Sean R. O'Bryan¹ (), William Kemball-Cook¹, Joo-Hyun Song¹; ¹Brown University

Task-evoked pupillary responses can provide a useful biomarker of dynamic internal cognitive states such as effort, attentional control, and surprise. These responses are typically quantified by subtracting pre-trial baseline pupil diameter (PD) from peak PD following a stimulus or motor response, corresponding to response amplitude. However, other physiological components such as baseline PD, dilation velocity, and peak latency all affect the shape of the canonical pupillary response, and moreover, each of these components can directly affect amplitude estimates. Here, we systematically tested the relationships between each of these measures by analyzing pupillometry data across three independent cognitive control experiments: a Stroop task (N = 87), a visual working memory task (N = 74), and a visuomotor adaptation task (N = 30). Furthermore, we examined how and whether each PD measure tracked cognitive load and individual differences in performance. Notably, subject- and trial-level correlation analyses revealed that measurements of amplitude were highly (negatively) correlated with baseline PD, and that the strength of these associations may lead to collinearity issues in studies seeking to separately test the contributions of PD amplitude and baseline to behavior. We also found that although amplitude and velocity measures were positively correlated, velocity explained less than 20% variance in amplitude at the trial level, such that velocity and amplitude measures are not interchangeable. Surprisingly, model comparison revealed that peak latency (i.e., time elapsed between stimulus onset and maximum PD) provided the best fit accounting for both task demands and individual differences in behavior across all three experiments: PD latency explained Stroop interference effects, accounted for load- and capacity-related differences in working memory, and tracked reach error under perturbed feedback conditions in visuomotor adaptation. Together, these results represent a critical step toward better understanding the capabilities (and limits) of pupillometry as a tool in the cognitive sciences.

Acknowledgements: NSF BCS-1555006