Plasma exchange has been reported to be efficacious in chronic inflammatory demyelinating polyradiculoneuropathy. We performed a prospective double-blind trial in which patients with static or worsening disease were randomly assigned to plasma exchange (n = 15) or to sham exchange (n = 14) for three weeks. After three weeks, we observed statistically significant differences in combined measurements of nerve conduction (total, motor, proximal, velocity, and amplitude) favoring patients who had received plasma exchange. Improvement to a greater degree than for any patient receiving sham exchange was detected in the neurologic-disability score in five patients (P = 0.025) and in subset scores for weakness and reflex in four patients (P less than 0.057). We conclude that for some patients with chronic inflammatory demyelinating polyradiculoneuropathy, plasma exchange has an ameliorating effect on neurologic dysfunction and nerve conduction, but in others no improvement is observed. Because plasma was replaced with normal serum albumin, a humoral factor or factors may have a role in the neurologic deficit of this disorder.