The Pigeon as Art Critic

Consider how mystifying a walk through an art museum can be. Although dramatic changes in style are simple for anyone to spot, imagine asking someone, on their very first visit: “which of these paintings are the best and which are the worst?” Yet a connoisseur, after learning to identify the finer points of various styles, would have no trouble picking out his or her favorite pieces.

A much easier task would be to identify good children’s artwork. Still, consider what a challenging task even this is for the brain -- taking in a complex field of colors and shapes, making some sense of it, and deciding “good” or “not.”

It’s an impressive task, but even a bird can do it. A recent experiment by Shigeru Watanabe showed that the utterly un-artistic pigeon could be taught to identify “good” and “bad” children’s artwork. How was the pigeon able to perform this feat and why should we care that it did?

Step back for a moment and look around your environment. Perhaps you’re sitting at a desk, by a table, or near a window. Outside, you might see trees, grass, other people, birds, and flowers. Despite the fact that you’re probably receiving only a small amount of visual information about many of these stimuli, it seems to be an effortless and unthinking process to classify these objects into their respective categories. Categorizing these patterns of visual stimulation allows you to make sense of an often complex and chaotic world.

It might very well be the case that much of our substantial neural machinery is built to address this complex problem of object recognition; nevertheless, identifying so many different kinds of objects must surely be a difficult computational task. After all, we rarely see the same version of any of these objects again. Still, the task may be simplified by constraints on the appearance of these objects. Most birds have beaks and wings; most trees have tall brown trunks and green leafy branches; etc. As well, the trees, flowers, and grass will stay relatively still, whereas the birds and people will move in ways that are true to their biological nature.

Remarkably, when pigeons are suitably trained, they too can learn a variety of visual categories. We train the animals to report these visual categories by placing them into an operant conditioning chamber (or Skinner box), a common psychological testing apparatus. There, the pigeons are shown photographs on a computer monitor and they receive grain when they correctly respond in a kind of computer game which requires that they peck one key when they are shown a picture of a cat, a second when they are shown a picture of a human, a third when they are shown a picture of a car, and a fourth when they are shown a picture of a flower. After training, the pigeons can also generalize these categorization responses to new photographs that they have never before seen, thereby revealing that they have extracted information about the characteristic features in each of the categories that they have been taught.

Understanding how animals categorize and process these visual stimuli has important implications for our view of both animals and people. These kinds of discriminations attest to the ability of animals to learn both categories and concepts, once thought to be uniquely human abilities. From other research in this field of comparative cognition, we now know that animals such as pigeons and monkeys respond to not only different perceptual categories of objects, like trees or cats, but also to relational concepts, such as sameness and differentness. Comparing the strategies that animals and people use to solve these complex tasks can reveal similarities, or disparities, in how we learn about the world and adapt to new problems, despite (or because of) what may be large neuroanatomical differences. 

To test the art discrimination abilities of pigeons, Watanabe first asked people to classify children’s artwork as either “good” or “bad.” Those pictures that were consistently classified as “good” were placed into one group of pictures; those that were consistently classified as “bad” were placed into a second group. The pigeons in this experiment were taught that pecking at the “good” set of pictures would yield a grain reward, whereas pecking at the “bad” set of pictures would not do so. In an average of only 22 sessions, all of the pigeons that were so trained learned to respond most of the time to “good” artwork and not to “bad” artwork.

How had the pigeons done it? Although people might not immediately suspect this strategy, one possible way to solve this kind of discrimination task is to simply memorize most or all of the various pictures in each set, paying little attention to the categories themselves. To test this possibility, Watanabe performed a generalization test, in which new, similarly classified artwork was shown that the birds had never before seen. All of the birds were able to appropriately peck at the new “good” pictures and to withhold pecking at the new “bad” pictures. So, simple memorization is probably not the main avenue by which the birds accomplished this complex visual discrimination.

To assess other possibilities, Watanabe gave other, more sophisticated tests. When the pictures were reduced in size, the pigeons were able to discriminate well. But, when the pictures were shown in grayscale, the pigeons discriminated quite poorly. In a third test, Watanabe presented “mosaicized” stimuli, with their color patterns preserved, but with their shape information destroyed; now, the birds could discriminate with slightly mosaicized stimuli, but they could not do so with still higher levels of mosaicization. Therefore, it seems that color information, shape information, and picture organization were critical for the pigeons sorting the pictures into categories, despite the fact that the pictures themselves contained no obvious grouping of colors or shapes and both “good” and “bad” artwork had been created in a variety of media (both pastels and watercolors).

Although we may be quick to praise each of the various doodles and paintings that our children make, we nevertheless do guide them toward creating certain kinds of artwork. We hire teachers and buy materials so that our children can create better, more expressive art. And, we do so despite the fact that the teachers’ own definition of “good” or “bad” artwork may be unstable, difficult to define, and even more difficult to explain. Yet, when we ourselves stand in front of a series of paintings at an art museum and we’re asked to pick the “good” art from the “bad,” we don’t have any clear answers either: we simply know it when we see it.