To assess the predictive value of prolongation of visual evoked response (VER) plus the presence of oligoclonal bands in the cerebrospinal fluid (CSF) in the diagnosis of multiple sclerosis (MS) in patients with isolated cord lesions, the false-positive rate for the two tests combined was determined by a prospective analysis of 42 patients with a structural spinal cord lesion. The false-positive rate for the combination of VER and CSF abnormalities was zero, but the individual false-positive rates were 10% for VER and 12% for CSF protein oligoclonal bands. According to Bayes' theorem, and taking into account the false-positive rates of the tests in the population studied and the prevalence of MS and the true-positive rates for the tests as derived from published reports, abnormalities of VER and CSF protein together gave a probability of MS of 100%. However, if only either the VER or the CSF were abnormal, the probability of MS was 44% or 49%, respectively. An abnormal result for both tests thus indicates a sufficiently high probability of MS to exclude myelography, but a positive result in only one test warrants the procedure.