The establishment of explicit associations between provenance statements, audit logs, or metadata across systems so that the lineage, transformations and custodial events affecting a resource can be traced, correlated and reconstructed.

Principle

Linking provenance relies on persistent identifiers, transformation metadata (who/what/when/how), and interoperable assertion formats so that discrete provenance records become navigable segments of a coherent provenance graph; trust and canonicalization rules determine when links constitute reliable causal lineage.

Demonstration

Illustrative scenario → Situation: A digital repository ingest record, a separate metadata-edit log, and a preservation workflow log exist in different systems. → Recognition: Each record references the same persistent object identifier; timestamps and operation types are present. → Action: Create cross-references between logs, normalize operation labels, and expose a combined provenance graph. → Consequence: A curator can trace when metadata was changed, by whom, and which preservation action produced derivative files, supporting reproducibility and accountability.

Misapplication

Interpreting mere temporal adjacency or matching timestamps across logs as evidence of causal transformation (error: conflating temporal correlation with provenance linkage) can assign incorrect lineage or responsibility.

Consequence

When correctly linked, provenance supports reproducibility, legal audits, rights attribution and error diagnosis; incorrect or overconfident linking can produce false attribution, misleading trust claims, and privacy exposure if links reveal sensitive operational detail.

Reversal

If systems lack stable identifiers, consistent timestamps, or interoperable assertion formats, automated linking may be impossible or unreliable; in such contexts provenance may require manual reconciliation or remain intentionally decoupled to protect privacy or reduce coupling risk.

Boundary

Clearly within: Creating persistent cross-references that join ingest logs, transformation records and custody events for a given digital object. Boundary case: Matching descriptive metadata 'source' fields across systems without operational metadata—useful for hints but insufficient for reliable lineage. Clearly outside: Simple provenance metadata recorded locally without cross-system correlation.

Semantic Tension

Transparency and traceability of lineage ↔ Privacy, minimal disclosure and operational decoupling that limit cross-system linkage.

Synthesis

Provenance linking converts isolated provenance assertions into a navigable lineage graph, but its validity depends on identifier stability, interoperable assertions and explicit trust rules; linking is as much an exercise in establishing reliable references as it is in joining records.