Abstract: Soil pollutin from mononuclear armatics due to fuel oil leakage has become a universal phenomenon and it is critical to precisely and efficiently determine volatile mononuclear aromatics in the soil. In Standard Method 5021 released by U.S. Environmental Protection Agency, headspace-GC-MS method is recommended to detect volatile mononuclear aromatics in the soil and the quantification library established based on aqueous solution is employed. But the accuracy of quantification results is still worth discussing. In this paper quartz sand and soil are added into the standard series to estabilish the quantification libraries and the slopes of calibration curves in the water, water-sand, water-soil are compared in order to study the matrix effect on quantification. The results indicate that, in general, the slope of mononuclear aromatics in soil matrix is less than that in quartz sand matrix, and the slope of mononuclear aromatics in water matrix is maximum. The slope of mononuclear aromatics in quartz sand matrix is approximately 1% (benzene)~9% (n-butylbenzene) less than that in water matrix. Only for benzene, the slope in soil matrix can be nearly equal to that in water matrix. Other components in soil matrix have slopes which are about 9% (propyl benzene)~24% (1,2,4-trichlorobenzene) less than that in water matrix. The relative slope of mononuclear aromatics in soil matrix negatively correlates to the boiling point of the compounds and the partition coefficient between octanol and water. The experiment results show that it is very critical to select suitable matrix to establish the quantification library for the calibration of soil samples.