The specificity of magnetic resonance (MR) imaging in the diagnosis of multiple sclerosis (MS) has not been measured systematically. Conventional MR head images with sagittal localizer and axial multiple-echo sequences with long repetition times were obtained in 92 patients with clinically verified MS (Schumacher criteria), 100 healthy volunteers, 60 subjects with hypertension, and eight patients with dementia. Two readers, without the aid of any clinical or demographic information, classified each of the 260 studies as MS or not MS. The readers classified the studies again after being supplied with the subjects' ages and sex. True-negative and true-positive diagnoses of MS were tabulated. The specificity of the MR diagnosis of MS (true-negative results in proportion to all non-MS studies) was 95%-99% with all the control groups included. There is a small risk of misinterpreting incidental periventricular white matter foci as plaques of MS in MR studies.