The authors investigate the reproducibility of metabolite signals measured with proton magnetic resonance spectroscopy (1H-MRS) acquired from the human hippocampus in controls and in a phantom. Two 1H-MRS studies separated by 3 weeks were performed in 8 healthy volunteers and in a phantom. N-acetyl compounds (NA), choline (Ch), and creatine (Cr) peak areas and ratios were measured and compared using percentage variation, and Pearson Correlation Coefficient at the level of every voxel, the level of 1 hippocampus (5 voxels), and the level of 2 hippocampi (10 voxels). Sensitivity for observing clinically significant between-session 1H-MRS changes was evaluated using the reliable change index. Reproducibility measures for metabolite peak areas were only moderately concordant with percentage variation ranging from 14% to 20% for NA, Cho, and Cr. Stability was much improved when NA ratios and sum of multiple voxels were considered. Between-session NA/(Cho + Cr) changes greater than 22%, 12%, and 10% in one given participant can be detected with a 90% confidence interval when considered at the single-voxel level, the level of a single hippocampus, or the level of both hippocampi, respectively. Left-right asymmetry indices showed similar and limited inter-hemispheric asymmetry in repeated examination. This study suggests that 1H-MRS reproducibility performance is adequate for the study and monitoring of human hippocampus function when NA ratios and the sum of multiple voxels are considered. Individual metabolite peaks and single-voxel measurements have low reproducibility at 1.5 T and should be used only with clearly established statistical parameters.