WHY ARE CONJUNCTIVE CATEGORIES OVEREXTENDED

Recent experiments have suggested that subjects tend to overextend conjunctive categories (Hampton, 1988). We present a series of four experiments that confirm this finding. In Experiment 1, we use Hampton's response scale, on which subjects rate both membership and typicality. In Experiment 2, we find that subjects still overextend their categories when they judge membership alone. In Experiment 3, we introduce more response options and conclude that the tendency to overextend is not an artifact of an insufficient range of possible responses. We propose an explanation of overextension that we term the compensation hypothesis: The more categories that make up a conjunction, the more leniently membership is judged. We argue that this is a result of having to make "best fit" judgments for multiple constraints in real life. We test this hypothesis in Experiment 4, which uses conjuncts of three categories. As predicted, we find that overextensions are greater with triple conjunctions than in appropriate controls in which two categories are combined or categories are judged alone. We consider the theoretical implications of a compensation strategy for categorization.