Watercolor Spreading Quantified by Matching and Cancellation

Abstract

When a dark chromatic contour (e.g., purple) surrounds a lighter chromatic contour (e.g., orange), the lighter color will assimilate over the entire enclosed area. This is known as the Watercolor Effect (see Pinna et al., 2001). Here we quantified the effect by asking subjects to either match or cancel the induced color. Sinusoidal bichromatic contours of approximately complimentary hues were used to induce color into a 1.7 central patch. Three sets of color pairs were used and each color within a pair was used for both the outer and inner contour in separate conditions. The contours had a luminance of 20 cd/m2 (outer contour) and 55 cd/m2 (inner contour) and the background was white (CIE xy 0.30 0.33) with a luminance of 80 cd/m2 A second set of stimuli was tested in which all colors were nominally isoluminant. Three color-normal subjects adjusted the chromaticity and luminance of a matching field so that it appeared identical to the induced color. In a second set of conditions, the subject's task was to cancel the induced color. There were small but reliable shifts in color measured using both tasks. The mean shifts in CIE u′v′ color space for the matching and cancellation tasks were generally in opposite directions and close to the directions in color space of the contours defining the stimulus. The magnitude of the color shifts was similar for all color directions tested. When the stimuli were isoluminant, the chromatic shifts were reduced substantially suggesting that luminance gradients play an important role in the Watercolor effect.

NIH A604058, Research to Prevent Blindness, Fondation des Aveugles de France