Abstract
Purpose. Internal knowledge of the physical world may play an important role in interpreting ambiguous sensory information. Computational vision models often impose physical constraints, such as surface smoothness or object rigidity, in order to achieve a stable interpretation. Gravity is a powerful physical constraint, since all objects receive 1 g on earth. Does the human visual system use any knowledge of gravity when predicting time-to-contact (TTC) and point-of-contact (POC) of an approaching object? The present study examines how gravitational acceleration affects the spatio-temporal prediction of an approaching object in a 3D environment. Methods. In the experiments, a computer-generated spherical object was displayed on a large binocular-stereo screen system and projected from a distant point toward a subject along a simulated parabolic trajectory. The subject tried to catch the virtual object, after the object disappeared at 2.4 m from the subject. The timing and location of each catch were recorded by a position sensor. To examine the effect of gravity on visual motion prediction, the value of the gravitational acceleration was varied. Results. The gravitational acceleration internally assumed by the visual system was estimated from the measured position and timing data. The results indicate that the estimated value of an internal gravitational acceleration did not change much even as the simulated gravity changed significantly. Conclusions. The results suggest that the human visual system does not directly estimate gravitational acceleration from retinal images, but uses an internal representation of gravity for visual motion prediction.