Emberson, L. 1, 2 & Aslin, R. . 1
1 University of Rochester
2 Princeton University
It is an open question how exposure to biased statistical information supports changes in perception during development. We employed fNIRS (a non-invasive optical method for recording the hemodynamic response similar to fMRI) and MR co-registration to record neural activity in the perceptual cortices as infants learned and responded to violations in cross-modal (auditory/visual) statistical information.
Infants received a period of familiarization where a sound always predicted a visual stimulus. After familiarization, to test whether the infant visual system changes its neural responses based on statistical learning, the predictive auditory stimulus was not followed by the expected visual stimulus on 20% of trials. These novel test trials overcome interpretative limitations of previous work: The unexpected absence of a stimulus cannot elicit a novelty response, but must be an adaptive change in perceptual cortex resulting from statistical learning.
Recordings from 15 6-month-olds revealed robust and similar increases in activity in the occipital cortex during both statistically-consistent (A+V+) and unexpected visual omission trials (A+V-), suggesting that statistical learning generates expectations even in early neural responses of the visual system. Recordings in 16 additional infants confirmed that without prior statistical learning the omitted visual stimulus (A+V- trials) produces no response in the occipital cortex.