Partition coefficients were measured for 47 compounds in the hexadecane/water (Phxd) and 1-octanol/water (Poct) systems. Some types of hydrogen bond acceptor presented by these compounds to the partitioning systems are not well represented in the literature of alkane/water partitioning. The difference, ΔlogP, between logPoct and logPhxd is a measure of the hydrogen bonding potential of a molecule and is identified as a target for predictive modeling. Minimized molecular electrostatic potential (Vmin) was shown to be an effective predictor of the contribution of hydrogen bond acceptors to ΔlogP. Carbonyl oxygen atoms were found to be stronger hydrogen bond acceptors for their electrostatic potential than heteroaromatic nitrogen or oxygen bound to hypervalent sulfur or nitrogen. Values of Vmin calculated for hydrogen-bonded complexes were used to explore polarization effects. Predicted logPhxd and ΔlogP were shown to be more effective than logPoct for modeling brain penetration for a data set of 18 compounds.