Distinct pupil features correlate with between-participant and across-session performance variability in a 16-week, longitudinal data set

Abstract

Pupillary response provides a rich and accessible source of insight into cognitive processes due to its link with neural circuits in the brain. However, the relationships between pupil diameter and specific cognitive constructs are not yet dissociable because of the complex temporal dynamics of cognitive processes and individual variability. While extant research has identified a number of features that correlate with cognition, group-level analyses are typically used, making it unclear whether these relationships would hold between participants and upon repeated measurements. To gain insight into the robustness of pupil features and cognition, we analyzed a longitudinal pupillometry data set, characterizing relationships between standard pupil features and cognitive performance between participants (subject-level), across sessions (session-level) and within sessions (trial-level). Participants (N=26) completed 8 bi-weekly sessions of a mental arithmetic task in which participants indicated whether modular arithmetic statements were true or false under easy and difficult conditions. In the present analysis, we focused on the relationship between response speed and three pupil features typically studied in the literature: pre-stimulus baseline, peak amplitude, and peak latency. Mixed-effects model results indicated that the pre-stimulus baseline was associated with response speed at the subject-level, whereas peak amplitude was associated with the trial-level type (i.e. easy vs. difficult trials), but not trial-level performance. Peak latency was robustly associated with response speed at the group-, subject-, and trial-levels. We show that distinct pupil features correlate with performance at dissociable levels of analysis and may also reflect distinct influences of underlying cognitive processes.