Three new strategies for sampling the conformation space accessible to a series of structurally diverse, flexible molecules are defined and compared to samples obtained using a fixed-grid torsion angle sampling strategy. A set of 28 potent inhibitors of angiotensin converting enzyme selected by Mayer et al. [J. Comput.-Aided Mol. Design, 1 (1987) 3] and the unrestricted active-site model proposed by Waller et al. [to be published] are used to produce a realistic experimental setting. We modified our Constrained Search algorithm [Dammkoehler et al., J. Comput.-Aided Mol. Design, 3 (1989) 3] to support these new sampling strategies, performing a series of 64 simulations (search experiments) and generating a large set of sterically allowed conformations. In each experiment, we systematically vary the internal torsion angles in each molecule using one of the sampling strategies. The common orientations of preselected functional groups thought to represent those dominating the interaction with the enzyme and presented by the set of molecules are classified and recorded for each experiment. Pairwise distances between groups are used to characterize the geometry of the common orientations. The results of each experiment, represented by a set of distance values, are compared and combined to evaluate the completeness of the conformational sampling. While no pure strategy or single search experiment was found to be adequate to fully explore the set of common sterically allowed conformations, a new sampling technique, called adaptive radial sampling, is shown to be significantly more complete than the commonly used fixed grid sampling.