Two 1981 papers, two unrelated fields: an oral historian and a cognitive psychologist both concluded that testimony's factual errors don't erase its deeper accuracy

Core claims

1. Alessandro Portelli found that oral testimonies about a 1949 killing had drifted to a different year — and argued the drift was itself the finding. Luigi Trastulli, a steelworker in Terni, Italy, was killed during a 1949 protest against the Atlantic Pact. Portelli documented that many workers' oral accounts relocated his death to 1953, the year of mass layoffs at the Terni steelworks — a year that "fit" the death's emotional logic better than the year it actually happened. Per Wikipedia's summary of Portelli's own words: "I was interested in narratives that were not factually true because it's one of the ways through which you can get at the meaning and the subjectivity as well as the facts of what actually happened." [unverified — needs primary text; the essay PDF returned 403 on two fetch attempts, so the exact 1949/1953 figures rest on aggregated secondary retellings, not a directly-read primary]

2. Ulric Neisser compared John Dean's sworn Watergate testimony against the actual Nixon tape transcripts and found accuracy at one level coexisting with distortion at another. Dean testified from memory about conversations that, unknown to him, had been recorded. Neisser's analysis found many specific conversational details wrong — Dean tended to inflate his own centrality — while the overall claim (that a cover-up existed and who was involved) held up. Neisser coined "repisodic" for this pattern: a memory that isn't a faithful single episode but a compressed, functionally-accurate blend of a repeated series of events. Source: Neisser, U., "John Dean's memory: a case study," Cognition 9(1):1-22 (1981) — citation confirmed via PubMed; specific findings above are secondary paraphrase since the full text was paywalled.

Why this was hop-worthy

Portelli (oral history, History Workshop Journal) and Neisser (cognitive psychology, Cognition) published in the same year, from opposite disciplinary directions, converging on the same structural claim: testimony checked against a document will often fail at the literal level and still be doing real epistemic work at another level — which is exactly the tension the night's seed question asked about, answered twice, independently, in 1981.

Further leads

• Neisser's broader complaint that laboratory memory research had ignored "important or socially significant" real-world memory (his 1978 provocation) — directly motivates why he went looking for a real-world test case like Dean's testimony. Not chased as its own claim tonight.
• Neisser & Harsch's 1992 Challenger-disaster flashbulb-memory follow-up (confidence reportedly uncorrelated with accuracy 2.5 years out) — genuinely on-theme, but the specific figures I found only clear Tier 3 science journalism, not a primary source, so left as a lead rather than a claim.
• Maurice Halbwachs, the sociologist who coined "collective memory" in the 1920s and died in Buchenwald in 1945 — a strong person-behind-the-thing hook (max_cosine 0.603, novel) saved for a future chain.

Hop chain

Hop 1: Oral-history methodology overview (WebSearch synthesis of Wikipedia/Springer/Smithsonian material on witness testimony vs documents) — no single URL, general search

• Hook type: the surprising claim
• Hook: Portelli's reframe that oral history's factual unreliability is "not a weakness but a strength"
• Why followed: directly inverts the seed question's premise (testimony vs documents as a reliability contest); vault_novelty scored 0.647, adjacent and not a duplicate of the vault's existing Ann Hardy oral-history question
• Key findings: Portelli's The Death of Luigi Trastulli documents Terni steelworkers relocating a 1949 killing to 1953 in memory, and argues the displacement reveals collective meaning rather than undermining the testimony.

Hop 2: Alessandro Portelli, Wikipedia summary — https://en.wikipedia.org/wiki/Alessandro_Portelli

• Hook type: cross-domain bridge (top priority per spec)
• Hook: the same structural question — testimony checked against a literal record — was being asked the same year (1981) inside cognitive psychology, via Ulric Neisser's analysis of John Dean's Watergate testimony against the actual Nixon tapes
• Why followed: cross-domain bridges are always-follow; oral history and cognitive psychology arriving at the same finding independently, same year, is a strong bridge; vault_novelty scored 0.660, adjacent
• Key findings: Neisser found Dean's testimony wrong on many conversational specifics but right on the deeper claim (the cover-up), coining "repisodic" memory for the pattern.

Hop 3: Ulric Neisser, "John Dean's Memory: A Case Study," Cognition 9(1):1-22 (1981) — https://pubmed.ncbi.nlm.nih.gov/7196816/

• Hook type: the person behind the thing / mechanism question
• Hook: why did Neisser go looking for a real courtroom-testimony case at all, instead of running a lab word-list experiment?
• Why followed: zooms into the theoretical motivation underneath hop 2's finding; vault_novelty scored 0.692, adjacent, closest neighbor was the vault's existing Bjork "desirable difficulties" memory-research note
• Key findings: Neisser was an outspoken critic of laboratory memory research's ecological validity, arguing psychologists had "hardly ever studied" the aspects of memory that actually matter in real life — the John Dean study was a direct response to that complaint.

Hop 4: Neisser's ecological-validity critique (secondary synthesis, no single primary URL retrieved)

• Hook type: mechanism question
• Hook: did Neisser ever test his "confident memories can still be wrong" claim on something bigger and more public than one man's Senate testimony?
• Why followed: closes the loop with a concrete, famous empirical test of the theory; vault_novelty scored 0.655, adjacent
• Key findings: Neisser & Harsch's 1992 study surveyed students about the Challenger disaster the morning after, then again 2.5 years later; per secondary science-journalism accounts, confidence in the later memories didn't track their accuracy — flagged [unverified-quant -- needs primary] since only Tier 3 sourcing was found tonight.

Saved hooks not followed:

• Flashbulb memory (Neisser & Harsch 1992, Challenger disaster) — from hop 4 — reason saved: promising but the quantitative figures I found only clear Tier 3 science journalism; needs the primary chapter (Winograd & Neisser, eds., Affect and Accuracy in Recall, 1992) before it can be a claim.
• Maurice Halbwachs / "collective memory" (sociology, 1920s, died in Buchenwald 1945) — from hop 1's broader reading — reason saved: strong person-behind-the-thing and WWII-resistance-adjacent hook (max_cosine 0.603, novel), but didn't fit this chain's alternation slot.
• Elizabeth Loftus / eyewitness memory and false-memory research — from hop 3's psychology-of-memory context — reason saved: max_cosine 0.692, but its nearest vault neighbors were squarely in the agent-memory/Anthropic-Dreams cluster this session was told to avoid engaging directly.
• R.G. Collingwood's "re-enactment" theory of historical knowledge — from hop 1's methodology search — reason saved: max_cosine 0.656, but again pulled toward the vault's schema-change/restructuring MOC, which sits adjacent to the AI-agent-epistemics cluster this session redirected away from.
• Marc Bloch / the Annales School's The Historian's Craft — from hop 1 — reason saved: rich person-behind-the-thing hook (Bloch was executed by the Gestapo in 1944 for Resistance work, book published posthumously and unfinished), but the chain had already committed to the psychology bridge by the time this surfaced.

post-worthy: maybe — the cross-domain convergence (1981, two fields, same structural finding) is a genuinely good bridge, but both core claims currently rest on secondary paraphrase rather than directly-read primary text (Portelli's essay PDF was blocked twice; Neisser's paper is paywalled), so it needs a primary-source pass before it's publication-ready.