Serum testosterone levels (ng/dl) were measured at 2-week intervals during the course of hospitalization in 35 male inpatients in the following four diagnostic groups: undifferentiated schizophrenia, paranoid schizophrenia, bipolar I disorder-manic, and major depressive disorder (endogenous type). The mean (+/- SE) testosterone levels during hospitalization were significantly higher (p less than 0.001) in the schizophrenic patients (510 +/- 38) than in the affective disorder patients (347 +/- 25). This difference persisted throughout hospitalization, being present in the first sample following admission (p less than 0.03) and the final sample before discharge (p less than 0.01). The above group differences were largely due to high testosterone levels in the paranoid schizophrenic subgroup (mean +/- SE level of 559 +/- 41). A longitudinal, as well as cross-sectional, view of the hormonal and clinical data suggests that the testosterone system is linked to both state and trait psychological factors, and this issue is discussed in the light of prior basic psychoendocrine research on this system. The potential application of these findings for new approaches to the development of biological criteria for psychiatric diagnosis is discussed.