DECISION-THEORETIC AND EMPIRICAL INVESTIGATION OF SOME PROBABILISTIC DISCRIMINATION LEARNING SITUATIONS
NORTH CAROLINA UNIV AT CHAPEL HILL PSYCHOMETRIC LAB
Pagination or Media Count:
A decision-theoretic analysis and experiment of three related choice situations is presented. The first situation is a standard probabilistic discrimination learning task. Each trial begins with the presentation of one of a set of stimuli. The subject must choose between two response alternatives to predict which of two events will occur on the trial. The second situation arises when the conditional probabilities, i.e., the probabilities of the stimuli given the events, are introduced to the subject at the beginning of the experiment. The third situation is like the second except that the subject is not told which event occurs on each trial. One group of subjects was run in each situation with the overall relative frequency of one event equal to .80. The performance of the subjects in the first and second situations was virtually identical, while the performance of the subjects in the third non-feedback was somewhat worse. The comparisons of the choice proportions for the first and second groups indicated that subjects in the second group did not integrate information concerning the overall relative frequencies of events and conditional probabilities. A large proportion of subjects in the third non- feedback group made every choice in agreement with the assumption that the overall relative frequency of one event was one-half.