OBJECTIVES: To investigate the consistency of physician judgments of treatment consent capacity (competency) for patients with Alzheimer's disease (AD) when specific legal standards (LS) for competency are used, and to identify the LS most clinically relevant to experienced physicians. DESIGN: Control and AD patient participants were videotaped being administered a measure of capacity to consent to medical treatment. Study physicians viewed videotapes of these assessments individually and made competency judgments for each participant under different LS followed by their own personal judgment of competency. SETTING: A university medical center. PARTICIPANTS: Participants were 10 older controls and 21 patients with AD (10 with mild and 11 with moderate AD). Five physicians with experience assessing the competency of AD patients were recruited from the geriatric psychiatry, geriatric medicine, and neurology services of a university medical center. MEASUREMENTS: The 31 participants were videotaped performing on a measure of treatment consent capacity (Capacity to Consent to Treatment Instrument) (CCTI). The CCTI consists of two clinical vignettes (A-neoplasm and B-cardiac) that test competency under five LS. Vignette A and B assessments were videotaped separately for each participant (total videotapes for sample = 62). Each study physician viewed each videotaped vignette individually, made judgments under each of the LS (competent or incompetent), and then made his/her own personal competency judgment. Physicians were blinded to participant diagnosis. Within participant group, consistency of physician judgments was evaluated across LS and personal judgments using percentage agreement and kappa. Agreement between personal and LS judgments for the AD group was evaluated for each physician using logistic regression. RESULTS: As expected, physicians as a group generally demonstrated very high percentage agreement in their LS and personal competency judgments for the control group. For the AD group, mean percentage judgment agreement among physicians ranged from a high of 84% (LS1) (evidencing a treatment choice) to a low of 67% (LS3) (appreciating consequences of treatment choice). Mean percentage agreement for personal competency judgments was 76%. For the AD sample, kappa analyses for physicians as a group demonstrated significant agreement not attributable to chance for LS5 (understanding treatment situation/choices) (k = 0.57, P =.001), LS4 (providing rational reasons for treatment choice) (k = 0.39, P =.04), and also for personal judgments (k = 0.48, P =.009). Analysis of LS judgment agreement within physician indicated that physicians applied the LS as discrete standards. Within-physician and for the AD sample, personal competency judgments were associated significantly with judgments on LS5 (P =.001), LS4 (P =.004), and LS3 (P <.04). See also p 919 and p 1014 CONCLUSIONS: Experienced physicians demonstrated significant agreement assessing competency in AD patients when judgments were based upon specific legal standards. Personal competency judgments of physicians showed a substantially higher level of agreement than found in a previous study, where specific LS were not used. These results suggest that consistency of physician competency judgments can be enhanced if they are guided by knowledge of specific LS. Physicians' personal competency judgments were most closely associated with comprehension and reasoning LS, the most conservative and clinically appropriate standards for deciding competency. J Am Geriatr Soc 48:911-918, 2000.