Selective Perception in Infants — Saxe 2020
Jul 19, 2020
Link to talk: https://videocast.nih.gov/summary.asp?live=35611&bhcp=1
Rebecca Saxe of MIT showcased research examining the extent of regional specialization of perceptual, social, and cognitive processing in human babies. Her work challenges the prevailing hypothesis that babies passively absorb information and develop top-down attentional mechanisms only after statistical patterns are learned over time.
Saxe’s first study involved fMRI of human infants during the observation of scenes and faces. The most obvious challenge is keeping babies still and calm during the scan. The way the researchers dealt with this problem was by sending a person into the scanner with the baby—an imperfect solution, no doubt. In any case, it became clear from the scans that certain regions were preferentially active in response to faces and others to scenes. Indeed, these were the same regions that showed preferential activation in adults: parahippocampal place area (for scenes), fusiform face area (faces), and posterior superior temporal sulcus (faces). In a follow-up study, babies were shown to also display preferential activation in the medial prefrontal cortex, which plays specific social and cognitive roles in human adults. Specifically, the mPFC is most active in contexts that are social, self-related, and/or high value.
To assess whether these same aspects drive baby medial prefrontal cortex (mPFC) activation, Saxe devised additional experiments. While brain activity was being recorded using near infrared spectroscopy, 6-12 month old babies would be shown videos of two people talking, one that talks in an exaggerated, infant-friendly way and the other that talks in a mumbling, unanimated way. Previous studies have shown that babies tend to look at the former more and the latter less. This fNIRS study found that left mPFC activity was higher when the baby was observing the infant-directed speaker. This relationship was not observed in the left lateral prefrontal cortex. Subsequent eye-tracking data showed that the degree of left mPFC activation was correlated with the degree of preferential observation of the infant-directed speaker. The degree of activation in the left lateral prefrontal cortex did not predict preferential observation for one speaker over the other.
The lateral prefrontal cortex (lPFC) is hypothesized to be involved in a different preference—one for novel statistical information. Babies can detect regularities in speech. Structured speech (e.g., if a speaker consistently follows “na” with “wo”) is more informative to babies than unstructured speech (e.g., “bababababa”—an expression with no informative patterns beyond pure repetition), and human infants have been shown to attend to the former more than the latter. Saxe sought to examine whether a particular brain region reflects this preference. Using the same fNIRS design of the previous study, she found that the left lateral PFC activation predicted a preference for structured speech over unstructured speech, and the same correlation was not observed in the mPFC. As such, a double dissociation was achieved; preferences for regularity and infant-directedness of speech are traceable to different regions of the infant brain.
Altogether, these findings suggest that infants are not passive recipients of experience. Rather, they exhibit systematic preferences; from a very early age, they selectively attend to phenomena with high social or informational value. These are just two preferences that Saxe studied, but she lists many others: direct versus averted gaze, rocking versus non-rocking motions, contingent versus non-contingent responding, etc. In terms of further research, Saxe points to the importance of studying interrelationships between the mPFC and attentional/representational mechanisms in infants. Perhaps in so doing, we may learn what babies are endogenously trying to find, learn, and do.