Combining pupillometry and drift-diffusion models reveals auditory category learning dynamics

Pupillometry and drift-diffusion models (DDM) have emerged as powerful tools to offer insights into learning processes and decisional dynamics. Specifically, pupillary dilation can serve as a metric of arousal and cognitive processing, and DDMs examine processes underlying perceptual decision-making using behavioral accuracies and response times. Methodological constraints have complicated the combination of the two methods in the study of learning. DDMs require precise response times, yet pupillary responses are slow and are impacted by motor movements. Here, we developed a learning task that separately optimized measurement of behavioral response times and pupil dilation during learning. We modified a standard learning paradigm to include trials with and without timing delays between stimulus presentation and response. Delayed trials optimized measurement of the pupil, while immediate trials optimized behavioral response times. We tested whether decision-making processes estimated with DDMs could be recovered from delayed trials where pupil response measurement was optimized. Our results indicated that pupil responses on delayed trials showed distinct markers of learning that were not present on immediate trials. DDM parameters from delayed trials also showed expected trends for learning. Finally, we demonstrated that DDM parameters elicit differential pupillary responses. Together, these results indicate that learning dynamics and decisional processes can be decoded from pupillary responses.