PURPOSE: To determine the optimal voxel volume covering the hippocampal area in single-voxel proton magnetic resonance (MR) spectroscopy and to evaluate the reproducibility of metabolite ratios of the spectra. MATERIALS AND METHODS: Localized single-voxel proton MR spectroscopy was applied to the right hippocampal area of five healthy volunteers at 1.5 T (Siemens Vision), using a standard head coil, and we employed the spin-echo or point resolved spectroscopy sequence. Voxel volume was changed from 1ml to 5ml but other operator-dependent measurement parameters were fixed, as follows: repetition time/echo time=1500/135 msec, number of scans=300. Using the same voxel volume, five consecutive measurements were obtained in each subject. Signal to noise ratio (SNR) of N-acetylaspartate (NAA), NAA / Choline containing compounds (Cho) and NAA / Cr (creatine and phosphocreatine) +Cho ratios were calculated for all 25 spectra. RESULTS: The SNR of NAA peaks increased significantly as voxel volume was increased to 3 ml (p < 0.01, Kruskal-Wallis test and Wilcoxon rank sum test). There were, however, no significant differences in values among the spectra obtained with voxel volumes of 3, 4 and 5 ml (p > 0.1) ; in those obtained with voxel volume of 2-4ml, the standard deviations of NAA / Cho (10.6-13.2% of mean values) were similar to those of NAA /Cr+Cho (8.5-13.2% of mean values). CONCLUSION: For spin-echo proton MR spectroscopy of the hippocampal area, the optimal voxel volume may be more than 3 ml in a setting of TR/TE=1500/135 msec and number of scans=300. In this situation, standard deviations of metabolite ratios may reach about 8-13% of mean values.