The aim of the present study was to identify determinants of attrition in a natural history study of a tertiary care sample of patients with Alzheimer disease (AD) and control subjects. A longitudinal study was performed with 978 patients with AD and 466 control subjects age 50 years and older enrolled at 25 sites of the Consortium to Establish a Registry for Alzheimer's Disease between January 1987 and January 1992; subjects were followed annually for up to 78 months. Both descriptive statistics and polytomous logistic regressions were run to identify determinants of attrition. Of the 1,444 subjects enrolled, 10.5% dropped out after initial evaluation, 31.0% provided at least two waves of data, and 58.4% provided complete follow-up. Inadequate involvement by the site, non-white status, and patient's spouse not enrolled in the study were predictive of dropout; cessation of participation because of death (which may have precluded dropout) predicted continuation in the study. Age, level of education, severity of dementia, and rapidity of progression of disease did not predict dropout. Level-of-site commitment was the most significant determinant of continued participation in this natural history study of AD, followed by white race, and the inclusion of both husband and wife where one is a patient and the other a control subject.