This study investigated the psychometric properties of the Mattis Dementia Rating Scale (DRS), including its internal consistency, sensitivity to age and education effects, pattern of change scores, convergent validity and predictive utility. Age and education effects and stability were assessed in a sample of 280 cognitively normal persons over age 55. All other features were assessed in a sample of 274 persons with cognitive impairment, including 221 newly diagnosed dementia patients and 53 patients with mild cognitive impairment. Both samples were drawn from the Mayo Clinic Alzheimer's Disease Patient Registry. Within the normal sample, Pearson correlation coefficients for DRS Total score with age, education and intelligence were all significant, p < .0001; MAYO Full Scale IQ (FSIQ) shared approximately 20% variance with DRS Total score. Internal consistency was greater than .7 for Construction, Conceptualization, Memory, and Total score, greater than .65 for Attention and less than .5 for Initiation and Perseveration. DRS subscale scores for Memory, Attention, and Conceptualization were significantly correlated with appropriate indices from the WAIS-R or Wechsler Memory Scale-Revised, as assessed in the Mayo Older Americans Normative Studies, providing convergent validity for these DRS scales. Importantly, proportional hazards modeling revealed initial DRS Total score to be a significant predictor of longitudinal institutionalization and mortality outcomes.