The Monkey Gaze

Andrew Sullivan —  Nov 12 2013 @ 8:09pm

Neuroscientist Asif Ghazanfar, whose lab uses monkeys to study auditory and visual perception, set up an experiment to test whether monkeys are capable of following filmic narratives the way people are:

Ghazanfar and his postdoc, Stephen Shepherd, tracked the eye movements of monkeys and people as they watched identical 3-minute clips from three films: the BBC’s “Life of Mammals,” Disney’s “The Jungle Book,” and Chaplin’s “City Lights.” The movies were converted into black and white and played without sound. As it turned out, humans and monkeys have similar cinematic tastes. … “There was a surprising degree of overlap,” Ghazanfar says. The gaze paths of humans and monkeys overlapped 31 percent of the time. A small part of this correlation is due to our shared visual reflexes: Both humans and monkeys are attracted to bright spots. But the bulk of the overlap was driven by the two species’ shared interest in complex scenes, particularly faces, body movements, and social interactions.

But the researchers also found two intriguing differences between the monkey and human gaze paths.

First, “humans appear to look at the focus of actor’s attention and intentions to a much greater extent than do monkeys,” Ghazanfar and Shepherd wrote in a fascinating review published in the film journal Projections. Second, “humans appear to pay attention to related details in a movie for much longer than monkeys do, suggesting that humans integrate events over time in a fundamentally different way.”

In other words, it seems that what makes people different is our ability to follow a narrative. Whereas monkeys look and react to scenes quickly, people fixate on one actor and integrate complex events over time. In a clip showing two monkeys, for example, people tended to look squarely on the monkey sitting quietly in the center of the screen. Monkeys, in contrast, looked at the more active second monkey, even thought it was jumping out of view of the camera. “Monkeys were reacting moment-by-moment instead of assembling and testing a narrative explanation for the scene before them,” the researchers wrote.