Is It Real?
Rummaging around your attic one day, you find a painting that looks like a Picasso. How do you know if it is real, and you are about to become very rich, or fake, and the item will be forever relegated to the attic?
The time-honored method of authentication is to call on experts. Someone who has devoted his or her life to studying Picasso’s brush technique, his methods of composition, his color palette, the sweep of his lines: this person will give you a reliable judgment.
You could also buttress the case with science. Carbon dating could reveal a forgery, as could spectral analysis of the pigments in use, or chemical verification of the oils, varnishes, and binders.
A new paper adds statistical analysis to the quiver. Writing in the Proceedings of the National Academy of Science (PNAS 107 [26 January 2010], 1279–1283), a group of mathematicians and computer scientists from Dartmouth and the Santa Fe Institute develop a quantitative method to classify artistic style. Focusing on Pieter Bruegel, the Dutch Renaissance painter, the group quantifies art based on a “sparse coding model” of the distinguishing features of sets of known works of an artist.
The result is a set of statistical coefficients that can distinguish known Bruegels from works by imitators of his style. The new method is superior to any previously applied techniques for recognizing forgeries.
Alas, the results don’t help you with your Picasso problem, since it isn’t known if more abstract art would be as susceptible to this type of analysis. It is amusing, though, that science can perform at least as well as art historians who have spent lifetimes understanding all the nuances of an artistic style.
But lest scientists become overly sure of themselves, we still need the art historians: after all, nuance is the very essence of art.