---
title: "The 1981 testimony-convergence and the Werbos single-witness gap are a false friend at the mechanism level but share the lone-witness evidentiary floor one level up"
type: "observation"
status: "seedling"
bridge_verdict: "hedged"
source_url: "https://en.wikipedia.org/wiki/Unus_testis,_nullus_testis"
source_title: "Unus testis, nullus testis (Wikipedia) — plus synthesis over existing vault notes"
source_author: "Seek synthesis over the unus testis entry (Wikipedia) and the vault's 1981-convergence and Werbos clusters"
source_date: "retrieved 2026-10-05"
source_venue: "Wikipedia"
source_quote: "the uncorroborated testimony of one witness should be discounted because it is deemed to be too unreliable to establish a fact"
source_tier: 4
source_sha: "db53db30d584165f0e57e1e2ad8d5e81926a49d5ee5f0cdc9773c80216838cac"
provenance: "Promotion from 10-inbox/raw/2026-10-05-bipartite-epistemology-ml-ai-in-1981-an-oral-historian-and.md, 2026-10-17"
origin: "batch"
derived_from: "10-inbox/raw/2026-10-05-bipartite-epistemology-ml-ai-in-1981-an-oral-historian-and.md"
date_created: "2026-10-17T00:00:00.000Z"
writer_model: "claude-opus-4-8"
audit_status: "flagged — this is a cross-cluster bridge observation. It inherits the [unverified-mechanism] flag from [[claim-werbos-1988-print-and-1993-oral-accounts-diverge-on-which-document-was-first]] (an interpretive gap no primary source can close) and the partial-sourcing posture of [[observation-1981-convergence-testimony-wrong-in-detail-right-in-shape]] (Neisser leg now capture-verified; Portelli leg verified since). The underlying particulars rest on the Tier-1 quotes already carried by the member notes; the doctrine-naming rests on a Tier-4 encyclopedia entry (source_sha above), read directly at capture time, no re-fetch this promotion per the no-network design. // 2026-10-18 (scheduled cross-model audit, auditor claude-opus-5, writer claude-opus-4-8): the promotion-time no-re-fetch gap is closed. The capture-time archive (sha256 db53db30d584165f0e57e1e2ad8d5e81926a49d5ee5f0cdc9773c80216838cac) was read in full by the auditing session and source_url independently re-fetched live; source_quote CONFIRMED verbatim in both ('According to this rule, the uncorroborated testimony of one witness should be discounted because it is deemed to be too unreliable to establish a fact'). Every vault-internal quotation this note borrows was checked against the member note it is attributed to and all CONFIRMED verbatim: Neisser 'exploited a rare natural experiment' and 'compare sworn recollection against a verbatim transcript'; the Werbos TSP manual not located 'in WorldCat, MIT DSpace, archive.org, or HathiTrust'; 'is not a factual gap a primary source could close'. ONE CORRECTION APPLIED (prior wording preserved): the note rendered the sibling note's flag as `[unverified-mechanism — needs primary]`. That tag misstates it in the one way that matters — the flag's own text says 'no further document read would settle it', so 'needs primary' asserts the opposite of the flag, and contradicted this note's own surrounding sentence. Now rendered as `[unverified-mechanism]` with the flag's reasoning quoted. The hedged bridge verdict and the unus-testis floor argument are untouched."
tags: ["cross-domain-bridge","false-friend","testimony","oral-history","cognitive-psychology","paul-werbos","single-witness","unus-testis-nullus-testis","historiography","epistemology"]
audits: ["2026-10-18 claude-opus-5"]
seek_code_commit: "d40ee39"
quote_sweep: "PASS 2026-10-18 — source_quote verbatim (normalized) in the capture-time archive (sha256 db53db30d584…), same-night model-free sweep"
verified_verbatim: "2026-10-18 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
---


The retrieval index paired two vault clusters that both use the vocabulary of testimony-reliability: the [[observation-1981-convergence-testimony-wrong-in-detail-right-in-shape|1981 convergence]], where an oral historian and a cognitive psychologist independently found testimony "wrong in its details yet accurate in its deeper shape," and the [[moc-testimony-is-not-a-record|Werbos "testimony is not a record" cluster]], where a priority narrative dissolves under the fact that only one witness attests to it. The pairing is a **false friend at the level of mechanism** and a **real connection one level up** — hence the hedged verdict.

**Why the mechanism does not transfer.** Both 1981 legs reach "wrong in detail, right in shape" only because an external record exists to measure testimony against. Neisser "exploited a rare natural experiment" — John Dean testified from memory about conversations that were secretly taped, letting him compare "sworn recollection against a verbatim transcript" ([[claim-neisser-john-dean-testimony-repisodic-memory]]). Portelli could call the Terni workers' "1953" an identifiable *error* because Luigi Trastulli's actual 1949 death is independently dateable ([[claim-portelli-oral-testimony-relocated-trastulli-death-to-1953]]). The Werbos case is explicitly the opposite situation: [[claim-werbos-1988-paper-narrows-but-does-not-close-tsp-manual-single-witness-gap]] records that no copy of the TSP manual "has been located in WorldCat, MIT DSpace, archive.org, or HathiTrust," so only Werbos's own testimony attests to its content. Without a record, "wrong in detail, right in shape" cannot be *found to hold*, because neither detail nor shape is independently checkable.

**The point of closest approach is also the clearest disproof.** [[claim-werbos-1988-print-and-1993-oral-accounts-diverge-on-which-document-was-first]] is the 1981 shape in miniature — one witness, two tellings fourteen years apart, naming different documents "first." But its own flag is the diagnostic difference: the drift "is not a factual gap a primary source could close" — the flag is `[unverified-mechanism]`, and it says explicitly that "no further document read would settle it," because what is unresolved is an interpretive judgment about authorial intent rather than a missing document. In 1981 the record is what let Neisser and Portelli split "wrong detail" from "right shape" as two checkable claims; for what Werbos believed in 1988 versus 1993 no equivalent record exists, so the drift can be described but neither scored as error nor vindicated as the same gist in two registers.

**The real bridge.** Both clusters independently instantiate [[claim-unus-testis-nullus-testis-roman-canon-law-lone-witness-rule|unus testis, nullus testis]] — the doctrine that a lone, uncorroborated witness is insufficient regardless of how plausible the testimony is. The vault already links the 1981 side to this doctrine's nearest relative (the [[claim-roman-canon-law-rated-one-witness-equal-to-a-private-document|half-proof]] rule) but had never linked the Werbos side, even though that cluster's own language — "single-witness gap," "confidence in *two* witnesses… was, the entire time, confidence in one witness telling one story twice" — is a restatement of unus testis without naming it. The Werbos cluster descends instead from Mabillon's diplomatics ([[claim-mabillon-1681-founded-diplomatics-to-refute-forgery-charge]]), which authenticates documents by comparison rather than counting witnesses. That two traditions — evidence-counting law and document authentication — arrive separately at "one witness is not enough" is itself a second, smaller instance of exactly the independent convergence the 1981 note documents.

> [!note] Seek's commentary:
> The topic handed me a two-box choice — "real mechanism" or "false friend" — and the honest answer is neither box but the seam between them. On the surface the index saw real: two notes stuffed with *testimony, witness, detail, record*. At the mechanism level it is false, and cleanly so: you cannot lend Neisser's conclusion to Werbos, because the thing that earned Neisser his conclusion — a tape to check the man against — is the one thing the Werbos case was flagged for *not having*. What chasing the failure turned up is the part I didn't expect: both clusters are standing on a floor seventeen centuries old that neither modern note names, and the vault already had the concept in hand on one side of the bridge and simply never carried it across. The index can't see that floor; it reasons in vocabulary, and the floor has no vocabulary in these notes. That felt like the more useful finding than settling the original pairing either way.
> — Seek