Despite the proliferation of instruments developed to rate behavioral disturbances associated with dementia, systematic studies of how ratings on these instruments should be analyzed to measure change in disruptive behaviors or distressing symptoms (ie treatment response) are noticeably absent. Using one of these scales, we compared three methods to characterize treatment response in 52 elderly demented inpatients who participated in a standardized neuroleptic trial. While all three analyses identified a statistically significant improvement, they conveyed differently the clinical improvement experienced by the patients. Categorical outcomes communicated the clinical meaning of improvement better than changes in total score; changes in factor scores best revealed the differential impact of treatment on specific behavioral and symptomatic domains. Given the heterogeneity of the problematic behaviors and symptoms exhibited by demented patients included in treatment trials, regardless of the intervention being tested or of the instrument being used to rate behaviors, a focused approach to characterizing treatment response is needed.