Cue Preference in a Multidimensional Categorization Task

Many natural categories vary along multiple dimensions. The present studies address two ma in questions underlying categorization with multiple dimensions. First, how well can humans perform in a categorization task consisting of five categories varying along nine continuously valued dimensions? Second, what are the properties of the cues preferred by humans if not all the available cues are used? Remarkably, participants not only learned to distinguish among the five categories, but they also learned to do so using only the relevant dimensions. A satisficing model of categorization was best able to account for participants’ responses. In addition, in a cue preference task, the results showed that nearly all participants preferred to use the dimension with the greatest variance when the number of dimensions available was restricted, in accord with predictions made by the satisficing model.