We conducted an extended clinical evaluation of localized proton magnetic resonance spectroscopy (MRS) of the brain, performed on various brain diseases using short stimulated echo times. Pathologies studied were mainly multiple sclerosis, stroke, leukoaraiosis, AIDS-related leukoencephalopathies and glial tumors. Other miscellaneous pathologies were also studied. Magnetic resonance examination of the brain was conducted on a Siemens Magnetom SP63 (equipped with a 1.5 T magnet). Localized proton MRS was performed on a routine basis immediately after imaging, using the STEAM (stimulated echo acquisition mode) with a short echo time (20 ms) combined with a CHESS (chemical shift selective excitation) sequence. One or two VOI (8 ml) were examined. Data on 125 spectra were processed by principal component analysis (PCA) and conventional variance analysis. The following metabolite resonances were studied: inositol-glycine, taurine-scyllo-inositol, choline derivatives, phosphocreatine-creatine, aspartate, glutamine glutamate, N-acetylaspartate, acetate and lactate. PCA demonstrates that the different metabolic variables are independent. The analysis of groups of spectra clearly demonstrates that the metabolic profiles detected by localized MRS in various pathologies (i) differ significantly from controls, and (ii) allow a metabolic discrimination between groups of pathologies. Results of PCA are confirmed by variance analysis. Strokes are characterized by an increase in lactate concentration and leukoaraiosis by a decrease in inositol-glycine resonance. AIDS-related leukodystrophies are characterized by increases in lactate and choline concentrations. Reduction in N-acetylaspartate which is observed in most pathologies is not significant in the small lesions of white matter. Lactate has often been found in MS plaques, but no variation in the choline/phosphocreatine ratio was observed. GABA was tentatively assigned in the spectrum of a patient with epilepsy under sodium valproate treatment. This study illustrates the clinical feasibility of the technique, the value of a multiparametric data analysis in the definition of the pertinent variables characterizing the metabolic impairment, and the impact of localized proton MR spectroscopy of the brain in the assessment of cerebral suffering.