Is there stereoscopic depth constancy?
Although the present findings support our claim that binocular stereopsis can support more accurate depth recovery than can structure from motion with geometrically similar displays, there exist reports of the failure of depth constancy in the literature that would seem to contradict our conclusions regarding the accuracy of depth from stereopsis generally (e.g., Tyler, 1983; Collett, Schwarz, and Sobel, 1991). Perhaps the clearest case of the failure of depth constancy is that demonstrated by Johnston (1991). She presented observers with random dot stereograms which simulated elliptical cylinders with their principal axis horizontal in the picture plane. The subjects’ task was to judge whether a presented cylinder appeared “squashed” or “stretched” in depth (relative to an imagined circular cylinder). Such data can be interpreted as a measure of a perceived depth to base ratio. The task was performed at three viewing distances, using five cylinder heights (vertical diameters). Her res
