A provenance evidence matrix is not a decorative appendix to a research file; rather, it is the structure through which a provenance argument can be read, tested and, where necessary, challenged without losing sight of the sources on which it depends. This distinction matters because provenance is frequently presented in narrative form, whereas its documentary basis is often discontinuous, uneven and evidentially heterogeneous. A work may be said to have passed from one owner to another, from one residence to another, or from one jurisdiction to another, until the sequence appears to constitute a coherent history. Yet the evidential basis for such a reconstruction may consist of very different kinds of material: an inventory entry, a notarial act, a sale reference, a frame label, an export file, an old photograph, a family tradition, a catalogue note, or a negative search in a repository where relevant documentation might reasonably have been expected. Each of these elements has a different evidential status. If they are absorbed too quickly into a continuous narrative, the resulting provenance may appear more stable than the underlying documentation actually permits. Archivitaly has previously discussed the movement from provenance claims to documentary evidence in relation to Italian archival sources, particularly in Provenance Research in Italy: From Claim to Evidence. The next methodological question is different. Once evidence has been identified, how should it be organised so that the resulting file remains usable for scholars, collectors, curators, advisers and legal reviewers? The issue is not simply one of presentation. It concerns the intellectual architecture of the file itself.
From narrative to evidential structure
A provenance narrative can be persuasive, but it is not always transparent. Its principal weakness lies in compression. A statement such as “the painting was probably in a Roman family collection by the late eighteenth century” may rest on several distinct evidential elements: a family inventory, a partial object description, a palace reference, an old attribution, a later sale catalogue and the absence of a competing record. Each element may be relevant, but not all elements support the same claim, nor do they do so with the same force. A provenance evidence matrix prevents this flattening. It separates the source from the claim, the claim from the inference, and the inference from the degree of confidence attached to it. It also makes visible what has not been found, which is especially important in Italian research, where a documentary trail may pass through family archives, notarial records, administrative files, Soprintendenza documentation, export records, libraries, photographic collections and private papers. A polished conclusion is useful only if the route by which it has been reached remains traceable. The purpose of such a matrix is not to replace interpretation with a mechanical table. Interpretation remains central to provenance research, particularly when sources are fragmentary, formulaic or materially difficult to identify. A structured file does, however, make interpretation accountable. It allows the reader to distinguish between direct evidence, circumstantial support, contextual relevance, plausible inference and unresolved gaps.
What the matrix should contain
A useful matrix begins with the source itself. This requires a full reference, not a simplified label. An archival citation should identify the institution, fondo, series, unit, folio or page where relevant, and any internal numbering that allows the item to be located again. A bibliographic, photographic or database reference should be treated with comparable precision. In provenance work, vague references create a particular kind of false confidence, because they make a statement appear documented while preventing its verification. The second element is the factual content of the source. This should be recorded as close as possible to what the document actually states. If an inventory lists “a portrait of a gentleman in a gilt frame”, the matrix should not silently translate that entry into “the present portrait”. If an export file names an applicant, the matrix should not automatically identify that person as the owner unless the document itself supports that conclusion. If a label gives a collection number, the number should be recorded before deciding which collection system, if any, it belongs to. The third element is the claim supported by the source. This is one of the points at which provenance files most often become unstable. A source may support object identification, ownership, possession, location, movement, attribution, legal export, exhibition history or only contextual plausibility. These categories may overlap, but they should not be treated as interchangeable. An inventory may place a similar object in a house without proving that it is the same object. A photograph may prove the presence of an object in a room at a certain moment without proving title. An export licence may document lawful movement without reconstructing the earlier ownership history. The fourth element is evidential weight. This does not require artificial quantification. In most cases, it is more useful to distinguish between direct evidence, strong circumstantial support, weak circumstantial support, contextual relevance, negative evidence and unresolved gap. Such distinctions do not weaken a provenance file. On the contrary, they strengthen it, because they prevent overstatement and show where the argument is secure, where it remains provisional, and where further research would be required.
Negative searches and evidential absence
A serious provenance file should also record negative searches, provided that they are described with sufficient precision. If a likely inventory was checked and the object was not identified, that result belongs in the matrix. If a cited shelfmark now leads to a different document, that too belongs in the matrix. If a family archive preserves papers for the relevant period but not the expected collection records, the absence is not meaningless; it is part of the evidential field within which the object’s history must be assessed. Negative evidence must, however, be phrased carefully. It is rarely adequate to write that “nothing was found”. The file should indicate what was searched, under which names, descriptions, variants, dates or institutional constraints, and with what documentary limits. A negative check in one inventory does not prove absence from a collection. It proves only that the object was not identified in that specific source under the criteria used. The distinction may seem narrow, but in provenance research it is often decisive. Recording negative searches also protects the research process itself. It prevents duplication, shows that plausible routes were tested, and clarifies why a conclusion remains open. In due diligence contexts, it may be as important to know which expected records are missing as to know which records survive.
Confidence, probability and restraint
One of the most important functions of a provenance evidence matrix is the separation of confidence from certainty. Provenance research often proceeds through degrees of support rather than absolute demonstration. A documented owner, a matched inventory number and an object photograph may together create a high-confidence identification. A generic subject description and a family tradition may create only a possible association. Both may be relevant to the history of the object, but they cannot responsibly be presented in the same language. This is particularly important for older Italian collections, where names, houses and family branches may be unstable across time. A reference to a noble family may conceal different lines of descent, different residences, different legal arrangements and different archival deposits. A place name may refer to a palace, a villa, a collection, a later market convention or a memory preserved in the art trade. A structured matrix forces these distinctions to remain visible. The same caution applies to modern administrative documentation. Export and circulation files can be powerful evidence, especially when they include photographs, measurements, ownership declarations or official decisions. Yet they must still be read according to what they actually prove. They may document movement, legal exposure, object identification or administrative review. They do not necessarily establish the full earlier ownership history of the object.
Why structure matters in art-market files
In an art-market context, provenance is often read under practical constraints. A collector, adviser, auction house, gallery or legal reviewer may need to know whether a claim can be used, whether a gap is serious, whether a document supports an attribution, or whether an export issue requires further attention. A matrix is useful because it does not ask the reader to trust the conclusion as a finished narrative. It shows the structure of the conclusion. This is also where structured provenance work differs from a merely polished dossier. A visually refined file may nevertheless remain methodologically weak if it absorbs uncertainty into a continuous narrative without identifying the evidential status of each statement. Conversely, a plain but carefully constructed matrix may be considerably stronger, because it preserves the relationship between source, claim, inference and limitation. For high-value objects, collection-history research and due diligence, the central question is not whether the provenance can be made to sound coherent. It is whether each statement can be traced back to a source, and whether that source has been asked to bear only the evidential weight it can reasonably carry. Resources such as the Getty Provenance Index are valuable in this respect because they show provenance as a field of ownership and market records that can be searched, compared and analysed, rather than merely repeated as inherited narrative. Yet database results, like archival discoveries, still require assessment, correlation and context. They are starting points for structured interpretation, not substitutes for it.
A file that remains usable
The best provenance evidence matrix is not a final ornament to be added after the research has been completed. It is a working instrument. It should remain useful when the project is reopened months later, when a curator asks for clarification, when a legal adviser reviews risk, when a catalogue entry must be revised, or when a newly identified source changes the interpretation. For this reason, the matrix should not only record positive findings. It should preserve the full audit trail: source references, photographs, transcriptions where necessary, translated summaries, claims supported, degree of confidence, negative searches, unresolved questions and recommended next steps. Such a file does not pretend that every provenance can be made complete. It makes clear what is known, what is probable, what is possible and what remains undocumented. That clarity is not a minor editorial preference. It is the difference between a provenance narrative that appears complete and a research file that can actually be used.
