Molecular connectivity indices (MCI), were used to predict log Kow of several classes of chemicals (mainly pesticides). The modeling capability was first evaluated for each class of compounds using simple and valence MCI up to the 7th order (nX andnXv). Moreover, different molecular descriptors, derived from MCI but accounting for non-dispersive intermolecular interactions (ΔX and ΔXv) were included in the regression models. A general equation of the type:proved effective in producing highly reliable log Kow values within homogeneous classes of chemicals. For heterogeneous groups, a similar equation proved useful at least for preliminary screening evaluations.