The search for biological markers of psychiatric disorders has traditionally involved univariate approaches, usually focusing upon one measure at a time, and to date has been primarily directed towards the assessment of depressive rather than psychotic illnesses. The present study explores a multidimensional psychoendocrine strategy, using a profile of five hormones, including cortisol, epinephrine, norepinephrine, testosterone, and free thyroxine, and is directed at the differentiation of two major psychotic illnesses, bipolar manic disorder and paranoid schizophrenia. When the levels of these hormones were assessed at admission and biweekly during hospitalization, the mean values for all five hormones were found to differ markedly between the two diagnostic groups. There was, however, always a zone of overlap in levels between the two groups when each of the five hormones were viewed individually, so that at best only about 70% of patients were correctly separated by diagnostic group using any single hormone alone. By contrast, multivariate approaches combining mean values of three or more hormones, using either stepwise discriminant analysis or multidimensional scaling, yielded 95% correct classification of the two diagnostic groups. Similar but not quite as great accuracy of classification was achieved with only the initial hormone sample obtained at the time of hospital admission. These preliminary findings provide encouragement for further exploration of multidimensional hormonal strategies in the search for useful biological criteria to assist in the diagnosis of psychiatric disorders.