Abstract
Objects hardly appear to change colour when the spectral power distribution of the illumination changes: a phenomenon known as colour constancy. Colour constancy could either be achieved by basing the perceived surface colour on properties that do not change with the illumination (such as spatial colour contrast), or by estimating and compensating for the chromaticity of the illuminant. The advantage of the second approach is that knowing the illuminant's colour can itself be useful. Here we examined whether subjects can judge the illuminant's colour well enough to account for their own colour constancy under the same conditions.
Subjects were very poor at matching the colour of the lamp illuminating a scene with the light from a single surface on a calibrated monitor. They were much better at choosing the sample that matches the colour of paint of a surface illuminated by the same lamp than one would predict from their judgments of the lamp's colour.
We conclude that colour constancy must be achieved by relying on relationships that are insensitive to the illumination, rather than by directly judging the colour of the illumination.