Journal of Memory and Language 44, 618–643 (2001) doi:10.1006/jmla.2000.2748, available online at http://www.academicpress.com on 0749-596X/01 $35.00 Copyright © 2001 by Academic Press All rights of reproduction in any form reserved. 618 The Role of Contrast Categories in Natural Language Concepts Timothy Verbeemen, Veerle Vanoverberghe, Gert Storms, and Wim Ruts University of Leuven, Leuven, Belgium In this paper, seven experiments are described in which the effect of contrast categories on the within-category structure of superordinate and basic level natural language concepts was studied. Intension-based and extension- based predictors originating from both the target category and a contrast category were used to predict typicality ratings and response times in two different speeded categorization tasks. Virtually no evidence for contrast cate- gory effects was found in any of the experiments. These findings contrast with what one would expect based on results reported by Rosch and Mervis (1975) and on many exemplar models, in which a contrast category effect is explicitly assumed. © 2001 Academic Press Key Words: natural language concepts; contrast categories; basic level concepts; superordinate level concepts. Since the growth of cognitive psychology in the 1970s, natural language concepts have been studied extensively. (For overviews of the litera- ture, see Komatsu, 1992; Smith & Medin, 1981.) One of the most important findings in these studies is that concepts are not defined accord- ing to a set of features that are singly necessary and jointly sufficient. Numerous experiments have demonstrated that concept membership is graded and that there is a stable within-category structure, usually described as the typicality gra- dient. Subjects can rate typicality in a very reli- able way for most natural language concepts, and rated typicality is influential in a wide vari- ety of cognitive tasks (Hampton, 1993; Malt & Smith, 1984). It predicts variables such as in- ductive inference (Osherson, Smith, Wilkie, Lopez, & Shafir, 1990; Rips, 1975), production tasks (Hampton & Gardiner, 1983), priming ef- fects (Rosch, 1975), semantic substitutability (Rosch, 1977), and memory interference effects (Keller & Kellas, 1978). Another measure of within-category structure is response time (RT) in speeded categorization tasks. In such tasks, participants are asked to de- cide, as quickly as possible, whether a presented stimulus word belongs to a target category. The advantage of RT as a within-category measure is that, unlike with typicality ratings, participants cannot deliberately bias RT. The disadvantage, however, is that the measure is usually less reli- able due to fluctuations in the attention of the participants. RTs have often been used as a meas- ure of within-category structure (e.g., Hampton, 1979; Larochelle & Pineau, 1994; Smith, Shoben, & Rips, 1974). Different models have been proposed to ac- count for the internal structure of concepts. These models can be divided into two classes, models primarily based on the intension of the concept and models primarily based on the ex- tension of the concept. An example of the former is Hampton’s (1979) feature-based prototype model, which assumes that a concept is repre- sented by a set of characteristic features that are stored with the concept label. Hampton studied eight concepts and showed that this model yields good predictions for typicality ratings and for RTs in a speeded categorization task. Extensional models, on the other hand, assume that concepts are represented by means of exem- plars (e.g., Medin & Schaffer, 1978; Nosofsky, 1986). Most evidence for exemplar models This project was supported by Grant G.0353.99 from the Science Foundation Flanders to Gert Storms and Paul De Boeck. We thank Paul De Boeck, Rob Goldstone, James Hampton, Lloyd Komatsu, Lance Rips, and one anonymous reviewer for their useful comments, and Michel Meulders for his help in the power analysis. Address correspondence and reprint requests to Gert Storms, Psychology Department, University of Leuven, Tiens- estraat 102, B-3000 Leuven, Belgium. E-mail: Gert.Storms@ psy.kuleuven.ac.be.