Abstract
What determines the colour appearance of lights and surfaces in a changeable visual environment? When the illumination on a scene changes, so do the visual signals elicited by that scene. During normal viewing, eye-movements additionally direct a succession of different images to the foveae. In spite of this flux in the visual input, the objects within a scene tend to remain constant in their apparent colour. In a series of experiments, we probe the mechanisms that govern colour appearance with particular focus on the way in which surface colour is extracted from a retinal input that changes over time. Our primary measures of colour appearance are the locations and steepnesses of boundaries between colour categories. On each trial, the observer is asked to classify the appearance of a test patch (e.g., as either “red” or “green”, or as “yellow” or “blue”). Colour boundaries are derived from repeated classifications, each obtained under precise control of the timing of the stimulus and response. We manipulate (a) the history of samples at a test location (temporal context) and (b) the composition of the surrounding area (spatial context). By (i) using a regression analysis to relate the observers‘ decisions to the recent history of test samples, and (ii) manipulating the time-course of changes in context, we are able to characterize the time-course of the mechanism that maintains a perceptual record of the temporal context. With a presentation rate of one sample per second, observers‘ judgements are influenced by the previous 5 to 10 samples, and additionally modulated by spatial context. We also test whether observers are able to collate information over time separately for different spatial regions. With two independent streams of test samples, observers‘ judgements are consistent with separable tracking mechanisms for the two streams. The ability to keep track of changes in chromatic context is important in a world with multiple illuminants and distinct sub-populations of reflectances.