The diagnoses provided by a standard indirect test of vasopressin function were compared with those obtained by radioimmunoassay of plasma vasopressin in 24 patients with nonglucosuric polyuria. All seven cases of severe neurogenic diabetes insipidus diagnosed by the indirect tests were confirmed by the vasopressin assay. However, two of six patients with partial neurogenic diabetes insipidus by indirect criteria had normal vasopressin secretion by the direct assay; one was found to have primary polydipsia, and the other nephrogenic diabetes insipidus. Moreover, three of 10 patients diagnosed as having primary polydipsia by the indirect test had clear evidence of partial vasopressin deficiency by the direct assay. The inability of the indirect test to distinguish accurately between partial neurogenic diabetes insipidus and primary polydipsia may be explained by increased sensitivity to low concentrations of vasopressin in the former disorder and a reduction of maximal concentrating ability in both. We conclude that the incorporation of a vasopressin assay improves accuracy in the differential diagnosis of polyuria.