Abstract
One can alter the colour appearance of an object either by painting it or by changing its illumination. Both material and lighting changes can result in a change of hue. We report on an experiment which shows that “material” hues are different from “lighting” hues. Two identical sets of Munsell papers (5R4/14, 5YR7/12, 5Y8/12, 5G6/10, 10BG5/8, 5PB5/12 and 10P5/12) were presented in two displays. In separate sessions of the experiment, the displays were illuminated independently by one of five lights: red, yellow, green, blue and purple, giving a total of 15 possible illumination conditions (red-red, red-yellow, etc). The lights were approximately equiluminant with CIE 1976 u′v′-coordinates (0.382, 0.488), (0.199, 0.530), (0.127, 0.532), (0.183, 0.210), and (0.259, 0.365). Dissimilarity judgments were made between papers in the two displays (as in asymmetric colour matching). Each pair was evaluated 6 times by ranking. As a standard pair, the paper 5Y8/12 lit by the yellow light and the paper 5PB5/12 lit by the blue light were presented at all times during the experiment to indicate the maximal rank. Two trichromatic observers participated in the experiment. The dissimilarity judgements were analyzed by using a non-metric multidimensional scaling technique. The output configuration was of a slightly distorted torus-like pattern (“doughnut”). When one changes the material (reflectance) property moving from paper to paper under the same light, one travels the circumference of the doughnut (referred to as material hue). When one changes the lighting property, moving from light to light for the same paper, one travels the cross-sectional circle of the doughnut (referred to as lighting hue). Thus, the material and lighting hues are found to be dissociated in the dissimilarity space. We conclude, contrary to general belief, that the manifold of object-colour hues is two-dimensional, being topologically equivalent to a torus.