Abstract
During natural viewing, the stimulus on the retina is never stationary, as small eye movements occur even during visual fixation. It is remarkable that we perceive a stable scene and do not see the fixational motion of the retinal image. When an object moves by the same amount, the resulting displacement of the retinal stimulus is instead clearly visible. How does the visual system identify and discard the fixational motion of the retinal image? It has long been debated whether such cancellation relies on information about the movement and/or position of the eye or whether eye motion is inferred directly from the visual signal. Retinal stabilization, a procedure in which fixational motion is eliminated, is a powerful technique for testing between retinal and extraretinal theories, as it decouples the motor signals related to eye movements from their associated visual changes. Here, we report the results of experiments in which we examined the influence of fixational eye movements on motion detection. In a forced-choice discrimination task, subjects reported whether a small, bright dot, briefly displayed on a CRT, was stationary or drifted with uniform motion at 30'/s. We compared performance measured during the normal instability of visual fixation to performance obtained under retinal stabilization. Whereas, in the normal condition, the stimulus was stationary or moved with uniform motion on the screen, in the stabilized condition, it was translated under real-time computer control to compensate for the subject's eye movements, so that it remained immobile or followed a linear trajectory on the retina. We show that the motion of the retinal image resulting from fixational eye movements is inferred from the visual signal and that no extraretinal information is used for this purpose. These findings provide support to Koffka's (1935) original proposal that spatial localization occurs within a framework.
Supported by NIH EY18363 and by NSF BCS-0719849.