Abstract
A substantial body of evidence supports the long-standing hypothesis that the visual system uses luminance modulations from eye movements to encode spatial information in the temporal domain. For simplicity, eye movements are commonly assumed to translate the image across the retina consistently, yielding luminance modulations with uniform statistics across the visual field. However, in reality, the optics and kinematics of the eye interact to yield a more complex pattern of image motion. Here we show that these considerations may have functional consequences for neural encoding and emmetropization. Using a detailed optical model of the rotating eye with a non-spherical retina, we examined two conceptually separate but geometrically related factors that contribute to retinal image motion: (1) Optical distortion, the mapping of visual space onto the retina, which is influenced by the eye’s refracting properties; and (2) motion transfer, the amount of motion at each point of the retinal image, which depends on the eye’s center of rotation. In an emmetropic eye accommodated to infinity, we find that retinal image motion increases nonlinearly across retinal eccentricity, with ~30% greater retinal image speed in the periphery compared to the fovea. This effect implies that the characteristics of luminance modulations induced by eye movements also vary with eccentricity. Specifically, during fixation, luminance modulations will deliver increasingly more power at low spatial frequencies as eccentricity increases. This transformation is altered in non-emmetropic eyes due to differences in eye shape, with each diopter of spectacle refraction resulting in a ~3% change in peripheral retinal image motion. Since there is greater motion in hyperopic eyes and less motion in myopic eyes relative to emmetropic eyes, the statistics of image motion provide a cue to the sign of blur, which can be accessed from the temporal content of neural signals.