The effective use of stereoscopic display systems requires the study of the perception in stereopsis especially to represent the physical objects as the realistic virtual scene. This paper investigated four factors that affect depth and size perception in stereoscopic displays generated by the binocular disparity principle; these factors are viewing distance, stimulus size, magnitude of disparity, and direction of disparity (crossed and uncrossed). We measured the magnitude of depth and size that the participants perceived under various combinations of levels of these factors and compared those with the depth and size predicted by the geometric model of stereopsis. The depth and size perceived became closer to prediction, as the viewing distance decreased or the stimulus size increased. In the aspect of disparity, the depth and size perceived became closer to prediction, as the magnitude of disparity decreased with the crossed disparity. The compensatory prediction model, which is the useful implement to depth and size scaling in stereopsis, was suggested to increase the accuracy of prediction in depth and size perception and compensate for the difference between perception and prediction by the geometric model of stereopsis.