In two studies, 145 psychiatric inpatients were each asked to say 100 numbers in random order, using the numbers 1 through 10. Compared with normative data, patients with personality disorders and neuroses were not impaired on the random number generation (RNG) task and patients with chronic alcoholism and primary affective disorder, depression, were significantly imparied, but not as much as those with schizophrenia and organic brain syndrome. The relationship between RNG performance and psychiatric diagnosis may reflect severity of disturbed cognitive functioning. The Randomization Index was sensitive to changes in symptoms during hospitalization. The RNG task provides a brief objective measure of those components of attention, cognitive capacity, and short-term memory that are affected by severity of psychopathology.