A quantitative model that incorporates information on both environmental temperatures ( T) and molecular structures, for logarithm of octanol–air partition coefficient to base 10 (log K OA) of polychlorinated biphenyls (PCBs) was developed. Partial least squares (PLS) analysis together with 16 theoretical molecular structural descriptors was used to develop the Quantitative relationships between structures, environmental temperatures and properties (QRSETP) model. The cross-validated Q cum 2 value for the optimal QRSETP model is 0.976, indicating a good predictive ability for log K OA of PCBs at different environmental temperatures. T, E LUMO (the energy of the lowest unoccupied molecular orbital), molecular size or average molecular polarizability ( α), and the net atomic charges on chlorine, hydrogen and carbon atoms of PCB molecules, are major factors governing log K OA. The lower the E LUMO, the greater the intermolecular interactions between octanol and PCB molecules, and thus the greater the log K OA values. Because of intermolecular dispersive forces, the more chlorine atoms in PCB molecules, the greater the molecular size or α, the greater the log K OA. The largest negative net atomic charge on a carbon atom ( q C −) and molecular size or average molecular polarizability ( α) are major factors governing temperature dependence of log K OA. PCB molecules with low q C − values and more chlorines (big size or α) tend to have strong temperature dependence, due to intermolecular electrostatic interactions and dispersive forces, respectively.