23 October 2012

Bees don’t know much about art, but they know what they like

Artistic “style” is often immediately recognizable, but – given the track record of successful forgeries of paintings – almost indefinable.

Could we turn to honeybees for help?

A new paper by Wu and colleagues asks if bees can learn to differentiate paintings based on artistic style. They’re not interested in detecting forged paintings; they’re interested in learning.

We know bees can learn to tell flowers apart, like flowers that give food rewards from those that don’t. But there’s some controversy on how sophisticated their learning abilities are. Some argue that the bees are looking for only simple primary cues, like "white” and “circle.”

Humans, on the other hand, routinely combine those primary cues into larger categories. For a long time, categorization was considered very high-level thinking, requiring intent. Philosophers of mind often touted categorizing as something that only humans could do, because animals and machines did not have intent (the argument went).

Wu and colleagues had a simple enough experiment: present the bees with two paintings. One was by Claude Monet (on the left in the pairs below), and one was by Pablo Picasso (on the right in the pairs below). They placed a food reward behind one, and let the bees try over and over again to find the food.

The bees were able to learn if a Monet meant munchies or if a Picasso portended a picnic. They never got perfect, though. They tended to top out at choosing the right master about 75% of the time. When given five pairs of pictures to learn, the bees continued to perform at about the same rate, getting to about 75% accuracy across all the pairs.

The bees showed no preference for the impressionistic Monet or the cubist Picasso. They started off choosing each about half the time.

But did the bees actually recognize the artists’ distinct styles, though? Would they generalize? The experimenters tested this by training the bees to several different sets of painting, again to where the bees were choosing better than chance. Then, they gave the bees new pairs of paintings by these artists that they had never seen before. If the bees had developed categories for style, those rewarded for going to a Monet should keep going to a new Monet.

The bees... were not great at this task. Their performance usually sank down to near chance success rates. In a few cases, the bees choose the new “correct” painting by the artist they had been trained to at about the same rate as they had been trained to previously. That they can do this at all suggests that the bees are paying attention to more than just simple properties of the visual stimuli. Bees seem to be able to learn categories, but they’re not great at it.

Wu and colleagues suggest that with time, bees could learn to do this task better. The problem is that honeybees are short-lived, and would not make it through the semester of an art appreciation class.


Wu W, Moreno AM, Tangen JM, Reinhard J. Honeybees can discriminate between Monet and Picasso paintings. Journal of Comparative Physiology A: in press. DOI: 10.1007/s00359-012-0767-5

Picture by celinecelines on Flickr; used under a Creative Commons license.

No comments: