Considering importance of developing selective estrogen receptor modulators (SERMs), the present paper explores selectivity requirements of tetrahydroisoquinoline derivatives for binding with ER(alpha) versus ER(beta) receptors using E-state index and physicochemical parameters. The best model [n=21, Q(2)=0.512, R(a)(2)=0.613, R=0.819, F=11.6 (df 3,17)] for ER(alpha) binding data obtained from radioligand binding assay showed importance of C(1), C(15) and lipophilicity (logP) while the best model [n=21, Q(2)=0.768, R(a)(2)=0.796, R=0.904, F=40.1 (df 2,18)] for ER(beta) binding data showed importance of C(1) and molar refractivity (MR). While modeling ER(alpha)/ER(beta) selectivity [n=21, Q(2)=0.695, R(a)(2)=0.739, R=0.882, F=19.8 (df 3,17)], C(1), C(15) and molar refractivity were found to be significant contributors. The data obtained from cellular transcription assay were also modeled. In case of ER(alpha), the best equation involving E-state values of C(1) and C(14) and logP explained 62.1% of the variance while the best equation for ER(beta) involving E-state values of C(1) and C(15) and MR explained 64.6% of the variance of the response variable. In case of ER(alpha)/ER(beta) selectivity, the best equation involving E-state values of O(8), C(14) and N(27) showed 48.3% explained variance, which increased to 63.5% on deletion of single outlier. From the analysis it appears that the nitrogen atom of the aminoethoxyphenyl substituent and 6-hydroxy substituent of the tetrahydroisoquinoline nucleus play important roles for ER(alpha)/ER(beta) selectivity in addition to R(1) and R(2) substituents.