OBJECTIVE--To determine whether, in the treatment of major depression in primary care, a brief psychological treatment (problem solving) was (a) as effective as antidepressant drugs and more effective than placebo; (b) feasible in practice; and (c) acceptable to patients. DESIGN--Randomised controlled trial of problem solving treatment, amitriptyline plus standard clinical management, and drug placebo plus standard clinical management. Each treatment was delivered in six sessions over 12 weeks. SETTING--Primary care in Oxfordshire. SUBJECTS--91 patients in primary care who had major depression. MAIN OUTCOME MEASURES--Observer and self reported measures of severity of depression, self reported measure of social outcome, and observer measure of psychological symptoms at six and 12 weeks; self reported measure of patient satisfaction at 12 weeks. Numbers of patients recovered at six and 12 weeks. RESULTS--At six and 12 weeks the difference in score on the Hamilton rating scale for depression between problem solving and placebo treatments was significant (5.3 (95% confidence interval 1.6 to 9.0) and 4.7 (0.4 to 9.0) respectively), but the difference between problem solving and amitriptyline was not significant (1.8 (-1.8 to 5.5) and 0.9 (-3.3 to 5.2) respectively). At 12 weeks 60% (18/30) of patients given problem solving treatment had recovered on the Hamilton scale compared with 52% (16/31) given amitriptyline and 27% (8/30) given placebo. Patients were satisfied with problem solving treatment; all patients who completed treatment (28/30) rated the treatment as helpful or very helpful. The six sessions of problem solving treatment totalled a mean therapy time of 3 1/2 hours. CONCLUSIONS--As a treatment for major depression in primary care, problem solving treatment is effective, feasible, and acceptable to patients.