The aim of this intervention was to evaluate the effect of training on patient-handling skills and prospectively to assess the effect of skill on subsequent back pain and back injuries in nursing. Of a total of 255 nurses, 199 were assessed for their skill in patient-handling. One-half (control group) received traditional training in patient-handling, and the other half (trained group) received a curriculum of instruction totaling 40 hours. The skills of both groups were assessed on graduation. The control group was rated as less competent in patient-handling. Nurses in both groups were questioned about the prevalence of back pain and incidence of back injuries in the first year after graduation. In multiple regression analysis, the major risk indicators for back injuries were poor patient-handling skill, low numbers of repetitions in the sit-up test, and high work-load scores. High score on the hysteria scale of Middlesex Hospital Questionnaire was a risk indicator for all kinds of back pain. Though back pain was independent of patient-handling skill, those rated as "bad" or "poor" had more back injuries (24%) than those who had been rated as "good" or "excellent" (2%) (P less than 0.001), but the difference between the trained and control groups was not statistically significant. It was concluded that back injuries may be prevented by the teaching of patient-handling skills.