Seventeen patients with presumed glial brain tumors were examined with proton chemical shift imaging and single voxel spectroscopy that used different echo times. Metabolite resonances were evaluated by metabolic ratios and absolutely by correcting for coil load and comparison to phantom measurements. Metabolic images were created to visualize the metabolic changes. All patients showed spectra that were different from those measured in healthy control subjects. Spectral changes were also present in normal-appearing matter (NAM) that was distant from lesions. The resonance at 3.55 ppm which is usually assigned to both myo-inositol and glycine, was the only one to allow a discrimination between healthy volunteers, astrocytoma grade II, and glioblastoma multiforme (GBM) (p < 0.02). From the different echo times used we conclude that an increase in this resonance has to be assigned to glycine rather than myo-inositol. This resonance might be used to grade human gliomas more reliably. Total creatine (Cr) decreased more drastically with malignancy than N-acetylated metabolites (NA). This led to a higher NA/Cr ratio in GBM compared to astrocytoma grade II. NA/Cr was thus pseudonormal in GBM due to a change in both nominator and denominator. This study reveals the importance of comparing magnetic resonance spectroscopy data of lesions to spectra measured in identical localizations in healthy control subjects instead of NAM and the importance of quantifying single metabolic peaks instead of creating metabolic ratios in clinical magnetic resonance spectroscopy.