Evidence architecture and provenance · Lesson 3

Reconciling claim-level conflicts

5 min read

Lesson objectives

  • Reconcile conflicting catalogues and oral accounts using claim-level confidence; produce a discrepancy register.
Reading tools

An older card may describe an object as an agricultural implement, a later catalogue as a maritime tool, and a witness as something used at home. The conflict is not settled by a vote among three voices or by always preferring the newest record. The sources might describe different uses of the same object, two identifiers might have been mixed during transcription, or the testimony might concern a similar object. Compare claims rather than labelling entire sources simply reliable or unreliable. Ask what each text literally says, which entity it identifies, who made it, when and for what purpose, and what checkable detail could distinguish one interpretation from another. An institution should not dismiss oral testimony merely because it differs from an older classification, nor turn every recollection into a verified date.

Consider three fictional records for tool S-9. A 1970 card says “small paddle, length 62 cm, from a fishing boat.” A 2005 catalogue says “wooden scoop, length 26 cm, for domestic use.” A narrator interviewed in 2024 says “I have seen this shape used to stir grain, but I have not seen the number S-9.” Direct measurement today gives 61.5 cm, and the later catalogue's image shows S-6 rather than S-9. It is reasonable to suspect that the 2005 description was attached to the wrong record: length, image and identifier all differ. That does not automatically prove use on a fishing boat. The 1970 card establishes its classification at that time, not that someone observed it at sea. The narrator supports a possible use for a similar tool, not the use of S-9. A provisional public description is “long wooden tool, earlier described as a small paddle; historical use under review,” with each reason attached to its claim.

Build a discrepancy register with columns for claim identifier, competing wordings, entity and measurement at issue, source and locator, source-to-claim relation, alternative explanation, discriminating test, public wording, reviewer and review date. Do not give one confidence mark to the whole source. The 1970 card may be strong evidence that a description existed then, moderate evidence of length if its measurement method is unknown, and weak evidence of function if it says nothing about who observed use. The 2005 catalogue may be right that a 26 cm scoop exists in the collection but have joined it to the wrong identifier. Such precision avoids discarding a record for one faulty field. Retain every version and add a visible linking correction rather than erasing the earlier catalogue.

When reading oral testimony, separate what a person witnessed from what they heard from others or inferred from an object's shape. Where permitted, preserve the interview conditions and the interviewer's question: a leading question may explain an apparently categorical answer when it is excerpted. If a narrator asks to review use of their voice or name, route that request to participation and access work rather than calling it a metadata error. UNESCO's guidance on intangible-heritage inventories addresses community participation and consent in that inventorying setting. It should not be recast as one ownership rule for every physical collection. Separating evidence from rights is necessary to prevent an editor from attempting to solve two different problems with one change.

Before publishing a recommendation, test a counterexample. If a dated photograph showed S-9 on a boat with its mark legible, support for boat use would improve, but the mistaken 2005 link would still need repair and the photograph would not establish every custody transfer. If a new measurement instead found 26 cm, the mixed-identifier hypothesis would need review; check both measurements and how they were taken before changing the description. A sound review says what would change the decision and what remains a separate issue. The institution need neither freeze its first interpretation nor rewrite every field whenever a new account arrives.

Prioritise the discrepancy by its consequence. A conflict that names a living person or changes access to sensitive material needs review before release. A difference in the description of a handle material may be recorded for the next cataloguing cycle if it does not affect interpretation or access. State the dependent decision and a review date.

Lesson summary

An older card may describe an object as an agricultural implement, a later catalogue as a maritime tool, and a witness as something used at home