We have studied the relation of occupational and recreational physical activity to fractures at different locations. All men born between 1925 and 1959 and all women born between 1930 and 1959 in the city of Tromsø were invited to participate in surveys in 1979-1980 and 1986-1987 (The Tromsø Study). Of 16,676 invited persons, 12,270 (73.6%) attended both surveys. All nonvertebral fractures (n = 1435) sustained from 1988 to 1995 were registered in the only hospital in the area. Average age in the middle of the follow-up period (December 31, 1991) was 47.3 years among men and 4501 years among women, ranging from 32 to 66 years. Fracture incidence increased with age at all locations among women, but it decreased with or was independent of age among men. Low-energetic fractures constituted 74.4% of all fractures among women and 55.2% among men. When stratifying by fracture location, the most physically active persons among those 45 years or older suffered fewer fractures in the weight-bearing skeleton (relative risk [RR] 0.6, confidence interval [CI] 0.4-0.9, age-adjusted), but not in the non-weight-bearing skeleton (RR 1.0, CI 0.7-1.2, age-adjusted) compared with sedentary persons. The relative-risk of a low-energetic fracture in the weight-bearing skeleton among the most physically active middle-aged was 0.3 (CI 0.1-0.7) among men and 0.9 (CI 0.4-1.8) among women compared with the sedentary when adjusted for age, body mass index, body height, tobacco smoking, and alcohol and milk consumption. It seems that the beneficial effect on the skeleton of weight-bearing activity is reflected also in the incidence of fractures at different sites.