Abstract
The perception of shape from binocular disparities is often not veridical (e.g. Johnston, 1991). We investigated the generality of this result for objects which varied in size and surface shape. Observers performed a shape judgement task for 3 differently shaped surfaces and 5 different sizes. Stimuli were random-dot stereograms. The shapes were a) elliptical hemi-cylinders, b) triangular ridges and c) frontoparallel rectangles, all horizontally oriented against a random dot background. All had a constant width but varied in height. Stimuli were presented with a range of depths and observers were required to judge whether the depth was less than or greater than the half-height. For the hemi-cylinder this was equivalent to judging whether it appeared squashed or stretched compared to a circular cylinder. For the triangular ridges this was equivalent to the task of judging whether the apex was greater than or less than 90 degrees. The point of subjective equality (PSE) was obtained for each shape at each size. When comparing across shapes, it was generally found that the disparity-defined depth at the PSE increased in the order: frontoparallel rectangles, triangular ridges, cylinders. In other words, cylinders were perceived as the flattest, then the ridges and the rectangles were perceived as having the greatest depth. Comparing across sizes showed that in most cases the PSE was larger for the bigger objects than the smaller objects, indicating that larger objects are perceived as flatter than smaller objects. Thus both size and shape influence the interpretation of disparity. Our results suggest a number of factors that affect performance in this task: availability of disparity gradients, cue combination with conflicting texture cues, the misperception of half-height with changing height:width ratios and individual differences in subject strategy.