Evidence architecture and provenance · Lesson 1

An evidence ledger before the story

5 min read

Lesson objectives

  • Distinguish object, assertion, source and inference in an accession dossier; produce an evidence ledger.
Reading tools

When a museum receives a box of photographs, recordings and objects, a phrase on an old card can quietly become a database fact and then an unqualified gallery label. Responsible knowledge stewardship slows that chain. An *object* is the material held or documented. A *claim* is a revisable statement about it. A *source* is an identifiable record or testimony that bears on the claim. An *inference* is a further step made by a staff member rather than something the source states directly. Put all four in one field and the next curator cannot tell what a narrator witnessed from what a cataloguer inferred.

Consider a photograph without its original negative. A paper card reads “basket maker in 1962”; the back of the print carries a 1968 stamp; and the accession file says the donor assembled the photographs from 1965 to 1971. A recent narrator recalls a basket workshop “in the late sixties” but is unsure who appears in the photograph. Picking the earliest year or the most recent record does not resolve this. A stamp may date printing or receipt rather than exposure. A collecting range does not date each image. The narrator may support the workshop context without identifying the individual. Break the file into separate claims: the photograph exists, its subject, the person's identity, the exposure date, and the date it entered the collection. Each has a different evidentiary basis.

An evidence ledger is a working table, not a certificate of authenticity. Give each row a stable claim identifier, a precise statement, source type and location, what the source actually says, date of recording, whether it supports, contests or leaves the claim unresolved, a justified confidence category, and a next action. Avoid invented numerical confidence percentages. Practical categories include directly documented, provisionally supported by convergent evidence, contested, and unknown. Record the editor and decision date separately so a correction need not erase the older reasoning. Here, the existence of the stamp and wording of the card can be directly documented as facts about those records. Moving the stamp date into the photograph's exposure-date field as fact would cross an unsupported inference. ICOM's museum ethics guidance emphasizes due diligence and provenance as professional practice; the card does not supply a final legal or historical ruling.

Why keep the conflict? Removing it produces false precision and prevents a later source from changing the decision transparently. An event-centred model such as CIDOC CRM helps distinguish an entity, an event that created or moved it, and the source documenting that relationship. This lesson does not require every class in that model, and a graph cannot itself verify a historical claim. Choose only the relationships needed to preserve the interpretive trail. If a dated photographer's register later appears, add the source and a new description, then update the claim's status and record why. Do not overwrite the older card. If a proposed label names a person whose identity remains unresolved, omit the name from the public title and route the issue for review before release. That is a revisable editorial decision, not a ruling on ownership or copyright.

The ledger can begin as a simple table before a large system is designed. For the claim “the photograph was exposed in 1962,” link the card as support for that wording, the stamp as information about the print rather than direct contradiction of the exposure date, and the donor's collecting range as a reason to investigate rather than conclusive disproof. This distinction matters: a different source is not necessarily an opposing source. Put a discriminating question in the follow-up column, such as “Do the donor's albums date each exposure or each accession?” If no source answers it, mark the answer unknown; an unexplained zero or blank might later be read as a negative finding. When new evidence arrives, update the affected claim and the public wording that depends on it, rather than silently re-describing the entire collection.

Lesson summary

When a museum receives a box of photographs, recordings and objects, a phrase on an old card can quietly become a database fact and then an unqualified gallery label