The main purpose of this study was to develop and to cross-validate an empirically derived psychosocial taxonomy of patients with diabetes. In the first study, 101 patients with Type I or Type II diabetes completed the Multidimensional Diabetes Questionnaire. Cluster analysis identified three clusters, labeled adaptive copers, low support-low involvement, and spousal overinvolvement. In the second study, the taxonomy was cross-validated using an independent sample of 132 patients with long-standing Type II diabetes. The results confirmed that the multivariate classification system was unique and highly accurate. External validation, using general psychological as well as diabetes-specific measures, supported the validity and distinctiveness of the patients' profiles. These findings help establish a multiaxial psychosocial taxonomy of diabetes and may have significant implications for the management of patients with diabetes.