Pupil dilation offers a time-window on prediction error

Abstract Task-evoked pupil dilation has been linked to many cognitive variables, perhaps most notably unexpected events. Zénon (2019) proposed a unifying framework stating that pupil dilation related to cognition should be considered from an information-theory perspective. In the current study, we investigated whether the pupil’s response to decision outcome in the context of associative learning reflects a prediction error defined formally as information gain, while also exploring the time course of this prediction error signal. To do so, we adapted a simple model of trial-by-trial learning of stimulus probabilities based on information theory from previous literature. We analyzed two data sets in which participants performed perceptual decision-making tasks that required associative learning while pupil dilation was recorded. Our findings consistently showed that a significant proportion of variability in the post-feedback pupil response during decision-making can be explained by a formal quantification of information gain shortly after feedback presentation in both task contexts. In the later time window, the relationship between information-theoretic variables and the post-feedback pupil response differed per task. For the first time, we present evidence that whether the pupil dilates or constricts along with information gain seems to be context dependent, specifically related to increasing or decreasing average uncertainty (entropy) across trials. This study offers empirical evidence showcasing how the pupil’s response can offer valuable insights into the process of model updating during learning, highlighting the promising utility of this readily accessible physiological indicator for investigating internal belief states.