A tracer kinetic method is developed for the in vivo estimation of high-affinity radioligand binding to central nervous system receptors. Ligand is considered to exist in three brain pools corresponding to free, nonspecifically bound, and specifically bound tracer. These environments, in addition to that of intravascular tracer, are interrelated by a compartmental model of in vivo ligand distribution. A mathematical description of the model is derived, which allows determination of regional blood-brain barrier permeability, nonspecific binding, the rate of receptor-ligand association, and the rate of dissociation of bound ligand, from the time courses of arterial blood and tissue tracer concentrations. The term "free receptor density" is introduced to describe the receptor population measured by this method. The technique is applied to the in vivo determination of regional muscarinic acetylcholine receptors in the rat, with the use of [3H]scopolamine. Kinetic estimates of free muscarinic receptor density are in general agreement with binding capacities obtained from previous in vivo and in vitro equilibrium binding studies. In the striatum, however, kinetic estimates of free receptor density are less than those in the neocortex--a reversal of the rank ordering of these regions derived from equilibrium determinations. A simplified model is presented that is applicable to tracers that do not readily dissociate from specific binding sites during the experimental period. In this instance, specific tracer binding may be accurately determined by measuring tissue ligand concentration at a single time point after bolus intravenous injection, providing that regional cerebral blood flow is known. This derivation has potential clinical application, because it will permit construction of quantitative pictorial maps of regional free receptor densities in the human brain by means of positron emission tomographic imaging.