Quantitative estimation of cerebral muscarinic receptors was investigated with the use of the antagonist [11C]tropanyl benzilate ([11C]TRB) and positron emission tomography (PET). Kinetic modeling alternatives were examined with the goal of identifying an analysis method providing stable receptor measures, yet avoiding biases from inappropriate reductions in model complexity. Dynamic PET scans were performed on six young normal volunteers. Several modeling approaches yielding relative receptor density measures were evaluated: (a) a single "late" scan using relative tracer concentration values; (b) a slope estimate from graphic analysis (Patlak plot); (c) a two-compartment, two-parameter model (transport and total ligand distribution volume); (d) a three-compartment, two-parameter model using the free+nonspecific distribution volume, DV', fixed to the cerebellar value; (e) an early scan for transport, a fixed value for DV', and a single late scan for the binding rate constant; and (f) a three-compartment, three-parameter model. Both computer simulations and PET scan results indicate all methods provide receptor density index measures with the same rank order as in vitro measures. Oversimplified approaches (methods 1 and 2) yield a more highly nonlinear relation between the estimated receptor density index and the known receptor density than do methods retaining greater model complexity (methods 3-6). However, noise propagation into the receptor measure is greater for the more complex methods. Reliable receptor density information can be obtained from kinetic [11C]TRB PET studies, with methods 3-5 providing the most appropriate levels of model complexity for estimates of relative muscarinic receptor density.