Classic and three-dimensional (3-D) QSAR analyses of 13 radical scavengers (1-13) were performed to derive two classic, two Apex-3-D and one comparative field analysis (CoMFA) models. Two classical models with predictive cross-validated r2 (Q2) over 0.96 indicated that the activity was attributed to the electronic COH and ELUMO, steric molar refractivity (MR) and lipophilic log P. Three-dimensional quantitative structure-activity relationship (3-D-QSAR) studies were performed by 3-D pharmacophore generation (Apex-3-D) and CoMFA techniques. For Apex-3-D studies, two best models with high Q2 (0.94 and 0.97) were yielded. Structural properties contributing to the activity were not only lipophilic but also the optimum steric property and geometry of side-chain composition. For CoMFA studies, the sp3 C(+1) probe provided the best Q2 of 0.79 with steric and electrostatic contributions of 42.3 and 57.7%, respectively. The activity of four new compounds (14-17) not included in the derivation were predicted with these models. Although the derived models were from limited data, the statistic relation was predictive. The linear correlations between the experimental IC50 values and the predicted values from classical and Apex-3-D models were found to be high and significant. The predicted activity of 17 from CoMFA was much lower than the experimental value; this deviation occurred according to the missing of hydrophobic field in standard CoMFA study. In vitro and ex vivo antilipid peroxidation in mouse brain and ESR studies of 14-17 were investigated for the radical-scavenging ability. The difference between the in vitro results, antilipid peroxidation and electron spin resonance (ESR) and ex vivo results in coumarin series was found. Thus, other properties for good bioavailability besides log P should also be taken into consideration.