Jaffe-Dax, S. 1 , Bejjanki, V. R. 2 & Emberson, L. 1
1 Princeton University
2 Hamilton College
The ability to predict upcoming events may be one of the principle ways that statistical learning supports development. Here, we investigate how the predictability of events modulates neural responses in young infants, and specifically, whether neural responses are consistent with a predictive coding framework. In this framework, top-down connectivity is used to compare sensory input against predictions, and neural responses to predictable stimuli are attenuated.
We examined prediction in the context of sequential learning in young infants. Using audiovisual stimuli (distinct shapes each with a non-verbal sound), we manipulated predictability within-subjects. In the Predictable condition, the temporal order of the stimuli was constant across trials (i.e., AAAB), while in the Unpredictable condition, the temporal order varied from trial to trial (e.g., BAAA, ABAA). During stimulus presentation, we used functional near-infrared spectroscopy (fNIRS) to investigate neural responses in perceptual (occipital, temporal) and associative regions (frontal, parietal).
Consistent with the framework of predictive coding, infants had a stronger neural response in all measured lobes in the Unpredictable condition. Moreover, we found greater functional connectivity towards sensory lobes in the Predictable condition. Thus, after learning of a short sequence, infants employ top-down connections to attenuate neural responses across the cortex.