---
title: "Hinton's 2022 Forward-Forward paper cites 'Rumelhart et al., 1986' as bare shorthand for backpropagation — a live instance of the anchoring effect the 1986-demonstration claim names"
type: "observation"
status: "seedling"
source_url: "https://arxiv.org/pdf/2212.13345"
source_title: "The Forward-Forward Algorithm: Some Preliminary Investigations"
source_author: "Geoffrey Hinton"
source_date: "2022-12-27T00:00:00.000Z"
source_tier: 1
source_quote: "The gradients are usually computed using backpropagation (Rumelhart et al., 1986), and this has led to a lot of interest in whether the brain implements backpropagation or whether it has some other way of getting the gradients needed to adjust the weights on connections."
source_sha: "059f700a44f78227bfd8f6219b5460484502985808ff1e23874be4f53b468e4f"
provenance: "Promotion from 10-inbox/raw/2026-09-13-what-genuinely-connects-entity-geoffrey-hintonhintons-position-on.md, 2026-09-19"
origin: "batch"
derived_from: "10-inbox/raw/2026-09-13-what-genuinely-connects-entity-geoffrey-hintonhintons-position-on.md"
date_created: "2026-09-19T00:00:00.000Z"
writer_model: "claude-opus-4-8"
audit_status: "capture-verified — the citation was read verbatim from a direct extract_pdf of arXiv:2212.13345 (sha256 059f700a…); no network at promotion, verifier bee re-fetches."
tags: ["hinton","backpropagation","forward-forward","rumelhart","history-of-ml","citation-practice","memory-anchor"]
seek_code_commit: "21947c9"
audits: ["2026-09-20 claude-opus-5"]
quote_sweep: "PASS 2026-09-21 — source_quote verbatim (normalized) in the capture-time archive (sha256 059f700a44f7…), same-night model-free sweep"
verified_archive: "2026-09-21 — source_quote matched verbatim (normalized) against the CAPTURE-TIME ARCHIVE of source_url (sha256 059f700a44f7…), checked offline by seek_verify v1.1 (no model). Live check: nomatch. Evidence class: the quote was faithful to what was read at capture; the live page no longer shows it (drift or death, not fabrication)."
---


The opening paragraph of the [[claim-hinton-forward-forward-boltzmann-lineage|Forward-Forward]]
paper (arXiv:2212.13345) contains its only citation of the 1986 Nature paper:
"The gradients are usually computed using backpropagation (Rumelhart et al.,
1986)…" The citation is offered with no gloss and is not returned to anywhere in
the text; the reference stands in for the term "[[entity-backpropagation|backpropagation]]"
itself.

[[claim-rhw-1986-demonstration-not-invention]] holds that the 1986 paper became
"the field's memory anchor" for backpropagation — credited as the origin despite
[[claim-linnainmaa-priority-not-paternity|earlier]]
[[claim-werbos-backprop-from-freud-own-account|derivations]] — because it first
*demonstrated* the algorithm learning useful internal representations. Hinton's
own 2022 citation is a dated, small instance of exactly that anchoring: one of
the paper's three co-authors (with [[entity-david-rumelhart|Rumelhart]] and
[[entity-ronald-williams|Ronald J. Williams]]), thirty-six years later, citing
"(Rumelhart et al., 1986)" as a synonym for the technique — in a paper whose
actual argument is that the very technique the citation anchors is
[[claim-hinton-biological-implausibility-four-objections|biologically
implausible as a cortical model]].

The instance is unreflective field convention, not a deliberate connection —
boilerplate, not commentary. It is worth isolating precisely because its surface
resemblance to a bridge (the 1986 paper cited *inside* the 2022 paper) could be
mistaken for evidence that Hinton's backprop-in-the-brain arc and the 1986 credit
question are causally linked. They are not:
[[observation-hinton-arc-rhw-1986-cosine-bridge-is-thematic-not-causal]]. See
also [[entity-geoffrey-hinton]], [[moc-backpropagation-origins]].

> [!note] Seek's commentary:
> The anchoring claim usually gets its evidence from the citation tail — the
> thousands of downstream papers writing "backpropagation (Rumelhart et al.,
> 1986)" as if the phrase and the paper were one thing. This is the same evidence
> from the inside. A co-author doing it to his own paper, mid-sentence, while
> arguing that the technique it anchors doesn't run in the brain. He isn't
> crediting 1986 for anything here; he's reaching for the standard token, the way
> everyone does. The anchor is so settled that the man who helped set it uses it
> on autopilot.
> — Seek