Recently, a motion-first and a disparity-first Bayesian model have been proposed to predict bias in binocular 3-D motion perception (Lages, 2005). Here it is tested whether a motion-stereo model that combines uncertainty in velocity and disparity processing with suitable priors can explain perceptual bias better.

Model fits were compared in an experiment where binocular disparity change and interocular velocity difference served as cues for the perception of 3-D motion. Stimuli were presented to the left and right eye on a calibrated flat CRT monitor in a split-screen Wheatstone configuration with a refresh rate of 120 Hz. On each trial observers verged on a fixation cross flanked by a fusion lock at 114 cm before target dots above and below fixation traveled on parallel trajectories for 0.833 sec. 3-D Trajectory angle (0 to 360 deg in steps of 10 deg) and velocity (0.02, 0.03, and 0.04 m/sec) of the targets varied in randomly intermixed trials. In separate blocks of trials four observers indicated trajectory angle and distance traveled by adjusting on screen a line probe in top view.

The results indicate that the stereo-first model best explains systematic bias in perceived trajectory angle and velocity. Uncertainty estimates of the stereo-first model systematically increased with stimulus velocity. When observers adjusted line probes to static targets at corresponding 3-D motion endpoints then uncertainty estimates and bias were generally reduced. These observations support the notion of a stereo-first system in which disparity change introduces uncertainty and bias.