Multimodal evoked potentials (EPs) were obtained serially in 57 patients with multiple sclerosis enrolled in a double-blind, controlled study of hyperbaric oxygen. We encountered drawbacks in the use of EPs in monitoring effects of a treatment for chronic multiple sclerosis. Limitations include a relatively arduous testing protocol and the significant occurrence of ambiguous and frankly unreproducible results. Test/retest variability is large in stable patients. There is poor correspondence between EP changes and changes of clinical function as well as among EP modalities. Whether EP measurements ultimately prove to be useful in testing new therapies awaits the development of a truly effective treatment, the effects of which can be shown despite these factors and limitations.