抄録
The purpose of this study is to examine that people seem to display implicit knowledge of Bayesian principles even if they do not display any explicit knowledge. We conducted two experiments replicating one performed by Gluck & Bower's (1998). In Experiment 1, participants were given 3 sessions of 100 symbol strings of 4 bits, were required to choose one of three categories they thought the strings related to, and were given feedback on the correct category. In Experiment 2, the procedure was almost the same as Experiment 1 except for the followings: (a) a training setting (replaced the strings and the categories with fictitious clients' conditions and effective psychotherapies respectively), and (b) the grades of students belonging to the psychology course. Overall, both experiments showed that the proportion of the participants' choices was getting closer to the true probabilities (i.e., Bayesian probabilities) not only with sessions but also with the grade going up. These findings suggest that existing knowledge has a facilitating effect on probability matching of psychotherapy selection without explicit memories of the frequencies of string-category or condition-psychotherapy combinations.
