Optimization calculations of 209 polychlorinated biphenyls (PCBs) were carried out at the B3LYP/6-31G* level. It was found that there is significant correlation between the Cl substitution position and some structural parameters. Consequently, Cl substitution positions were taken as theoretical descriptors to establish a novel QSPR model for predicting -lgSw of all PCB congeners. The model achieved in this work contains four variables, of which r2 = 0.9527, q2 = 0.9490 and SD = 0.25 with large t values. In addition, the variation inflation factors (VIFs) of variables in this model are all less than 5.0, suggesting high accuracy of the -lgSw predicting model. And the results of cross-validation test and method validation also show that the model exhibits optimum stability and better predictive capability than that from the AM1 method.