The emergence of organized complexity in the physical and biological worlds is best understood as the product of the selection and retention of undirected variation. This process is exemplified in such diverse areas as the building of heavier elements through stellar fusion and the evolution of species through natural selection. The major goal of research in this laboratory is to characterize the biobehavioral processes whereby the environment selects individual behavior and, thereby, produces complex behavior. The principle that summarizes the action of these processes is the principle of reinforcement, and the outcome of its action is the environmental guidance of behavior.
Based upon experimental and theoretical research from this and other laboratories, a principle of reinforcement has been formulated that informs much of our current empirical and theoretical work. This principle -- the Unified Reinforcement Principle -- provides a unified account of conditioning in that learning in both the classical (Pavlovian) and operant (instrumental) procedures is interpreted as the outcome of a common set of biobehavioral processes.
John W. Donahoe
Biobehavioral Mechanisms of
Learning: Experimental Analysis
and Neural Network Simulations
of Reinforcement and Stimulus
Control
Learning: Experimental Analysis
and Neural Network Simulations
of Reinforcement and Stimulus
Control
The principle provides a single conceptual framework that accommodates "biological constraints on learning" (e.g., stimulus-reinforcer and response-reinforcer interactions) as well as phenomena otherwise interpreted as evidence for different "associative structures" in learning (e.g., stimulus-reinforcer and response-reinforcer "associations").
Experimental work on the reinforcement principle primarily uses an avian preparation in which respondents and operants are simultaneously measured in the same organism.
The respondents are water-elicited swallowing and airpuff-elicited movement of the eyelid and nictitating membrane; the operants are various topographies of movement of the legs. In addition to this preparation, more standard techniques such as keypecking in pigeons and leverpressing in rats are used to study other problems in reinforcement and stimulus control. The stimulus control research typically employs "molecular" measures of responding (e.g., interresponse times) rather than more "molar" measures (e.g., mean response rate). Studies are conducted, and data are recorded and evaluated, by means of on-line computer systems.
Theoretical work on the reinforcement principle is carried out by means of computer simulations of real-time adaptive neural networks. The simulation research is informed by findings at the neuroanatomical and cellular as well as the behavioral level. The method for modifying the strength of connections within networks exploits a biologically plausible mechanism involving diffusely projecting monoaminergic systems of the forebrain. A genetic, neurodevelopmental algorithm is used to construct the architecture (neuroanatomy) of the networks.
Back to the Faculty Index