A census was taken on 26 long-term care facilities in the Ottawa-Carleton Region comprising 3, 825 residents to estimate the point prevalence of cognitive impairment or dementia. Nurses familiar with the residents identified those who had either memory impairment or dementia (according to DSM III criteria). A validation of the data obtained through the census was carried out on a stratified, random sample of residents. The long-term care facilities were stratified by level of care and by size. From each of the three strata, 11 facilities were randomly selected then a 10 per cent random sample with replacement (411 residents) was determined. The validation survey consisted of interviewing residents using two screening instruments, interviewing the nurse using two behavioural instruments and a chart audit. On the census the prevalence for cognitive impairment in the surveyed group was 67 per cent compared to 61 per cent for the whole resident population. The sensitivity and specificity for cognitive impairment was calculated at 84 and 70 per cent respectively. The prevalence of dementia in the surveyed group was 61 per cent compared to a census estimate of 52 per cent for the whole resident population. The overall sensitivity of the census for dementia was 95 per cent and the specificity was 83 per cent. The results indicated that a census undertaken with standardized definitions, by knowledgeable staff directly involved with patient care is a reasonably effective method for estimating the prevalence of persons with dementia in long-term care facilities.