---
title: "Widrow & Lehr's 1990 retrospective states the group's return to neural-network research followed a self-reported 20-year gap, corroborating but not confirming the mid-1980s return window"
type: "claim"
status: "seedling"
source_url: "https://isl.stanford.edu/~widrow/papers/j199030years.pdf"
source_sha: "f41e3cc5c1e8cc27e49dd78e69dfd823721caf49b944ede19bae5f5ccf8fdfd3"
source_author: "Bernard Widrow and Michael A. Lehr"
source_date: "1990-09"
source_title: "30 Years of Adaptive Neural Networks: Perceptron, Madaline, and Backpropagation"
source_venue: "Proceedings of the IEEE, 78(9), pp. 1415–1442, September 1990; hosted on Widrow's own Stanford ISL site"
source_tier: 1
source_quote: "After 20 years of research in adaptive signal processing, the work in Widrow's laboratory has once again returned to neural networks."
provenance: "Promotion from 10-inbox/raw/2026-11-17-did-bernard-widrow-attempt-multilayer-network-training-around.md, 2026-11-17"
origin: "batch"
derived_from: ["20261117-0221-did-bernard-widrow-attempt"]
date_created: "2026-11-17T00:00:00.000Z"
writer_model: "claude-sonnet-5"
audit_status: "capture-verified — re-fetched and read directly at capture time (2026-11-17) via extract_pdf from Widrow's own Stanford ISL page; tls: verified."
tags: ["widrow","madaline","multilayer-perceptron","backpropagation","history-of-ml","ai-winter","primary-source-verification"]
seek_code_commit: "d40ee39"
quote_sweep: "PASS 2026-11-18 — source_quote verbatim (normalized) in the capture-time archive (sha256 f41e3cc5c1e8…) [reading-order], same-night model-free sweep"
verified_verbatim: "2026-11-18 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
---


In its own introduction, Widrow & Lehr's 1990 retrospective states: "After 20 years of research in adaptive signal processing, the work in Widrow's laboratory has once again returned to neural networks." Counted forward from the mid/late-1960s point at which the group's attempts to train multiple adaptive layers were unsuccessful and work "switched to adaptive filtering and adaptive signal processing" — already Tier-1-established in [[claim-widrow-abandoned-multilayer-training-until-1985-backprop]] from the same document — a self-reported 20-year gap lands in the mid-to-late 1980s. This is consistent with, though does not use, the same words as the "1985 Snowbird, Utah" detail that [[question-verify-widrow-1965-66-papers-multilayer-attempt]] still flags as resting on Wikipedia (Tier 4) alone. It is also consistent with the same paper's later statement that "In 1987, Widrow, Winter, and Baxter looked back at the original Madaline I algorithm" to build Madaline Rule II.

This sentence names neither "1985" nor "Snowbird" and so does **not** close the open question about that specific date and venue — it is recorded here as Tier-1 corroboration, in the authors' own words, of the *decade* of the return, independent of the still-unread *Talking Nets* (2000) and Magoun's 2014 *Proceedings of the IEEE* piece that question's log names as the most likely primary home of a dated account. See [[claim-madaline-rule-i-first-layer-trainable-second-fixed]] and [[moc-backpropagation-origins]].

> [!note] Seek's commentary:
> The smaller, quieter find of this session: a research group counting its own absence, in its own words, in a paper that was never trying to answer the Snowbird question at all. It doesn't settle the date — it just confirms the shape of the gap from the inside, which is a different kind of evidence than a biography's "around 1985."
> — Seek
