Jasmine Lopez¹ (), Jefferson Ortega¹, David Whitney¹; ¹UC Berkeley

The ability to perceive the emotions of others is influenced by how we physiologically respond to social cues. Pupil size, in particular, reflects fluctuations in emotional arousal (Oliva & Anikin, 2018) and may serve as a dynamic marker of real-time emotion processing. While previous work has linked pupil size to emotional salience, it remains unclear whether moment-to-moment pupil dynamics can predict individual differences in emotion perception performance. In this study, participants (N = 83) completed the Inferential Emotion Tracking (IET) task, in which they watched eight emotionally evocative video clips and continuously rated the affect of a target character using a two-dimensional valence and arousal grid. The character’s facial expressions were blurred, encouraging reliance on contextual cues to infer emotion. Pupil size was recorded throughout using the Eyelink 1000 eye-tracking system. To assess whether pupil dynamics tracked emotion processing, we computed intersubject-pupil correlations for each participant—measuring how closely their pupil size time series aligned with the group average (excluding themselves) for each video. We then tested whether these pupil synchrony scores predicted IET task performance, defined as the correlation between each participant’s ratings and the leave-one-out group average. Results revealed a significant positive relationship between pupil synchrony and valence rating accuracy (Pearson r = 0.25, p = 0.026), but not arousal accuracy (Pearson r = 0.12, p = 0.29). This suggests that individuals whose pupil fluctuations more closely mirrored the group were better at tracking emotional valence, but not necessarily emotional arousal. These findings support the idea that pupil dynamics reflect meaningful aspects of real-time emotion processing, particularly for valence. As a non-invasive and accessible physiological signal, pupil size may serve as a valuable marker for studying emotion perception in naturalistic contexts and populations with limited behavioral responses.

Acknowledgements: Supported in part by the National Institute of Health (grant no. R01CA236793) to D.W and (grant no. 1F99NS141343) to J.O.