Allen Neuringer - US grants

Operant behaviors are learned and repeated when they are followed by reinforcers. For example, the learning by a student in a classroom or job performance by a worker in a factory partly depend upon reinforcing consequences. The question asked in the present research is whether variability is a dimension of behavior that is also controlled by contingent reinforcers. If, for example, only unpredictable patterns of activity are followed by reinforcement, will behavior become unpredictable and such unpredictability be maintained? Little is known about the control exerted over behavioral variability by contingent reinforcers and, indeed, some researchers doubt that such control is possible. Previous research in Dr. Neuringer's laboratory showed that variability can be reinforced. That research involved an animal model to explore these basic questions. Dr. Neuringer found that sequences of responses on two operanda by pigeons became highly variable when reinforcement depended upon such variability. The present research continues to test whether variability is, in fact, an instrumental dimension of behavior controlled by contingent reinforcement. If it is, then a number of predictions should be supported. For example, reinforcement should exert precise control over the degree of variability in behavior; animals should be able to choose between different degrees of variability, with these choices controlled by reinforcers; and discriminative stimuli should produce variability or stereotypy, depending upon the contingencies. Dr. Neuringer's research also will compare the reinforcement of repetitive and habitual behaviors with the reinforcement of variable behaviors in seeking to distinguish between laws of habit and laws of variation. Other studies will explore whether and how the reinforcement of variability contributes to learning generally. These basic experiments should help to define important functions of instrumental variability. They may help researchers and educators produce variability when it is desired, for example, in learning, problem-solving, creativity-demanding, or competitive situations. And the studies may provide assessment methods to help physicians and health-care workers factor out the relative effects on behavioral variability versus learned habits of drugs, brain damage, motivational states, and psychological deficiencies.

DESCRIPTION (provided by applicant): Behavioral variability plays an important role in operant learning, problem solving, and creativity. Abnormal levels of variability are characteristic of some psychopathologies, with high levels seen in attention deficit hyperactivity disorder and low levels in depression and autism. Much research has been devoted to identifying sources of variability and ways in which it is controlled. This grant studies a potentially important source, one that has received relatively little attention in the past, namely reinforcers that are directly contingent upon variability itself. The working hypothesis is that individuals learn how, what, when and where to vary. Such learning is controlled by reinforcing consequences and discriminative stimuli: in some situations, it is adaptive to behave unpredictably or do novel things, in others, to respond repetitively and predictably. Animal models -- rats and pigeons -- are used (i) to explore the generality of reinforced variability; (ii) to test for transfer of training; and (iii) to use reinforcement-of-variability techniques to study dynamic changes that occur within operant classes. To test generality, one study compares different types of reinforcers, food versus water versus social, and asks whether levels of variability correlate with speed of learning. Another compares how positive versus negative reinforcers influence variability. And a third asks whether differences between aged and young animals can be meliorated by explicit reinforcement of variations. A theoretically important question concerns whether animals and people learn to vary as a general competency, one that can transfer across a wide array of responses and contexts, or whether specific variations are limited to specific cases. If an organism can learn "to vary," and this comes under stimulus control, then many aspects of learning and skills training could be facilitated. Lastly, to study intra-operant effects, a model is established in which two or more operant classes share a subset of response instances. Effects of intra-operant sharing on response rates, probability distributions, choices, and skills acquisition are studied. These experiments analyze an important, but still not well-understood, aspect of behavior, namely, the control over behavioral variations exerted by reinforcing consequences.