The precision of cerebral proton magnetic resonance spectroscopy (MRS) measurements is critical both in the clinical setting and for research purposes. Marshall et al. have recently concluded that "disappointing in vivo repeatability...is likely to limit" the ability of MRS to detect modest changes. We present here a comprehensive study of the precision of short- and long-term metabolite peak area ratios and water referenced metabolite peak areas for long echo time point resolved spectroscopy (PRESS) spectra (repetition time (TR) = 2000 ms, echo time (TE) = 136 ms) acquired from the occipital lobes of normal volunteers and a phantom using a conventional whole body 1.5 T MR system and conventional acquisition and analysis protocols. Short-term in vitro precision determined by five repeat scans on five occasions was excellent as measured by a mean coefficient of variation (NAA/Cho = 1.3%, NAA/Cr + PCr = 1.0%, Cho/Cr + PCr = 1.6%, NAA/H2O = 0.5%, Cho/H2O = 1.2%, Cr + PCr/H2O = 0.8%). Long term in vitro precision using 100 spectra acquired over 2 years was also very good (NAA/Cho = 2.7%, NAA/Cr + PCr = 1.4%, Cho/Cr + PCr = 2.2%, NAA/H2O = 1.5%, Cho/H2O = 2.4%, Cr + PCr/H2O = 1.5%). Short-term in vivo precision determined by five repeat scans in a single scanning session on eight subjects was also excellent (NAA/Cho = 5.2%, NAA/Cr + PCr = 3.0%, Cho/Cr + PCr = 6.6%, NAA/H2O = 1.4%, Cho/H2O = 4.9%, Cr + PCr/H2O = 2.7%) and only worsened slightly for long-term in vivo precision determined by five repeat scans on eight subjects over 3 months (NAA/Cho = 5.2%, NAA/Cr + PCr = 4.8%, Cho/Cr + PCr = 7.7%, NAA/H2O = 2.5%, Cho/H2O = 6.4%, Cr + PCr/H2O = 3.8%). We attribute the excellent precision reported here to the use of highly automated techniques for voxel shimming, water suppression and peak area measurements. These results allow us to repudiate Marshall's assertion regarding disappointing repeatability of in vivo MRS.