Abstract
Purpose: Scene geometry affects color appearance, but the range of scene factors that drive this interaction is not understood. We want to identify stimulus conditions that produce robust effects of the perceived depth of a surface on its color appearance. The broader goal is to then exploit these conditions to study the phenomenon in parametric detail. For this reason, we wanted to produce the effects using rendered stimuli displayed on a dual CRT haploscope, so that we would then have convenient software control over the stimulus configuration. Methods: Left and right eye images were generated using Radiance, together with custom software that enabled wavelength-by-wavelength rendering. The rendered images were displayed on a computer-controlled haploscope. We used a forced-choice procedure to measure the achromatic locus of flat test patches embedded at different depths in various contextual scenes, and sought conditions that led to a change in the achromatic locus with rendered depth. Results: Experiment 1 was a close replication of an experiment reported by A. Werner (Perception, 2006), in which the degree of color constancy decreased when the depth between a test patch and a background plane increased. Here we found a small effect of test patch depth on the achromatic locus; this effect was marginally significant (p < 0.10), and was only present for one of the two illuminant changes we studied. Experiments 2 and 3 studied more complex scenes, in which two contextual planes with different chromaticities were presented at different depths and the test patch was presented co-planar with one or the other. There was a fronto-parallel gap between the test patch and closer of the two contextual planes. In Experiment 3, ancillary cues were added to indicate an illumination gradient between the two contextual planes. In Experiment 2, there was no effect of test patch depth. In Experiment 3, there was a small effect at most, and this was not reliable from subject to subject. Conclusion: To date, we have found that obtaining large effects of test patch depth on color appearance is elusive, at least when the stimuli are presented using a haploscope based on conventional CRTs.