Is the National Gallery Samson and Delilah a fake?

My talent is such that no undertaking, however vast in size, (…) has ever surpassed my courage“. So wrote about himself Pieter Paul Rubens, also known as “Pietro Paolo” since he had lived in Italy for 8 years and in Italy developed his style and his “trademark”.

Well, yes, you can talk about Rubens using industry terms, as if he was managing a factory, given the incredible organization that he imprinted on his very crowded workshop. In fact, he relied on a atelier with dozens of students that the master guided and directed like an orchestra conductor.

Rubens  was to fascinate the clients with his speeches, then probably made a drawing for them, then a monochrome sketch and then a colored one, all to get their approval. At that point, the sketches passed into the large workshop on the ground floor of his noble home in Antwerp, where the “students” brought back the inventions of the master on the chosen support, canvas or wood, enlarging them, with Rubens sometimes adding only the last retouching.

With this organization of the work, in fact, different types of artworks could be created, from the totally autograph paintings to replicas with a more or less extensive intervention by the master. You could have those painted by him only in some essential parts, and others in which he the simply revised the artwork. The workshop itself produced paintings of very different quality, also depending on who was assigned to the job, and the given subject: from those elaborated from the master’s design to the mere copy.

However, apparently, the famous painting Samson and Delilah that the National Gallery of London bought in 1980 for 2.5 million pounds does not fall into any of these categories, meaning that it is neither an original nor a copy, nor a workshop work. It would simply be a fake, perhaps even make in the 20th century. An Artificial Intelligence algorithm seems to have established it after a very long series of tests, coming to the conclusion that the probability that it is false is greater than 91%! Of course, now there is an open confrontation between scholars and this authentication method, which is based on algorithms specially trained to capture the details of the artists’ style and brushstroke, then evaluated by a sophisticated “convolutional neural network”.

Actually, this “Samson and Delilah” painting is not new to this type of controversy related to the attributions to the Flemish master. Quite the opposite. Since its appearance on the market in Paris in 1929 it has been linked to several names, and the definitive one, Rubens, was proposed by Ludwig Burchard, an expert on the Flemish master, in the 1950s. But when the scholar died in 1960, some unclear documents emerged, according to which Burchard authenticated several works for his own economic advantage.

However, there is a rather detailed report of the studies conducted between 1980 and 1983 by Joyce Plesters, published in the National Gallery Technical Bulletin, which report of an X-ray and numerous microsamples analyzed where no anachronism is found.

What is the credibility of the results obtained with artificial intelligence? Not much, apparently. The National Gallery doesn’t seem very worried by the outcomes of the algorithm, and continues to exhibit the painting as genuine.

Filippo Melli