REINFORCEMENT CONTINGENCY AND RESPONSE DIFFERENTIAION

点击翻译按钮获取中文摘要

The proposed research examines the generality of a quantitative model of associative contingency in operant response differentiation. In this model, the degree of response differentiation is related to the phi coefficient, derived from three experimentally-controlled parameters; w, the probability of a super-criterion responses and u and v, the probabilities of reinforcement following super- and sub-criterion responses, respectively. Control of w is achieved using percentile schedules of reinforcement, which maintain a fixed probability of super-criterion responses by definite responses with respect to their rank order in the subject's "current" response distribution. Both spatial- and temporal-response differentiation will be studied. The former involves shaping pigeons to respond to a "target" location chosen arbitrarily from one of 72 equally-spaced locations forming the perimeter of a four-foot diameter circle. An invariance property unique to phi will allow further experimental tests of its utility as a metric of reinforcement contingency, as well as establishing a degree of species generality over previous work using phi as a measure of reinforcement contingency. Phi will also be used to examine interresponse-time differentiation in a free-operant situation. Different degrees of contingency, as measured by phi, will be programmed in such a fashion as to hold the overal rate of reinforcement constant. By extending phi to responses with temporal extent, the probability-based phi coefficient can be extended to the free-operant situation which traditionally has exployed rate measures as both dependent and independent variables. To the degree phi successfully reflects the degree of both spatial- and temporal-response differentiation, quantification of the law of effect will be increased. If phi is an appropriate metric, the present results will demonstrate an extensive degree of response-and species-generality in both discrete-trial and free-operant situations. Finally, the research proposed herein utilizes exclusively within-subject design, thus providing data on the control of an individual subject's behavior, augmenting the predominantly between-subjects comparisons conducted previously.