A new method for ascertaining dementia in epidemiologic research and the results of a study to evaluate it are described. The method relies on an expert panel of clinicians reviewing clinically relevant information collected by lay evaluators to arrive at a diagnosis based on DSM-III-R criteria. The approach was developed to study dementia in a statewide sample of over 2400 new admissions to 59 nursing homes in Maryland. Expert panel ascertainment of dementia was compared to that obtained by direct clinical evaluation for 100 nursing home residents. Agreement between the panel and direct assessment was 76% (kappa = 0.59) using a three-category classification of dementia, no dementia, and indeterminate. This ascertainment strategy provides an alternative to methods currently in use and is particularly well-suited for populations with a high prevalence, in those dispersed over large geographic areas, and when timely, cost-effective evaluations are required.