This study tests the association between residents' cognitive impairment and nursing homes' pain management practices. We used chart abstraction to collect data on 551 adults in six North Carolina nursing homes. From the standard data collected in the Minimum Data Set, 24% of residents experienced pain in the preceding week. Reports of pain decreased as cognitive abilities declined: nurses completing the Minimum Data Set reported pain prevalence of 34%, 31%, 24%, and 10%, respectively, for residents with no, mild, moderate, and severe cognitive impairment (P<0.001), demonstrating a "dose-response"-type result. Eighty percent of cognitively intact residents received pain medications, compared to 56% of residents with severe impairment (P<0.001). Cognitively impaired residents had fewer orders for scheduled pain medications than did their less cognitively impaired peers. Yet the presence of diagnoses likely to cause pain did not vary based on residents' cognitive status. We conclude that pain is underrecognized in nursing home residents with cognitive impairment and that cognitively impaired residents often have orders for "as needed" analgesics when scheduled medications would be more appropriate.