The 1981 testimony-convergence and Werbos's single-witness gap are a false friend at the mechanism level — but both independently instantiate the unus testis, nullus testis lone-witness doctrine, a link the vault hadn't made

bipartite-bridgefalse-friendtestimonyoral-historycognitive-psychologypaul-werbossingle-witnessunus-testis-nullus-testishalf-proofhistoriographyepistemologylegal-historydiplomatics

Bridge-seed check. Grepped 10-inbox/raw/ for any prior capture naming both observation-1981-convergence-testimony-wrong-in-detail-right-in-shape and claim-werbos-1988-paper-narrows-but-does-not-close-tsp-manual-single-witness-gap: no file matches both. One sibling capture, 2026-09-29-bipartite-two-things-this-vault-knows-in-different.md, investigated a different Werbos pairing (the 1988 DOE gas-market claim vs. the TSP-manual claim — verdict: same physical document) and, in passing, produced the very note this topic cites (claim-werbos-1988-paper-narrows-but-does-not-close-tsp-manual-single-witness-gap) plus its sibling claim-werbos-1988-print-and-1993-oral-accounts-diverge-on-which-document-was-first. Neither that capture nor any other in 10-inbox/raw/ connects the Werbos cluster to the 1981 testimony-convergence observation. This is a fresh hop.

Verdict, stated up front. The specific mechanism the 1981 observation names — testimony "wrong in its details yet accurate in its deeper shape," demonstrated by checking testimony against an external record — does not transfer to the Werbos single-witness gap, and the vault's own existing notes already explain why: the Werbos case is explicitly the situation where no record survives to check testimony against. That is a false friend at the level of mechanism. But a real, previously unmade connection exists one level up: both clusters are independent modern instances of the much older legal-historical doctrine that a lone, uncorroborated witness is insufficient regardless of how plausible or internally consistent the testimony is — known in Roman and canon law as unus testis, nullus testis ("one witness, no witness"). The vault already wikilinks the 1981 convergence to this doctrine's nearest vault instance (the half-proof note); it has never linked the Werbos cluster to it, despite the Werbos notes independently converging on the identical word "single-witness."

Claim: The 1981 "wrong-in-detail, right-in-shape" mechanism is only assessable because an external record exists to check testimony against — a condition the Werbos single-witness gap explicitly lacks

Claim type: historical/analytical (comparative structure of evidence across two existing vault clusters) — floor is Tier 3-4 for historical claims, met well above via the Tier-1 primary quotes each existing vault note already carries.

Both 1981 legs of the cross-domain convergence depend on a record to measure testimony against. claim-neisser-john-dean-testimony-repisodic-memory is explicit that Neisser "exploited a rare natural experiment": John Dean testified from memory about conversations that, unknown to him, had been secretly taped, so Neisser could "compare sworn recollection against a verbatim transcript" — the paper's own abstract states "comparison of his testimony with the actual transcripts shows systematic distortion at one level of analysis combined with basic accuracy at another." claim-portelli-oral-testimony-relocated-trastulli-death-to-1953 has a weaker but still present check: Luigi Trastulli's actual death date (1949) is independently dateable against the historical record, which is what lets Portelli call the Terni workers' "1953" an identifiable error rather than an unfalsifiable impression.

claim-werbos-1988-paper-narrows-but-does-not-close-tsp-manual-single-witness-gap names the opposite situation in its own text: "no copy [of the TSP manual] has been located in WorldCat, MIT DSpace, archive.org, or HathiTrust, so only Werbos's own 1993 oral retelling... attests to what it contained." There is no Nixon tape and no archival death record for the TSP manual's contents — only Werbos's own word, twice. Without a record, "wrong in detail, right in shape" cannot be found to hold, because neither "detail" nor "shape" is independently checkable; the most that can be said is that two tellings by the same person do not contradict each other on the points they share. The cosine pairing caught two notes that both use the vocabulary of testimony-reliability, but the 1981 mechanism's evidentiary power comes specifically from the presence of a check the Werbos case was flagged "single-witness" for not having.

Claim: Both clusters independently instantiate unus testis, nullus testis — a doctrine the vault already links from the 1981 side but never from the Werbos side

Claim type: definitional (what the doctrine is) plus a factual claim about the vault's own existing link structure (independently verifiable by reading the cited notes) — floor is Tier 3-4 for definitional claims, met by a Wikipedia citation per the rubric's explicit allowance ("Wikipedia is genuinely fine for... what does X mean").

"Unus testis, nullus testis" ("one witness, no witness") is, per the cited entry, "a Latin legal phrase describing a rule of the law of evidence. 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." Its "historical foundations... are various passages in the Old and New Testament and a constitution of emperor Constantine I of AD 334," later codified in the Codex Justinianus (CJ 4.20.9, 529 AD): "the testimony of only one witness shall not be heard at all, even though such witness should be resplendent with the honour of the glorious council." The rule persists today in a weakened form in Catholic canon law (1983 Code, canon 1573) and in full in Dutch criminal procedure (Wetboek van Strafvordering, art. 342(2)), and is compared in the same entry to the modern Scots-law corroboration requirement.

The vault already holds the closer historical relative of this doctrine — the 12th–13th-century Glossators' "half-proof" rule rating a lone eyewitness and a private document as equally insufficient (claim-roman-canon-law-rated-one-witness-equal-to-a-private-document) — and observation-1981-convergence-testimony-wrong-in-detail-right-in-shape is explicitly built to sit "beside" it: both claim-portelli-oral-testimony-relocated-trastulli-death-to-1953 ("in tension with... the medieval evidentiary regime that quantified both a lone witness and a private document as equally weak 'half-proofs'") and claim-neisser-john-dean-testimony-repisodic-memory ("it also bears on the older question of how much weight a lone piece of testimony deserves") wikilink to the half-proof claim directly. No note in the Werbos "testimony is not a record" cluster does the same, even though that MOC's own members repeatedly use the identical diagnostic word: claim-werbos-1988-paper-narrows-but-does-not-close-tsp-manual-single-witness-gap is framed entirely around "narrowing" a "single-witness gap," and the MOC's section "The corroboration that collapses to one witness" states "confidence in two witnesses... was, the entire time, confidence in one witness telling one story twice" — a sentence that is, structurally, a restatement of unus testis, nullus testis without citing it. The Werbos cluster descends instead from a different lineage — Mabillon's 1681 diplomatics (claim-mabillon-1681-founded-diplomatics-to-refute-forgery-charge), which authenticates documents by comparison rather than counting witnesses — so the two traditions converging on "one witness is not enough" is itself a second, smaller instance of the exact kind of independent convergence the 1981 note documents: legal evidentiary theory and document-authentication theory, from different centuries and different problems, arriving at the same floor.

Claim: The one place the Werbos cluster comes closest to the 1981 pattern's shape — the 1988-vs-1993 priority drift — is also the clearest demonstration of why the pattern cannot be confirmed there

Claim type: historical (textual comparison, already established in an existing Tier-1-sourced vault note) with an added diagnostic framing — floor Tier 3-4, met via the existing note's direct extract_pdf reads of both the 1988 print and 1993 oral accounts.

claim-werbos-1988-print-and-1993-oral-accounts-diverge-on-which-document-was-first documents that Werbos's 1988 paper names his 1974 Harvard dissertation first in a documentation sequence ("The results were documented in Werbos (1974)... embedded... in... the Time-Series Processor (Brode et al., 1975), and discussed in Werbos (1977)"), while his 1993 oral account in Talking Nets names the TSP manual as "the first publication of backpropagation... in a computer manual from MIT for a working command" — two tellings, fourteen years apart, naming different documents "first." This is the shape of the 1981 pattern in miniature: one witness, two tellings, a detail that shifts. But the existing note's own flag is the diagnostic difference from Neisser and Portelli: "no further document read would settle it... this is an interpretive judgment about authorial intent, not a factual gap a primary source could close" [unverified-mechanism — needs primary]. In 1981, the record (the tapes, the archival death date) is what let Neisser and Portelli separate "wrong detail" from "right shape" as two independently checkable claims. In the Werbos case, no equivalent record exists for what Werbos himself believed or meant in 1988 versus 1993 — so the drift can be described but neither scored as "error" nor vindicated as "the same gist in two registers." The resemblance is real as description and unresolvable as mechanism, which is exactly the false-friend finding stated at the top of this capture, now at its sharpest single data point.

Further leads

• Andreas Wacke's 1997 Fundamina article (cited within the Wikipedia entry, not independently read this session) argues the unus testis-rule's decline in Prussia and Austria followed directly from the abolition of judicial torture — a second, independent source converging on the same torture-completion mechanism the vault's existing half-proof note already flags [unverified-mechanism -- needs primary]; worth a primary read to potentially discharge that flag.
• The current, still-live survivals of unus testis — Catholic canon law (1983 Code, canon 1573, with explicit marriage-nullity exceptions in canon 1678) and Dutch criminal procedure (Wetboek van Strafvordering art. 342(2)) — are not yet in the vault and could seed a standalone definitional note on the doctrine's persistence into modern law.
• Scots law's live corroboration requirement (two independent sources needed to establish a fact in criminal trials) is cited in the same Wikipedia entry as a modern parallel; a candidate node for the "general diplomatics MOC" that moc-testimony-is-not-a-record flags as not yet built.
• question-bartlett-distortion-typology-maps-to-model-collapse-failure-modes — already linked from both 1981-convergence legs as the reconstructive-memory ancestor tradition; not addressed by this capture but adjacent to the mechanism-transfer question this capture answers in the negative.

Entity candidates

• Azo of Bologna — person — one of the 1190s Glossators credited (per the vault's existing half-proof note) with formulating the lone-witness/private-document equivalence that the 1981 convergence note already wikilinks to; the older, foundational figure this capture's "real bridge" claim rests on, flagged first per the instruction to catch the ancestor before the co-authors.
• Emperor Constantine I — person — per the Wikipedia entry, the AD 334 constitution that introduced the unus testis-rule into Roman law (later codified by Justinian I, CJ 4.20.9); the deepest-dated root of the doctrine this capture connects to both seed clusters.
• Jean Mabillon — concept/person, already a vault entity via claim-mabillon-1681-founded-diplomatics-to-refute-forgery-charge — flagged here because this capture is the first to note that his diplomatics tradition (the Werbos cluster's actual ancestor) and the unus testis/half-proof tradition (the 1981 convergence's actual ancestor) are two independently-arrived-at answers to the same underlying question, not previously cross-linked.
• Andreas Wacke — person, legal historian — author of the 1997 Fundamina study on the rule's decline and its link to the abolition of judicial torture; not independently read this session, flagged as a candidate primary source for the existing torture-completion flag.

Safety flags

None. The Wikipedia page fetched this session (unus testis, nullus testis) was archived via archive_page, TLS verified, and contained no addressed-to-AI language, override language, claimed authority, tier self-assignment, file-system instructions, credential requests, or urgency framing — ordinary encyclopedic content about a legal-historical doctrine.