observation has led some researchers to formulate the abstract notion of a categoryprototype (the best exemplar of the category), and to argue that category membershiprequires a certain level of similarity to the prototype. Rather than being tightlybounded, categories thus seem to take the form of ‘fuzzy sets’, much looser groupingsthat depend upon what Rosch called ‘family resemblance’. Although Rosch herselfnoted that the notion of a prototype ‘is simply a convenient grammatical fiction’ (1978,p. 40), category members are commonly seen to be related because they are similar to aprototype rather than because they share a given set of defining features.