REINFORCEMENT OR REWARD IN LEARNING

[Reinforcements and rewards drive learning. They can add effect to otherwise neutral percepts with which they coincide. They can alter the probability of behaviors that precede them, as Thorndike captured in his Law of Effect. How reinforcers and rewards exert these effects is the topic considered in the following four sections.

The first section considers theA natomicalpathways of the brain that mediate reinforcement. This section takes the perspective that pathways for reinforcement are part of a larger brain system that supports adaptive, goal-directed behaviors, including feeding, drinking, and reproduction. From this perspective, reinforcement-driven learning and memory are major influences on the likelihood of production of goal-directed behaviors. The pathways that mediate reinforcement have been delineated by a combination of traditional anatomical tracing techniques and the use of electrical stimulation to identify areas where neuronal activation substitutes for natural reinforcements.

A brain system in which reinforcement learning plays a strong role involves pathways through the cerebellum that mediate classical conditioned motor responses. As summarized inC erebellum, conditioned stimuli and reinforcement signals critical to classical motor conditioning converge in the cerebellar cortex and deep nuclei, and compose the memory trace that mediates this form of learning. In this situation, the reinforcing stimulus is viewed as a "teaching" input that directly elicits the motor response and, when paired with a neutral stimulus, confirms the predictive relationship between the two stimuli. Multiple lines of evidence have been brought to bear on the nature of this interaction and its pathways. These studies have revealed that the dorsal olivary nucleus is the source of the reinforcing input to the cerebellar cortex in support of conditioned motor responses.

The section onE lectricalS elf-S timulationfocuses on the use of electrical stimulation to identify brain areas and pathways that mediate reinforcement. These studies have revealed that brief trains of electrical stimulation at specific sites serve as reinforcements for goal-directed behaviors. This artificial activation of the reinforcement pathway is perceived as intensely rewarding, and can support a variety of behaviors. Studies using this technique, combined with other approaches, have been used to characterize a widespread brain system that supports reinforced behaviors. This pathway involves the mesolimbic dopamine system that arises in the ventral tegmental area and projects to the forebrain in the nucleus accumbens. The outputs of this system strongly influence the extrapyramidal motor system, as well as limbic cortical and subcortical areas involved in the assessment of stimulus contingencies and emotional behavior.

Another brain system in which reinforcement learning plays a critical role is the extrapyramidal motor system, and theS triatumis a major component of that system. Neurons in the striatum detect both the delivery of rewards and the presentation of stimuli that predict rewards. Some cells fire at specific steps in a sequence of events leading to rewards. In addition, the activity of striatal neurons can reflect general states of expectation of rewards predicted by previous experience. The striatum receives signals from the orbitofrontal cortex about reward preferences, and receives global reinforcement signals from the mesolimbic dopamine system. How the striatum integrates this information to make predictions about future rewarding events constitutes a major area of study on reinforcement mechanisms.

Together these sections provide an overview of different aspects of reinforcement and reward as a major influence in the formation of memory traces that support acquired behavioral responses.]