https://scienceblog.com/m-pigeons-were-trained-to-tell-a-monet-from-a-picasso-and-then-did-something-odder-shown-paintings-they-had-never-seen-they-sorted-cezanne-with-the-monets-and-braque-with-the-picasso/
Pigeons were trained to tell a Monet from a Picasso, and then did something odder — shown paintings they had never seen, they sorted Cezanne with the Monets and Braque with the Picassos.
The birds had grasped something deeper than memorization—a visual principle that let them recognize artistic style in works they'd never encountered.
Apigeon sits in a box, facing a small screen. A Monet appears, the bird pecks a key, and a pellet of food drops. A Picasso appears, the bird pecks, and nothing comes. Do that a few hundred times and the pigeon learns the rule: peck for a Monet, ignore a Picasso. That much is ordinary animal training. The strange part came later, when the birds were shown paintings nobody had ever put in front of them.
We are not psychologists or scientists, and what follows is a reading of one experiment and the debate around it, not a claim about how minds work. It’s a single study from 1995, taken seriously but not treated as the last word.
What the pigeons were trained to do
The experiment came from Keio University in Tokyo. In 1995, Shigeru Watanabe, Junko Sakamoto and Masumi Wakita published a study in which pigeons learned to tell color slides of Monet paintings from Picasso paintings. It later won a 1995 Ig Nobel Prize in psychology, the prize that goes to work that makes people laugh and then makes them think.
The setup was simple. Eight pigeons were split into two groups. One group got food for pecking at a Monet and nothing for a Picasso; the other group had it the other way around. Each artist was represented by ten paintings in training. The birds learned it fast, reaching over 90 percent accuracy.
But if that were the whole story, it would just show that a pigeon can be trained to react to a specific set of images. Neat, but not puzzling.
The part that’s genuinely odd
When the researchers showed the birds Monet and Picasso paintings that had never appeared in training, the pigeons still sorted them correctly. So they weren’t just memorizing ten pictures.
Then it went further. Shown work by painters the birds had never seen at all, the pigeons filed Cezanne and Renoir with the Monets, and Braque and Matisse with the Picassos.
That grouping isn’t random. Cezanne and Renoir belong to the broad family of Impressionism, alongside Monet; Braque helped found Cubism with Picasso. The birds put each newcomer on the side an art historian would choose, without ever being told a name or a movement.
What the birds were sorting on
Not names. Not intent. Not the story behind a French garden or a fractured guitar. A pigeon has no access to any of that. Whatever it was doing had to run entirely on what was visible on the surface of the image: edges, color, brightness, texture, how shapes were arranged.
One extra finding sharpens this. When the researchers turned Monet’s paintings upside down, the birds got worse at telling them apart. Flip a Picasso, and they didn’t care. The authors’ best guess, again cautious, was that the birds may have been picking up on recognizable objects in the Impressionist paintings, objects that look wrong upside down, while the Cubist images were already jumbled enough that orientation made no difference.
Whether any of this counts as real “categorization” is what other researchers questioned. In a 1998 response, Jos Monen, Eli Brenner and Jenny Reynaerts argued that studying categorization only makes sense if the animal could actually have the information used to define the groups. Their point: a bird sorting by color and brightness is not the same as a bird grasping a concept called “Impressionism.” You have to rule out simple visual cues, they said, before claiming the bird learned a concept. This is a live disagreement, not a settled result.
What this does and doesn’t tell us about style
It would be easy to read the study as a put-down, as if it reduces a Monet to a bundle of edges a bird can peck at. We don’t think that’s the fair reading. What the pigeons found is that the surface of a Monet and the surface of a Cezanne share something measurable, something steady enough to carry over to a painter they’d never seen. That shared signal is real. It’s in the paint. The birds found it without any of the meaning we attach to it.
The same trainability later turned to a serious use. In 2015, a team including Richard Levenson and Edward Wasserman trained pigeons to tell harmless breast tissue images from cancerous ones. According to the University of Iowa, individual birds reached nearly 85 percent accuracy, and a group of four pooled together did even better. A bird that can spot the visual signature of a cancer cell is running the same basic trick as a bird that files Braque with Picasso: pulling a pattern out of images and applying it to new ones.
Which leaves a good question open. When we say two paintings share a “style,” part of what we mean is a pattern in the image itself, a fingerprint of color, edge and texture that repeats from one canvas to the next. A pigeon can find that fingerprint. What it can’t find is everything else we mean by style: the intent, the history, the argument a painter is having with the painters before them. The experiment didn’t shrink the art. It showed how much of style sits right there on the surface, available to any eye careful enough to look, even one that will never know whose name is in the corner.
https://scienceblog.com/m-pigeons-were-trained-to-tell-a-monet-from-a-picasso-and-then-did-something-odder-shown-paintings-they-had-never-seen-they-sorted-cezanne-with-the-monets-and-braque-with-the-picasso/
