The efficacy and safety of risperidone, haloperidol, and thioridazine for treating dementia-associated behavioral disturbances were evaluated in a retrospective study of 186 patients aged 65 years or older with DSM-III-R or DSM-IV diagnoses of Alzheimer's dementia, senile dementia NOS, or organic brain syndrome. Study patients were selected from the charts of 12,000 residents of 60 long-term-care facilities in New Jersey if they were treated with one of the three agents for behaviours dangerous to themselves or others. The 186 selected patients included 60 treatedwith risperidone (mean, 1 mg/day), 83 with haloperidol (mean, 2 mg/day), and 43 with thioridazine (mean, 33 mg/day). Target behaviors were violence (74 patients), shouting (31), delusions (26), paranoia (19), pacing (3), and mixed behaviors (33). Target behaviors improved in 94% of patients given risperidone, 65% given haloperidol, and 67% given thioridazine (p < .001). Treatment failures (treatment discontinued in patients because of side effects or no improvement) were more frequent in patients started on haloperidol (28) or thioridazine (15) than on risperidone (3). Extrapyramidal symptoms were reported in 7% of patients taking risperidone, 22% taking haloperidol, and 18% taking thioridazine. Safe, effective doses are readily achieved with risperidone but difficult to achieve with haloperidol or thioridazine because their effective doses often cause unacceptable side effects. These data are only suggestive because no guidelines exist for defining or measuring behavioral disturbances or for how they are affected by social, psychological, or environmental factors.