Fig. 4.

Bars show the percentages of variance explained by the best-fitting Munsell models for pair preference (Experiment 1), pair harmony (Experiment 2), two-color similarity (Experiment 3), and figural color preference against colored backgrounds (Experiment 4). Stripes within each bar show the percentage of variance explained by each factor in the order with which they were entered in to the regression model (bottom to top). The sign before each term indicates whether the factor was positively or negatively weighted in the corresponding regression equation (e.g., “+∑Cool” indicates that the sum of the coolnesses of the component colors was positively related to rated preference, harmony, and similarity, whereas “-|∆Hue|” indicates that the absolute value of their difference in hue was negatively related to these ratings)