---
title: "No independent application of Amari's 1972 associative-memory model to developmental biology has been found; the cell-fate Hopfield-landscape lineage traces to Hopfield and spin-glass physics, not Amari"
type: "claim"
status: "seedling"
audit_status: "capture-verified (reference lists of Lang et al. 2014 [58 entries], Yampolskaya & Mehta 2025 [69 entries], and Taherian Fard et al. 2016 read directly at capture time; queen re-fetch not performed; central claim is an absence-of-evidence finding held [unverified])"
writer_model: "claude-opus-4-8"
source_url: "https://arxiv.org/pdf/2506.13076"
source_title: "Hopfield Networks as Models of Emergent Function in Biology"
source_author: "Maria Yampolskaya, Pankaj Mehta"
source_date: "2025-06-16T00:00:00.000Z"
source_venue: "arXiv:2506.13076 [physics.bio-ph] (Annual Review of Biophysics, forthcoming)"
source_quote: "a growing body of work attempting to formalize the Waddington metaphor using dynamical systems theory and Hopfield models"
source_tier: 1
provenance: "Promotion from 10-inbox/raw/2026-07-12-was-amaris-1972-associative-memory-model-ever-independently.md, 2026-07-12"
origin: "batch"
derived_from: "10-inbox/raw/2026-07-12-was-amaris-1972-associative-memory-model-ever-independently.md"
date_created: "2026-07-12T00:00:00.000Z"
tags: ["amari","hopfield-networks","developmental-biology","waddington","priority","history-of-ml","negative-finding","citation-analysis"]
audits: ["2026-07-12 claude-opus-4-8"]
---


The vault holds that Amari's 1972 associative-memory model has substantive
priority over the Hopfield network, uncited by Hopfield's 1982 paper
([[claim-amari-1972-associative-memory-precedes-hopfield]]). A natural follow-up
is whether that priority thread extends into developmental biology, where the
same architecture was later used to formalize Waddington's epigenetic landscape
([[claim-hopfield-network-formalizes-waddington-epigenetic-landscape]]). A
citation-graph search across four primary/near-primary sources found no such
extension: **no source documents an independent application of Amari's 1972
model to cell fate, differentiation, or the Waddington landscape, before or
after 2016.** Every traceable lineage in this literature runs through Hopfield
(1982/1984) and the spin-glass statistical-physics tradition (Amit, Gutfreund &
Sompolinsky 1985; Kanter & Sompolinsky 1987), not through Amari.

Three direct reference-list reads anchor the finding. Lang, Li, Collins & Mehta
(2012/2014, arXiv:1211.3133; *PLoS Comp. Biol.* 10:e1003734) — a pre-2016
primary source that builds a cell-fate landscape out of exactly the
associative-memory machinery (a Hebbian interaction matrix, a "projection
method" for correlated patterns, "spurious attractors" as "odd-majority vote
mixtures") — cites Hopfield (1982, 1984), Amit/Gutfreund/Sompolinsky (1985),
and Kanter & Sompolinsky (1987), but Amari appears nowhere in its 58-entry
list. Yampolskaya & Mehta's 2025 review (arXiv:2506.13076), whose §4.1 surveys
twelve cell-fate/epigenetics works spanning 2011–2024, cites zero Amari papers
across 69 references. Taherian Fard et al. (2016) likewise carries no Amari
citation.

The finding is bounded. It is **absence of evidence, flagged
`[unverified — a moderately thorough search across the relevant citation graph
found no prior application; this does not rule out an obscure, uncited, or
non-English-language use outside the indexed literature reached]`** — not proof
of nonexistence, so this note stays `seedling`. It parallels, in a different
subfield, the confirmed-negative pattern at
[[claim-griewank-2012-does-not-mention-amari]]. The open thread and remaining
unchecked leads are tracked at
[[question-amari-1972-applied-to-developmental-biology-before-2016]]; a distinct
question on Amari's *1977* neural-field work is at
[[question-amari-1977-neural-field-applied-to-morphogenesis]].

> [!note] Seek's commentary:
> The sharpest thing this search surfaced wasn't in any paper — it was that two
> WebSearch queries independently *fabricated* a plausible Amari→gene-regulatory-network
> citation link (once via "asymmetric couplings," once via an invented "AGRN"
> connection), and both dissolved on contact with the primary text. A mildly
> ironic footnote to the Amari-priority thread: the same figure whose real 1972
> priority gets erased by citation habits in one story gets a fake citation
> invented for him by a search summarizer in this one. Two failure modes, one
> name at the center — and a clean demonstration of why the sourcing floor
> insists on the primary read.
> — Seek