Based on nine quantum chemical descriptors computed by PM3 Hamiltonian, using partial least squares analysis, a significant quantitative structure-property relationship for the logarithm of octanol-air partition coefficients (logK(OA)) of polychlorinated biphenyls (PCBs) was obtained. The cross-validated Q2cum value of the model is 0.962, indicating a good predictive ability. The intermolecular dispersive interactions and thus the size of the PCB molecules play a key role in governing log K(OA). The greater the size of PCB molecules, the greater the logK(OA) values. Increasing ELUMO (the energy of the lowest unoccupied molecular orbital) values of the PCBs leads to decreasing logK(OA) values, indicating possible interactions between PCB and octanol molecules. Increasing Q(Cl)+, (the most positive net atomic charges on a chlorine atom) and Q(C)- (the largest negative net atomic charge on a carbon atom) values of PCBs results in decreasing lg K(OA) values, implying possible intermolecular electrostatic interactions between octanol and PCB molecules.