Wall Labels and the Politics of Attribution

The label is the shortest consequential document in the building

A wall label carries perhaps thirty words, and one of them decides everything. Rembrandt and Workshop of Rembrandt describe the same object, hung in the same room, in the same condition, and they are separated by an order of magnitude in market value and by a complete difference in scholarly status. The label is the institution's public position on a question that specialists may still be arguing, and unlike a catalogue entry it carries no footnotes, no dates and no record of who decided.

That compression is what makes attribution changes politically awkward for museums. A demotion is an admission that the institution, its predecessors, its donors and often its published catalogues were wrong. It reduces the value of a gift and can complicate the museum's relationship with the family that gave it. A promotion is an unambiguous good-news story. The asymmetry in how the two are communicated is visible in the records themselves, and it is not a matter of conspiracy so much as of institutional incentive.

The Rembrandt corpus is the standard case because it is the largest and best documented reattribution event in the history of the discipline, and because the same body of scholarship first cut the oeuvre and then partly restored it.

The corpus that halved and then grew back

Abraham Bredius's catalogue of 1935 numbered the autograph paintings from 1 to 630. Horst Gerson's revision of 1968 numbered them from 1 to 420. That reduction of a third, achieved by one scholar in a single revision, established the direction of travel and the method: the working assumption shifted from accepting a traditional attribution to requiring a positive case for it.

The Rembrandt Research Project was founded in 1968 by Josua Bruyn, Bob Haak, Simon Levie, Pieter van Thiel and Ernst van de Wetering, and set out to separate Rembrandt's own hand from the very large body of Rembrandtesque painting produced by his pupils and followers. The first three volumes of A Corpus of Rembrandt Paintings appeared in 1982, 1986 and 1989, arranged chronologically and covering 1625 to 1642. Entries were sorted into categories, with rejected attributions and works by followers grouped together and a middle category for pictures the group could not resolve. That middle category attracted the sharpest criticism, because it left paintings publicly unattributed on the authority of a committee whose internal reasoning was not published.

The project fractured. In 1993 Bruyn, Haak, Levie and van Thiel resigned together, unable to agree with van de Wetering, who continued alone and switched the remaining volumes from a chronological to a thematic arrangement: the self-portraits in 2005, the small-scale history paintings in 2010, and in October 2014 volume six, Rembrandt's Paintings Revisited: A Complete Survey.

Volume six accepted 340 autograph paintings, against the 280 counted by Christian Tumpel in 1986. Van de Wetering was explicit that the additional sixty were not discoveries. They were pictures that had been rejected previously, and he attributed those rejections to his predecessors' wish to purify Rembrandt's oeuvre, adding that the arguments for those purifications were sometimes poor. A project that had built its reputation on subtraction ended by reversing a substantial part of its own subtraction.

Two records, two directions, two kinds of announcement

The Man with the Golden Helmet in Berlin is the demotion everybody knows. Wilhelm von Bode acquired it for the Gemaldegalerie in 1897, having identified it as a Rembrandt, and the purchase was decisive both for the painting's fame and for Bode's standing as a connoisseur. Doubts surfaced in the mid-1980s, and in 1986, after technical and art-historical investigation, the attribution was withdrawn. The museum's own catalogue entry, inventory number 811A, 67.5 by 50.7 centimetres, now records it as circle of Rembrandt, made in Amsterdam between 1650 and 1660. The entry states that the attribution was withdrawn but does not name who withdrew it.

The Ashmolean's Head of a Bearded Man ran in the other direction. Percy Moore Turner bequeathed it to the museum in 1951 as a Rembrandt. In 1982 the Rembrandt Research Project rejected it, calling it the best of a number of copies of a lost original, and the picture went into store. The museum commissioned the dendrochronologist Peter Klein, who established that the oak panel came from a Baltic tree felled between 1618 and 1628, and from the very same tree that supplied the panels for Rembrandt's Andromeda Chained to the Rocks and Jan Lievens's Portrait of Rembrandt's Mother, both of about 1630. Allowing a minimum seasoning period, the painting dates to 1620 to 1630. That does not prove Rembrandt's hand, and the Ashmolean did not claim it did: the label now reads Workshop of Rembrandt Harmensz. van Rijn, about 1630, with discoloured varnish and later overpaint still obscuring the surface.

The communication is the tell. The Ashmolean issued a press release on 31 August 2020 and put the picture on display on 2 September. The Berlin demotion of 1986 lives in a catalogue entry. Neither institution behaved improperly, and both records are public, but a reader who relies on press coverage will systematically over-count promotions and under-count demotions.

What a demotion does to everything downstream

When a museum changes a label, three things move. The first is the literature: every catalogue, textbook and monograph that reproduced the old attribution becomes, silently, unreliable on that point, and there is no mechanism for correcting them. The second is the market: comparables built on the demoted picture lose their footing, and any work whose attribution rested on stylistic comparison with it inherits the doubt. The third is the object itself, which usually leaves the gallery. The Ashmolean picture spent nearly forty years in store on the strength of a 1982 judgement.

The technical evidence has shifted the balance in a specific way. Dendrochronology can establish that two panels came from one tree, which is a fact about a workshop's timber supply rather than about a hand, and it cannot resolve authorship. What it can do is fix a date range and a place, which removes the copyist hypothesis from the table. That is why the Ashmolean's revised label says workshop rather than Rembrandt: the science answered the question it could answer, and the museum declined to answer the one it could not.

The honest reading of the Rembrandt century is that connoisseurship produced 630, then 420, then a committee that could not agree in public, then 340 arrived at by the last member of that committee reversing his colleagues. The number has never been stable and there is no reason to expect the current one to be.

How to read a label like a specialist

Treat the label as a dated institutional position, not as a fact. Look for the museum's own online catalogue entry, which will usually carry more qualification than the gallery text and sometimes the date of the last revision. Where an entry says an attribution was withdrawn, note whether it says by whom and on what evidence; the Berlin entry's silence on that point is typical rather than exceptional.

Then check the entry against the standing literature. For a Rembrandt, that means asking which Corpus volume covers the picture, whether it was assessed before or after 1993, and whether volume six revisited it. A rejection published in 1982 and never revisited is a weaker statement in 2026 than it was in 1985, precisely because the body that made it has since disowned its own severity. A museum that has kept a picture at studio of for forty years without re-examining the panel has made a decision about resources, and that is worth knowing before treating the label as settled.