We evaluated patient-proxy agreement in a population of veterans with Parkinson's disease and compared levels of agreement by patient subgroups. Patient and caregiver pairs completed questionnaires composed of standard measures and additional demographic and activity questions. Participants completed the Center for Epidemiologic Studies Depression Scale (CES-D), the PD Questionnaire 39 (PDQ-39), and three questions regarding physical activity. Caregivers completed proxy forms of the PDQ-39 and the physical activity questionnaire. The proxy forms asked caregivers to choose the answers that best described their "friend's/patient's situation." The results of our comparison of patient and proxy reports of quality of life were consistent with findings in other diseases. On average, proxies rated patient disability higher and quality of life lower than did patients. However, our comparison of patient and proxy reports of frequency of exercise diverged from previously published work. Less agreement was observed between patient and proxy reports of physical activity, even though this is a more objective variable than are the domains measured by the PDQ-39. Proxy reports may diverge appreciably from patient self-reports. These differences should be considered in research design and clinical decision making. Alternative approaches to the measurement of patient relevant outcomes could supplement traditional, retrospective self-reports.