Survivors from a nationally representative sample of elderly people originally screened in 1985 were reassessed in 1989 and again in 1993. On each occasion respondents were rated as cognitively impaired, borderline impaired or unimpaired (using a brief information/orientation scale), with the validity of these ratings assessed in subsequent clinical interviews. Where follow-up screening was not possible, information was derived from death certificates and hospital case-notes. Over 8 years (1985–93) the overall incidence rate per person–year at risk was 1·58%, giving age-specific rates of 0·72, 1·32, 1·63, 3·46, 2·55 and 1·41% for the age groups 65–69, 70–74, 75–79, 80–84, 85–89 and ≥ 90 respectively. Of 43 individuals classified at screening as borderline impaired in 1985 and 1989, 19 were diagnosed as demented at clinical interviews conducted within 16 weeks of screening. Four-year follow-ups among the remaining 24 showed that 15 had died, while 6 showed a worsened cognitive status. Controlling for both age and sex, aggregated 4-year mortality was significantly higher among those defined at screening in 1985 and 1989 as either impaired or borderline, when compared with the unimpaired.