OBJECTIVE: To estimate the prevalence and examine the course of mild cognitive impairment (MCI), amnestic type, using current criteria, within a representative community sample. METHODS: Retroactive application of MCI criteria to data collected during a prospective epidemiologic study was performed. The subjects were drawn from voter registration lists, composing a cohort of 1,248 individuals with mean age of 74.6 (5.3) years, who were nondemented at entry and who were assessed biennially over 10 years of follow-up. The Petersen amnestic MCI criteria were operationalized as 1) impaired memory: Word List Delayed Recall score of <1 SD below mean; 2) normal mental status: Mini-Mental State Examination score of 25+; 3) normal daily functioning: no instrumental impairments; 4) memory complaint: subjective response to standardized question; 5) not demented: Clinical Dementia Rating Scale score of <1. RESULTS: At the five assessments, amnestic MCI criteria were met by 2.9 to 4.0% of the cohort. Of 40 persons with MCI at the first assessment, 11 (27%) developed dementia over the next 10 years. Over each 2-year interval, MCI persons showed increased risk of dementing (odds ratio = 3.9, 95% CI = 2.1 to 7.2); 11.1 to 16.7% progressed to Alzheimer disease and 0 to 5.0% progressed to other dementias. Over the same intervals, 11.1 to 21.2% of those with MCI remained MCI; of 33.3 to 55.6% who no longer had MCI, half had reverted to normal. CONCLUSIONS: In this community-based sample, 3 to 4% of nondemented persons met MCI operational criteria; despite increased risk of progressing to dementia, a substantial proportion also remained stable or reverted to normal during follow-up. Amnestic MCI as currently defined is a high-risk but unstable and heterogeneous group.