An attempt was made to assess in detail subjects screened as suffering from dementia among a North London community sample of elderly people. Forty-eight (80%) subjects were interviewed, 7 (12%) were found to have died and 5 (8%) either refused interview or were lost to follow-up. By clinicians' diagnosis of the 48 interviewed, 22 subjects (46%) had probable Alzheimer's disease, one had multi-infarct dementia, five had mixed dementia, five had secondary dementia, 10 had a dementia which could not be further classified and 5 were not demented. No subject had a reversible condition. The prevalence rate for clinical dementia was 6.1%, and for Alzheimer's disease 3.1%. According to AGECAT diagnosis the prevalence of organicity was 5.7%. The AGECAT diagnoses and psychiatrists' diagnoses were significantly associated (P less than 0.003) and AGECAT was more likely to identify as organic those subjects with dementia diagnosed by psychiatrists as Alzheimer's disease, than those not so diagnosed (P less than 0.04). A short psychometric battery, including the MMSE in two versions, was administered and its acceptability to a community sample evaluated. This detailed clinical investigation showed that the Dementia Diagnostic Scale of the Short-CARE was a specific predictor of clinical dementia or death at the time of follow-up, whereas the more inclusive Organic Brain Syndrome scale was a more satisfactory first phase screening instrument.