We performed a double-blind, placebo-controlled trial of intrathecal baclofen (ITB) in stiff-man syndrome. Three patients, unresponsive to current therapy, received 50 mu g of ITB or placebo on sequential days. Following ITB, all patients demonstrated improvement in reflex EMG activity. The mean reduction in total EMG activity (from all muscles) following stimulation of the medial plantar nerve (cutaneous flexor reflex) was 72% following 50 mu g of ITB compared with 18% following placebo (ANOVA: significance of F, p less than 0.0001). The mean latency to onset of the response was also significantly prolonged for all muscles following ITB (ANOVA: significance of F, p less than 0.05). Although reflex EMG activity was reduced in all patients, clinical improvement was evident in only one patient, who differed from the others studied by a longer duration of disease, greater severity of stiffness, less fear of falling, and greater electrophysiologic improvement.