A series of new derivatives of 3-(1,2,5,6-tetrahydropyridin-4-yl)indole (4-THPI) has been synthesized, and their dissociation constants at the 5-HT1A and 5-HT2 serotonin (5-HT) receptor subtypes have been determined. The new data were combined with similar binding data on a related set of THPI analogs reported previously (Taylor et al. Mol. Pharmacol. 1988, 34, 42-53) and used to develop 3-dimensional quantitative structure-activity relationships (3-D QSARs) for these compounds at the 5-HT1A and 5-HT2 receptor sites, by the method of comparative molecular field analysis (CoMFA). Since the previous study included several conventional QSARs obtained by Hansch analysis, and the new compounds in some cases fall within the congeneric series used in those analyses, we were able to make a direct comparison of the predictive capabilities of CoMFA and Hansch analysis using identical training and test data sets. The overall quality of actual predictions of activity by both methods appears to be about the same, as assessed by the root mean square (rms) residuals between actual and predicted pKi values. On the one hand, the compounds most poorly predicted by the Hansch analysis were 34, 35, and 37, while compounds 30-33 were relative poorly predicted by CoMFA. However, a clear advantage of CoMFA is the ability to include diversely substituted or noncongeneric analogs that must be omitted from conventional QSAR analysis. Using the entire data set of 45 THPI analogs reported here, pKi predictions for six additional compounds having 5-heteroarylindole substituents gave rms residuals of 0.46 and 0.36 for the 5-HT1A and 5-HT2 models, respectively; this is close to the experimental error of the binding data. The significance of the CoMFA field graphs in terms of molecular features required for activity and selectivity at these 5-HT receptor subtypes is discussed.