Pupil Dynamics Reflect Behavioral Choice and Learning in a Go/NoGo Tactile Decision-Making Task in Mice

The eye's pupil undergoes dynamic changes in diameter associated with cognitive effort, motor activity and emotional state, and can be used to index brain state across mammalian species. Recent studies in head-fixed mice have linked arousal-related pupil dynamics with global neural activity as well as the activity of specific neuronal populations. However, it has remained unclear how pupil dynamics in mice report trial-by-trial performance of behavioral tasks, and change on a longer time scale with learning. We measured pupil dynamics longitudinally as mice learned to perform a Go/NoGo tactile decision-making task. Mice learned to discriminate between two textures presented to the whiskers by licking in response to the Go texture (Hit trial) or withholding licking in response to the NoGo texture (Correct Reject trial, CR). Characteristic pupil dynamics were associated with behavioral choices: large-amplitude pupil dilation prior to and during licking accompanied Hit and False Alarm (FA) responses, while smaller amplitude dilation followed by constriction accompanied CR responses. With learning, the choice-dependent pupil dynamics became more pronounced, including larger amplitude dilations in both Hit and FA trials and earlier onset dilations in Hit and CR trials. A more pronounced constriction was also present in CR trials. Furthermore, pupil dynamics predicted behavioral choice increasingly with learning to greater than 80% accuracy. Our results indicate that pupil dynamics reflect behavioral choice and learning in head-fixed mice, and have implications for understanding decision- and learning-related neuronal activity in pupil-linked neural circuits.