Cholecystokinin 2 receptor antagonists encompass a wide range of structures. This makes them unsuitable candidates for existing 3D-QSAR methods and has led us to develop an alternative approach to account for their observed biological activities. A diverse set of 21 antagonists was subjected to a novel molecular field-based similarity analysis. The hypothesis is that compounds with similar field patterns will bind at the same target site regardless of their underlying structure. This initial report demonstrates a linear correlation between ligand similarity and biological activity for this challenging data set. A model generated with three molecules was used to predict the activity of 18 test compounds, with different chemotypes, with a root-mean-square error of 0.68 pKB units. The ability to automatically derive a molecular alignment without knowledge of the protein structure represents an improvement over existing pharmacophore methods and makes the method particularly suitable for scaffold-hopping.