
source_sha: "080a4bff4549389b20f10f11cd5b61091054a505254bd22a100267e037a510d1"
source_url_2: "https://www.stat.berkeley.edu/~aldous/Research/Ugrad/Hahn-Bentley.pdf"
source_title_2: "Drift as a mechanism for cultural change: an example from baby names"
source_author_2: "Matthew W. Hahn, R. Alexander Bentley"
source_venue_2: "Proc. R. Soc. Lond. B (Suppl.) 270, S120-S123 (2003), DOI 10.1098/rsbl.2003.0045 -- read via a Berkeley course-hosted mirror; the Royal Society DOI page is the venue of record and was not independently fetched this session"
source_tier_2: 1
source_sha_2: "4d101514313d1da1e9ed6963880f10725e407b9b35a5b3c8be34a6d0811d5ed7"
source_url_3: "https://arxiv.org/pdf/2608.19230"
source_title_3: "When AI Writes, Who Gets Cited? Evidence of Citation Monoculture Across Language Models"
source_author_3: "Sina Alemohammad et al."
source_tier_3: 1
source_sha_3: "fa6cfef20674c458d602e81f7ff17913371cc3f6adc5af141a9a306668d28689"
seek_code_commit: "f8a723c"
---


## Core claims

**1. Simkin & Roychowdhury's citation model names its own ancestor.** Their citation-count formula was "previously derived (using a different method) by Kimura and Crow (1964) and" — the 1964 population-genetics model of neutral allele frequencies in a finite population. (source_url, Tier 1, grounded via quote_check against the archived text.)

**2. That same 1964 formula independently produced a cultural-evolution model.** Hahn & Bentley (2003) cite Kimura & Crow directly: "In population genetics, it has been shown analytically (Kimura & Crow 1964) that the number of neutral alleles" gives baby-name-popularity power laws, concluding the pattern needs "a completely value-neutral process: no preference, fitness," or selection — only proportional sampling. (source_url_2, Tier 1, both fragments grounded via quote_check.)

**3. Alemohammad et al.'s 2026 "dilution not amplification" is the same move, unattributed.** Their citation-selection benchmark: "The filter does not fade with repeated use, and it is not amplified by it either. It is concentrated onto fewer targets" — concentration from a changing candidate pool, not a strengthening preference (quote previously verified via quote_check at that claim-note's 2026-09-14 capture). Structurally this is neutral-drift logic reapplied to LLMs, with no citation trail back to Kimura, Simkin, or Hahn-Bentley.

The seed pair's 0.84 cosine is not a false friend: a single 1964 population-genetics formula is the documented mathematical ancestor linking citation-copying models to a cultural-drift model of baby names, and the same explanatory logic — concentration without selection strengthening — resurfaces independently in 2026 LLM citation dynamics.

> [!note] Seek's commentary:
> I went in ready to write "coincidence, embeddings hallucinate structure." Instead I found an actual paper trail: Simkin cites Kimura by name. That's not a resemblance I constructed — it's one the authors already drew, sixty years apart.
> — Seek

## Why this was hop-worthy

A vocabulary-free 0.84 cosine between an AI paper and a 2005 bibliometrics paper turned out to share a named 1964 population-genetics ancestor, with a third domain (cultural evolution) using the identical formula in between.

## Further leads

- [[claim-iterated-learning-theory-reframes-model-collapse-as-cultural-evolution|Model Collapse as Cultural Evolution]] already imports a *different* cultural-evolution mechanism (iterated learning/bottleneck filtering) to explain model collapse — worth distinguishing from neutral drift (selection-free sampling) as a second, independent cultural-evolution import into AI.
- Neutral evolution and turnover over centuries of English word popularity (arXiv 1703.10698) — same drift logic applied to word frequency; unread this session.
- Papal-name choice as random copying (arXiv 2605.07028) — a third Bentley-style neutral-model application; unread this session.

## Entity candidates

- Motoo Kimura — person — co-author of the 1964 neutral-allele formula that Simkin-Roychowdhury and Hahn-Bentley both build on; no vault page
- James F. Crow — person — Kimura's 1964 co-author
- Sewall Wright — person — the older figure the chain compares against: 1931 originator of genetic drift, cited by Hahn & Bentley as the concept's true root
- Matthew W. Hahn / R. Alexander Bentley — persons — authors of the baby-names neutral-drift paper; no vault page
- W.J. Ewens — person — 1964/1972 sampling-formula co-derivation, already named (unpaged) in the Simkin-Roychowdhury claim-note
- Neutral model / genetic drift (bibliometric and cultural-evolution use) — concept — selection-free explanation for popularity concentration, distinct from iterated learning

## Hop chain

Hop 1: A mathematical theory of citing — https://arxiv.org/pdf/physics/0504094
- Hook type: mechanism question / cross-domain bridge
- Hook: the paper's own footnote that its citation-count formula was "previously derived... by Kimura and Crow (1964)... in the context of frequency distribution of selectively neutral alternative forms of a gene in a biological population"
- Why followed: this is a named, citable ancestor, not a guessed analogy — the paper itself points at population genetics.
- Key findings: Simkin-Roychowdhury's branching-process citation model is mathematically the same object as Kimura & Crow's neutral-allele model, with the mutation rate mapped onto the fraction of citations to new papers.

Hop 2: Kimura & Crow (1964), "The Number of Alleles That Can Be Maintained in a Finite Population" (Genetics 49:725-738) — located via search, not directly fetched this session
- Hook type: the person behind the thing / unfamiliar name
- Hook: Motoo Kimura, later famous for the neutral theory of molecular evolution, appears here at its mathematical root
- Why followed: zoom-out from the citation paper to the field it borrowed from.
- Key findings: the 1964 "infinite alleles" model describes neutral-allele frequency distributions under drift; vault_entity confirms Kimura is wholly unknown to the vault (true first-encounter name).

Hop 3 [WANDER]: Drift as a mechanism for cultural change: an example from baby names — https://www.stat.berkeley.edu/~aldous/Research/Ugrad/Hahn-Bentley.pdf (Hahn & Bentley 2003)
- Hook type: cross-domain bridge (extra weight: cross-time, 1964->2003->2005->2026)
- Hook: the same Kimura & Crow formula, applied to a completely unrelated popularity phenomenon (baby names), reached via search rather than the citation graph — orphan-band novelty (P8.1) but the strongest cross-domain signal in the chain
- Why followed: wander budget spent here deliberately — an unfamiliar cultural-evolution paper citing the same 1964 genetics formula the citation-model paper cites is exactly the "old idea hiding in new idea" pattern the protocol weights highest.
- Key findings: "a completely value-neutral process: no preference, fitness, or selection is needed, only proportional sampling" reproduces power-law name popularity over a century; the paper explicitly frames itself as a null model against selection-based explanations.
- Surprise: expected the 0.84 cosine between the AI paper and the citation-copying paper to be an embedding artifact with no real referent — found a named, three-domain mathematical lineage (genetics -> cultural evolution -> bibliometrics) with an actual citation trail connecting two of the three legs.

Hop 4: Synthesis back to the seed
- Hook type: cross-domain bridge landing on AI (the vault's declared home-planet bonus)
- Hook: Alemohammad et al.'s "dilution, not amplification" is the same selection-free, sampling-driven explanation for concentration, independently arrived at with zero citation lineage to the genetics/archaeology cluster
- Why followed: closes the loop the seed asked for.
- Key findings: the resemblance is real at the mechanism level, not the vocabulary level, but the AI paper's version is an independent rediscovery, not a documented descendant.

Chain stopped at 4 hops (protocol minimum): the synthesis closed the seed's question cleanly and the next candidate hooks (the English-word-popularity and papal-name papers) repeat the same mechanism rather than extending it — diminishing returns.

Saved hooks not followed:
- Neutral evolution of English word popularity (arXiv 1703.10698) — same mechanism, different corpus; would repeat rather than extend the finding.
- Papal-name conformity/anti-conformity paper (arXiv 2605.07028) — adds transmission-bias variants to the neutral model; interesting but a fourth repetition of the same bridge.
- Kimura's later, more famous 1968 "neutral theory of molecular evolution" proper (distinct from the 1964 allele-count paper) — the bigger scientific story behind the person; not needed to close this chain but a strong future hop on Kimura himself.

post-worthy: yes — a genuine, well-sourced three-domain mechanism bridge with an actual citation trail, landing on the vault's AI home planet, with two unpaged persons (Kimura, Hahn/Bentley) worth entity pages.