Several reasons can exist for a change in reversed-phase selectivity, and several separation conditions can be changed for this purpose. In the present investigation, a change in column is considered for the improved separation of non-ionized solutes only. Differences in column selectivity (and the selection of "orthogonal" columns) can be predicted on the basis of the hydrophobic-subtraction model. For ionized solutes, the F(s)-parameter is used to predict orthogonality. For use with non-ionized solutes, it is suggested that the cation-exchange term of the model be dropped [F(s)(-C)] for better predictions. For samples containing both ionized and non-ionized solutes, F(s) and F(s)(-C) should be used together for the best results. Predicted separations involving 64 non-ionized solutes and maximally different columns from a 400-column database were used to validate this procedure.