It has been difficult to establish a strong correlation between total brain T2-weighted lesion volume on MRI and clinical disability in multiple sclerosis, in part because of the lack of pathological specificity of T2-weighted MRI signal changes. Proton magnetic resonance spectroscopy studies have shown that measurements of the resonance intensity of N-acetylaspartate (which is localized exclusively in neurons and neuronal processes in the mature brain) can provide a specific index of axonal damage or dysfunction. Here we report a 30-month longitudinal study of 29 patients with multiple sclerosis who had either a relapsing or a secondary progressive clinical course. Conventional brain MRI and single-voxel proton magnetic resonance spectroscopy examinations were obtained at intervals of 6-8 months with concurrent clinical evaluation. At the onset of the study, the brain N-acetylaspartate:creatine resonance intensity ratio was abnormally low for the whole group of patients (control mean = 2.93 +/- 0.2, patient mean = 2.56 +/- 0.4, P < 0.005). There were no significant differences between the relapsing and secondary progressive subgroups. Over the follow-up period, there was a trend towards a decrease (8%) in the brain N-acetylaspartate:creatine ratio for the 11 relapsing patients and a significant (P < 0.001) correlation between changes in the brain N-acetylaspartate:creatine ratio and expanded disability scale scores for the patients in this group. This correlation was even more evident for the patients who had clinically relevant relapses during the 30 months of follow-up (seven of 11 patients). Increases in T2-weighted lesion volumes (35% in 30 months for the group as a whole, P < 0.0001, without differences between the subgroups) did not correlate with disability either in the group of patients as a whole or in the different subgroups. We conclude that indices of axonal damage or loss such as brain N-acetylaspartate may provide a specific measure of pathological changes relevant to disability. Total T2-weighted lesion volumes, although more sensitive to changes with time than brain N-acetylaspartate, may be less relevant to understanding the progression of disability.