We examined six systems or scales designed to distinguish melancholia from residual nonmelancholic depressive disorders in a sample of 305 patients. A count of the number of significant psychosocial risk factors showed that a clinical diagnosis was the most differentiating (19 significant risk factors), followed by the Newcastle index (13), DSM-III (10), and the CORE system (10)--the last essentially assessing psychomotor change; Research Diagnostic Criteria (RDC) (7) and an endogeneity symptom scale (2) were the least differentiating. A subsample of "composite melancholics" was derived, comprising 138 who met "melancholia" criteria for DSM-III, RDC, and CORE, and they were contrasted with residual depressives. The composite melancholics were older, had had a briefer depressive episode, and differed significantly on 12 risk factors, essentially being less likely to report deprivational experiences such as deficient parenting and dysfunctional marital relationships. We suggest that such a risk factor strategy is of potential use in refining the clinical definition of melancholia.