The hypothesis that the relaxant action of many drugs on vascular and other smooth muscle is mediated by increases in intracellular cGMP, the "cGMP hypothesis," is gaining wide acceptance. While much information supporting this idea can be found in the literature, there is also a significant amount of information indicating that an elevation in the tissue content of cGMP is by itself insufficient to cause smooth muscle relaxation. The literature is reviewed with reference to the criteria that need to be fulfilled to consider cGMP as the second messenger mediating relaxation of smooth muscle by a drug; i.e., activation of guanylate cyclase, elevation of tissue content of cGMP, potentiation by phosphodiesterase inhibitors, antagonism by inhibitors of cGMP synthesis, and production of relaxation by cGMP analogues. For each criterion, key observations supporting the hypothesis are considered, followed by examples of important observations not consistent with the hypothesis. It is concluded that in some smooth muscles, for example, rat myometrium and vas deferens, cGMP is not a mediator of drug-induced relaxation. In other smooth muscles, including vascular smooth muscle, cGMP appears to play an important role in the relaxation process; but current evidence suggests that other factors are also important and that the cGMP hypothesis may need to be modified.Key words: cGMP, vascular relaxation, smooth muscle relaxation, vasodilators.