What experience tells us about students sharing test information
Over a period of about five years, courses enrolling a large number of junior level students used computer-based testing. Early in this process, testing took place over several days. Concern about cheating changed this policy so that testing for each of the four courses was completed in just one day. A few students completed a makeup test about a week after the course testing day. Careful question sampling was used to control release of test information by early testing students. Periodically a statistical summary of the scores by testing period was undertaken. In most cases the average score of students testing in the earliest period tended to be the highest of all similar periods. The mean score in the latest several testing periods tended to be the lowest among all testing sessions. The lowest average scores were almost always for the students taking a makeup test that was given one week later than the regular testing. A lower average score for later testing students gives some indication that early testing students either do not tend to share usable information to later testing students or if they do, its effect on test performance is generally small. What these studies do not describe is the extent to which some individual students (probably small in number) were able to use information gained from early testing students to improve their score. So while the general effect of later testing appears to be slightly lower average performance, some later testing individuals still may have been able to improve their test score by using information from early testing students.