Among dementia patients in long-term care facilities, neuroleptics (NL) are frequently prescribed for the treatment of agitation. Although good clinical practice and federal law mandate attempts at withdrawal of these medications, empiric data regarding the cessation of NL treatment are limited in this population. The objective of the present study was to assess through direct observation the effects of short-term NL withdrawal on physically aggressive behavior and other aspects of agitation. We carried out a randomized, double-blind, baseline NL-controlled 4-week trial of the effects of NL withdrawal in 36 institutionalized patients with possible or probable Alzheimer's disease. Patients were directly observed for four 2-hour sessions during baseline, a prerandomization week, and during weeks 1, 2, and 4 of the double-blind portion of the study. Completion of the 4 weeks of double-blind medication and number of observed episodes of physically aggressive behavior (PAB) were the two primary outcome measures. Of the 22 patients who were withdrawn from NL, 20 (91%) completed the 4-week double-blinded withdrawal. Two patients were discontinued from the study due to unacceptable levels of agitation at the request of their nursing staff. Of the 14 patients not withdrawn, all completed the 4-week trial. The chi-square test for the difference between groups was not significant (P > .05). Based on the observed instances of PAB, there was no significant difference (P > .05) between withdrawn and not-withdrawn subjects. Half of the withdrawn patients remained off NLs for an extended period of time after the end of the study, even after the blind was broken. Withdrawing institutionalized dementia patients from NLs was successful in most but not all study patients. Generalization of these results is limited by the highly selected nature of the participants. Unmasking of unmanageable agitation and physical aggressiveness in a small minority must be weighed against the benefits of removing unnecessary medication in the majority of dementia patients in whom NL withdrawal is attempted. PAB itself should not drive continuing NL use without regard to objective assessment of efficacy of the NL treatment.