The prediction of event frequency: The frequency-learning, confidence-matching, and subjective-probability distribution hypotheses

Three hypotheses were employed to predict the results of a forced-choice test of two events differing in their frequencies of occurrence. The frequency-learning hypothesis proposes that choice data can be predicted by a negatively accelerated function of the difference between frequencies of occurrence of two events. The confidence-matching hypothesis suggests that choice data can be predicted by the utilization of one's confidence in choosing the item with the higher frequency of occurrence. And the subjective-probability distribution hypothesis proposes that choice data can be predicted by the utilization of the subjective probabilities of occurrence which subjects ascribed to all possible frequencies of occurrence of an event. Of the three hypotheses, it was the frequency-learning hypothesis which best predicted choice data. Furthermore, this hypothesis quite accurately predicted certain frequency data reported in the literature.