Despite much debate as to the validity of different models of agonism, there is apparently very little to choose between them in practice. However, most experiments to test the various models have used classical isolated tissue systems. In these systems, the response that is measured is far removed from receptor activation, and it is likely that the properties of the events at the receptor may be masked by more distal processes. In this article, Mary Keen examines the effects of agonists on responses that follow, very closely, activation of G protein-coupled receptors. The data appear to be inconsistent with any of the current models of agonism.