Task-evoked pupillary responses track precision-weighted prediction errors and learning rate during interceptive...

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Human-Computer Interaction

Background:

• Active inference theory posits perception and action minimize sensory surprise.
• Oculomotor learning, particularly anticipatory eye movements, is crucial for tasks like hand-eye coordination.
• Understanding the physiological encoding of surprise is key to explaining learning and action control.

Purpose of the Study:

• To investigate the link between physiological surprise encoding and anticipatory eye movement learning.
• To test if eye movement updates align with Bayesian principles during a virtual reality interception task.
• To determine if pupil dilation, a marker of surprise, correlates with learning rates and prediction errors.

Main Methods:

• A hierarchical Bayesian inference model was fitted to anticipatory eye movement data.
• Participants performed a virtual reality interception task with predictable and unpredictable ball bounce profiles.
• Task-evoked pupil responses were recorded and analyzed alongside eye movement data.

Main Results:

• Pupil dilation tracked prediction errors and learning rates, but not beliefs about ball trajectories or environmental volatility.
• Anticipatory eye movements were updated in line with Bayesian principles, reflecting learned expectations.
• The study successfully estimated individual prediction errors and belief updating rates.

Conclusions:

• Findings partially support active inference models by linking pupil dilation to surprise encoding during learning.
• Physiological responses, specifically pupil dilation, serve as a marker for prediction errors and learning rates.
• This research elucidates how the brain's surprise encoding mechanisms shape the control of goal-directed actions.