Babies' long looks: Can researchers trust them?

Researchers studying babies to discover the origins of perception and cognition measure what babies look at, or what captures their attention. But scholars have long wondered how much they can rely on those long looks.

New work that evaluates decades of infant cognition work validates the classic research technique. The findings are published in Nature Human Behaviour.

"Looking behavior is the most common measure in studies of early cognitive development, but what are we really measuring?" said author Shari Liu, a Johns Hopkins University assistant professor of Psychological and Brain Sciences specializing in infant cognition. "This project was a way for us to get at that question."

Infant cognition researchers have used looking behavior to demonstrate that babies have expectations about the physical and social world. Researchers typically present babies with a situation that should violate these expectations—like a ball appears to pass through a wall, or float in midair. But some researchers wonder if it's the mere visual novelty of these things that captures babies' attention rather than a violation of their expectations about the world.

"When babies look longer at the ball floating, what are they responding to: the novelty of the visual stimulus, or the fact that it violates Earth physics?" Liu says. "Answering this question is critically important for our understanding of infant minds."

The team performed a meta-analysis of data from roughly 20 years of infant cognition research that measured both types of looking behavior. These studies were set up to measure looking for unexpected events (e.g. the ball floating in midair, versus resting on a platform), but also incidentally measured looking to visually novel events (e.g. the ball resting on a platform, versus preceding trials where it rolled across the floor). If there was no difference between babies looking at novelties and babies looking at things that defied their expectation, these two effects should be predicted by the same factors across studies. But that's not what the team found, suggesting these are two truly separate behaviors.

The age of the infants was the best predictor of how long the babies looked at unexpected events, Liu said. By contrast, how quickly infants got bored earlier in the experiment was the best predictor of how long they looked at visually novel events.

"We found that an answer was sitting right in front of us, hidden in data that have been painstakingly collected over the last 20 years," Liu said. "This work is one example of work in a recent movement in the field to gain cumulative insights about infant development rather than performing a series of 'one-off' studies and then moving onto the next question."

Authors include Linette Kunin, Sabrina Piccolo and Rebecca Saxe, all of Massachusetts Institute of Technology.

The work was supported by National Institutes of Health grant F32HD103363, DARPA grant CW3013552, and the MIT Undergraduate Research Opportunities Program.