The three-dimensional molecular shape analysis-quantitative structure-activity relationship (3D-MSA-QSAR) technique has been applied to develop correlations between the calculated physicochemical properties and the in vitro activities of a series of 3-(acylamino)-5-phenyl-2H-1,4-benzodiazepine cholecystokinin-A (CCK-A) antagonists. 3D-MSA-QSARs were developed for varying subsets of 53 analogs (J. Med. Chem. 1988, 31, 2235-2246). An active conformation is hypothesized for these compounds using the loss in biological activity-loss in conformational stability principle. After placing all compounds in the active conformation and performing pairwise molecular shape analysis, it was determined that not any one analog serves as the best shape reference compound. Nonidentical volumes of allowed receptor space are mapped out by different antagonists. A shape reference compound that consists of selected overlapped structures expands the definition of the accessible receptor space. This type of mutant improves the predicted activity of analogs over the value predicted if only one compound is chosen as the reference. Molecular shape, as represented by common overlap steric volume and nonoverlap steric volume, is the major factor contributing to the affinity of this class of compounds. Intramolecular conformational stability, as measured by the difference in energy of the active conformation and the global minimum energy conformation, is also important. It is further concluded from the 3D-MSA-QSAR models that part of the binding pocket for the 3-amido substituent has a preference for lipophilicity. The method used in this study of fragmenting the antagonist into spheres of varying radii and measuring lipophilicity isolates the substructure with highest probability of interacting with the receptor. Two indicator variables marking the presence of an N-methyl group and an o-fluoro atom on the 5'-phenyl substituent of the benzodiazepine ring structure also contribute significantly to the 3D-MSA-QSAR models. The 3D-MSA-QSAR results have led to the proposal of a 3D pharmacophore model for the benzodiazepine CCK-A antagonists.