Recent studies have demonstrated that activation of the same G-protein coupled receptor can generate different agonist pharmacology depending on the signaling pathway(s) to which it couples. Two types of behavior have been exemplified; differences in affinity order, and differences in efficacy order with the same affinity order. The two-state model of receptor activation cannot explain these data, since a single active receptor state cannot couple differently to the two response pathways for different ligands. We have therefore extended the two-state model to a three-state model in which receptors exist in three states: an inactive state, R, and two different active states, R* and R**. The model has two modes, the 'intact mode', in which all the equilibria are linked; and the 'isolated mode' in which the two response pathways are isolated from each other, giving effectively two separate two-state systems. In the 'intact mode' the same agonist affinity order is predicted for both response pathways, but a different efficacy order. In the 'isolated mode', since the equilibria are no longer linked, the model predicts that a different affinity order may be obtained for the two pathways. Owing to the linkage of all the equilibria in the intact three-state model the level of constitutive activity through one pathway can affect the direction of agonism through the other pathway, resulting in the conversion of an inverse agonist into a positive agonist. This change in the direction of agonism is also predicted to occur when the two response pathways are isolated. The three-state model therefore predicts that agonists, acting at the same receptor, may show different affinity orders and different efficacy orders depending upon which response is measured and the assay system used, and also predicts that inverse agonism may be system dependent.