Objective: Limited work has addressed the construct validity of the Cohen-Mansfield Agitation Inventory (CMAI) in nursing home patients. The authors tested the CMAI factor structure by use of data from three samples of nursing home patients categorized a priori as having at least mild behavioral and psychological symptoms of dementia. Methods: CMAI data were from the baseline assessment of three randomized, placebo-controlled trials of risperidone for treating elderly nursing home patients. Exploratory factor analyses were conducted on two trials (N = 304; N = 344), and the results of these exploratory factor analyses were then tested with confirmatory factor analysis by use of data from a third trial (N = 617). Results: Principal-components analysis suggested the presence of four factors: 1) Aggressive Behavior (hitting, kicking, scratching, biting, pushing, grabbing, throwing things, cursing or verbal aggression, spitting, tearing things/destroying property, hurting self or others, screaming); 2) Physically Non-Aggressive Behavior (pacing, trying to get to a different place, general restlessness, inappropriate dressing or disrobing, handling things inappropriately, performing repetitious mannerisms); 3) Verbally Agitated Behavior (complaining, constant requests for attention, repetitive questions, negativism); and 4) Hiding and Hoarding. Confirmatory factor analysis indicated that the four-factor solution was a reasonable fit to the data. Conclusion: Four factors emerged on the CMAI in nursing home patients with behavioral and psychological symptoms of dementia. The results obtained converge reasonably with previous publications concerning the factor structure of the CMAI, which, taken together, suggest a fairly robust factor structure for the instrument.