The Mini-Mental State Examination (MMSE), a brief test of cognitive function, has been widely used to screen for dementia. We administered the MMSE to 74 community-dwelling patients meeting criteria for probable Alzheimer's disease (AD) and 74 age- and education-matched controls. Twenty-four patients with AD performed in the nondemented range by scoring above the recommended cutoff point of 23 of a possible 30 on the MMSE. We compared the scores for items of the MMSE in controls and subjects with AD and used logistic regression to model a shorter MMSE that retained the accuracy of the complete test. A score summing tests of recall and orientation for place had similar sensitivity to the full MMSE. Adding a verbal fluency test to the MMSE reduced the error rate by improving the accuracy of diagnosis of patients with AD scoring in the nondemented range.