<p>Four experiments were undertaken to identify properties of differential reinforcement relevant to response differentiation. Differential reinforcement was given for the length of "runs" of pigeons' key pecks, the length of a run being defined as the number of responses on a key preceding a spatially different terminal response.</p> <p>In Experiment I, a criterion run-length was assigned each subject and each subject was reinforced only following emission of a run-length which exceeded its criterion. Criterion run-lengths of 5, 10 and 20 responses were assigned to different groups. Mean run-lengths (N) were related to criteria (n) by a power function:N = 2.63 n^0.69. The fractional exponent of this function implies that the proportion of responses exceeding criterion was lower for higher criterion values. Because reinforced response values were confined to an extreme of the dis tribution, a lower proportion of reinforced response values indicated a greater difference between the distribution of values of the response emitted by a subject and the distribution of values reinforced. Since higher criterion values produced higher asymptotic mean run-lengths, the differential between emitted and reinforced run-lengths was also related to asymptotic run-length.</p> <p>A similar relationship between mean run-length and the proportion of responses exceeding a criterion was obtained in Experiments II and III. These experiments used a differential reinforcement paradigm in which thecriterion was continuously adjusted for each subject so that a constant proportion of its distribution of run-lengths exceeded its criterion. In Experiment II, the proportion of the distribution exceeding the criterion was set at 15%, 30% or 45%, for different groups. However, only a random half of the run-lengths exceeding the criterion for animals in the 30% group, and one-third of those for the 45% group were actually reinforced, so that the overall probability of reinforcement was constant at 0.15 for all groups. Increases in the mean length of runs were obtained for all groups, with the 15% group exhibiting a more rapid change and a higher asymptotic mean run-length than the 30% group, which in turn exceeded the 45% group.</p> <p>In Experiment III, pigeons were reinforced only for run-lengths in the most extreme 30% of their run-length distributions. One group received reinforcement for a random half of these run-lengths, while others received reinforcement for all of the run-lengths in this range. The relative extremeness of run-lengths selected for reinforcement was therefore constant (30%) 'vhile the overall probability of reinforcement was either 0.30 or 0.15. Increases in mean run-length were obtained, but no differences in the rate of change or asymptotic mean run-length were found between groups.</p> <p>The results of Experiments II and III suggest that the relative extremeness of run-lengths selected for reinforcement, that is the differential between emitted and reinforced run-lengths, determines the rate and extent of changes in values of a response. The relationship between the relative extremeness of reinforced response values and the rate and extent of changes in emitted response values was further</p> <p>demonstrated in Experiment IV. Reinforcement was provided for a fixed area of the distribution of run-lengths closest to a specified target value. For one group, the area of the distribution reinforced was 15%, while another group was reinforced for a random half of run-lengths in the closest 30% to the target. Reinforcement probability was thus held constant at 0.15. Mean run-lengths changed in the direction of the target at a rate dependent on the difference between the mean run-length and the target run-length. An asymptote was reached (rate of change = 0) when mean run-length approached the target value (differential = 0). The relationship between asymptotic mean run-length and target run-length was nearly linear for both 15% and 30% groups.</p> <p>The results of this series of experiments were discussed in terms of a hypotnetical differentiation process emphasizing the differential between the central tendencies of the distributions of emitted and reinforced response values, expressed in percentage units. When a differential exists, a change in the distribution of response values occurs in a direction which tends to reduce the differential, and at a rate proportional to the magnitude of the differential. When no differential exists, no change occurs. Since change in many dynamic response properties is resisted in a way which suggests an opponent process (e.g.iLaw of Least Effort), the asymptote reached after differential reinforcement represents an equilibrium between differential reinforcement and its opponent process. The increased effectiveness of differential reinforcement when relatively more extreme response values are reinforced would thus not only result in greater rates of change, but would also overcome the resistance of the opponent process to a greater extent,shifting the equilibrium point to a higher value.</p> / Doctor of Philosophy (PhD)
Identifer | oai:union.ndltd.org:mcmaster.ca/oai:macsphere.mcmaster.ca:11375/12827 |
Date | 10 1900 |
Creators | Webster, Bryant John |
Contributors | Platt, J. R., Psychology |
Source Sets | McMaster University |
Detected Language | English |
Type | thesis |
Page generated in 0.0075 seconds