During 1976, 80 patients with a primary depressive illness were selected from a consecutive series of referrals to the Royal Edinburgh Hospital for a short-term follow-up study. This paper concerns the 12-year longitudinal assessment of the survivors. During the follow-up, mortality risk for the sample was almost doubled. Of the series, 35% had experienced a recurrence within 2 years of the initial interview, and just over 60% within the entire study time of 12 years. The risk of recurrence was enhanced for those aged over 45 years at index, for those with a history of depressive disorder and for those who had not engaged in parasuicidal behaviour during the index episode. The observed effect of previous episodes was largely accounted for by age. The chance of recurrence was not affected by whether the treatment of the index episode included ECT, or by whether diagnosis was 'endogenous' or 'neurotic'. Application of the Lee-Murray outcome criteria showed that about one-third of the Edinburgh series experienced a very poor outcome. Results are presented concerning the prediction of long-term outcome as represented by the Depression Outcome Scale (DOS), a measure specially constructed for this study.