How to Read a Provenance Line

The notation is published, and almost nobody reads the key

A provenance line looks like a list of names and dates. It is a compressed legal history written in a notation that most institutions publish openly and that most readers never look up. The National Gallery of Art in Washington sets out its conventions in a page called Notes to the User. The Association of Art Museum Directors and the American Alliance of Museums have issued guidance since 1998 and 1999 respectively. The Carnegie Museum of Art went further and published a formal provenance standard that quotes the AAM guide in full.

The consequence of not reading the key is not confusion. It is false confidence. A reader who does not know that the punctuation is doing work will read a broken chain as a continuous one, because the names look continuous. That reading is available to anyone who spends ten minutes with the published conventions, and it is the difference between treating a line as an introduction and treating it as evidence.

The guidance is also explicit that the key should be supplied. The AAM text instructs that in publishing provenance information, a museum should include an explanation of its format. Where an entry gives no such explanation, the reader has been handed a coded document and no cipher, and the reasonable response is to distrust the coding rather than to assume it follows a standard.

The punctuation carries the transactions

The core rule is narrow and it is the same at the two institutions that state it most plainly. The National Gallery of Art writes that a semicolon between two names indicates a direct transfer of ownership of the painting, whereas a period indicates uncertainty about the chain of ownership and the whereabouts of the object between two documented owners. The AAM formulation is the same in substance: a semicolon indicates that the work passed directly between two owners, and a period separates two owners if a direct transfer did not occur or is not known to have occurred.

That last clause deserves attention. The full stop does not assert that the object changed hands irregularly. It concedes that the cataloguer cannot show it changed hands at all. A line with three full stops in it contains three admissions, and the Carnegie standard states the working assumption baldly: if uncertain, assume that the transfer is not direct.

The second rule concerns agents. The National Gallery gives the names of dealers, agents and auction house sales in parentheses, and the AAM guide gives the same rule, so that a commercial intermediary is visibly not a private owner. This matters when the question is who held title in a given year, because a picture sitting in a dealer's stock is in a different legal position from a picture hanging in a house.

The conventions are not uniform, and two major museums invert each other

Anyone who learns the notation from one museum and applies it to another will misread entries. Under the AAM guide, life dates of owners are enclosed in brackets and dealers, auction houses or agents are enclosed in parentheses. The Los Angeles County Museum of Art publishes the exact inverse: life dates of owners, when known, are enclosed in parentheses, and dealer, auction house or agent transactions are enclosed in brackets.

The Philadelphia Museum of Art uses the semicolon and full stop rule and puts life dates in parentheses, and it adds a term of its own that has no analogue elsewhere. It records that the word 'with', as in 'with Ambroise Vollard', is a blanket term used where the precise relationship between the object and the party is not established. That is a candid device, and it is invisible to a reader who has not seen the museum's own explanation of it.

The practical instruction is therefore not to memorise a single convention. It is to find the publishing institution's own key before reading its lines, and to treat any entry that supplies no key as unannotated. Auction house entries frequently supply none. They are researched documents written by the seller's agent under a catalogue deadline, and they compress the same information without committing to the museum notation at all.

The divergence has a second consequence that matters for anyone assembling a record from several institutions. A provenance stitched together from a Washington entry, a Los Angeles entry and a Philadelphia entry is written in three dialects, and the marks that survive the stitching will mean different things in different clauses of the same line. The only safe procedure is to normalise to one convention, state which one has been adopted, and footnote every place where a source institution used the opposite sense.

Only two of the hedges have a published definition

Cataloguers hedge on a scale, and the scale is much less formalised than its use suggests. The AAM guide defines exactly two terms: uncertain information is indicated by the terms 'possibly' or 'probably' and explained in footnotes. That is the whole of the published vocabulary. The National Gallery of Art's own ladder of qualifications, which runs attributed to, studio or workshop, follower of, circle of and style of, governs authorship rather than ownership.

'Said to be' has no published definition in any of the standards. It is nonetheless in constant use, and its function can be read from its position: it records that a claim was made and that the cataloguer could not confirm it. Because it is undefined, it is also unfalsifiable, which is why it appears most often at exactly the links a house would prefer not to defend.

The Carnegie standard shows how far the formalisation can be pushed when an institution decides to be exact. An uncertain party name takes a trailing question mark, so the entry reads 'George Strait?' rather than 'Possibly George Strait', and uncertain life dates are written as [1880?-1990]. That is a machine-readable admission of doubt. Most catalogue prose is not machine-readable, and the doubt survives only in adverbs that a hurried reader can skim past.

What an honest uncertain line looks like

The Mauritshuis entry for Vermeer's Girl with a Pearl Earring, inventory 670, is the model. Four consecutive early owners are each prefixed with a bracketed question mark: Pieter Claesz van Ruijven before 1674, his widow Maria de Knuijt from 1674 to 1681, their daughter Magdalena van Ruijven with Jacob Dissius from 1681, and Jacob Dissius to 1695. The museum is not asserting that chain. It is reporting it as a hypothesis.

The entry then does something rarer. For the Dissius sale in Amsterdam on 16 May 1696 it gives not one lot but three: no. 38 at 36 guilders, no. 39 at 17 guilders, or no. 40 at 17 guilders. The painting is routinely said to have appeared in that sale. The museum's own record declines to say which lot it was, and the three prices differ by a factor of two, which means the identification carries no reliable evidence about what the picture was then worth.

The documented history begins much later and much smaller. The picture was bought at the Braams sale in The Hague in 1881, day and month unknown, for 2.30 guilders, by Arnoldus Andries des Tombe. Essential Vermeer records the figure with its components separated: two guilders with a thirty cent premium, from a sale at the Venduhuis der Notarissen in the Nobelstraat. Des Tombe died on 16 December 1902 and the bequest reached the Mauritshuis in 1903, a date the Frick Collection gives independently. Even at two guilders, hammer and premium are two different numbers.

Read the gaps, then ask who cited what

The most informative part of a line is usually the part that is absent. Most gaps are ordinary record loss and two centuries of silence on a Dutch cabinet picture is unremarkable. One span is different. An object of European origin with no documented owner between 1933 and 1945 requires an explanation, and the AAM standards adopted in 1999 and amended in 2001 direct museums to document their research into the Nazi-era provenance of acquisitions, of loans and of objects already in their collections.

The second informative gap is the one immediately before a work appears on the market. An object whose first documented owner is recorded in the year it was consigned has, in effect, no provenance, however respectable the consignor. Neither gap is an accusation. Both are statements about what can be inferred.

The final test is citation. The AAMD task force report of June 1998 concedes that provenance research may take years, may be inconclusive and may require additional funding, which is an institution saying in advance that honest failure is a permitted outcome. The National Gallery of Art's own practice follows through: its notes provide sources, details of research and discussion of outstanding questions. An unbroken chain with no notes has been written to look finished. A chain with a hole in it and a footnote at every link has been researched, and the second is worth more.