Pigeons Learned to Tell Monet From Picasso, and Won an Ig Nobel for It

In 1995, three psychologists at Keio University in Tokyo reported that ordinary pigeons could learn to tell a painting by Claude Monet from one by Pablo Picasso, and then correctly sort paintings they had never seen before. The study, by Shigeru Watanabe, Junko Sakamoto and Masumi Wakita, ran across ten pages of the Journal of the Experimental Analysis of Behavior, volume 63.

The training itself was plain behavioral science. The birds were shown color slides of works by the two artists and earned food rewards for responding to the right painter. Over time they learned to separate the two styles reliably, without anyone ever explaining what impressionism or cubism is.

The interesting part came afterwards. When the researchers showed them Monets and Picassos that had never appeared during training, the pigeons still sorted them correctly. That rules out simple memorization of individual pictures: the birds had picked up something general about the way each artist painted.

The team then pushed further with painters the birds had never seen at all. Works by Paul Cézanne and Pierre-Auguste Renoir were grouped with Monet, while works by Georges Braque and Henri Matisse were grouped with Picasso. In other words, the pigeons drew a line that roughly matches the one art historians draw between impressionist landscapes and the more fragmented modern styles of the early twentieth century.

One more trick revealed how the birds were looking. Turning Monet's paintings upside down disrupted their performance, but inverted Picassos caused no trouble. The authors suggested that the pigeons were partly recognizing real objects in the impressionist scenes, such as water, trees and boats, which look wrong when flipped, while the cubist canvases offered no stable objects to begin with.

The work earned Watanabe, Sakamoto and Wakita the 1995 Ig Nobel Prize in Psychology, awarded for their success in training pigeons to discriminate between the paintings of Picasso and those of Monet. The Ig Nobels honor research that first makes people laugh and then makes them think, and this one fits both halves.

It was not a one-off. Back in 1964, Richard Herrnstein and Donald Loveland showed in Science that pigeons could learn a complex visual concept: spotting whether a person appeared in a photograph. Later work found that the birds can memorize more than 1,800 images.

The most striking follow-up came in 2015, when Richard Levenson and Edward Wasserman's team trained pigeons on breast tissue slides in PLOS ONE. In about 15 days, accuracy rose from 50 percent, pure chance, to roughly 85 percent at telling benign from malignant tissue, and it held on new images. Pooling the votes of four birds pushed the score close to perfect, although the pigeons failed at the hardest task, judging suspicious masses on mammograms.

None of this means pigeons enjoy art. What the experiments show is that recognizing a painter's style, something we usually treat as educated human taste, can be learned from visual patterns alone by an animal whose visual system shares many basic features with ours.

Sources: Watanabe, Sakamoto & Wakita, Journal of the Experimental Analysis of Behavior (1995); Full text at PubMed Central; Improbable Research, Ig Nobel Prize winners; Herrnstein & Loveland, Science (1964); Levenson et al., PLOS ONE (2015). Cover: Rock dove by Alan D. Wilson.