One quarter of all injuries in children of school age occur at school. The prevention of these injuries presupposes some kind of reporting system. In this study the reliability of routine reporting of injuries at school was analyzed. Only those injuries requiring treatment by a physician were included. All studies of injuries at school indexed in MEDLARS since 1970 were reviewed. In school-based systems carried out on a routine base, an average of 17 injuries per 1,000 student-years were recorded. Yet, both in hospital/health center-based systems and in school-based studies carried out during a limited time, conspicuously higher frequencies were recorded: 38 and 71 injuries per 1,000 student-years, respectively. In the empirical study all injuries were recorded in the compulsory schools of Linköping (about 13,000 students) during two periods. During a first period of three years, injuries were routinely reported by the school nurses. During a second period of one year, all staff at school were repeatedly encouraged to report injuries. In the first period 9.8 injuries per 1,000 student-years and in the second period 17.7 injuries were recorded. These findings indicate that routine reporting of injuries at school results in substantial underreporting.