This study investigates how a holistic color interval i.e. the nondirectional color difference between a pair of colors in a CIELAB uniform color space, influences perceived color harmony. A set of 1035 test color pairs displayed on a CRT was evaluated for the degree of harmony. These test color years consist of pairs combined from among the selected 46 test colors evenly distributed in color space. The subjects were asked to select their three preferred colors from these 46 test colors and then to evaluate the degree of harmony of the test color combinations. The color intervals (DeltaE(ab)*) of each test color combination were calculated and treated as values of a independent variable. In addition, the evaluated degrees of color harmony were considered as values of a dependent variable, in which statistical analysis confirmed the relationship: the degree of harmony is a cubic function of the color interval. Moreover; the plot of this relationship allowed us to identify four color intervals: roughly corresponding to the regions of first ambiguity, similarity, second ambiguity, and contrast in Moon and Spencer's model. However, our results indicated that Moon and Spencer's principles for classifying harmonious/disharmonious regions in ten,ls of the color interval for three color attributes-lightness, chroma and hue-may be inappropriate in predicting perceived color harmony. As for the color intervals between a pair of colors considered as a function of the three attributes, the interval for lightness may have a predominant effect on color harmony, expressed in terms of a cubic relationship Results of the study further demonstrated that the subject's choice of colors significantly influences perceived color harmony. (C) 2001 John Wiley & Sons, Inc.