A cohort of 38-year-old men and women were studied for leisure time physical exercise in relation to low back pain (LBP), education, work, social class and smoking by a self-administered questionnaire. At the age of 14 years, the subjects had been interviewed by their school doctor regarding history of LBP and radiographs of the thoracic and lumbar spine were taken. The results show no positive correlation between radiographic changes and LBP in the adolescent period and decreased physical activity in adulthood. Physical activity for at least 3 h/week reduces the risk of LBP measured as lifetime, 1-year and point prevalence. Eighty-five percent of the subjects who reported taking physical exercise for at least 3 h/week had participated in sports activity almost constantly since their school days and these reported being in better condition than the rest of the cohort. Otherwise they did not have a healthier mode of life. No physical exercise during leisure time was associated with a short school education, unskilled work, unemployment and sickness, low social class, divorce, living in an apartment and smoking. Sixty percent had never or not for many years been interested in participating in sports. Badminton and tennis were the most common sports practised (36%), followed by gymnastics (32%), ball games-soccer and team handball-(25%), running (20%) and swimming (18%). Gymnastics and swimming seem to reduce LBP significantly. Our results show a falling interest in participating in sports activities over time, with 68% of the subjects being members of an athletic association previously, but only 29% currently. Women were more physically inactive during leisure time, probably because of their dual role. Logistic regression analysis indicates that physical activity is related to a long school education, high social class and regular sports activity over time.