[PS-3.7]Examining the online mechanisms of visual statistical learning using pupillometry

Zhang, F. & Emberson, L.

Princeton University

Adults are able to use statistical learning to generate predictions about future sensory inputs, but how does learning and prediction change on a trial-by-trial basis? Changes in pupil diameter, a marker of learned uncertainty and surprise, is closely related to learning and prediction. Therefore, we compared the pupil dilation response (PDR) for participants that performed well versus poorly on a visual statistical learning task, as measured with a surprise memory test. Results revealed a significant increase in PDR throughout the experiment when participants are most able to predict the upcoming image (before the third image of a triplet, R = 0.30, p = 0.002) but not when they are less able to predict (e.g., before the first, R = -0.04, p = 0.69 or second image, R = -0.013, p = 0.89). Splitting subjects based on accuracy, we find that this correlation is driven by participants who performed poorly (R = 0.27, p = 0.007 vs. R= 0.19, p = 0.06). These preliminary results suggest that more protracted increases in PDR to predictable items reflects poor performance, whereas good learners reach their asymptote earlier. These findings make a link between early prediction, as measured by PDR, and post-test memory performance.