The hypothesis that mild cognitive impairment (MCI) represents an early stage of Alzheimer's disease (AD) was investigated by reviewing recent research from three sources: asymptomatic and symptomatic individuals carrying mutations that cause AD, hospital-samples of non-demented patients with MCI at the initial examination that are followed longitudinally, and community-based incident cases of AD. Studies with asymptomatic mutation carriers of the amyloid precursor protein and presenilin 1 gene have shown a linear and disease-related decline in most cognitive functions that begins approximately 10 years before the expected clinical onset of AD. However, there is considerable overlap between the level of impairment for mutation carriers and non-carriers of the same age during the early preclinical stage of AD. Hospital-based longitudinal studies have shown that non-demented individuals with isolated mild episodic memory impairment may develop clinically diagnosed AD with widespread cognitive deficits in a few years time. Community-based epidemiological studies on the incidence of AD demonstrate that indices of episodic memory, in addition to measures of general cognitive functioning, are useful in predicting early AD. In contrast, subjective memory impairment or age-associated memory impairment are less powerful predictors of future dementia development. In summary, there is converging evidence to demonstrate that preclinical AD is characterized by a common behavioral phenotype, with cognitive decline in several domains, predominantly in episodic memory. The decline appears to start many years before the clinical onset of AD. Moreover, the progression of the impairment appears to be continuous. Finally, this pattern of performance generalizes across etiology of AD (familial or sporadic), clinical onset (early or late), sample composition (hospital or community), and method of assessment.