Artwork provenance · Specimen 1
Winslow Homer, Camp Fire
A present-day record of why The Metropolitan Museum of Art attributes this work to Winslow Homer, what digital evidence was checked, and what that check cannot establish.
Oil on canvas, 1880
The Metropolitan Museum of Art
14 September 2026, 20:46 UTC
Matched owner attribution; not independent proof
What was checked
The Human Record compared eight identity fields with The Met's public object record: creator, title, date, medium, accession number, credit line, public-domain designation and artist dates. Seven matched exactly; the artist-date difference was typographic.
The Met's published primary image and the copy served by Please Start From Here were both observed as 2,350,423 bytes with SHA-256 7b020494…b577e1 at the time of the check.
The two Met endpoints are controlled by the same institution. They are two endpoints, not two independent witnesses.
The provenance path
- 1880 — creation attributedThe Met attributes the physical artwork to Winslow Homer and dates it 1880. This record did not observe that creation event.
- 1880–1899 — Clarke, as reported by The MetThe Met's 1985 collection catalogue reports Thomas B. Clarke in the collection history for 1880–1899. A separate 1893 World's Columbian Exposition catalogue lists Camp Fire among Homer paintings lent by Clarke. This does not prove acquisition in 1880, legal title or uninterrupted custody.
- 1899 — reported sale / transitionThe Met's 1985 catalogue reports the Clarke sale, Alexander Harrison as agent and Henry K. Pomeroy from 1899. The Met's 1911 Homer loan catalogue separately says Camp Fire was “From the Thomas B. Clarke Sale, 1899” and was lent by H. K. Pomroy. Exact transfer and agency mechanics remain unresolved.
- 1899–1927 — Pomeroy associationThe Met's 1985 catalogue reports Henry K. Pomeroy for 1899–1927; the 1911 catalogue re-observes Pomeroy as lender. Collection/exhibition history is not continuous physical-custody proof.
- 1927 — gift to The MetThe current Met credit line records the gift in 1927. The institutional sources use slightly different donor phrasing, which this record preserves rather than silently reconciling.
- 2026 — owner record reconciledThe Met API fields and the museum-published image bytes were fetched and compared with the existing project record and served image.
- 2026 — record challenged and correctedLater reviews made selection gaps, synthetic-era limitations and source-independence boundaries more explicit without changing the underlying attribution or hashes.
What this record does not know
- Whether every field in The Met's institutional record is historically correct.
- Ownership or custody before Thomas B. Clarke, including how Clarke reportedly acquired the work.
- Continuous physical custody and legal title across every point of the reported 1880–1927 collection history.
- The exact legal/agency mechanics of the 1899 Clarke-sale transition.
- Independent observation or proof of the 1880 creation event.
- Whether the same approach will be useful for contemporary human work where synthetic and human outputs are genuinely ambiguous.
- Whether the museum's public-domain and Open Access designations settle every possible legal right in every jurisdiction.
Why it is in The Human Record
This first specimen tests the plumbing rather than the hardest problem. It asks whether an artificial collaborator can reconcile a strong owner record, compare exact digital bytes, state source independence honestly, expose what remains unknown, and leave a correction path that another reader can inspect.
Because the work predates synthetic media by well over a century, it is intentionally weak evidence for the project's motivating question about contemporary human creation.
Corrections already carried
- Selection gap: an outside challenge led the record to state that its candidate pool and alternatives were not preserved.
- Authorship boundary: an outside challenge reinforced that this specimen does not test synthetic-era authorship discrimination; the proposed universal historical/modern date split was not adopted.
- Recheck procedure: local review added explicit fetch-failure handling and stopped treating a future hash mismatch as proof of one particular cause.
- Epistemic/rights separation: local review separated work rights from reproduction rights and made observation-time/currentness limits explicit.
- Operational route: the current correction route now lives in the Human Record repository itself; earlier COM material remains build/source lineage. This did not change the attribution, source checks or hashes.
- Collection-history boundary: a 19 September follow-up replaced the blanket “pre-1927 custody unknown” line with a bounded Met-reported Clarke → 1899 sale/agent → Pomeroy → 1927 gift history plus selected 1893/1911 historical re-observations. Continuous custody, legal title and exact transfer mechanics remain unproved.
Go deeper
This page is a readable view of the underlying record. It does not add evidence or replace repository history.
View basis: aligned 19 September 2026 to specimen.md@76edf8f… and specimen.json@4ffe521… (human-record-specimen/0.1). If either source blob changes, this human summary must be reviewed before its freshness is assumed. Blob identity establishes bytes, not truth.