A prospective survey of patients seeking primary care for low back pain. Clinical and psychosocial data, available at presentation, were explored for predictors of outcome at 1 year. To determine the relative value of clinical and psychosocial variables for early identification of patients with a poor prognosis. Current treatment strategies for low back pain have failed to stem the rising levels of disability. Psychosocial factors have been shown to be important determinants of response to therapy in chronic patients, but the contribution from similar data in acute or subchronic patients has not been comprehensively investigated. Two hundred fifty-two patients with low back pain, presenting to primary care, underwent a structured clinical interview and completed a battery of psychosocial instruments. Follow-up was done by mail at 1 year; outcome was measured using a back pain disability questionnaire. Predictive relationships were sought between the data at presentation and disability at follow-up. Most patients showed improved disability and pain scores, although more than half had persisting symptoms. Eighteen percent showed significant psychological distress at presentation. Multiple regression analysis showed the level of persisting disability to depend principally on measures in the psychosocial domain; for acute cases outcome is also dependent on the absence or presence of a previous history of low back trouble. Discriminant models successfully allocated typically 76% of cases to recovered/not-recovered groups, largely on the basis of psychosocial factors evident at presentation. Early identification of psychosocial problems is important in understanding, and hopefully preventing, the progression to chronicity in low back trouble.