Variability in occurrence estimates is one of the basic features of the epidemiology of dementia and mild cognitive impairment (MCI). This review will cover two levels of variability that affect epidemiological research on dementia and MCI: the conceptual and the operational level. More specifically, it is highlighted how the lack of a precise definition of MCI leads to a greater variability in the occurrence estimates of this condition, when compared to dementia. Variability will decrease only when more precise criteria and aims of the concept "MCI" will be specified.