Abstract
Two kinds of constancy underlie the everyday experience of surface color: its constancy under changes in illuminant and its constancy under changes in position. These two constancies seem to place conflicting demands on the visual system: to both take into account the local chromatic context and yet discount it. An experiment was performed in which observers binocularly viewed computer simulations on a color monitor of two illuminated Mondrian patterns, presented side by side, each consisting of 49 (7 × 7) abutting 1-deg square surfaces each drawn at random from the Munsell set. The two patterns, each subtending 7 deg × 7 deg when viewed at 100 cm, were illuminated by different uniform daylights of 6700 K and 25 000 K. The central and one other square surface in one pattern were interchanged in the other pattern, where they were illuminated by an independent local illuminant instead of the uniform illuminant. The observer adjusted the local illuminant so that the two patterns appeared to comprise the same surfaces but uniformly illuminated by different lights. Patterns were presented continuously. The degree of observers' color constancy under simultaneous changes in illuminant and surface position was as good as under illuminant changes alone. This equality held with or without the presence of a bright surface in the scene that could have been used to estimate the change in illuminant. The visual system seems able to extract information about color relations over widely separated regions in an image under simultaneous changes in illuminant, and this information may be more important than local chromatic context in determining surface color perception.