Behavioral and Modeling Evidence that Eye Movements Bias Self-motion Perception

Abstract

To navigate the world, humans must integrate what they see with how they move. As the body moves, for example, the eyes rotate to explore the environment; these eye rotations, in turn, alter the visual signals used to judge body motion. Yet the practical impact of gaze dynamics on self-motion perception remains poorly understood. We tested how gaze position and gaze velocity shape self-motion perception--specifically heading direction--from visual signals in two behavioral experiments. In Experiment 1, we directly manipulated gaze position and velocity; in Experiment 2, a range of task demands evoked distinct gaze patterns. Across both experiments, heading estimates showed systematic, gaze-dependent errors: these estimates shifted toward the direction of eye motion and grew with horizontal gaze eccentricity and speed. A Bayesian ideal observer reproduced these error patterns across participants and tasks when it included three known features of visuomotor processing: (i) encoding retinal motion with eccentricity-dependent noise, (ii) underestimating eye-rotation speed, and (iii) a prior for moving straight ahead. These results reveal a lawful coupling between oculomotor behavior and heading perception. They further suggest that natural gaze strategies, such as keeping gaze near the optic-flow singularity and limiting pursuit speed, help mitigate gaze-dependent biases during navigation.