Application of a statistical model to simple discrimination learning in human subjects.

Abstract

Human subjects (Ss) learned a successive discrimination between 2 sets of signal stimuli. The task was to predict which of a pair of "reinforcing" lights would appear following the signal on each trial. Uniform reinforcement (conditioning phase) or non-reinforcement (extinction phase) was given in the presence of one stimulus set and 50% random reinforcement in the presence of the other. Variability of the signal pattern was a 2d independent variable. Curves of responding in the presence of the 2 stimulus sets taken separately were generally similar to curves reported by previous investigators for learning under similar conditions of reinforcement in nondiscriminative situations. Effects of stimulus variability were insignificant during conditioning but large and significant during extinction, with the more constant stimulus yielding the faster extinction. An attempt was made to predict the detailed results of this study by means of a statistical model developed in connection with earlier studies of simple learning under random reinforcement. Correspondences of theory and data, although by no means perfect, tended to support the view that discrimination learning in this situation is a simple resultant of effects of reinforcement and nonreinforcement.