Abstract
Information gain models attempt to explain the typical pattern of responses in the Wason selection task by assuming that the probability of selecting a card increases with the scaled information measure about the rule to be tested in the task. Because such models assume that humans could detect differences in the scaled information measure between the cards, this study investigates how sensitive humans are to such differences by employing a newly devised “ordering task”. The result indicates that the observed sensitivity severely restricts the range of possible values for the parameters assumed in information gain models.
