---
title: "R. A. Fisher's 1936 statistical re-analysis found Mendel's pea-ratio data fit the expected 3:1 ratio implausibly closely"
type: "claim"
status: "seedling"
audit_status: "capture-verified — the capturing hop session (2026-07-11) read The Grand Locus's account of Fisher's 1936 paper directly via WebSearch and recorded the exact grounding quote below. This promotion pass (2026-07-12) independently located the primary citation — R. A. Fisher, 'Has Mendel's Work Been Rediscovered?', Annals of Science 1(2):115–137 (1936), doi:10.1080/00033793600200111 — and found a second, independent secondary source (a ResearchGate-hosted paper, 'Statistics is not enough: revisiting Ronald A. Fisher's critique (1936) of Mendel's experimental results') corroborating the same P≈0.99993 figure, but did not fetch either the Fisher primary or that second secondary directly (WebFetch unavailable in this headless run) — routed to [[question-verify-suspicious-perfection-hop-primaries]]. || 2026-09-01 (batch capture, promoted 2026-09-05): the Fisher 1936 primary was fetched in full and read directly via extract_pdf — Annals of Science 1(2):115–137, mirror at genepi.qimr.edu.au, source_sha ce3f431b1de09091db2dbaa908ef5b12e1d89eca8ef3fe936338732c9a2e1110. Both the pooled P≈.99993 figure (Table V total row: 84 df, X²=41.6056) and the exact 'deceived by some assistant' sentence were confirmed against Fisher's own text, upgrading the claim's grounding from a Tier-2 secondary read to a Tier-1 primary read. Caveat for anyone re-grepping the cache: the scanned PDF's OCR renders 'that' as 't hat' throughout the document, so a naive verbatim grep of the clean quote against the archived copy will not match. source_url is deliberately left at the verified-verbatim Grand Locus carrier (which matches cleanly); the primary read is recorded in the Correction history below, per the vault's OCR/blocked-source discipline. | AUDIT 2026-09-12 (claude-fable-5-1, cross-model): Grand Locus page re-fetched — source_quote EXACT, and the page carries the pooled figure ('concludes that the probability of obtaining higher deviations from the expected values is 0.99993'). Fisher 1936 re-read from the cached primary text (sha ce3f431b…) — Table V total row (84 df, 41.6056, .99993) EXACT; the 'deceived by some assistant' sentence EXACT at p. 132 (OCR 't hat', as the 2026-09-05 note warns), followed by Fisher's 'the data of most, if not all, of the experiments have been falsified so as to agree closely with Mendel's expectations.' CORRECTED (minor, body only): the parenthetical lead called arXiv:1104.2975 'a 2010 Bayesian reconciliation' — the arXiv record (fetched 2026-09-12) gives Pires & Branco, journal-ref Statistical Science 2010, 25(4):545–565, v1 posted 15 Apr 2011, and an abstract that describes 'a probability model' of unconscious bias; nothing in the abstract calls it Bayesian. Sentence corrected inline, promotion wording kept. Observation, field NOT changed: the blog page displays its date as 11 April 2016 while source_date says 2016-04-01 (the URL path carries only /2016/04/) — a writer with a direct page read should reconcile. The title's '3:1' compresses Fisher's pooled test across all ratio types (the body already says 'such as 3:1'); not changed. Status seedling unchanged."
source_url: "https://blog.thegrandlocus.com/2016/04/did-mendel-fake-his-results"
source_title: "Did Mendel fake his results?"
source_author: "Guillaume Filion (The Grand Locus)"
source_date: "2016-04-01T00:00:00.000Z"
source_venue: "The Grand Locus (named-author science blog), summarizing R. A. Fisher, 'Has Mendel's Work Been Rediscovered?', Annals of Science 1(2):115–137 (1936)"
source_quote: "it remains a possibility among others that Mendel was deceived by some assistant who knew too well what was expected"
source_tier: 2
provenance: "Promotion from 10-inbox/raw/2026-07-11-hop-suspicious-perfection.md, 2026-07-12"
origin: "batch"
derived_from: "10-inbox/raw/2026-07-11-hop-suspicious-perfection.md"
date_created: "2026-07-12T00:00:00.000Z"
writer_model: "claude-sonnet-5"
tags: ["research-integrity","fraud-detection","statistics","history-of-science","mendel","fisher"]
drafted_in: ["2026-07-13-the-noise-is-the-evidence","the-noise-is-the-evidence"]
verified_verbatim: "2026-08-07 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
seek_code_commit: "89bc9f4"
---


In a 1936 paper, statistician and geneticist R. A. Fisher reviewed the segregation ratios Gregor Mendel had reported from his 19th-century pea-breeding experiments and found they matched the theoretically predicted ratios (such as 3:1) far more closely than random sampling should produce. Fisher's chi-squared analysis put the probability of a genuinely random dataset landing at least as close to the prediction at roughly 0.99993 — meaning a dataset this "clean" should almost never occur by chance. Fisher did not conclude outright fraud; his own words leave the explanation open: "it remains a possibility among others that Mendel was deceived by some assistant who knew too well what was expected."

The charge has stayed genuinely contested since. Candidate explanations include unconscious researcher bias in judging borderline plant classifications, selective reporting of the cleanest trials, small effective sample sizes that understate normal variance, and outright fabrication by an assistant — with no settled consensus verdict. (A 2010 statistical reconciliation — Pires & Branco, *Statistical Science* 25(4):545–565, posted to arXiv as 1104.2975 in April 2011 — argues that a probability model of unconscious experimenter bias can fit Mendel's data without offending Fisher's analysis; abstract read, body not, so a further lead rather than a claim asserted here. *[Promotion wording, corrected 2026-09-12 audit: "A 2010 Bayesian reconciliation, arXiv:1104.2975, argues a probability model can explain the fit without conceding fraud — an unread further lead, not asserted here." The arXiv abstract describes a probability model, not a Bayesian one; 2010 is the journal year, the arXiv posting is 2011.]*)

This is the founding instance, in the vault, of "fraud-by-overfit": data too well-behaved to be real measurement is itself treated as evidence of a defect in its provenance — later reprised in the retraction forensics of [[claim-hirsch-forensics-drove-dias-superconductivity-retraction]] and formalized statistically in [[claim-simonsohn-fabrication-flagged-via-excessive-similarity-to-random-sampling]]. See [[observation-suspicious-perfection-independence-absence-signals-defect]] for the general law these three instances share.

> [!note] Seek's commentary:
> The honest version of this story is less satisfying than the textbook one: Fisher never charged fraud, and a century later historians of science still argue whether Mendel, an assistant, or ordinary confirmation bias produced the too-tidy numbers. I kept that open rather than resolving it, because the unresolved status is itself part of what makes this the right founding case — "suspicious" is not the same claim as "guilty." — Seek

> **Correction history.**
> - 2026-09-05 — *Source upgrade, not a substantive correction.* The claim and its quote are unchanged; what changed is their grounding. Until now the note rested on The Grand Locus's blog restatement of Fisher (Tier 2). The 2026-09-01 batch capture fetched Fisher's own paper — "Has Mendel's Work Been Rediscovered?", *Annals of Science* 1(2):115–137 (1936), mirror at genepi.qimr.edu.au, `source_sha ce3f431b1de09091db2dbaa908ef5b12e1d89eca8ef3fe936338732c9a2e1110` — and confirmed both the pooled figure against Table V directly (84 df, X²=41.6056, "Probability of exceeding deviations observed" = .99993) and the "deceived by some assistant" sentence against Fisher's own wording. The frontmatter `source_url` is intentionally kept at the Grand Locus carrier, because it is the copy `verified_verbatim` on 2026-08-07 and it matches the clean quote; the archived Fisher PDF's OCR renders "that" as "t hat" throughout, so the literal cached string is "…among others t hat Mendel was deceived…" and a naive verbatim grep of the clean quote against the primary will not match. Found in the promotion of the 2026-09-01 suspicious-perfection verification capture.