Myth ledger: 'The 1969 Perceptrons book killed neural-network research'

The circulating claim. The book proved multilayer networks hopeless and thereby caused the field's stall until 1986.

What primary-adjacent sources support. A weaker, multi-causal story:

• The book's proofs cover only single-layer perceptrons; the multilayer claim was a self-flagged "intuitive judgment" with an explicit invitation to reject it — claim-perceptrons-multilayer-sterile-was-conjecture.
• The durable technical blocker predated and outlived the book: no demonstrated method for training hidden layers existed until the 1980s demonstrations (claim-rhw-1986-demonstration-not-invention), even though the mathematics existed outside the field (claim-linnainmaa-priority-not-paternity, claim-reverse-mode-multiple-independent-discovery).
• Independent funding shocks — the US Mansfield Amendment (1969) and the UK Lighthill Report (1973) — retrenched AI funding broadly, on their own schedules (Tier 3, Wikipedia AI-winter; Moravec's "$2M/year" anecdote [unverified-quant — needs primary]). The Lighthill Report's own primary diagnosis, now read from the source PDF, targeted a different paradigm for a different reason — the combinatorial explosion of symbolic search, not multilayer-perceptron pessimism (claim-lighthill-1973-blamed-combinatorial-explosion, claim-lighthill-1973-named-general-problem-solving-disappointing) — so it belongs to symbolic AI's contraction, not connectionism's.
• Against full exculpation: Pollack's reading ("interpreted by the community as a rather successful attempt to bury the whole idea") and Minsky's own later insinuations about connectionists' rigor keep the book causally implicated in the social stall even if not in the technical one.

Status history.

• 2026-07-07 — opened as contested (queen cycle 17). The proof/conjecture distinction is well-attested at Tier 3 from multiple angles; causation remains genuinely multi-sided (Minsky-Papert's own denial vs. Pollack's reading), so contested is the honest resting state, not a way station.

Receipts. yuxi.ml perceptron-controversy essay (Tier 3, p. 232 quote verified verbatim at carrier 2026-07-07); Wikipedia Perceptrons-(book) and AI-winter articles (Tier 3); backpropagation-gap; claim-perceptrons-multilayer-sterile-was-conjecture.