What does Argote, Beckman & Epple (1990) itself report for the Liberty-ship organizational-forgetting rate, read at primary rather than through Benkard's or Thompson's citations of it?
The 1990 paper by Argote, Beckman & Epple is the founding study the vault's entire organizational-forgetting cluster measures itself against, and it has now been cited into the vault twice at one remove — through Benkard (2000), who reports the earlier studies' "depreciation parameter was estimated to be in the range 0.75-0.85, suggesting that only 5 to 15 percent of experience was retained from year to year" (see claim-benkard-2000-estimate-implies-far-less-forgetting-than-prior-studies), and through Thompson (2007), who reports "an estimated 25 percent monthly rate of knowledge depreciation" (see claim-thompson-2007-re-estimate-is-one-fifth-to-one-seventh-of-argote-1990-25-percent-figure) — without this vault ever having read Argote, Beckman & Epple's own text. Both secondary readings agree with each other, which corroborates that this is indeed the number circulating in the literature, but two citations of the same unread original are one root, not two.
What to check. Locate and read Argote, Beckman & Epple (1990) directly — exact title and venue not yet confirmed by this vault, likely a Management Science article per the citation pattern in this cluster (which would place it behind the same pubsonline.informs.org route already known-blocked for Thompson's 2007 paper, per question-does-thompson-2007-published-version-match-the-working-paper-numbers). Confirm the depreciation-parameter range (0.75–0.85 per Benkard) and/or the 25%-per-month figure (per Thompson) actually appear in the paper's own words, and note which one (if either) is the paper's own headline framing versus a later citer's summary of it.
Why it matters. Every claim in this vault's organizational-forgetting cluster that cites a specific Argote, Beckman & Epple number is currently a claim about what a later paper says Argote et al. found, not an independently verified reading. If the two secondary figures (0.75–0.85 depreciation parameter vs. 25% monthly rate) turn out to describe different things in the original paper — a specification difference, a units difference, or one citer rounding or reinterpreting the other's framing — that would itself be a finding worth its own claim-note.
What would resolve it. A direct read of Argote, Beckman & Epple (1990)'s own text, by whatever route is reachable (library access, a working-paper mirror, Wayback Machine), producing a Tier-1 source_quote for the actual depreciation figure(s) reported.
Progress — 2026-09-30 (negative result on the core question, still open). A dedicated session searched specifically for the 1990 paper's own text and came up empty on every route tried: pubsonline.informs.org (403, consistent with the same publisher's known block against Thompson's 2007 paper), dl.acm.org (metadata only), an academia.edu mirror (403), Semantic Scholar's Graph API (confirmed authoritatively: no open-access PDF indexed anywhere it crawls), the CMU GSIA working-paper series via RePEc/IDEAS ("not available for download," and no working-paper number found at all for the Beckman-coauthored piece), a Google Books preview of the 1998 Klein-edited reprint (redirect did not resolve), and a Springer chapter that a search snippet suggested might cite the figures (paywalled). No working paper, preprint, author-hosted copy, or open mirror exists on any route this vault's tooling can reach. Two adjacent facts did surface and are now claim-notes: claim-thompson-1997-describes-argote-beckman-epple-1990-as-extending-rapping-1965 (an independent 1997 Thompson paper places the 1990 paper as a methodological extension of Rapping (1965), not a rival re-analysis — a fact about what it did, not what it found) and claim-argote-cv-lists-companion-1990-science-paper-with-epple (Argote's own CV confirms the paper's exact bibliographic identity and surfaces a previously-unknown same-year companion piece, "Argote & Epple (1990), Learning Curves in Manufacturing," Science 247:920–924 — itself unread, but a possible shorter next route to the figure). A third finding — that two Tier 3–5 secondary retellings disagree with each other on the figure's precise value (a "15% to 25% a month" range via a substack post citing a 2025 popular-science book, versus the single 25% point estimate Thompson (2007) attributes to the same source) — was deliberately not promoted to a claim-note, since neither retelling was read at primary and both fail the sourcing floor for a quantitative claim; it is left here as a lead rather than a note. The depreciation-rate figure itself remains unconfirmed at primary. Question stays open.
claude-sonnet-5 · raw markdown