Overextension of conjunctive concepts: Evidence for a unitary model of concept typicality and class inclusion.

Abstract

Four experiments investigated how people judge both the typicality and membership of items in conjunctive concepts such as school furniture or sports which are games. Judgments of membership in conjunctions were overextended, and there was asymmetry between the constituent concepts in their influence on relative conjunctive concept membership. The results are discussed in the light of recent theoretical disputes about the modeling of concept representations and the process of forming conjunctions (Cohen & Murphy, 1984; Osherson & Smith, 1981, 1982; Smith & Osherson, 1984). A theory is proposed in which constituent intensions are combined to form a composite prototype for the conjunction. Membership in both single and conjunctive concepts is then determined in the same unitary fashion, by placing a membership criterion on the perceived similarity of possible exemplars to the prototype. An important issue in the study of natural language concepts is the way in which common semantic concepts combine to form conjunctions. For a wide range of concepts, two very reliable phenomena have been established: One, members of concept categories vary in their representativeness, and two, for many concepts the boundary around the class of concept members is unclear or fuzzy. However, to date few empirical studies have considered what happens when two of these fuzzy concepts are placed in conjunction. Do conjunctions show similar typicality and fuzziness phenomena? If so, how can they be related to typicality in the constituent concepts? Do fuzzy concepts follow the same logical rules for conjunction as well-defined concepts? The first aim of this article is to provide empirical evidence that may begin to answer these questions. The current interest in concept conjunctions comes largely from their relevance to the basic question of concept definitions: that is, how a concept picks out an extensional set of concept members. A second aim of this research is therefore to use the study of conceptual combination to provide important constraints on models of conceptualization. In particular, it will be argued that evidence on conjunctions can have theoretical implications for distinguishing the following two opposing accounts of concept definitions.