We compared the diagnostic capabilities of MRI to CT, evoked potentials (EP), and CSF oligoclonal banding analysis in a prospective evaluation of 200 patients with suspected multiple sclerosis (MS). MRI was the best method for demonstrating dissemination in space. An abnormal appropriate EP in monosymptomatic disease was usually supported by MRI and CSF analysis as being predictive of MS as a clinical diagnosis. A normal appropriate EP study was not satisfactory because MRI and CSF analysis often did not support a diagnosis of non-MS. When there is agreement between three of these paraclinical studies, the diagnosis of MS is probably unequivocal. For use in research studies, laboratory-supported definite MS (LSDMS) could be diagnosed in 85 patients of the total 200 (42.5%), in 19/38 (50%) of optic neuritis (ON) patients, and in 24/52 (46%) of chronic progressive myelopathy (CPM) patients. MRI was 100% successful in identifying patients who qualified for LSDMS in the ON and CPM groups. In a short follow-up (less than 1 year), 19/200 (10%) went on to develop clinically definite MS (CDMS), and MRI predicted that diagnosis in 18/19 (95%). Only long-term follow-up will show how well these studies and the category of LSDMS predict the development of CDMS. The clinical diagnosis of MS (CDMS), even though only 95% accurate, must remain the gold standard.