The Dealer as Maker of Taste, and the Ledgers He Left

The dealer did not sell the picture, he sold the certainty

In 1921 Henry and Arabella Huntington bought Gainsborough's Blue Boy from Joseph Duveen, who had acquired it from the Duke of Westminster. The Huntington records the price at 728,800 US dollars, the highest ever paid for a painting at that date. Before the picture left England, Duveen arranged for it to be shown at the National Gallery in London and then at his own New York showroom.

Those exhibitions are the transaction. A private sale between a British duke and a Californian railway fortune needed no public viewing at all. What the viewings produced was consensus: newspaper coverage, queues, a national argument about export, and a picture that arrived in San Marino already famous for having been bought. Duveen's clients were not buying connoisseurship they lacked. They were buying the assurance that the object would be recognised, and recognition is manufactured.

This is the mechanism by which a dealer makes taste, and it is worth stating without moralising. Between roughly 1900 and 1940 a small number of firms decided which European pictures crossed the Atlantic, in what order, and at what price, and the American museum map still reflects those decisions. Frick, Mellon, Huntington, Widener and Kress did not independently converge on the same schools. They were sold them.

The connoisseur was on the payroll, and the terms are the point

In 1907 Bernard Berenson entered into a private agreement with Joseph Duveen. The reported terms were 25 per cent of the profit on any picture Duveen acquired on his advice, in exchange for which Berenson represented Duveen's interest in the Italian market and remained available for authentication. Both men wanted the arrangement kept quiet: Berenson did not wish to be seen as an employee, and Duveen wished his clients to believe that his pictures were certified independently.

The existence of the arrangement is not seriously disputed, and it ran for about a quarter of a century. What is disputed is its consequence. The strongest claims, that Berenson knowingly issued false attributions to famous names in order to earn his percentage, come principally from Colin Simpson's Artful Partners of 1986, written with access to Duveen records, and they have been contested by scholars sympathetic to Berenson ever since. The figure often quoted for Berenson's total earnings from the partnership, around 150,000 dollars, circulates in secondary summaries and should be checked against the Duveen accounts before use.

An auction specialist does not need the moral question resolved to act on it. What is needed is a procedural rule: a Duveen-period attribution supported by Berenson, in a picture Duveen sold, is an interested opinion. It may well be right, and often is. It is not independent, and any catalogue that presents it as independent is misdescribing its own evidence.

The ledgers outlived the firms and became the evidence

The dealers' persuasive apparatus, the certificates and the photographs and the letters to clients, is exactly the part of the record a researcher should trust least. The part to trust is the bookkeeping, because it was written for the partners and the tax authorities rather than for the buyer.

The Getty Research Institute holds the Duveen Brothers records, 1876 to 1981, with the bulk falling between 1909 and 1964: stock books, sales books, invoice books, shipping receipts, customer ledgers, loose photographs, some two thousand glass negatives and hundreds of correspondence files. The Getty acquired the M. Knoedler and Co. archive, covering approximately 1848 to 1971, in 2012. It contains stock books, sales books and commission books, correspondence with collectors, artists and dealers, photographs of works sold, records from the London and Paris offices, exhibition files, framing and restoration records and the print department's papers.

The Getty has since published the contents. An initial release put nearly 24,000 records online from painting stock books one to six, covering 1872 to 1920, with books seven to eleven following. The dataset now exceeds 40,000 records transcribed from the eleven painting stock books and enhanced from the twenty-one paintings and watercolours sales books of the New York office, 1872 to 1970, with artist names validated against an authority file. The transcriptions are also published as open data.

The Wildenstein material followed a different route. In 2016 the Wildenstein family and the Hasso Plattner Stiftung created the Wildenstein Plattner Institute, which received the archives in 2017 and launched its digital archives on 23 October 2020 with more than 50,000 digital resources; the digital library now offers more than 25,000 digitised items and the institute reports over 100,000 digitised archival materials online, including dealer stock books, artists' papers, historic photographic collections and the working dossiers of suspended catalogue raisonne projects.

Three of the firms that redirected the European picture market into American collections are now, in effect, public research infrastructure.

A stock book settles what a catalogue entry cannot

The clearest demonstration is the Hermitage transaction. Between 1928 and 1931 the Soviet government sold roughly 250 paintings out of the Hermitage to raise foreign currency for the first Five Year Plan. Andrew Mellon acquired twenty-one of them in 1930 and 1931 for a total of 6,654,000 US dollars, negotiated through a consortium of M. Knoedler and Co., P. and D. Colnaghi and the Matthiesen Gallery in Berlin. Raphael's Alba Madonna alone accounted for 1,166,400 dollars in April 1931 and van Eyck's Annunciation for 502,899. Mellon gave the group to the United States government in 1937, and it became the nucleus of the National Gallery of Art in Washington.

The Getty, describing its own Knoedler holdings, points to a sales book entry of January 1931 recording the commission the firm charged on two of those pictures, Botticelli's Adoration of the Magi and Rembrandt's Joseph and Potiphar's Wife. That single ruled line does work no press account can do. It fixes a date, identifies the intermediary, and states what the intermediary was paid, in a document written contemporaneously and for internal purposes. It is also a reminder that the Soviet sales were kept quiet: the transactions did not become public knowledge until November 1933.

The same archives close the loop on the older English dispersals. Van Dyck's Philip, Lord Wharton of 1632 hangs in Washington as accession 1937.1.50 in the Andrew W. Mellon Collection, and its recorded provenance runs from Robert Walpole through the Hermitage and then through Colnaghi and Knoedler. A picture bought by a British prime minister in the 1730s reaches an American national museum through a dealer's ledger, and the ledger is the only continuous witness.

Knoedler also proves the opposite, which is why the distinction matters

M. Knoedler and Co. closed abruptly in 2011 after 165 years in business. The gallery had sold around 60 million dollars of forged paintings presented as works by Mark Rothko, Robert Motherwell and Jackson Pollock. In 2013 Glafira Rosales pleaded guilty to federal tax evasion and money laundering and admitted that forty paintings she had supplied to the gallery over some fifteen years were forgeries, painted by an artist working in Queens.

The forgeries were sustained by a story about ownership. A collector referred to as Mr X, said to have bought directly from the artists, was Rosales's invention. When buyers wanted a chain of custody, the artist Alfonso Ossorio was attached to the works as the intermediary who had guided the purchases; when the Ossorio connection was rejected, his name was replaced with that of David Herbert, a dealer who was conveniently dead. Ten former clients brought civil claims; most settled, including the De Sole case, which settled during trial in 2016.

The two Knoedlers are the same firm, and the contrast is exact. What survives as evidence is the ruled, numbered, contemporaneous stock book, written to account for money. What collapsed was narrative provenance, assembled after the fact to satisfy a purchaser, resting on names rather than entries. The forgeries failed the moment anyone asked which ledger line recorded the acquisition, because there was none.

That is the practical rule the dealer archives teach. Ask what document generated the provenance statement and who was paid at the moment it was written. A line that can be traced to a firm's own books is evidence. A line that can be traced only to what the firm told a client is advertising, whether it was produced in 1921 or in 2007.