The structural variation of lead compounds often includes the variation of substituents at a particular site. A common difficulty, in quantitative structure-activity relationship (QSAR) studies, is to select the best substituents to obtain, with a minimum number of experiments, a major change in some impor. tant properties of the molecule. To improve a proposed modified Topliss method, 187 different substituted compounds belonging to 70 series were studied. A single linear regression was carried out between the experimental activity of each series of compounds and the molecular structure parameters pi, sigma, Es and MR, along with some combinations of them, in order to select the best substituents for constructing the "reduced" correlation. This selection considered the scaled smallest residue average of the substituent, the number of times that the substituent is inserted in the 70 series and its quadrant in the Craig graph. The results obtained in this study with this simple method showed a good predictive capacity when compared with the multiple regression analysis, using the Hansch method.