Evoked potentials were recorded to the occurrence of a disparate stimulus in dynamic random dot stereograms. Seven adult males, all of whom had vision which was normal or corrected to normal, participated in the experiment. Subjects viewed 100 ms duration stimuli which embodied 30 arc min of either crossed or uncrossed disparity under four conditions of spherical overcorrection: -0.25, +1.0, +2.0, +3.0 dioptres. The first condition, essentially normal refraction, yielded reliable behavioural reports of the stimulus and clear evoked potentials to both crossed and uncrossed disparity. With increasing overcorrection the behavioural reports became less reliable, and the evoked potentials were degraded for both conditions of disparity. The responses to the crossed disparity condition, however, showed significantly less degradation in both behavioural and electrophysiological measures. The implications of this finding may be that there are separate cortical subsystems for the processing of crossed and uncrossed disparity and that the former is more robust under non-ideal viewing conditions.