We performed electrophysiological and serological testing within 15 days of symptom onset on 369 patients with Guillain-Barré Syndrome (GBS) enrolled in a trial comparing plasma exchange, intravenous immunoglobulin, and both treatments. Patients were classified into five groups by motor nerve conduction criteria; 69% were demyelinating, 3% axonal, 3% inexcitable, 2% normal, and 23% equivocal. Six of 10 (60%) patients with axonal neurophysiology had had a preceding diarrheal illness compared with 71 of 359 (20%) in other groups. Antiganglioside GM1 antibodies were present in a higher proportion of patients with axonal physiology or inexcitable nerves than other patients. The number dead or unable to walk unaided at 48 weeks was greater in the group with initially inexcitable nerves (6 of 12, 50%) compared with the rest (52 of 357, 15%), but was not significantly different between the axonal (1 of 10, 10%) and demyelinating (44 of 254, 17%) groups. Sensory action potentials and clinical sensory examination were both normal in 53 of 342 (16%) patients, and these "pure motor GBS" patients were more likely than other GBS patients to have IgG antiganglioside GM1 antibodies and to have had preceding diarrhea but had a similar outcome. The axonal group was more likely than other groups to have normal sensory action potentials. The outcomes in response to the three treatments did not differ in any subgroup (including patients with pure motor GBS or preceding diarrhea) or any neurophysiological category.