Various researchers have suggested that traditional psychiatric nosologies may not be as useful in understanding and planning for persons with disordered behavior as might be categories based on empirical research. Although a number of cluster analytic studies have suggested that this approach can generate meaningful typologies of behavior among psychiatric patients, little attention has been paid to the issue of categorization of psychiatric disorders among older chronic inpatients. This paper presents results of a preliminary study targeted at developing a categorization system with such patients. A six-cluster solution was obtained using K-means clustering. The six clusters were "Paranoid Symptoms," "Motor Symptoms," "Depressive Symptoms," "Nonsymptomatic," "Positive Symptoms," and "Negative Symptoms." Results of this study of patients whose psychiatric disorders arose early in life are contrasted with studies of patients whose disorders occurred late in life.