Abstract
Pupil size responds to a plethora of cognitive processes and is known to be a reliable marker of arousal. Accordingly, pupil size increases in response to surprising or salient events. However, it remains unclear whether these pupil dilations correspond to responses to centralized, behaviorally relevant events or to prediction errors (PE; i.e. the difference between expected and actual stimulus) distributed across the brain, in accordance with Predictive Coding frameworks. Here, we ran 2 series of experiments in order to tease these hypotheses apart. In the first series, we tested 1) how pupil responds to non-frequent statistical regularities (transitions between stimuli), 2) how attention modulates these responses and 3) the relation between these responses and the level of awareness of the transition statistics. We show that pupil responds to rare transitions even when unconscious, suggesting that relevant events do not need to reach consciousness to trigger pupillary responses. In the second series of experiments we tested whether the pupil responds to perceptual PE (i.e. the temporal or spatial predictability of a stimulus), by showing participants full-screen kinematograms with low or high levels of predictability. Our preliminary results suggest that pupil does not respond to perceptual PE per se. Ongoing experiments are investigating whether, rather than absolute PE, pupil size could be viewed as a marker of precision-weighted PE, which depends crucially on context.
