Objective: to determine whether medication use patterns in community-dwelling elderly people vary with level of cognitive function—dementia, cognitive impairment (but not dementia) and intact cognition. Design: cross-sectional survey. Setting: a five-county area of central North Carolina, USA. Participants: 520 members of the Duke Established Populations for Epidemiologic Studies of the Elderly. Measurements: medication use in the previous 2 weeks was ascertained during a interview in the patient's home and was coded as to prescription and therapeutic class status. Cognitive status, the primary independent variable, was divided into: (i) dementia (n = 100); (ii) cognitive impairment but not dementia (n = 117); and (iii) cognitively intact (n = 303). The dependent variables were any prescription or over-the-counter (OTC) medication use (vs nonuse); number of prescription or OTC medications used; and prescription and OTC use combined within major therapeutic classes. Multivariate analyses controlled for socio-demographic characteristics, health status, functional status and access to health care. Results: the use of any prescription medication was similar in the three groups. The demented were significantly less likely than cognitively impaired people to use any OTC medications (OR = 0.65, 95% CI = 0.45, 0.93), cardiovascular medications (OR = 0.70, 95% CI = 0.49, 0.99) and analgesics (OR = 0.54, 95% CI = 0.39, 0.75). As a combined group, those who were demented and cognitively impaired were less likely than the cognitively intact group to use any OTC medications (OR = 0.78, 95% CI 0.65, 0.92). Compared with the cognitively impaired subjects, the demented group took fewer prescription medications (β coefficient = −0.31, 95% CI = −0.59, −0.03) and similar numbers of OTC medications. Compared with those who were cognitively intact, the combined group of demented and cognitively impaired subjects took fewer OTC medications (β coefficient = −0.14, 95% CI = −0.23, −0.05) and similar numbers of prescription medications. Conclusion: increasing level of cognitive dysfunction is associated with decreased use of OTC, cardiovascular and analgesic medications and the use of fewer prescription medications. These results suggest important differences in medication use patterns among community-dwelling elderly people who vary in cognitive status.