Despite the advance represented by nuclear magnetic resonance imaging, the diagnosis of multiple sclerosis (MS) remains clinical. It has been our impression that a diagnosis of MS reduces further critical neurological thought and assumes an air of certainty after a number of visits in the absence of any new evidence for the diagnosis. One of us (E.G.S-W.), as part of a nationwide epidemiological study, reviewed 387 patients in Western Australia who had been diagnosed as having MS. Three diagnostic sub-groups were identified: group A--318 patients considered to have MS, group B--32 patients with an alternative chronic neurological disorder, and group C--35 patients not considered to have a neurological illness. Patients in group C were predominantly females (89%) and nurses (34%) with an earlier age of onset of symptoms (mean 28 years, range 12-46 years). Frequent clinical features included monocular diplopia, tunnel vision, 'give-way' weakness and hemisensory loss of all modalities of sensation. The major clinical features of patients in group B, which were atypical for MS, included onset of non-remitting, progressive disease before age 35 years, localised disease, and normal optic nerves and eye movements. The results of this study serve to promote critical analysis of the accuracy of diagnosis of MS in individual cases.