Integrating Behavioral and Biological Models of Classical Conditioning

This chapter presents several classical conditioning models within a common formalism based on adaptive networks for associative learning. Behavioral models can serve as the concise embodiments of multiple constraints imposed by the behavioral phenomena on a biological model. The key ideas are embodied by three influential models of classical conditioning within this adaptive network formalism: (1) the Rescorla–Wagner model, which can be taken as a version of the least mean squares (LMS) algorithm of adaptive network theory and shown to account successfully for a variety of phenomena of stimulus selection, as well as conditioned inhibition in the Pavlovian conditioning literature, (2) the priming model developed by Wagner and colleagues, which describes the ways in which unconditioned stimulus (US) processing can be modulated by antecedent stimuli, and (3) the sometimes opponent process (SOP) model that is a quantitative, real time, connectionist model, which addresses a broad range of phenomena in the Pavlovian conditioning and habituation literatures. The chapter also focuses on neural circuits in the Aplysia and mammalian cerebellum, which have been identified as critical to the learning and expression of classically conditioned behaviors.