G protein-coupled receptors (GPCRs) represent the largest family of signal-transducing molecules known. They convey signals for light and many extracellular regulatory molecules. GPCRs have been found to be dysfunctional/dysregulated in a growing number of human diseases and have been estimated to be the targets of more than 30% of the drugs used in clinical medicine today. Thus, understanding how GPCRs function at the molecular level is an important goal of biological research. In order to understand function at this level, it is necessary to delineate the 3D structure of these receptors. Recently, the 3D structure of rhodopsin has been resolved, but in the absence of experimentally determined 3D structures of other GPCRs, a powerful approach is to construct a theoretical model for the receptor and refine it based on experimental results. Computer-generated models for many GPCRs have been constructed. In this article, we will review these studies. We will place the greatest emphasis on an iterative, bi-directional approach in which models are used to generate hypotheses that are tested by experimentation and the experimental findings are, in turn, used to refine the model. The success of this approach is due to the synergistic interaction between theory and experiment.