Abstract: The multiple linear regression analysis was applied to process the gas chromatographic data of Polychlorinated Biphenyls (PCBs) standard samples and actual soil samples, in order to quantitatively and objectively estimate the PCBs composition in the soil. It is shown that the calculated compositions of standard PCBs sample (Aroclor 1248) and the mixture of two standard samples prepared in laboratory are in line with those from labels. The absolute analysis error is smaller than 2%. Two soil samples polluted by PCBs from two cities of China, Beijing and Lanzhou, are analyzed by this method. The soil sample from Beijing is polluted by Aroclor 1242. The soil sample from Lanzhou is polluted by 49.0% of Aroclor 1242 and 51.0% of Aroclor 1254. The method described also provides a quality control approach to distinguish the interference compounds in the samples.