Neural dynamics of attentionally modulated Pavlovian conditioning: Conditioned reinforcement, inhibition, and opponent processing

Abstract

A real-time neural network model is developed to explain data about the acquisition and extinction of conditioned excitors and inhibitors. Systematic computer simulations are described of a READ circuit, which joins together a mechanism of associative learning with an opponent-processing circuit, called a recurrent gated dipole. READ circuit properties clarify how positive and negative reinforcers are learned and extinguished during primary and secondary conditioning. Habituating chemical transmitters within a gated dipole determine an affective adaptation level, or context, against which later events are evaluated. Neutral conditioned stimuli can become reinforcers by being associated either with direct activations or with antagonistic rebounds within a previously habituated dipole. Neural mechanisms are characterized whereby conditioning can be actively extinguished, and associative saturation prevented, by a process called opponent extinction, even if no passive memory decay occurs. Opponent extinction exploits a functional dissociation between read-in and read-out of associative memory, which may be achieved by locating the associative mechanism at dendritic spines. READ circuit mechanisms are joined to cognitive-emotional mechanisms for associative learning of conditioned reinforcers and of incentive motivation, and to cognitive—in particular, adaptive resonance theory—mechanisms for activating and storing internal representations of sensory cues in a limited-capacity short-term memory (STM); for learning, matching, and mismatching sensory expectancies, leading to the enhancement or updating of STM; and for shifting the focus of attention toward sensory representations whose reinforcement history is consistent with momentary appetitive requirements. This total neural architecture is used to explain conditioning and extinction of a conditioned excitor; conditioning and nonextinction of a conditioned inhibitor; and properties of conditioned inhibition as a “slave” process and as a “comparator” process, including effects of pretest deflation or inflation of the conditioning context, of familiar or novel training or test contexts, of weak or strong shocks, and of preconditioning unconditioned-stimulus-alone exposures. The same mechanisms have elsewhere been used to explain phenomena such as blocking, unblocking, overshadowing, latent inhibition, superconditioning, inverted U in conditioning as a function of interstimulus interval, anticipatory conditioned responses, partial reinforcement acquisition effect, learned helplessness, and vicious-circle behavior. The theory clarifies why alternative models have been unable to explain an equally large data base. © 1987, Psychonomic Society, Inc.. All rights reserved.