Normal spouses may be an excellent source of control subjects in research on noninstitutionalized Alzheimer's disease (AD) cases. To determine to what extent spouses might differ from nonkin controls, we compared information on 145 spouses of AD patients with 158 nonkin community volunteers enrolled by 23 centers of the Consortium to Establish a Registry for Alzheimer's Disease. Chi square and 1-way ANOVA analyses indicate that neither at entry nor over the next two years did the two groups of controls differ significantly on demographic characteristics, health status, or performance on neuropsychological measures. However, hierarchical logistic regression showed that, after controlling for demographic characteristics and physical status, dropout was nearly twice as high among spouse controls. Dropout was related to nonreturn of the AD case. Thus, as controls, spouses are viable as long as the cases to whom they are married remain in the study.