Discrimination and Generalization in Identification and Classification: Comment on Nosofsky

Nosofsky's (1986) exacting tests corroborate fundamental aspects of Shepard's earlier theory of identification and classification. However, his results also appear to disconfirm two specific assumptions, which have hitherto received wide empirical support and which are deducible from a cognitive theory of generalization recently proposed by Shepard (1984) -namely, assumptions (a) that generalization falls off exponentially with psychological distance between stimuli, and (b) that psychological distance approximates the city-block metric if the dimensions of the stimuli are separable. It is conjectured that Nosofsky's discordant results may depend on his use of extensive discrimination training with highly similar stimuli and, therefore, reflect the limiting noise of discriminal processes more than the postulated cognitive mechanism of generalization.