For agitated dementia showing insufficient response to conventional antipsychotics, the feasibility of transition to atypical agents remains unknown. Sixty-two Chinese inpatients with dementia and disruptive behaviors were recruited into an 8-week screening trial of haloperidol. Thirty-five (56%) of them responded insufficiently. They then entered a prospective, 16-week, open-labeled study. Haloperidol was abruptly shifted to risperidone 0.5 mg/day at weeks 1 to 4 and then 1 mg/day at weeks 5 to 12. At weeks 13 to 16, the regimen was shifted back to haloperidol at previous doses, mostly 1 mg/day. Safety, efficacy, cognition, and moods were evaluated at least every 4 weeks. Generalized estimating equation methods were used for determining the effects of the prognostic variables on the outcome values. Risperidone, particularly at 0.5 mg/day, was generally tolerable. The Brief Psychiatric Rating Scale (BPRS) score decreased progressively under risperidone treatment; at week 12, 16 (46%) patients showed response (≥25% reduction in the BPRS). Patients with vascular dementia were more likely to respond than those with Alzheimer’s disease (p= 0.02). Haloperidol reinstitution resulted in no further improvement, except trend increments in motor symptoms. Risperidone also tended to benefit the performance on the Behavioral Pathology in Alzheimer’s Disease Rating Scale. Six (17%) patients improved on moods and self-care with risperidone. These preliminary results suggest that crossover from haloperidol to risperidone is generally safe and effective and may produce favorable moods in agitated dementia patients. Vascular dementia is a predictor of treatment response. In contrast to the dose (1 mg/day) recommended for most white individuals, 0.5 mg/day could be tried at first in Chinese patients. Because of the design’s limitations, further controlled studies are warranted.