In a study of 226 elderly residents in a retirement village in Sydney, Australia, general practitioners' opinions about dementia status had high positive and negative predictive values and high specificity, but low sensitivity when evaluated against the mini-mental state examination and the Blessed dementia rating scale. General practitioners were found to disagree with these two measures more often when patients were in advanced old age, and when they considered the patients to be depressed. We conclude that the general practitioner can increase his or her sensitivity to dementia in the elderly by use of either measure.