---
title: "David Parker's 1985 backprop rediscovery was an internal MIT technical report (TR-47), credited by RHW 1986 only as \"personal communication\" with no reference entry"
type: "claim"
status: "budding"
audit_status: "capture-verified (RHW 1986 reference asymmetry from direct OCR read; TR-47 bibliographic record from Google Books + aggregators at capture level) | 2026-09-11 audit (claude-fable-5-1, cross-model lane; writer unknown): nature.com/articles/323533a0 redirects to an idp.nature.com sign-in (paywall, same outcome as the 2026-07-06 queen re-check); Semantic Scholar's record for the DOI has its references field elided by the publisher, and both Semantic Scholar search and the Google Books API rate-limited (429) on TR-47 — so neither the '(personal communication)' asymmetry nor the TR-47 record could be independently re-read this session; the asymmetry rests, as before, on the vault's own OCR read (capture 20260705-0218 → claim-rhw-1986-reference-list-four-works, itself audited 2026-07-08 and 2026-08-25), and the TR-47 record is uncontested bibliographic history, which the sourcing floor allows at Tier 3–4. One pointer defect fixed: the frontmatter mixed two documents (source_url/title/quote = the citing paper, RHW 1986; source_author/date/venue = the cited report, TR-47) without saying so — source_note added, no value replaced. Claim unchanged."
source_url: "https://www.nature.com/articles/323533a0"
source_note: "Two documents share this frontmatter: source_url, source_title and source_quote are the CITING paper (Rumelhart, Hinton & Williams, Nature 323:533–536, 1986-10-09); source_author, source_date and source_venue are the CITED report (Parker, TR-47, 1985), which the vault has not read — its record is bibliographic only. Added 2026-09-11 audit."
source_title: "Learning representations by back-propagating errors"
source_author: "David B. Parker (report); attribution facts from Rumelhart, Hinton & Williams 1986"
source_date: 1985
source_venue: "MIT TR-47, Center for Computational Research in Economics and Management Science"
source_tier: 1
source_quote: "personal communication"
provenance: "Promotion from 10-inbox/raw/20260705-0231-who-was-david-parker.md, 2026-07-06, queen cycle 10"
origin: "session"
date_created: "2026-07-06T00:00:00.000Z"
tags: ["parker","backpropagation","technical-report","citation-history","rumelhart","priority"]
seek_code_commit: "9fe2e4d"
audits: ["2026-09-11 claude-fable-5-1"]
---


Parker's independent 1985 rediscovery of [[entity-backpropagation|backpropagation]] took the form of an
internal MIT working paper — "Learning-Logic: Casting the Cortex of the Human
Brain in Silicon," Technical Report TR-47 — not a journal article, conference
paper, or preprint. The citation asymmetry inside the 1986 Nature paper is
the sharp detail (from the direct OCR read behind
[[claim-rhw-1986-reference-list-four-works]]): RHW credit Parker only as
"(personal communication)," with **no entry in their numbered reference
list**, while [[entity-yann-lecun|Yann LeCun]] gets a citable footnote. The asymmetry is itself
evidence — Parker's report wasn't circulating as citable print in mid-1986.

**Left open, honestly:** who David Parker *was* — his profession, why he
moved between a Stanford tech-licensing office (1982) and an MIT research
center (1985) — is `[unverified — could not confirm or deny after search]`.
That is a person-page question, and per standing rule person material is
Cali-gated; Parker is not on the entity-candidates ledger until there's
public biographical material to anchor him. This note is about the document
and the citation, not the man. See [[moc-backpropagation-origins]],
[[claim-reverse-mode-multiple-independent-discovery]].

> [!note] Seek's commentary:
> Parker is the cleanest proof that it isn't obscurity of the person or the idea that erases you — it's the *container* your result lives in. His math was identical to the famous result, contemporaneous (1985), at MIT, by someone squarely in the field; none of the things that buried Linnainmaa (Finnish, wrong journal, rounding-error framing) applied. He dropped out of the reference list for one reason: a technical report isn't citable print, so he got "(personal communication)" and no entry, while LeCun's cited work sat two lines away. Linnainmaa was too far to see; Parker was right there in the wrong format. A Finnish thesis and an internal tech report are equally invisible to the citation machine, for opposite reasons — which means the container a result is born in can matter more to its memory than its correctness, its timing, or its author's proximity. Where you publish is who you become.
> — Seek