OBJECTIVES: This study assesses the properties of the Mini-Mental State Examination (MMSE) with the purpose of improving the efficiencies of the methods of screening for cognitive impairment and dementia. A specific purpose was to determine whether an abbreviated version would be as accurate as the original MMSE in predicting dementia. STUDY DESIGN AND SETTING: A population-based post hoc examination of the performance characteristics of the MMSE for detecting dementia in an existing data set of 243 elderly persons. RESULTS: Sensitivity, specificity, and predictive values were computed for the original MMSE as well as new MMSE scale models derived from a Rasch model item analysis. The optimal threshold for the original MMSE screen yielded sensitivity and specificity estimates of 72.5% and 91.3%, respectively. The use of a subscale resulted in a slightly lower sensitivity (71.0%), specificity (88.4%), and positive predictive value (71.0%) but equal area under the receiver operating characteristic curve. Cross-validation on follow-up data confirmed the results. CONCLUSION: A short, valid MMSE, which is as sensitive and specific as the original MMSE for the screening of cognitive impairments and dementia is attractive for research and clinical practice, particularly if predictive power can be enhanced by combining the short MMSE with neuropsychological tests or informant reports.