Abstract
By determining boundaries between colour categories, we tracked changes in colour appearance as a function of the simulated illumination, and of the reflectance composition of the background. A 3-degree square test-patch was presented on a variegated background of random ellipses. Test materials were simulated with reflectance spectra chosen from measurements of natural and man-made objects. Background ellipses were randomly assigned reflectances from one of three subsets: neutral, red-biased, and green-biased. For illuminants we used the spectra of direct sunlight and zenith skylight. For each background-illuminant condition, observers categorized test materials as either reddish or greenish, or as yellowish or bluish, in separate sessions. Under prolonged adaptation to a single illuminant, observers demonstrated a high degree of classical, appearance-based colour constancy: a change in illuminant produced a substantial change in the chromaticity that elicited the percept of neither red nor green, while classification of materials was largely unaffected. Secondly, the chromatic bias of the background had only a small effect on the classification of test materials. Our third and critical manipulation was to use different illuminants for the test and background. Under these conditions, the spatial context provides information about the background-illuminant, but in single trial, the observer has no information about the test-illuminant. This is available only by collating information over time. Observers continued to demonstrate reasonable colour constancy. The percentage of materials near the border that retained their original classification can be used as an index of relative performance across our three manipulations: 80% under a simple illuminant change, 76% under a background change, and 69% under an illuminant change on the test only, which is significantly different from that expected if the visual system discounted the (conflicting) background illuminant. These data suggest that temporal and not only spatial context is important in colour constancy.