Abstract
Compensation for a change in distance in the amount of perceived depth (depth constancy) has been found to be substantially smaller (35–40%) than compensation for the flatness of the frontal plane (∼100%) (Rogers and Bradshaw, Perception, 1994; Bradshaw et al. Vision Research, 1995). We attributed this to the fact that an explicit estimate of distance is not required for frontal plane scaling because local visual information (e.g. HSR = VSR2) could be used to solve the task (see also Gärding et al, 1994). However, this cannot be the whole story because manipulating vergence angle (which provides only distance information) also has a significantly greater effect on frontal plane scaling. There is an alternative possibility - poor depth constancy may be an artefact of the shape/depth tasks typically used. To investigate this possibility, observers were presented with large field (70deg × 50deg) stereoscopic images at 57 or 172 cm distances. Vergence and differential perspective cues were manipulated to depict surfaces in front of or behind the display screens. In a frontal plane task, observers adjusted the pattern of horizontal disparities until the surface appeared flat. As well as measuring depth and shape constancy using traditional methods, observers adjusted the slants of (i) a 3-D pyramid or (ii) a series of intersecting horizontal strips with alternating disparity gradients until they appeared to meet/intersect at 90deg. Constancy was close to 100% in the new tasks (and in the frontal plane task), suggesting that incorrect estimates of distance are not necessarily the cause of poor constancy.