Abstract
Craven & Foster (1992) found that observers could distinguish illuminant chromaticity changes from surface color changes in Mondrian scenes. The change in illuminant did not affect the photoreceptor excitation ratios across edges; they concluded that the visual system detected surface changes as ratio constancy violations. In 3D scenes, with multiple illuminants and matte surfaces at many orientations, changes in the positions or chromaticities of light sources need not leave edge ratios invariant. We examine how well observers can distinguish changes in light source position from matched changes in achromatic surface colors in such scenes. Stimuli: We rendered stereo pairs of scenes containing 4-sided achromatic pyramids with randomized heights and facet albedos under a combination of a collimated light source and a diffuse light source. On each trial, observers viewed a set of pyramids illuminated by a light source perpendicular to the ground plane for 1500ms, immediately replaced by a second scene that differed in collimated source direction or surface albedos. On light change trials, the collimated source could move 2, 4, 6 or 8 degrees. Surface change levels were matched in magnitude of space average and variance in color signals, as were ratio changes between adjacent facets. There were no visible cast shadows. Task: Observers judged whether a light or surface change had occurred. Analysis: We computed d' estimates for each of four light and surface change conditions (240 trials/estimate). Results: Three naïve observers completed the experiment. Estimates of d' increased significantly from a mean of 0.7 at 2 degrees to 2.6 at 8 degrees, significantly greater than zero even with a 2 degree positional change. Conclusion: In 3D scenes with non-homogeneous illumination, observers can discriminate very small changes in the spatial distribution of light from matched surface changes despite the lack of ratio constancy between adjacent surfaces.