Serial MRI and clinical testing was performed on 45 well-defined untreated multiple sclerosis patients in different categories of disease (relapsing-remitting, progressive, stable). Up to 24 MRIs were scheduled over a 1-year period for each patient. Clinical evaluation was performed monthly and at times of attacks using the Expanded Disability Status Scale (EDSS) and the Ambulation Index (AI). MRI scans were performed both with and without gadolinium enhancement. MRI lesion volume was determined by computerized analysis and gadolinium-enhancing lesions were counted by radiologists. We observed an increase in lesion volume over 1 year in all patient groups except those classified clinically as stable. In relapsing-remitting patients there were correlations between increases in the number of gadolinium enhancing lesions and increases in EDSS and the occurrence of attacks. In chronic progressive patients, increases in lesion volume were correlated with both increases in EDSS and AI. These results demonstrate a linkage between MRI and clinical disease that depends both on the stage of MS and the MRI measures used and support the use of MRI as a surrogate marker of clinical disability in the study of multiple sclerosis.