To assess empirically the competency of patients with Alzheimer's disease (AD) to consent to medical treatment under different legal standards (LSs). Comparison of normal older subjects and patients with AD on measures of competency to consent to medical treatment. University medical center. Normal older control subjects (n=15) and patients with probable AD (n=29 [15 with mild and 14 with moderate AD]). Two specialized clinical vignettes were developed that test a subject's capacity to consent to medical treatment under five wellestablished LSs for this competency: LS1, evidencing treatment choice; LS2, making the reasonable choice; LS3, appreciating consequences of choice; LS4, providing rational reasons for choice; and LS5, understanding treatment situation and choices. Performance on the LSs was compared across control and AD groups using Student's ttest, X2, and analysis of variance. Demented subjects were categorized as competent, marginally competent, or incompetent under each LS by using a cutoff score derived from normal control performance. No differences between groups emerged for LS1 and LS2. Control subjects performed significantly better than patients with mild AD on LS4 and LS5, and significantly better than patients with moderate AD on LS3, LS4, and LS5. Patients with mild AD performed significantly better than patients with moderate AD on LS4 and LS5. With respect to competency status, patients with AD showed a consistent and progressive pattern of compromise (marginal competence or incompetence) related to dementia severity and stringency of the LS. A reliable prototype instrument validly discriminated the competency performance and classified the competency status of control subjects and patients with mild and moderate AD under five LSs for competency to consent to medical treatment. While the groups performed equivalently on minimal standards requiring merely a treatment choice (LS1) or the reasonable treatment choice (LS2), patients with mild AD had difficulty with more difficult standards requiring rational reasons (LS4) and understanding treatment information (LS5), and patients with moderate AD had difficulty with appreciation of consequences (LS3), rational reasons (LS4), and understanding treatment (LS5). The results raised the concern that many patients with mild AD may not be competent to consent to treatment and supported the value of standardized clinical vignettes for assessment of competency in dementia. © 1995 Arch Neurol All rights reserved.