Lawful kinematics link eye movements to the limits of high-speed perception

Abstract

Perception requires active sampling of the environment. What part of the physical world can be perceived is limited by the sensory system’s biophysical setup, but might be further constrained by the kinematic bounds of the motor actions used to acquire sensory information. Here, we tested this fundamental idea for humans’ fastest and most frequent behavior—saccadic eye movements—which entail incidental sensory consequences (i.e., swift retinal motion) that rarely reach awareness in natural vision. Using high-speed video projection, we display rapidly moving stimuli that faithfully reproduce, or deviate from, saccades’ lawful relation of velocity, duration, and amplitude. For each stimulus, observers perform perceptual tasks for which performance is contingent on consciously seeing the stimulus’ motion trajectory. We uncover that visibility of the stimulus’ movement is well predicted by the specific kinematics of saccades and their sensorimotor contingencies, reflecting even variability between individual observers. Computational modeling shows that spatiotemporal integration during early visual processing predicts this lawful relation in a tight range of biologically plausible parameters. These results suggest that the visual system takes into account motor kinematics when omitting an action’s incidental sensory consequences, thereby preserving visual sensitivity to high-speed object motion.

Publication
In Nature Communications

In this paper, we report a mysterious finding. When detecting rapid stimulus motion of a Gabor stimulus oriented orthogonal to its motion direction, it is not simply its absolute velocity that determines its visibility, but a combination of velocity and movement distance. Curiously, the specific combination that predicts velocity thresholds follows an oculomotor law - the main sequence, an exponential function describing the increase of saccadic velocity with growing amplitude. My proud contributions to this paper feature the masking experiment, the modeling of saccade trajectories which ultimately revealed significant correlations between saccade metrics and velocity thresholds, and most importantly, the early vision model to predict the measured psychophysical data - without fitting and based only on the trajectory of the stimulus. Finally, I evaluated the timing of the motion stimulus using photometric measurements using the LM03 lightmeter.

Richard Schweitzer
Richard Schweitzer
Postdoc in Vision Science

Passionate about psychophysics, eye tracking, M/EEG, and computational modeling.