The limits of perceptual robustness in perspective distortion

Pictures often do not appear distorted even when viewed at oblique angles. Three hypotheses have been put forth to explain this robustness of virtual space toward affine transformations. First, array specificity holds that the perception of depicted space is fully specified by the information available at the point of observation. Second, the notion of a compensatory mechanism involves an unconscious recreation of the scene according to the original viewpoint. Third, the indiscrimination hypothesis denies the ability of the visual system to resolve or detect affine transformations up to a certain degree. Three experiments were conducted to investigate these claims. Using a double projection technique devised by Cutting (1987), Experiment I showed that observers are able to discriminate and compensate, to some degree, for affine transformations if information about the projection surface is available. However, observers relied on relative image velocities rather than reconstructing the object. In Experiment 2 additional observer motion was simulated. In single and double projection trials that required more difficult judgments of object rotation, compensation was poor and observers seemed to rely on local cues. Finally, real and simulated rotation of the projection surface revealed that observers are able to compensate for only one primary projection surface slant. The results reject the indiscrimination hypothesis and support the notion of array specificity.