Abstract
Several recent studies support the proposal that the visual system uses luminance modulations from eye movements to encode spatial information in the temporal domain. This proposal makes specific quantitative predictions about the dependence of visual sensitivity on the characteristics of the visual flow delivered by eye movements to the retina. Here we tested these predictions by measuring contrast sensitivity in human observers, while manipulating the visual input via gaze-contingent display control. Subjects (N=7) reported the orientation (±45 deg) of 16 cycles/deg gratings, while exposed to retinal image motions replicating the signals delivered by fixational eye drifts with different amplitudes. As predicted, sensitivity was directly proportional to the spatiotemporal power of the luminance flow released by eye movements at nonzero temporal frequencies. This is particularly striking given the complex, non-monotonic relation between drift amplitude and the power of luminance modulations. We then examined whether active changes in ocular drift affect sensitivity. Here subjects (N=5) were exposed to normal retinal image motion, and performance compared in the naturally occurring trials in which ocular drift was larger and smaller than average. We measured contrast sensitivity at two spatial frequencies that bear contrasting predictions (1 and 10 cycles/deg). A Brownian motion model of eye drift predicts that, as the amount of drift increases, the luminance flow contains more power at low spatial frequencies and less power at high spatial frequencies. In keeping with this prediction, trials with larger drifts were associated with better performance at 1 cycle/deg and worse performance at 10 cycles/deg. These results provide strong support to a theory of active space-time encoding. They show that oculomotor-induced luminance modulations drive visual sensitivity. They also raise the possibility that humans use eye movements to control the effective contrast of the stimulus on the retina.