Many of the problems associated with the clinical application of event-related potentials (ERPs) are well recognized. These include the broad inter-subject variability of peak latencies and amplitudes in control groups, as well as the apparent non-specificity of some ERP tests. In addition, it is often the case that the same test yields a different proportion of 'hits' (in terms of identification of abnormality in patients classified by other means as having dementia) when used by different research groups. This is almost certainly due to differences in the disease state of the patients being tested. The difference between hit rates identifies another important factor which is not often considered in the application of ERP tests in clinical medicine. This is that although an ERP test may identify abnormality in the later stages of disease, the true task for the test is to identify abnormality in early disease when evidence of cognitive dysfunction will be different and less frank than in later stages of the disorder. It is therefore, necessary to identify the cognitive deficits associated with the early stages of dementia and devise appropriate ERP tests to detect these deficits. A recent paper by Fox et al. (Fox, N.C., Warrington, E.K., Seiffer, A.L., Agnew, S.K., Rossor, M.N., 1998. Presymptomatic cognitive deficits in individuals at risk of familial Alzheimer's disease. A longitudinal prospective study. Brain 121, 1631-1639.) suggests that ERP tests of verbal memory function may be the most sensitive for detecting the early stages of Alzheimer's disease.