Abstract
Humans use differences in the visual input to the two eyes to extract depth information, a process known as stereopsis. The visual system is exquisitely sensitive to these differences and able to detect disparities smaller than the photoreceptor spacing. This accomplishment is even more remarkable considering that the eyes drift incessantly during fixation, resulting in retinal image motions that are largely uncorrelated in the two eyes. Since these drifts cover tens of photoreceptors and continually change the correspondence between retinal points, one might expect they would hinder stereopsis. However, we hypothesized that the temporal modulations they produce are in fact beneficial. To test this, we examined the consequences for stereopsis of the retinal image motion resulting from eye drifts. Subjects (N=7) were asked to discriminate the orientation (±10deg) of a sinusoidal depth corrugation (1 cycle/deg) created by modulating the disparity of random dot stereograms. Combination of binocular high-resolution eye-tracking with real-time gaze-contingent display enabled precise control of the retinal stimulus delivered to each eye. We show that fixational eye movements are beneficial to stereopsis. In all observers, stereoscopic discrimination was greatly impaired when the images in both eyes were continually adjusted to counteract the visual consequences of eye drifts and eliminate retinal image motion (a three-fold difference in d’; p=0.03, signed-rank test). To determine whether this effect originated from a monocular reduction in contrast sensitivity or temporal modulations in binocular disparity, we selectively eliminated the version and vergence components of retinal image motion. The resulting signals conveyed similar power in their luminance modulations but only vergence motion affected disparity. Discrimination was normal in the presence of vergence motion and greatly impaired in its absence. These findings extend dynamic theories of vision to depth perception. They suggest that stereoscopic perception relies on transient disparity signals produced by fixational vergence eye movements.