Dundee Discussion Papers in Economics 150: Positive confirmation in rational and irrational learning

An experiment is reported which tests for positive confirmation bias in a setting in which a person can make decisions about which evidence to select to test a rule. The experiment reveals strong evidence of positive confirmation bias and a tendency not to choose the Bayesian- optimal selection of information. There is also evidence that the information, once selected, is used in a pattern of reasoning which, while sub- optimal, is internally coherent. These results are consistent with and expand on previous results on the subject.