Can Technology Tell Who's Laughing Now?

June 23, 2008 — -- Perhaps it was the reassuring smile. Or was it something else?

One can only guess what it was that convinced a buyer to drop $4.5 million on a "laughing" Rembrandt portrait that experts had panned as a fake.

The painting, which sold in October at a British auction house, had been considered the work of a follower, rather than the master, and was valued at $3,100. But last week, an investigation by the Rembrandt Research Project confirmed that it was indeed a self-portrait by the legendary painter himself.

Although the high-stakes process of authenticating famous artwork is far from foolproof, modern technology has allowed investigators to probe masterpieces in novel ways. X-rays, for instance, revealed another painting beneath the self-portrait. The hidden image exhibited qualities similar to other Rembrandt paintings -- evidence that evaluators were unable to discern.

And now researchers at Dartmouth University hope to take it a step further by developing a computer program that may someday distinguish clever imitations from the real deal.

Museums have typically determined authenticity by polling historians and scholars until a consensus is reached. Experts, however, can sometimes disagree, get it wrong or end up stumped. During the 1930s and '40s, a painter named Han Van Meergeren showed just how tenuous this method was by passing off his forgeries as the work of well-known artists and making millions in the process.

"We rely on connoisseurship incredibly, but it's also the most fallible of all things," says Scott Schaefer, a curator at the J. Paul Getty Museum in Los Angeles. "And in the balance hangs thousands, sometimes millions of dollars."

Forensic science emerged as way to better spot knock-offs. The combination of carbon dating, X-rays and other hi-tech tools can pinpoint the age of the canvass, the chemical makeup of color pigments and purity of the materials -- all of which should coincide with the purported era the piece was created.

Still, historians say much of the headache with attribution can be blamed on a common practice during Rembrandt's time in which students were encouraged to emulate their mentors. Nearly everything -- from the teacher's technique to the types of paint used -- was often copied with the instructor's approval. Artists would sometimes even sign an apprentice's work to earn extra cash.

So while a scientific analysis can confirm that the painting was of a certain time period, that might not get you very far, says Schaefer.

Dan Rockmore, a mathematician at Dartmouth University, believes that digital technology may change all that.

His idea came during a trip to New York's Metropolitan Museum of Art back in 2001, where he caught a glimpse of how curators judged whether a piece of artwork was bogus. As Rockmore toured the exhibit, which featured the drawings of Pieter Bruegel the Elder, he observed the way they highlighted certain details that set Bruegel's strokes apart from his imitators, and pondered: "Can we quantify what a connoisseur does?"

He soon put the question to the test by first digitizing slides from the exhibit, eight authentic Bruegel sketches and five drawn by other artists. From there, the pictures were converted into statistical representations and fed into software programmed to identify subtle patterns, such as the one-of-a-kind swoosh of an illustrator's pen.

"Different artists simply have different ways of generating lines. Some will make few or denser lines than others," he explained. "You can see statistical differences in that."

The findings, reported in the Proceedings of the National Academy of Sciences, demonstrated that the computer was able to correctly pick out the actual Bruegel drawings based on those differences.

Even as he looks to build on the study's promising results, the rational-minded professor thinks it will be a while before we can start installing digital connoisseurs in galleries. The computer program has yet to fare as well outside of drawings and analyzing complex works like paintings, he says, would require "significantly more development."

So for now, Shaeffer says, "it often falls on more mundane things like two eyes."