A year ago, I blogged about David Hockney's controversial book -- which argues that the secret behind the explosion of realism in the Renaissance was the camera obscura. Hockney thinks the old masters were using cameras obscura to project their subjects onto canvases, and then traced them from life.

But now there's a rebuttal, from two computer scientists. David Stork and Antonio Criminisi did a close analysis of the chandelier in "Portrait of Arnolfini and His Wife" by Jan van Eyck. Their verdict? Not that realistic, according to The New York Times:

The problem, as Mr. Stork and Mr. Criminisi see it, is that the chandelier in the Arnolfini portrait is hardly painted in perfect perspective. Relying on digital image registration techniques, Mr. Stork and Mr. Criminisi applied projective geometry to one of the arms of the Arnolfini chandelier to see what the others should look like given the various angles that the painter would have seen them from. Then, as Mr. Criminisi described it during a phone interview yesterday, each painted arm was compared with its ideal perspectival projection; they were not identical.

You can read Stork and Criminisi's paper here online -- and check out their way-kewl illustrations showing how the vanishing point for the perspective on the chandelier is slightly off.