Verify the three primary sources behind the 'suspicious perfection' cluster: Fisher (1936) on Mendel, Maimonides on Sanhedrin unanimity, and Simonsohn's fabrication forensics

research-integrityfraud-detectionstatisticslegal-historyhistory-of-sciencemendelfishersimonsohntalmudsource-verificationindependence

All three anchors of observation-suspicious-perfection-independence-absence-signals-defect were previously sourced to secondary or tertiary accounts read only via WebSearch (per each claim-note's own audit_status). This pass fetched all three directly. Two of the three (Fisher, Simonsohn) are confirmed and upgraded to Tier 1 on a better venue than the vault had before. The third (Maimonides/Sanhedrin 9:1) confirms the underlying legal doctrine but surfaces a genuine mismatch: the quote currently recorded in claim-sanhedrin-unanimous-guilty-verdict-acquits-the-defendant does not match the wording of the primary translation this pass could actually fetch.

Claim: Fisher's 1936 paper, read directly, confirms both the P≈.99993 pooled figure and the exact "deceived by some assistant" sentence

Claim type: quantitative (the .99993 figure) and a direct quotation. Sourcing floor requires Tier 1-2 for quantitative claims — cleared at Tier 1 via a direct full-text read of the primary, not a secondary paraphrase.

R. A. Fisher's "Has Mendel's Work Been Rediscovered?" (Annals of Science 1(2):115-137, 1936) was fetched in full via extract_pdf (23-page scan, pdftotext, tls: verified). Table V ("Deviations Expected and Observed in all Experiments") sums chi-squared values across all of Mendel's reported segregation experiments: 84 total degrees of freedom, X² = 41.6056, with a stated "Probability of exceeding deviations observed" of .99993 — the founding figure behind the whole "too good to be true" reading, confirmed against Fisher's own table rather than a blog's restatement of it. Immediately after presenting this analysis, Fisher writes the sentence the vault already carried: "Although no explanation can be expected to be satisfactory, it remains a possibility among others that Mendel was deceived by some assistant who knew too well what was expected. This possibility is supported by independent evidence that the data of most, if not all, of the experiments have been falsified so as to agree closely with Mendel's expectations."

This upgrades claim-fisher-1936-flagged-mendels-pea-data-as-improbably-close-fit from a Tier-2 secondary read (The Grand Locus's blog account) to a Tier-1 direct read of the primary. One transcription note for anyone re-grepping the cached text: the scanned PDF's OCR consistently renders "that" as "t hat" (a broken ligature throughout the whole document, not just this sentence) — the quote above is given in its correct English form; the literal OCR string is "...among others t hat Mendel was deceived..." and is what a raw grep against the cache will find.

Sourcing floor check: cleared, Tier 1.

Claim: Simonsohn's own writing confirms "excessive similarity" is his own diagnostic term, and names a second, independent case beyond Sanna and Smeesters

Claim type: specific technical-mechanism claim (how the fabrication-detection method works). Sourcing floor requires Tier 1-2 — cleared at Tier 1 via Simonsohn's own blog, not the BITSS org's secondary summary the vault previously relied on.

The vault's existing claim-simonsohn-fabrication-flagged-via-excessive-similarity-to-random-sampling sourced its two grounding phrases to a bitss.org page (Berkeley Initiative for Transparency in the Social Sciences) describing Simonsohn's 2013 Psychological Science paper "Just Post It" — an organizational summary, not Simonsohn's own words. This pass located and fetched Simonsohn's own venue directly: Data Colada, the blog he co-hosts, post [1] ("'Just Posting It' works, leads to new retraction in Psychology," 2013-09-17, byline "by Uri Simonsohn"). In his own words: "Fabricated data often exhibit a pattern of excessive similarity (e.g., very similar means across conditions). This pattern led to uncovering Sanna and Smeesters as fabricateurs (see 'Just Post It' paper)." This confirms "excessive similarity" as Simonsohn's own coinage, not a paraphrase layered on top of it, and it names the same two cases (Lawrence Sanna, Dirk Smeesters) the vault's note already carries.

The post also supplies a third, independent instance of the same method, applied later the same year: a "coin-size estimation" study (corresponding author Wen-Bin Chiou) was flagged because per-condition value distributions across two experiments were "almost identical" — Simonsohn gives the actual reported value lists and states the bootstrap-simulation result directly: "The hypothesis that these data came from random samples is rejected, p<.000025." The paper was subsequently retracted; Chiou's own emailed response (quoted in the post) attributes the pattern to a research assistant sorting the data, which Simonsohn rebuts in the same post ("the similarity of data is based on the frequency of values across samples, not their order, so sorting does not explain that the data are incompatible with random sampling"). This is a fourth concrete fabrication case for the vault's forensics cluster, independent of Sanna/Smeesters and of the CSH superconductivity case already mapped in moc-too-clean-to-be-real.

The exact phrase "inconsistent with random sampling" (also currently recorded in the vault's note) was not found verbatim in this Tier-1 source; it remains sourced only to the Tier-3 BITSS summary and should be treated as [unverified-quote — needs direct read of the 2013 Psychological Science paper itself], which this pass could not obtain (SSRN, the publisher's SAGE page, and Simonsohn's own paper-index page all point to it, but none served a fetchable PDF this session — see Further leads).

Sourcing floor check: cleared for "excessive similarity" and the Chiou case, Tier 1. The "inconsistent with random sampling" phrase stays flagged rather than silently carried forward.

Claim: the primary Sefaria text of Maimonides' Sanhedrin 9:1 confirms the unanimous-acquittal rule, but in wording that does not match the vault's currently recorded quote

Claim type: historical/legal-doctrinal claim ("what does this ancient legal text say"). Sourcing floor allows Tier 3-4 for uncontested historical/definitional claims; this clears Tier 2 via a direct fetch of a named, published translation in a dedicated primary-text repository — but the finding itself is a discrepancy, not a clean confirmation.

Sefaria's rendered page for Mishneh Torah, "The Sanhedrin and the Penalties within Their Jurisdiction" 9:1 loads its text via JavaScript and returns no readable content to a plain page fetch; Sefaria's own text API (/api/texts/...) was fetched instead and returns the underlying data directly. The primary (English) version recorded there — Eliyahu Touger's translation, published by Moznaim Publications, marked as the top-priority English version on the page — reads: "When all the judges of a Sanhedrin begin their judgment of a case involving capital punishment and say that the defendant is liable, he is exonerated. There must be some who seek to exonerate him and argue on his behalf, but yet the majority hold him liable. Only then he is executed."

This confirms the same rule claim-sanhedrin-unanimous-guilty-verdict-acquits-the-defendant describes — a unanimous first-vote conviction voids the proceeding, and execution requires that some judges argued for acquittal before a majority still convicted — but the wording is materially different from the quote currently recorded in that note ("In a Sanhedrin where all judges saw fit to convict a defendant in a capital case, they acquit him."), which was sourced only to a WebSearch summary of the hop session, never to a direct read. Neither this pass nor the original hop identified where that exact earlier wording comes from; it does not match the Touger/Moznaim translation Sefaria serves as primary. The existing claim-note's quote should be treated as unconfirmed against any primary translation actually read, pending either a match against a different named Sefaria version or a correction to the wording above.

Sourcing floor check: the underlying doctrine clears the floor (Tier 3-4 sufficient for uncontested historical/legal claims; this reaches Tier 2, a direct fetch of a named published translation). The specific prior wording in the existing claim-note does not clear verification and should be corrected or re-sourced at promotion.

Further leads

• Ellis, Hofer, Swain & van Dijk, Hereditas (October 2019) reportedly re-audit Fisher's own statistical method and argue his pooling of chi-squared values across non-comparable experiments (trait linkage, batch effects, classification noise) invalidates the 1936 "too good" verdict on its own terms — found only via a WebSearch summary this session, not fetched or read; would directly complicate Claim 1 above. [unverified — needs direct read of the Hereditas paper].
• Ephraim Glatt, "The Unanimous Verdict According to the Talmud: Ancient Law Providing Insight into Modern Legal Theory," 3 Pace Int'l L. Rev. Online Companion 316 (2013) — author and full citation now confirmed via the landing page at digitalcommons.pace.edu/pilronline/35/, but the article's own full-text PDF returned HTTP 403 to both archive_page and WebFetch this session.
• arXiv:2407.07293, "Optimal Decision Mechanisms for Committees: Acquitting the Guilty" — a modern game-theoretic paper that appears (per search snippet only) to engage with the Talmudic unanimity rule; not opened this session.
• The R.A. Fisher Digital Archive at the University of Adelaide (digital.library.adelaide.edu.au) — Fisher's own institutional collected-papers repository, and the "original venue" this capture would have preferred over the QIMR mirror — is a JS-rendered DSpace application that returned 404/empty content to both archive_page and WebFetch on every URL form tried this session (the old /dspace/handle/ path and the newer /items/ path alike).

Entity candidates

• Gregor Mendel — person — the earlier, foundational figure the whole Fisher claim is measured against: Fisher's 1936 paper is entirely an analysis of Mendel's own reported 1866 pea-hybridization numbers, so Mendel — not just Fisher — is the ancestor this cluster's priority claim rests on.
• R. A. Fisher — person — statistician/geneticist, author of the 1936 "Has Mendel's Work Been Rediscovered?" critique.
• Uri Simonsohn — person — co-founder of Data Colada, author of "Just Post It" (2013) and the coin-size fabrication post.
• Maimonides (Rambam) — person — author of the Mishneh Torah, source of the Sanhedrin 9:1 unanimity rule.
• Ephraim Glatt — person — author of the Pace International Law Review Online Companion piece on the Talmudic unanimous-verdict rule.
• Wen-Bin Chiou — person — corresponding author of the 2013 "coin-size" study Simonsohn's post flagged and which was subsequently retracted; a fourth concrete "excessive similarity" case for the vault.
• Noel Ellis (with Julie Hofer, Martin Swain, Peter van Dijk) — persons — authors of the 2019 Hereditas re-audit reportedly disputing Fisher's own statistical method; the missing "who challenges the challenger" figure for this cluster.
• Dirk Smeesters / Lawrence Sanna — persons — the two researchers whose fabricated data Simonsohn's "excessive similarity" method originally caught; already the concrete subjects behind the vault's existing Simonsohn claim-note.