Objective: To identify cognitive predictors of competency performance and status in Alzheimer's disease (AD) using three differentially stringent legal standards for capacity to consent. Design: Univariate and multivariate analyses of independent neuropsychological test measures with three dependent measures of competency to consent to treatment. Setting: University medical center. Subjects: 15 normal older controls and 29 patients with probable AD (15 mild and 14 moderate). Main Outcome Measures: Subjects were administered a battery of neuropsychological measures theoretically linked to competency function, as well as two clinical vignettes testing capacity to consent to medical treatment under five legal standards (LSs). The present study focused on three differentially stringent LSs: the capacity simply to 'evidence a treatment of choice' (LS1), which is a minimal standard; the capacity to 'appreciate the consequences' of a treatment choice (LS3), a moderately stringent standard; and the capacity to 'understand the treatment situation and choices' (LS5), the most stringent standard. Control subject and AD patient neuropsychological test scores were correlated with scores on the three LSs. The resulting univariate correlates were then analyzed using stepwise regression and discriminant function to identify key multivariate predictors of competency performance and status under each LS. Results: No neuropsychological measures predicted control group performance on the LSs. For the AD group, a measure of simple auditory comprehension predicted LS1 performance (r2 = 0.44, p < 0.0001), a word fluency measure predicted LS3 performance (r2 = 0.58, p < 0.0001), and measures of conceptualization and confrontation naming together predicted LS5 performance (r2 = 0.81, p < 0.0001). Under discriminant function analysis, confrontation naming was the best single predictor of LS1 competency status for all subjects, correctly classifying 96% of cases (42/44). Measures of visuomotor tracking and confrontation naming were the best single predictors, respectively, of competency status under LS3 (91% [39/43]) and LS5 (98% [43/44]). Conclusions: Multiple cognitive functions are associated with loss of competency in AD. Deficits in conceptualization, semantic memory, and probably verbal recall are associated with the declining capacity of mild AD patients to understand a treatment situation and choices (LS5); executive dysfunction with the declining capacity of mild to moderate AD patients to identify the consequences of a treatment choice (LS3); and receptive aphasia and severe dysnomia with the declining capacity of advanced AD patients to evidence a simple treatment choice (LS1). The results offer insight into the relationship between different legal thresholds of competency and the progressive cognitive changes characteristic of AD, and represent an initial step toward a neurologic model of competency.