Five male patients participated in a pilot open-label study of dose-related aspects of response to intracerebroventricular bethanechol in Alzheimer's disease. No patient had remission of symptoms, but three patients improved symptomatically and on tests of memory. Improvement was evident over a restricted range of doses for each subject, and symptoms were worse at doses below and above the optimal range. There was little overlap in the range of doses producing improvement among these three. Two patients had no consistent improvement in memory, and agitation, depression, paranoia, and seizures developed during treatment. Qualitative differences and variability in dosages producing responses complicate the identification of true drug response in the treatment of Alzheimer's disease.