In 1965 and 1966, a consecutive series of 89 patients admitted to the Maudsley Hospital, London, England, with depressive illness were interviewed, and various personality questionnaires were administered; 18 years later, they were followed up and reinterviewed. Then, on the basis of the index data alone and without knowledge of their eventual outcomes, they were subtyped according to the Research Diagnostic Criteria, DSM-III, Newcastle Index, and Present State Examination diagnostic criteria. Patients who met the various subtype criteria at index were compared with those who did not in respect to their long-term outcome. Subtyping had little prognostic utility except for three endogenous criteria that were all associated with poor outcome. In addition, DSM-III melancholia had an interactive effect with the personality measure neuroticism, so that those melancholic patients who at index had high neuroticism scores were very likely to have a poor outcome.