Psychophysiological dynamics of pupil diameter during category learning in humans

Psychophysiological research links pupil size to neuromodulatory systems like the Locus Coeruleus, yet its role in tracking abstract category learning remains unclear. We used the Weather Prediction Task (WPT) to investigate whether trial-by-trial pupil dynamics reflect this process. Fifty-three participants categorized geometric patterns while pupil diameter was recorded at 500 Hz. Learning was quantified via behavioral performance and strategy analysis, which identified the cognitive approaches utilized (e.g., optimal abstract integration). Results showed significant improvement in 96% of participants, with 81% adopting an optimal strategy. Using a time-resolved analysis of the pupillary signal, we dissociated two distinct, temporally separated dilations: one during the categorization decision and another during feedback processing. These dilations differed for correct and incorrect trials during specific temporal windows within the trial structure. Linear mixed-effects modeling revealed that the integrated pupil responses could be modeled as a function of trial number but not performance, suggesting that pupil size tracks task progression. These findings suggest that pupil size serves as a non-invasive index of the internal dynamics of category learning, reflecting changes in both decision-making and feedback processing.