Category representations and their implications for category structure

In a series of experiments and reanalyses of previous research, we tested the hypothesis that categories that are primarily represented by extrinsic features (i.e., those that are relations between two or more entities) would yield more graded structures than would categories primarily represented by intrinsic features (i.e., those features true of an item considered in isolation). These predictions were confirmed. Extrinsically represented categories showed (1) less agreement across subjects on membership judgments, (2) more graded membership in a membership judgment task, and (3) smaller differences between gradients of typicality and of membership judgments