Minimal short-term benefits with oral physostigmine have been reported in Alzheimer's disease. We examined long-term oral physostigmine therapy in 14 patients with probable Alzheimer's disease in an extended double-blind crossover trial in which they received physostigmine for five intervals of 4 to 6 weeks in length and placebo for one interval determined at random. At each interval a Selective Reminding Test was administered and three indexes were examined: total recall, long-term recall, and intrusions. As a group, patients' memory test scores were significantly better during the drug period. Two patients who previously had performed worse on drug than on placebo in a short-term trial continued to do so. Nine of the remaining 12 patients performed better on two or more measures on physostigmine compared with placebo. In some cases scores improved up to 50% over placebo values. This improvement was not predicted by response to medication in the earlier short-term trial. Seven patients completed an additional six-interval crossover trial. All but one continued to have improved performance with oral physostigmine and, in grouped data, memory measures remained significantly improved. These results suggest that long-term administration of physostigmine in Alzheimer's disease may be more effective than short-term. This benefit can be sustained for up to a year in some patients.