The idea that opioid receptors may be organized dimers, together with theconcept of "bridging" such neighboring receptors by igands containing two pharmacophoresinked through a spacer of appropriate length, led to the design of bivalent ligands with greaty enhanced potency. The finding that both opioid agonist pharmacohores in the bivalent ligand bound to neighboring opioid recognition sites provided circumstantia evidence for dimeric receptors. The recent biochemical characterization of opioid receptor dimers has provided more definitive evidence for their existence. Our mode for dimeric receptors has provided a structural basis for rationalizing data that was not easily interpretable using classical receptor models. Based upon our evidence for dimers and the large difference in the abiity of μ agonists and antagonists to protect μ receptors against irreversible blockage by β-FNA 6, we have proposed that opioid agonists and antagonists bind to separate allosterically coupled recognition sites on a receptor dimer. Such cooperativity is consistent with the results of recent studies in the porcine ileum which contains co-localized κ and δ receptors. The ability of the κ antagonists has suggested a κ-δ heterodimer model in which norBIN antagonizes δ agonists via a separate recognition site that is κ-selective. Also, the apparent transition from μ to δ agonism upon chronic exposure of mice to μ agonists may reflect a shift in the the distribution of μ and δ receptors to μ-δ heterodimers. In such case, it is conceivable that a μ agonist may be antagonized by a δ antagonist via the allosteric mechanism that has been disscussed. The existence of receptor dimers may also explain how lipophilic ligands that become localized in the cell membrane lipid bilayer can gain access to the central cavity of opioid receptors when the receptors are organized as interlocking dimers. Access via the lipid bilayer calls into question the significance of "affinity" when a high concentration of ligand is localized in the membrane. On the other hand, hydrophilic peptidic ligands would be expected to access the recognition site directly from the aqueous phase. Differential access would account for the generally higher apparent "affinity" of lipophilic opioid ligands relative to peptides. Exploration of the relationship between spacer length and antagonist potency of bivalent ligands led to the development of the first highly selective and potent κ opioid antagonist, norBNI 8. κ selectivity was found to be due to a cationic "address" moiety. The concept of a "message" and an "address", as originally proposed for rationalizing the SAR of peptide hormones, was subsequently employed to design several other selective non-peptide opiates that include NTI 12, GNTI 14a, NTB 16, BNTX 20, BSINTX 22, and SIOM 21b. These ligands are presently employed as pharmacologic tools. Our approach for the design of the κ opioid antagonist, GNTI 14a, required a detailed analysis of the SAR of the series, a knowledge of the amino acid sequences for all three opioid receptor models. Taken together, our studies have revealed that the high potency and selectivity of a ligand is not merely a a consequence of enhanced affinity for the target receptor. The origin of such selectivity changes have been found to be due to a combination of molecular exclusion of the ligand a lower affinity receptors and increased affinity at the high affinity receptor. Finally, structure-activity relationships are rarey straightforward and often more complicated than they appear. For this reason, the use of site-directed mutagenesis as a complementary tool to analyze SAR has been invaluable. Indeed, site-directed mutagenesis is a usefu technology that is adaptable to a medicinal chemistry environment. Given the paucity of high-resolution crystal structures for membrane-bound receptors, the use of a coordinated 'two-dimensional' paradigm, greater insight into the molecular recognition process may be obtained, particularly when combined with molecular modeling. It therefore behooves the medicinal chemist to employ such technology to analyze molecular recognition from the perspective of both the ligand and the receptor, in an effort to obtain a more penetrating analysis of the SAR. The power of this approach has been illustrated in the analysis of molecular recognition of selective opioid antagonists at opioid receptors.