A model for stimulus generalization in Pavlovian conditioning.

Abstract

A selective review of experiments that can be said to demonstrate the effects of generalization decrement in Pavlovian condition is presented, and it is argued that an adequate theoretical explanation for them is currently not available. This article then develops a theoretical account for the processes of generalization and generalization decrement in Pavlovian conditioning. It assumes that animals represent their environment by a stimulus array in a buffer and that this array in its entirety constitutes the conditioned stimulus. Generalization is then held to occur whenever at least some of the stimuli represented in the array on a test trial are the same as at least some of those represented in the array during training. Specifically, the magnitude of generalization is determined by the proportion of the array occupied by these common stimuli during training compared to the proportion of the array they occupy during testing. By adding to this principle rules concerning excitatory and inhibitory learning, it is proposed, the model can explain all the results that were difficult for its predecessors to account for. A fundamental property of conditioned responding is that it may occur in circumstances that differ from those prevailing during acquisition. This phenomenon is referred to as generalization, and the term generalization decrement is used when this transfer is less than complete. These effects have been revealed with both Pavlovian and instrumental conditioning, using a variety of species and a wide range of conditioned stimuli (CS) and unconditioned stimuli (US). As iar as instrumental conditioning is concerned, a considerable number of studies have been directed at examining the factors that influence the magnitude of generalization from one situation to another (for reviews see Honig & Urcuioli, 1981; Mackintosh, 1974). These results also have been the focus of considerable theoretical analysis. In contrast, much less attention has been paid to the role of generalization in Pavlovian conditioning. Only a few studies have been concerned specifically with demonstrating the generalization decrement of a Pavlovian conditioned response (CR) as a result of modifying the properties of the CS (Hofiman & Fleshier, 1961; Hovland, 1937; Moore, 1972; Pavlov, 1927; Siegel, Hearst, George, & O'Neal, 1968). Furthermore, only a few theorists have been specifically concerned with the analysis of these and related effects (Hull, 1943; Pavlov, 1927). An alternative approach has been to show how a particular theoretical analysis of the Pavlovian learning process can be extended to account for generalization and generalization decrement (cf. Rescorla, 1976). The argument presented in this article is based on the premise that an adequate account for the effects of generalization must lie at the core of