Abstract
Goal: Pupil size is an easily accessible, noninvasive online indicator of various cognitive processes. Pupil measurements can reveal continuous processing dynamics throughout an experimental trial, including anticipatory responses. However, the relatively sluggish (~1 s) response dynamics of pupil dilation makes it challenging to connect changes in pupil size to events occurring close together in time. Researchers have begun to use models to link changes in pupil size to specific trial events, but such methods have not been systematically evaluated. Here we developed and evaluated a model-based procedure that estimates pupillary responses to multiple events within an experimental trial. Methods: The mean pupil area timeseries across trials was modeled as a linear combination of pupil dilations to each trial event (general linear model). We evaluated the model using a sample dataset in which multiple sequential stimuli –precue, two sequential grating stimuli, and response cue– were presented within 2-s trials. We compared alternative models to determine model parameters, performed parameter recovery to validate fitting procedures, and used bootstrapping to determine the reliability of parameter estimates for our sample dataset. Results: We found that the best model provided robust estimates of pupil response amplitude and latency even for trial events separated by only 250 ms. Importantly, two timing parameters not previously modeled –pupil response latency and time-to-peak– improved fits. Estimated response latencies indicated anticipatory pupil responses to predictable trial events. Pupil response dynamics varied substantially across observers but were consistent for a given observer. Conclusions: A general linear model with specific parameters can estimate separate pupil responses to events in a rapid sequence for each individual. We provide our pupil modeling pipeline as a freely available software package (Pupil Response Estimation Toolbox, PRET) to facilitate the estimation of pupil responses and the evaluation of the estimates in other datasets.
Acknowledgement: NIH NEI R01 EY019693, F32 EY025533, T32 EY007136