A cohort of 2,342 cases constituting a random sample of all occupational back injuries compensated in Quebec (Canada) during 1981 was observed prospectively for 3 years to study associations between cumulative duration of absence from work and sex, age, site of symptoms (cervical, thoracic, and lumbar), and occupation. Of the cohort members, 227 (9.7%) cumulated 6 months of absence or more. A logistic regression model showed age and site of symptoms to be the two most important risk factors associated with absences of 6 months or more. No association was found with sex or occupation. The results showed discrepancies between measures of frequency distribution of back injuries and duration of absence from work, an important finding in terms of identifying health priorities.