ABSTRACT
Many studies of category learning have emphasized a single role of the concept that is learned -namely, the concept as a mechanism for classifying objects and discriminating them from members of other categories. Recently, researchers have noted that concepts have many purposes besides classification-prediction, communication, explanation, goal attainment, and so on. This paper presents a study that varied the roles of concepts during a classification learning task. Specifically, one group of subjects (the discrimination group) was given standard instructions to learn about pairs of categories. A second group of subjects (the goal group) was given these instructions but also informed about the functions of the categories. The results of the study suggest that the two groups formed different concepts, even though they saw the same examples of the categories. The concepts of the discrimination group were based on those features in the examples that had predictive value-features with high cue and category validity. In contrast, the concepts of the goal group were based on predictive features and features that were important to the function of the category (called “core” features). Relative to the discrimination group, the goal group placed less emphasis on predictiveness. The results are discussed in terms of their implications for standard classification tasks in psychology and explanation-based and similarity-based approaches in machine learning.
