Problems consisted of two categories of questions, of general knowledge and forecasting future events. Given each question, the subjects chose a more likely answer from the given two alternatives, rated their own confidence on the correctness of the choice, and then, assessed the hit-rate of the classmates. The major result is as follows. The difference between average confidence and average hit-rate was small, namely, calibration was good, for problems of both general knowledge and familiar future events. On the other hand, calibration was poor for problems of accidental future events. In other words, the more available knowledge, the better calibration is. In discussion we proposed “a model of retrieval and generation”, which could explain our results for the problems of general knowledge. Results for problems of future events suggest that the subjects possibly used a certain model to make their probability judgments. On the basis of our results and with our discussion, we found the phenomenon that people believe themselves to make mental judgments better than the average. We call the phenomenon “self-superiority phenomenon”.
