Abstract
To examine how color affects the identification of flavors, we created a new experimental design, measuring flavor identification while manipulating the probabilities that color cues are associated with flavor outcomes. Stimuli were mixtures with varying relative proportions of two fruit flavorants, either banana+peach (five subjects) or orange+watermelon (another five subjects). The mixtures could be colored either green or red, such that color was a reliable, although not perfect, predictor of the greater component in the mixture. In one condition, red and green were equally associated with their flavor outcomes, while in the other two conditions, one color or the other appeared more frequently. We found, in line with predictions, that subjects used color to improve identification rates. Moreover, also in line with predictions, flavor identification rates were associated with changes in the underlying joint probabilities of color cues and flavor outcomes in the three conditions. From the perspective of statistical learning, these results are consistent with the view that the effect of color on subject’s identification of flavor depends on contiguity (conjoint probability) or sensitivity (conditional probability of associated color, given specific flavor), rather than validity (conditional probability of associated flavor, given specific color).