The Classical View of Categorization

Traditionally, categories were assumed to be isomorphic with respect to the definite structures, properties, and interrelationships of the "real" world. An old and influential idea is that the properties contained in these categories are singly necessary and jointly sufficient to define membership. Although other properties may be allocated randomly, natural categories will always include a set of "essential" or "criterial" features. 

Categorization refers to the extraction of these "critical" features by the application of a predefined interactive algorithm of the mind. According to this classical or "Lockean" position, categorization is conceptual; it is an inferential process that works on the arbitrary perceptual categories of the world. Needless to say, such categorizing powers were assigned to the human mind alone.

In the midst of our century, the dominance of the classical view has been seriously undermined by the collective work of philosophers (e.g. Wittgenstein, 1953; Ryle, 1949), linguists, and psychologists (Rosch, 1975; Rosch, 1978; Rosch & Mervis, 1975). Wittgenstein, for example, reflected on the polymorphism of common-sense language concepts, while Rosch and colleagues were able to empirically support the concept of family resemblance by testing the classification performance of humans with basic level concepts. The probabilistic or fuzzy view of categorization that emerged in this study was thought to be a means of grouping objects that are characterized by large clusters of attributes. These attributes do not occur in an all-or-none fashion, but recur approximately in subsequently perceived instances and are oftly perceptually salient.

Comparative psychologists and ethologists studying the conceptual behavior of animals do not believe in any form of the classical view. Categorization is assumed to be perceptual, and all animal experimenters have denied that the boundaries of natural categories are clear-cut, that membership can be defined by necessary and sufficient conditions, and that logical rules guide the grouping of objects or events. Even for humans, the available evidence suggests that group membership is graded and assessment is based upon characteristic or probable properties alone.