Structure-activity relationships (SAR) modeling of separation efficiencies of 17 arylhydroxamic acids tested as mineral collectors is attempted using topological indices, partition coefficients and Connolly molecular surface areas and solvent-excluded volume. Several of the computed parameters are found to fit in a second-order polynomial regression with the separation efficiencies (E-s) of the collectors. Soil-water partition coefficient, logK(oc), gives the best correlation with a standard error of estimate (SEE) of 4.95 and correlation coefficient, r=0.9635. The data set is split into N-aryl-C-alkyl and N-aryl-C-aryl types based on the structure of the hydroxamic acids, and the regression analyses were repeated. In the case of N-aryl-C-alkyl compounds, Connolly solvent-excluded volume gives the best fit with an SEE of 2.88 and correlation coefficient of 0.9936, and several other variables used in the study are found to have r>0.97. First-order valence connectivity index and logK(oc) give the best fit for N-aryl-C-aryl compounds, and the SEE in either case is 4.14 and r=0.9756. These results indicate that any one of the four classes of parameters considered in the study can be used for the interpretation of variation of separation efficiencies with the change in the hydrophobic part of a series of flotation collectors. It is inferred that lower-order valence connectivity indices are the good starting point to SAR modeling of flotation collectors. (C) 2003 Published by Elsevier Science B.V.