Abstract
A common assumption is that people have a single set of stable color preferences. We investigated color preferences in different contexts to find out whether this assumption is justified and, if not, what sort of systematic regularities and/or differences might hold. We tested participants' preferences for 37 colors sampled systematically from the Munsell space: 8 hues (unique red, green, blue, and yellow plus their approximate angle bisectors of orange, purple, cyan, and chartreuse) at 4 brightness/saturation levels, plus 5 achromatic colors at corresponding lightnesses. Subjects rated their preference for each color in each of a variety of imagined contexts: room walls, couches, dress-shirts/blouses, t-shirts, etc., in addition to context-free ratings of pure color preference. Cross-context correlations for the 37 colors varied widely, but many of the differences were systematic. Regression analyses showed that models based on colorimetric variables (participant-rated redness/greeness, blueness/yellowness, lightness/darkness, and saturation) predicted, on average, 65% of the variance across contexts, but different variables were differentially important in predicting color preferences, depending on the task. For example, hue dimensions (blue/yellow and red/green) explained the majority of the variance for the context-less ratings of pure color, with little effect of either lightness or saturation. In all other contexts, however, one or both of the latter variables was important. For example, saturation accounted for the most variability in both wall and couch color preference, with a clear preference for less saturated colors in both tasks. Analyses of individual differences showed variability in the importance of these factors as well. Our results show that although people do not have a single set of stable color preferences across tasks, their preferences are nevertheless orderly and can be predicted from colorimetric variables.