Proton magnetic resonance spectroscopy (MRS) and MRI were carried out in 11 patients with multiple sclerosis who had clinical evidence of severe cerebellar involvement, 11 multiple sclerosis patients (of similar age and disease duration) who had minimal or no signs of cerebellar disease, eight patients with autosomal dominant cerebellar ataxia (ADCA) and 11 healthy controls. In all subjects MRS was localized to cerebellar white matter (volumes of interest 3-6 ml). Apparent metabolite concentrations were calculated using the fully relaxed water spectrum as an internal standard of reference. The patients also underwent MRI to assess cerebellar volume and (in the two multiple sclerosis groups) lesion volume within the posterior fossa. Magnetic resonance spectroscopy from cerebellar white matter showed a highly significant reduction in the concentration of N-acetyl groups (NA) [which consists predominantly of N-acetylaspartate (NAA), a neuronal marker] in the multiple sclerosis group with cerebellar deficit compared with the multiple sclerosis group with minimal or no signs of cerebellar involvement, and healthy controls. Follow-up MRS performed in six of the multiple sclerosis patients 9 months later showed no change in the median NA concentration. The ADCA group showed a significant reduction of NA from a region of cerebellar white matter and also a reduction in the concentration of choline-containing compounds. The multiple sclerosis group with severe cerebellar deficit and the ADCA group both had significant cerebellar atrophy (suggesting nerve cell body and axon loss) compared with the multiple sclerosis patients with minimal or no signs of cerebellar deficit and healthy controls. The multiple sclerosis patients with cerebellar deficit had a significantly greater lesion volume in the posterior fossa, although the proportion of the spectroscopic voxel occupied by lesions was small, suggesting that axonal loss from normal appearing white matter also contributes to the observed reduction in NA. These results support the hypothesis that axonal loss is important in the development of persistent clinical disability in multiple sclerosis.