Metabolomic analysis of human fecal water recently aroused increasing attention with the importance of fecal metabolome in exploring the relationships between symbiotic gut microflora and human health. In this study, we developed a quantitative metabolomic method for human fecal water based on trimethylsilylation derivatization and GC/MS analysis. Methanol was found to be the best solvent for protein precipitation and extraction of fecal water metabolome. Within the optimized linear range of sampling volume (less than 50 mu L), compounds showed a good linearity with a correlation coefficient higher than 0.99. The developed method showed good repeatability for both sample preparation and GC/MS analysis with the relative standard deviations lower than 10% for most compounds and less than 20% for a few other ones. The method was further validated by studying analytical variability using a set of clinical samples as well as a pooled sample. The pH value and matrix effects were the main factors affecting the accuracy of quantitative calibration curves. The increased pH value decreased the loss of short chain fatty acids during lyophilization. Spiking fecal water to a standard mixture significantly enhanced the accuracy of quantitative calibration curves, probably due to the inhibition of volatile loss during lyophilization and the increase of compound solubility in the derivatization medium. A strategy for calibration curve preparation was proposed in order to avoid the effects of pH and matrix. Totally, 133 compounds were structurally confirmed from a set of clinical samples, and 33 of them were quantified, which demonstrates the suitability of this method for a quantitative metabolomic study of human fecal water samples.