---
id: "20260705-0231-who-was-david-parker"
title: "Who was David Parker, and what exactly did his 1985 independent rediscovery of backpropagation consist of — paper, report, or preprint?"
type: "capture"
status: "promoted"
date_promoted: "2026-07-06T00:00:00.000Z"
promoted_to: ["30-notes/claim-parker-1985-technical-report-not-cited.md"]
not_promoted: ["S81-64 invention-report number + Oct-1982 date — [unverified-quant] honored","Who-Parker-was biographical half — [unverified], person-gated; not queued"]
origin: "batch"
model: "claude-sonnet-5"
date_created: "2026-07-05T00:00:00.000Z"
provenance: "batch run 2026-07-05; web research via WebSearch + WebFetch, plus a direct full-text read (via the Read tool against a fetched PDF) of the primary Nature 1986 letter's mention of Parker; harvested from 2026-06-29-what-did-the-1986-rumelhart-hinton-williams-nature-paper-add-that-earlier-formulations-did-not"
derived_from: []
tags: ["backpropagation","david-parker","learning-logic","independent-discovery","history-of-ai","mit","technical-report","primary-source-verification"]
primary_sources_consulted: [{"url":"https://gwern.net/doc/ai/nn/1986-rumelhart-2.pdf","author":"David E. Rumelhart, Geoffrey E. Hinton, Ronald J. Williams","date":"1986-10-09","tier":1,"note":"Direct scan/mirror of the actual Nature letter 'Learning representations by back-propagating errors,' Nature 323, pp. 533-536 (received 1 May, accepted 31 July 1986). WebFetch's summarization pass could not extract readable text from the compressed scan, but the Read tool applied directly to the saved PDF binary returned clean OCR'd text of all four pages, including the numbered reference list. Quoted verbatim below."},{"url":"https://books.google.com/books/about/Learning_Logic.html?id=2kS9GwAACAAJ","author":"Google Books (catalog record, publisher-supplied metadata)","date":"retrieved 2026-07-05","tier":3,"note":"Library-style catalog record for Parker's report itself, directly fetched and resolved. Gives title, author, publisher chain, page count (47 pp.), and series/volume number."},{"url":"https://www.scirp.org/reference/referencespapers?referenceid=1827753","author":"Scientific Research Publishing (automated reference-list aggregator)","date":"retrieved 2026-07-05","tier":4,"note":"Directly fetched and resolved. One of many independent citation-aggregator pages reproducing the identical Parker (1985) 'Learning-Logic' citation; used only as corroboration for an uncontested bibliographic/historical claim, not as sole evidentiary basis."}]
access_failures: [{"url":"https://www.worldcat.org/title/learning-logic-casting-the-cortex-of-the-human-brain-in-silicon/oclc/17489314","reason":"HTTP 403 Forbidden, then HTTP 429 (rate-limited) on repeated WebFetch attempts. A WorldCat catalog listing for this exact title (OCLC 17489314) was located via search-engine snippet but never independently opened and read in this session."},{"url":"https://people.idsia.ch/~juergen/who-invented-backpropagation.html","reason":"DNS resolution failure (getaddrinfo ENOTFOUND) on WebFetch; Schmidhuber's own detailed 'Who Invented Backpropagation?' page could not be checked directly for its account of Parker."},{"url":"https://www.nature.com/articles/323533a0","reason":"Redirects to an authentication/paywall gate (idp.nature.com); the publisher's own hosted copy could not be read directly (the gwern.net mirror above was used instead, and is the same document)."},{"url":"https://arxiv.org/pdf/1404.7828","reason":"PDF fetch/summarization returned only compressed-stream noise; the ar5iv HTML mirror (https://ar5iv.labs.arxiv.org/html/1404.7828) was used instead and only yielded the short in-text citation '(Parker, 1985; LeCun, 1985, 1988)', not the underlying reference-list entry."}]
---


**Short answer on the document type: a technical report, not a peer-reviewed paper and not a preprint.** Multiple independent bibliographic sources converge on the same description: an internal MIT working paper, "Learning-Logic: Casting the Cortex of the Human Brain in Silicon," Technical Report TR-47. Direct verification of the primary 1986 Nature letter (via full OCR read of the actual document, not a summarization pass) adds an important wrinkle: Rumelhart, Hinton & Williams credit Parker's independent discovery only via "(personal communication)," with **no entry for Parker anywhere in their numbered reference list** — unlike Yann LeCun, who gets a citable footnote to a 1985 proceedings paper. That asymmetry is itself evidence about how (non-)public Parker's report was as of mid-1986. **On who David Parker actually was as a person** (profession, role, why he moved between a Stanford tech-licensing office in 1982 and an MIT research center by 1985) — no source located in this session establishes this, and that half of the question is left `[unverified -- could not confirm or deny after search]`.

---

## Claim: Parker's 1985 rediscovery took the form of an institutional technical report, not a journal paper

**Claim type:** Historical/bibliographic claim about what kind of document exists ("who did what," specifically "what did Parker produce"). Uncontested across independent sources — Tier 3–4 acceptable per the sourcing floor — and additionally corroborated by the claim below at Tier 1.

The report is consistently cited, across independent catalog and reference sources, as:

> "Learning Logic: Casting the Cortex of the Human Brain in Silicon" — David B. Parker — Publisher: Center for Computational Research in Economics and Management Science, Alfred P. Sloan School of Management, Massachusetts Institute of Technology — Date: 1985 — Page Count: 47 pages — Series: Volume 47 of *Technical report / Center for Computational Research in Economics and Management Science, Alfred P. Sloan School of Management, Massachusetts Institute of Technology*

(quoted from the Google Books catalog record, directly fetched and resolved). This matches the "TR-47" designation ("Technical Report," internal series number 47) that recurs across independent citation-aggregator pages, e.g. scirp.org's reproduction of the same reference: "Parker, D.B. (1985) Learning-Logic Casting the Cortex of the Human Brain in Silicon. Technical Report Tr-47, Center for Computational Research in Economics and Management Science. MIT Cambridge, MA." No source located in this session describes it as a peer-reviewed journal article, a conference paper, or a preprint (the latter term, in the sense of a pre-journal-submission circulating manuscript, is anachronistic for the pre-arXiv era in the way it's used today, but nothing found suggests Parker's report was ever submitted to or accepted by a journal under this title). It reads, across every source found, as an internally-numbered institutional working paper.

**Sourcing floor check:** Cleared for a historical/bibliographic claim (Tier 3–4 acceptable when uncontested); the specific "document type = technical report" fact is additionally corroborated at Tier 1 by the claim immediately below.

| Field | Value |
|---|---|
| source_url | https://books.google.com/books/about/Learning_Logic.html?id=2kS9GwAACAAJ |
| source_author | Google Books catalog record (publisher-supplied metadata for D.B. Parker's report) |
| source_date | retrieved 2026-07-05; underlying report dated 1985 |
| source_tier | 3 |
| exact_quote | "Learning Logic: Casting the Cortex of the Human Brain in Silicon" ... "Center for Computational Research in Economics and Management Science, Alfred P. Sloan School of Management, Massachusetts Institute of Technology" ... "Volume 47 of Technical report" |

---

## Claim: the Rumelhart, Hinton & Williams (1986) Nature letter credits Parker's discovery only as "personal communication," with no citable reference — while LeCun's independent work gets a full bibliographic citation

**Claim type:** Specific technical/historical claim about the primary document's own treatment of Parker. Requires Tier 1–2 per the sourcing floor. **Cleared at Tier 1** — verified by reading the actual OCR'd text of the letter directly (not a tool's summary of it).

The letter's body text reads, in full: "To break symmetry we start with small random weights. Variants on the learning procedure have been discovered independently by David Parker (personal communication) and by Yann Le Cun³." The superscript "3" attached to Le Cun's name points to a numbered reference — "3. Le Cun, Y. *Proc. Cognitiva 85*, 599–604 (1985)" — in the letter's four-item reference list on p. 536. **David Parker's name carries no superscript and has no corresponding entry anywhere in that reference list.** The full reference list (all four entries) reads: "1. Rosenblatt, F. *Principles of Neurodynamics* (Spartan, Washington, DC, 1961). 2. Minsky, M. L. & Papert, S. *Perceptrons* (MIT, Cambridge, 1969). 3. Le Cun, Y. *Proc. Cognitiva 85*, 599–604 (1985). 4. Rumelhart, D. E., Hinton, G. E. & Williams, R. J. in *Parallel Distributed Processing: Explorations in the Microstructure of Cognition*. Vol. 1: *Foundations* (eds Rumelhart, D. E. & McClelland, J. L.) 318–362 (MIT, Cambridge, 1986)." Parker is absent.

This directly bears on the "paper, report, or preprint" question: as of the letter's acceptance (31 July 1986), its authors did not treat Parker's work as something with a citable, publicly circulating reference — only as something conveyed to them privately. That is consistent with Parker's report being a limited-circulation internal technical report at that point, rather than a document in general scholarly circulation the way LeCun's 1985 proceedings paper already was.

(Note: an earlier general web-search pass in this same session had produced a plausible-looking but **incorrect** synthesized claim that "the references section of this paper includes Parker, D. B. (1985)... Cambridge, MA: Massachusetts Institute of Technology, Center for Computational Research in Economics and Management Science." Direct reading of the primary document's actual reference list, reproduced in full above, shows this is false — Parker does not appear in the reference list at all, only as an unfootnoted "personal communication" in the body text. This is recorded here as a caution about trusting search-engine-synthesized "quotes" of primary documents without independently reading the source.)

**Sourcing floor check:** Cleared — Tier 1, the primary letter itself, read in full and quoted directly.

| Field | Value |
|---|---|
| source_url | https://gwern.net/doc/ai/nn/1986-rumelhart-2.pdf |
| source_author | David E. Rumelhart, Geoffrey E. Hinton, Ronald J. Williams |
| source_date | 1986-10-09 |
| source_tier | 1 |
| exact_quote | "Variants on the learning procedure have been discovered independently by David Parker (personal communication) and by Yann Le Cun³." — plus the full 4-item reference list on p. 536, which contains no Parker entry. |

---

## Claim: Parker's rediscovery had an earlier, non-published precursor — an internal Stanford invention disclosure, reportedly dated October 1982

**Claim type:** Historical claim carrying a specific document number and month/year date — treated cautiously per the sourcing floor's quantitative-adjacent guidance for specific dates and identifiers.

Multiple secondary sources describe Parker as having filed an "invention report" titled "Learning – Logic" (Invention Report S81-64, File 1) with Stanford University's Office of Technology Licensing in October 1982 — roughly 2.5 years before the MIT technical report and about 4 years before the Rumelhart/Hinton/Williams Nature letter. This detail recurs in near-identical wording across many independently-hosted pages, suggesting a single underlying origin (plausibly Werbos's historical writing on the priority question, though that origin could not be directly located and read in this session — see access_failures). No Tier 1–2 source was independently opened and read to confirm the exact report number or month.

`[unverified-quant -- needs primary: the specific invention-report number "S81-64" and the month "October" 1982 have not been traced to a directly-read Tier 1-2 source in this session; the general fact that Parker disclosed the idea at Stanford before his 1985 MIT report is repeated consistently but not independently verified against a primary record]`

| Field | Value |
|---|---|
| source_url | (no single Tier 1-2 source directly read; claim recurs identically across multiple secondary aggregator pages surfaced by search) |
| source_author | unattributed / recurring secondary paraphrase |
| source_date | claim describes October 1982; retrieved 2026-07-05 |
| source_tier | 4 (or lower — could not pin to a specific, directly-verified page) |
| exact_quote | not independently confirmed; recorded and flagged rather than presented as verified |

---

## Claim: Parker separately published later, related, peer-reviewed conference work in 1987 — but this is not the 1985 rediscovery itself

**Claim type:** Historical/bibliographic. Tier 3–4 acceptable (uncontested, consistent bibliographic citation across independent aggregator pages).

Independent citation records consistently describe a later work: D. B. Parker, "Optimal algorithms for adaptive networks: second order back propagation, second order direct propagation, and second order Hebbian learning," in *Proceedings of the IEEE International Conference on Neural Networks* (1st ICNN, San Diego, June 1987), Vol. II, pp. 593–600. This is a distinct, extended piece of work on second-order training methods, not a republished or peer-reviewed version of the original 1985 "Learning-Logic" rediscovery — it is included here only to establish that Parker's *known, formally citable* publication record on this topic begins in 1987, two years after the internal MIT report and after the Rumelhart/Hinton/Williams Nature letter had already been submitted.

| Field | Value |
|---|---|
| source_url | (multiple independent citation-aggregator pages; not individually re-verified beyond consistency across several) |
| source_author | recurring secondary citation |
| source_date | conference held June 1987; retrieved 2026-07-05 |
| source_tier | 4 |
| exact_quote | "Optimal algorithms for adaptive networks: second order back propagation, second order direct propagation, and second order Hebbian learning" — Proc. IEEE ICNN, Vol. II, pp. 593–600 (1987) |

---

## What could not be established: who David Parker actually was

No source located in this session gives David Parker's profession, academic status, or role — e.g., whether he was a graduate student, a research staff member, an industry engineer, or a consultant, at either Stanford (1982) or MIT's Sloan-affiliated Center for Computational Research in Economics and Management Science (1985). He does not appear among the interview subjects in Anderson & Rosenfeld's *Talking Nets: An Oral History of Neural Networks* (MIT Press, 2000), which does include contemporaries such as Rumelhart, Hinton, Werbos, and Widrow — his absence from that roster was confirmed via the book's listed table of contributors, but this is negative evidence (an absence), not a positive biographical fact, and is noted rather than treated as a claim. Searches for a biography, obituary, retrospective profile, or first-person account by Parker himself did not surface one.

**This half of the core question is `[unverified -- could not confirm or deny after search]`.**

> [!note] Seek's commentary:
> The most interesting find here wasn't a fact about Parker but a fact about the Nature letter — the missing reference-list entry. It reframes "independent rediscovery" as a spectrum rather than a binary: LeCun's independent work had already cleared the bar of being a citable, publicly-circulating document by mid-1986, while Parker's apparently had not, at least not from Rumelhart/Hinton/Williams's vantage point. That's a genuinely load-bearing distinction for a note on "what did his 1985 rediscovery consist of" — it suggests the honest answer isn't just "a technical report" but "a technical report that, a year later, was still circulating narrowly enough that the most consequential paper in the field's history could only gesture at it as private correspondence." Also worth flagging procedurally: a general web-search pass produced a fluent, specific-sounding "quote" of the Nature paper's reference list that turned out to be fabricated once the actual OCR'd text was read directly. That's the second time in this general research thread (see the 2026-07-05 sibling capture on symmetry-breaking) that a WebFetch/WebSearch summarization has confidently reported page content that isn't there. Direct reads of saved primary-source PDFs, not summarization passes over them, are the only way this got caught.

**Central question status:** Partially confirmed. The "paper, report, or preprint" half is answered with reasonable confidence: **a technical report** (MIT TR-47, "Learning-Logic: Casting the Cortex of the Human Brain in Silicon," Center for Computational Research in Economics and Management Science / Sloan School of Management, 1985, 47 pp.), preceded by an even more informal, unpublished 1982 Stanford invention disclosure `[unverified-quant -- needs primary]`, and — per direct primary-source verification — still uncited/uncitable as of the Nature letter's mid-1986 acceptance. The "who was David Parker" half is `[unverified -- could not confirm or deny after search]`.