Historical story

A dive under the work of the Dutch masters through X-ray scans of the canvas in which threads are counted

Research on paintings often takes place on the surface, but the deeper layers of paint and the canvas contain a wealth of information. X-ray scans of Vermeer's paintings now show which works come from the same roll of canvas. A possible gold mine for art historians.

In the Mauritshuis I am surrounded by portraits, still lifes and views from a distant past. Canvases from the seventeenth century, the Golden Age, also with regard to Dutch painting. One of the best-known works is that of the Girl with a Pearl Earring, which Johannes Vermeer painted around 1665. But something strange is going on:where the back of the girl's head should be, I look straight through the painting on the naked canvas.

I look at a composite of a photograph and an X-ray scan of the painting. In front of the large poster is Rick Johnson, professor of Computational Arts and Humanities from Cornell University in the United States. Over the past ten years, he has used X-ray scans to research the canvases under the work of great painters, from Vermeer to Van Gogh. A technique that provides surprising clues about the origin of the paintings and the painting practice of the masters themselves.

Johnson stands out. Good build, big glasses and gray beard. As an engineer, he successfully infiltrated the museum world over the past ten years, looking for X-ray images of the canvases. He looked for high-resolution scans so that the threads of the canvas – usually less than a millimeter wide – could be seen. He wanted to count those wires – not himself, but with a computer program developed by him and colleagues. He suspected that the pattern in the wire structure contains important information. For example, it should be possible to determine whether works come from the same roll of canvas.

Featured by the editors

MedicineWhat are the microplastics doing in my sunscreen?!

AstronomySun, sea and science

BiologyExpedition to melting land

Johnson was an outsider to the museum world. Experts were skeptical about the project. Unjustly, as it turns out now that he has published the results of Vermeer's paintings in collaboration with the Netherlands Institute for Art History. Forty percent of Vermeer's paintings can be linked to a work by the same master on the basis of the canvas. Furthermore, the technique showed (once again) that a previously doubted painting is a real Vermeer. The canvas contains a wealth of information.

Automatically count wires

X-ray scans of old masterpieces are not new. As early as the 1930s, paintings were screened to unravel painting techniques. Under the top layers of paint, 'earlier versions' of the work emerged, and information was gained about how the artist applied the paint to the canvas. The effects of the framework and nails on the canvas and 'invisible' damage were also seen.

Johnson came into contact with the practice of scanning artworks about ten years ago as an engineer. “That was a pretty 'classic' approach at the time,” he says. “The curator had X-rays that hung on a light box like in a hospital. Thread counting was already done, but manually for small pieces of paintings. This way you could determine whether two paintings came from different rolls of cloth. Unfortunately, the method was not precise enough to confirm that they came from the same canvas.”

With a background in electrical and information technology, Johnson immediately saw that there was profit to be made here. A human is able to count only a small number of threads in a canvas before losing concentration and perhaps making mistakes. But a computer can tirelessly map the density of an entire canvas, with great precision, canvas after canvas. Perhaps it was possible to determine whether canvases actually came from the same roll. He set to work with the support of the American Cornell University.

Based on existing image processing techniques, Johnson's team developed a fully automatic system that turns a digital X-ray image into a map of the 'density' of a canvas. In a typical Vermeer, there are about twelve or thirteen wires next to each other in a centimeter. But due to imperfections in the canvas, that number varies in certain places.

Vermeers pair

Once it was clear that the technique worked, it could be unleashed on old works. Johnson decided to focus on the paintings of Johannes Vermeer, the world-famous Dutch painter with a modest oeuvre of 34 paintings. “The small number of works meant it had to be relatively easy to analyze the entire oeuvre,” says Johnson. “I also had connections with a number of museums, including the Rijksmuseum, which together own about half of the Vermeers. It took years, but we eventually got x-rays of all his works.”

Was he allowed to take the paintings to his laboratory? “Certainly not!” says Johnson. “Such a request is never honored. We used images that already existed, mostly made by the museums, for research that they do themselves.”

In the end, Johnson managed to find eight 'couples' in Vermeer's oeuvre based on the canvases. Thus, 'Seated woman at the virginal' and 'The lacemaker' appear to have the same role, as do 'Sitting virginal player' and 'Standing virginal player', and 'The astronomer' and 'The geographer'. There was also a kind of special jigsaw puzzle in the works:with some turning and sliding, four works turned out to be from the same canvas.

An interesting case was the above-mentioned 'Woman sitting at the virginal'. Experts now agree that this painting was made by Vermeer, but this was previously doubted. The research by Johnson and colleagues now proves that the painting comes from the same role as an "undoubted" Vermeer. Whether this method provides conclusive proof? Johnson is careful. “The assignment of works is complicated. The X-ray scans can seem convincing, but you really need more,” he says. “I think that historians will always disagree on points, but maybe after my research it turns out to be more difficult to disagree.”

How is it possible that paintings by Vermeer come from the same roll of canvas, while according to historians they were made years apart? “Maybe he kept the canvas for a long time,” Johnson suggests, “or maybe the dating of these paintings should be different. But that remains the task of historians. That's not me. With this technique we only provide indications that the experts can use for (new) insights."

Chinese painting on silk

With the 34 works by Vermeer, the researchers show that the technique works. But it doesn't stop there. The work of Vincent van Gogh, Claude Monet, Dirk Bouts, Diego Velázquez and Nicolas Poussin also contain couples from the same canvas. And Johnson thinks further.

He now wants to focus on Chinese paintings from the 12th and 13th centuries, which are often painted on silk. That research is a challenge, because those threads are much closer together than those in canvas. Not ten per centimetre, but a hundred. “Nobody is going to count that by hand,” Johnson says. “Perhaps our automation of the technology can provide new insights there. And perhaps in other areas. In principle, you can examine anything made of fabric in this way, such as historical flags or clothing.”

It seems like a powerful new tool in the emerging field of forensic art history. But why didn't we start with this sooner? “People often say that anyone could have done this, and that's true to a degree,” Johnson says. “In this case, it was essential to have someone who speaks both the language of engineers and that of art historians. Fortunately, I had enough knowledge of both worlds to bridge this gap.”