Abstract
Stereopsis arises from the successful matching of corresponding features in the two retinal images. Recent studies, however, have shown that unmatched features can also play a significant role in this process. The present study demonstrates that stereoscopic depth can be perceived without any point-by-point retinal correspondence. The stimuli represented vertically oriented wedges that appeared either concave (central edge backward) or convex (central edge forward). These wedges were briefly presented behind a specially designed occluder (the crooked window). Observers fused the binocular display with a modified Wheatstone stereoscope. Various disparities were introduced at the central edge to define the convexity and sharpness of the wedge. The crooked window occluded the central edge in such a way that it was visible only partially and in a mutually exclusive manner in the two eyes. The edge, and the wedge that it defined, could therefore only be perceived amodally. For each trial, the edge was randomly located along the wedge and the latter was also randomly located behind the occluder in order to eliminate any possible monocular cues. Observers had to report whether the wedge was perceived as concave or convex. The frequency of correct responses increased with the disparity of the central edge. Accuracy was defined as the slope of the derived psychometric function. In a control condition, the left and right images were interchanged so that disparities indicated that the wedge was in front of the occluder (invalid occlusion). Although the overall available information was identical to that of the original (valid occlusion) condition, the observers performed at chance level. These results demonstrate that the visual system can extract the three-dimensional structure of objects in the absence of point-by-point retinal correspondence. This experiment substantiates the view that surface recovery may precede binocular matching in stereoscopic processes.
Supported by Human Frontier Science Program RG0109/1999-B.