Concern over safety has resulted in legislation by, for example, the Commission of the European Union, to limit the number of hours worked at night. There is, however, no direct evidence that injuries are more frequent at night. We analysed all 4645 injury incidents reported for a year on a rotating three-shift system in a large engineering company where the a-priori accident risk appeared to be constant. The relative risk of sustaining an injury was 1.23 (95% CI 1.14-1.31) higher on the night shift than on the morning shift, which showed the lowest incidence. The type of work (machine versus self-paced) affected the pattern and nature of injuries; for self-paced work the relative risk of more serious injury on the night shift compared with the morning shift was 1.82 (1.30-2.34).