A random sample including 2342 cases representative of all occupational back injuries in Quebec (1981) was followed up prospectively over three years to assess the recurrence rate of back problems (lumbar, thoracic, and cervical). Each medical and accident report was reviewed to obtain the site of symptoms and occupation. Age, sex, industrial sector, and number of episodes of absence from work were abstracted from the computerised Quebec Compensation Board files. The recurrence rate was 20.0% at one year follow up and 36.3% at three years. A multivariate analysis using a Poisson regression, was performed to model the risk of recurrence over time. Men had a higher chance of recurrence (risk ratio = 1.85, 95% CI = 1.50-2.27) but among recurrent cases, the average total number of episodes was comparable between men and women. Age showed a protective effect on the probability of recurrence (10 years: RR = 0.93, 95% CI = 0.88-0.98) due to the lower recurrence rate in the 45-64 year old group (31.8%). Cervical and lumbar symptoms had identical recurrence profiles whereas thoracic symptoms had a significantly lower recurrence rate. Drivers had the highest recurrence rate (42.1%) and nurses had the highest average number of recurrences (2.03) among recurrent cases. Both occupations had statistically significant excesses after controlling for the other variables.