Psychological aptitude tests have been carried out on train operation staff in an attempt to improve safety. We searched for a test that would enable us to predict the human errors that are most likely to occur under current working conditions. First, we interviewed railway company experts to identify those errors. Next, we carried out personal computer experiments to simulate occurrences of these errors then applied 20 kinds of tests to 79 subjects. Consequently, we found a correlation between the error occurrence tendencies in the experiments and the actual test results. Based on these results, we have proposed effective tests according to the type of tasks undertaken.