Rate of verbal conditioning in relation to stimulus variability.

Abstract

Effects of stimulus variability were studied in a conditioned key-pressing experiment. Subjects were volunteer undergraduate students. On each of 120 trials, the subject predicted, at a signal, which of two lights would come on by pressing a key beneath it. The two lights were presented in a 90-10 random sequence, independently of the subject''s response. The signal also was a sample of lights from a panel; and three signal probabilities, defining three experimental groups, were used in determining whether each of these lights would be presented on any trial. In accordance with predictions from statistical learning theory, the signal probability had no effect on the rate of predicting the less frequently reinforced light during the later trials but was inversely related to the cumulative number of such predictions. Statistical learning theory was used to predict the behavior of one of the groups, using parameters evaluated from the other two.