To assess the reliability of injury registration and to determine the incidence of injury in intramural (program of the institution) and extramural (leisure time) sports activities, we prospectively recorded information on sports injuries for 4 years in a group of 136 physical education students. Registration forms were completed every 3 weeks, and data on medical consultations were recorded. During the last year, we asked 59 students to recall all injuries sustained to establish the reliability of retrospective injury registration. In the prospective study, 525 sports injuries were recorded: 58% during intramural activities and 42% during extramural. The incident rate per 1000 hours of intramural activities (1.26) was significantly lower than that for extramural activities (1.77). A gradual decline in response rate from 98.4% the 1st year to 87.7% in the final year indicated a decreasing compliance during the study period. Eighteen percent of all injuries with recorded medical consultations were not reported by the students. At the retrospective injury registration in the final year of the study, students had forgotten 54% of the recorded injuries and 50 injuries (20% of all recorded injuries) were mentioned that had not been recorded during the previous years. We conclude that, even in a well-supervised population, prospective injury registration is not complete, and the reliability of retrospective injury registration is even poorer.