Recent views of conceptual structure. 论文

摘要

This article reviews theories of concept structure proposed since the mid-1970s, when the discov-ery of typicality effects led to the rejection of the view that instances of a concept share necessary and sufficient attributes. To replace that classical view, psychologists proposed the family resem-blance and exemplar views (and hybrids of the 2), which argue that instances of a concept share a certain level of overall similarity, rather than necessary and sufficient attributes. These similarity-based views account for much of the typicality data but fail to provide an adequate explanation of the coherence of conceptual categories and of various context effects. Recently proposed explana-tion-based accounts address these issues but raise further questions about the distinction between concept-specific information and general knowledge and about the relationship between concep-tual knowledge and various forms of inference. Psychologists have traditionally equated knowing the mean-ing of a word with knowing (or perhaps more accurately, having) the concept labeled by a word (e.g., Ogden & Richards, 1956; but see Clark, 1983). In this approach, a concept is assumed to be the mental representation of a category or class (Gleitman, Armstrong, & Gleitman, 1983; Medin & Smith, 1984). The con-tents of such a mental representation (i.e., the intension of a word), in concert with certain assumptions about how those contents are processed, have been taken to explain a wide vari-ety of phenomena, including people's knowledge of linguistic relations (e.g., synonymy, antynomy, hy ponomy), how people rec-ognize the objects, events, and so on properly labeled by the word (i.e., the extension of the word), how people understand novel combinations of the word with other words, and the infer-ences people are able to make about an object, event, and so on, properly labeled by the word (Johnson-Laird, Herrmann, &