To the Editor: Dr. Frisoni and colleagues report that in Alzheimer’s patients with behavioral disturbances, a program of nonpharmacologic intervention in special care units was successful. Whereas this approach may indeed work in many patients with dementia, methodologic flaws in their study make it difficult to draw any firm conclusions. Patients were not randomly assigned to the special care units or to traditional nursing homes, and this probably contributed to the patients in the special care units having significantly more severe behavioral disturbances than the patients in traditional nursing homes. The well-known statistical phenomenon of regression to the mean, whereby outliers at one assessment time tend to drift closer to the mean at the next assessment, may partly explain the observed clinical improvement, especially because the group in the special care units started with more severe behavioral disturbances and, hence, was more likely to move down toward the mean. Also, placebo-controlled pharmacotherapy trials of behavioral disturbances in patients with dementia consistently report placebo response rates ranging from 20%–50%, and it is common for patients taking placebo to show considerable improvements in behavioral symptoms (1–3). Therefore, in the absence of an adequate control group, the nearly identical decrease in symptoms in the special care unit and nursing home groups (38% and 41%, respectively) is difficult to interpret.