Pattern recognition methods were carried out on a sample of 80 depressed men, assessed by means of 14 items relevant to depressive symptomatology of the Structured Clinical Interview for DSM-III-R. 1985 edition (SCID). A cluster analysis generated two classes, which were described as a vital (n = 35) and a nonvital cluster (n = 45). Vital depressives were characterized by psychomotor disorders, loss of energy, cognitive disturbances, a distinct quality of mood, early morning awakening and nonreactivity (the "vital" symptoms). Our findings support the descriptive validity of the DSM-III melancholia diagnostic category, although the DSM-III criteria are too conservative and include nonrelevant symptoms (e.g., diurnal variation, anorexia-weight loss) whilst excluding some important items (e.g., loss of energy, cognitive disorders). Vital depressed men were significantly older, more severely depressed and they exhibited biological disturbances (abnormal dexamethasone suppression test, lower basal thyroid secreting hormone) as opposed to nonvital depressives. There are several arguments to support the possibility that both clusters constitute relevant stages in the overall severity of illness continuum, whilst showing qualitative differences with regard to the vital symptoms. In other words, both clusters are continuous categories within the overall severity of illness continuum and form discrete categories with regard to the vital symptoms. By merging the dimensional and categorical hypotheses, we were able to construct a new integrated threshold model: unipolar depression in men is probably a homogeneous disease with reference to overall severity of illness, but--as severity increases--vital symptoms emerge, grouping together into a distinct profile, i.e., vital depression.