Partial activation of membrane receptors using specific partial agonists may be particularly beneficial to the treatment of some chronic diseases like congestive heart failure, because receptor partial activation usually avoids overstimulation and also minimizes desensitization. At present, partial agonists are obtained almost exclusively by screening numerous newly-synthesized compounds, and little is known about the rational design of partial agonists with desired potencies and maximum activities. It is proposed in this study that a proper covalent attachment of a receptor full agonist to an antagonist (competitive or non-competitive) of the same receptor type may form a hybrid compound with a reduced maximum activity. In theory, the potency (EC50 value) and the reduced maximum activity of such an agonist-antagonist hybrid are predictable, based on the receptor binding affinities of the incorporated agonist and antagonist. The agonist-noncompetitive antagonist hybrids produce bell-shaped concentration-response curves, whereas the agonist-competitive antagonist hybrids produce concentration-response curves with reduced but plateaued maximum responses. The agonist-competitive antagonist hybrids can be very useful for therapeutic purposes because they enable us to stably activate a specific type of receptor at desired submaximal set-points. In this study, it is postulated that some partial agonists may also contain both agonistic and antagonistic moieties, analogous to the proposed hybrid compounds. This hypothesis is supported by data from modeling studies showing that several unique pharmacological characteristics of known partial agonists exactly match the projected features of the agonist-antagonist hybrid compounds. Modeling studies based on the hybrid compound model also explain why the reduced maximum response elicited by some partial agonists can be increased or even fully restored in the presence of spare receptors. As indicated by modeling studies, a reduced maximum activation of receptors may, at least in some cases, not result from a partial activation of each individual receptor being occupied by the partial agonist, but rather may be due to a reduced maximal receptor occupancy by the agonistic moiety in the co-presence of an antagonistic moiety. The agonist-antagonist hybrid proposed in this study not only enhances our understanding of the possible structural requirements and rational design of receptor partial agonists, but it also enhances our understanding of the possible molecular basis of receptor partial activation. This new model also provides an alternative mathematical framework for modeling partial agonist data.