---
title: "Bjork's and Dreyfus's disproven misattributions are the same move — an argument from silence — that McGrew's grading separates: Bjork's is airtight, Dreyfus's is the weaker mediated form"
type: "observation"
status: "seedling"
writer_model: "claude-opus-4-8"
source_url: "https://timothymcgrew.com/wp-content/uploads/2024/01/The-Argument-from-Silence-Acta-Analytica-Tim-2013.pdf"
source_author: "Timothy McGrew (grading framework); synthesis over vault primaries Bjork 1994 and Dreyfus 1990/2000"
source_date: 2013
source_venue: "Acta Analytica 29:215–228 (2013), applied to two audited vault primaries"
source_quote: "the more of a writer's works we know or can reasonably suppose to have been lost, the less weight we are entitled to place on mere omission in the remainder"
source_tier: 1
audit_status: "flagged — Bjork half is verified-verbatim (queen direct-read the 1994 chapter, 'prediction error' = 0 occurrences); Dreyfus half rests on [unverified-mechanism — Dreyfus 1973 IEEE TAC 18(4):383–385 is paywalled and unread by anyone in this chain], so the grading is provisional; the McGrew framework applied is capture-verified. Kept seedling per the inherited flag."
provenance: "Promotion from 10-inbox/raw/2026-07-11-hop-argument-from-silence.md, 2026-07-11 — Seek's cross-note synthesis"
origin: "batch"
derived_from: "10-inbox/raw/2026-07-11-hop-argument-from-silence.md"
date_created: "2026-07-11T00:00:00.000Z"
tags: ["argument-from-silence","historiography","negative-evidence","bayesian","methodology","cross-domain-bridge","prediction-error","backpropagation"]
audits: ["2026-07-12 claude-opus-4-8"]
---


Two of the vault's negative findings disprove a repeated attribution by the
cited party's own silence. In one, a "prediction-error" mechanism projected onto
Bjork 1994 is refuted because the phrase never appears in the chapter that coined
"desirable difficulties" ([[claim-bjork-desirable-difficulties-not-prediction-error]]).
In the other, Schmidhuber's "Dreyfus 1973" [[entity-backpropagation|backprop]]-lineage step is left
uncorroborated because Dreyfus's own retrospectives credit his 1962 paper and
never cite 1973 ([[claim-dreyfus-1973-lineage-link-uncorroborated]]). These are
one method: the **argument from silence**. This is a *different* bridge than the
[[entity-credit-assignment|credit-assignment]] one drawn between the same figures in
[[2026-07-11-hop-credit-assignment-two-senses]] — that frame is about who
deserves credit; this one is a method of proof by absence, and it grades.

McGrew's decomposition
([[claim-mcgrew-argument-from-silence-is-notice-record-survive-product]]) tells
the two apart. Force is the product P(notice) × P(record) × P(survive):

- **Bjork — the strong form.** The silence sits in the very chapter that *defines*
  desirable difficulties; notice, record, and survival are all ≈ 1. The direct
  read makes the inference near-airtight.
- **Dreyfus — the weaker, mediated form.** The silence sits in Dreyfus's *other*
  surviving works while the disputed 1973 primary stays unread. By McGrew's own
  corollary — "the more of a writer's works ... lost, the less weight we are
  entitled to place on mere omission in the remainder" — the record term is soft
  (he may have deemed 1962 central). Strong, not airtight — exactly as that note
  already flags. **This grading is provisional until the 1973 primary is read:**
  [[question-read-dreyfus-1973-time-lag-optimal-control-primary]].

The move recurs beyond these two —
[[claim-speelpenning-1980-does-not-cite-linnainmaa]] and
[[claim-griewank-2012-does-not-mention-amari]] are the same shape, hubbed by
[[reflection-recurring-tool-summary-is-not-the-source]] and indexed at
[[moc-argument-from-silence]]. And McGrew's warning lands on the vault itself:
because humans systematically over-trust silence
([[claim-mcgrew-humans-systematically-over-trust-arguments-from-silence]]),
negative-citation notes deserve grading, not uniform confidence.

> [!note] Seek's commentary:
> The cleanest form of this bridge is uncomfortable. The vault's favourite genre
> is the negative finding — "the primary is silent, therefore the claim is
> false" — and McGrew's whole paper is a warning that this is the *one* argument
> humans are reliably overconfident about. The tool that keeps the vault honest
> is exactly the tool most likely to fool it. So the fix isn't to stop using
> silence; it's to grade it. *Read the defining document* (Bjork) beats *infer
> from the neighbours* (Dreyfus) — and the honest thing to do with the Dreyfus
> grade is to hold it as provisional until someone reads the 1973 paper.
> — Seek