In vivo 1 H NMR spectroscopy was used to obtain the neurochemical profile in the posterior parts of the brain, the cerebellum and the medulla oblongata in comparison to the hippocampus and the thalamus. Using small voxel sizes between 16 and 32 ml to avoid partial volume effects, most metabolites demonstrated significant regional differences except acetate, g-aminobutyric acid, and phosphorylcholine. Noticeable regional differences in metabolite concentrations were the significant increase of total creatine in the cerebellum and the substantial decrease of taurine in thalamus and medulla oblongata. In particular, the glycine concentration in the medulla oblongata was determined to be 4.37 6 0.68 mmol/g (Cramé r-Rao lower bounds 7%) and thus significantly higher than in the other regions, consistent with findings reported in both in vivo 1 H NMR spectroscopy and in vitro biochemical assays. Intraindividual reproducibility and interindividual variability were investigated by acquiring spectra from the thalamus of the same rats in two sessions. No prominent influence on measurement session was observed in metabolite concentrations with coefficients of variations being below 20% in 16 metabolites. Magn Reson Med 66:11-17, 2011. V C 2011 Wiley-Liss, Inc. Key words: in vivo 1 H NMR spectroscopy; STEAM; ultra-short TE; CRLB; CV; regional differences; reproducibility Localized in vivo 1 H NMR spectroscopy of the brain is able to provide invaluable comprehensive and noninvasive information for investigating the healthy and diseased brain. In preclinical research, it has helped to understand pathologic changes in various disease models. Lately, the information content of in vivo 1 H NMR spectroscopy has been boosted by increasing magnetic field strengths available in modern MR instruments, making it possible to accurately quantify as many as 20 metabolites in the rat brain.An important factor for determining the applicability of in vivo 1 H NMR spectroscopy in preclinical studies is the quality of the localization, as metabolite concentrations have been shown to differ between tissues types. Thus, contamination from adjacent brain structures has to be strictly avoided. A double localization strategy, combining a stimulated echo acquisition mode volume selection with outer volume suppression (1), has been established as a reliable technique for many in vivo 1 H NMR spectroscopy studies, efficiently suppressing contributions from unwanted signal, while maximizing signal-to-noise ratio (SNR) and information by using short echo time (TE).Avoiding partial volume effects, however, is still difficult in many applications due to the large voxel sizes necessary to gather sufficient signal within acceptable measurement time. Thus, the selected volumes often contain more than one tissue type, which affects both the information content of the data as well as the reproducibility, as slight changes in the positioning of the voxel between sessions cause variations in the results. Thus, voxel size and accurate positioning are important f...