In 1907, Alzheimer's single case report of presenile dementia led to the condition being recognized as a rare disease of middle-aged people. In 1964, Roth and colleagues reported an epidemiologic survey showing that the same condition was in fact a relatively common disease of the elderly. In 1987, Katzman introduced the concept of brain reserve, suggesting how individuals could have Alzheimer disease pathology in their brains without clinically manifesting the disease. In 1999, Petersen described mild cognitive impairment (MCI) as an amnestic state, since broadened to include other cognitive deficits, which, in the majority of patients, was a prelude to the development of full-blown Alzheimer disease or other dementia. MCI today is in some ways analogous to Alzheimer disease a century ago. We recognize and describe MCI clinically, among patients with memory complaints, but we do not know its distribution, outcomes, and risk factors in the elderly population at large. All population-based studies to date have found MCI to be an unstable and heterogeneous entity with a far wider range of outcomes than in the clinical setting, including reversion to normal in a substantial proportion. However, these studies thus far have been purely descriptive, and retro-fitted on to existing studies of dementia, using measurements not necessarily appropriate for the assessment of MCI. This review will address what is known and what has yet to be determined about MCI from the perspectives of the classic 7 uses of epidemiology.