An 8-year longitudinal study of elderly people has provided data concerning age-associated memory impairment (AAMI). In 1985 a random sample of 146 persons aged 65 years or more, living in their own homes, were assessed using the Guild Memory Test, the Mini-Mental State Examination (MMSE) and other ratings. After excluding 21% of the sample because they scored less than 24 on the MMSE, and another 34% who fulfilled other exclusion criteria, some 48% of the remainder (22% of the total sample) clearly fulfilled NIMH criteria for AAMI and a further 36% (16% of the total sample) were recorded as forgetful. The NIMH criteria are appropriate for certain research purposes but not in assessing prevalence of memory disorders. Follow-up interviews were conducted after 2, 4, 6 and 8 years. The mortality rate and development of dementia among those fulfilling criteria for AAMI appeared similar to the other non-demented groups of subjects; the mortality rate of those with MMSE scores below 24 was significantly higher. Guild test results at 2-yearly intervals showed considerable changes; half of those scoring least well who were retested showed improvement.