ESR spectra from protein-containing lipid dispersions have been interpreted in the past primarily in terms of two nitroxide species related respectively to "fluid" and "immobile" phospholipid environments. In this report we consider interpretations based primarily on a single type of lipid, for the two cases of spin-probe-doped membranes and of chemically labeled membrane proteins. The doped membranes are viewed as bilayer fragments with considerable local order but with macroscopic disorder of such fragments in the dispersion. In this "one-site" model all lipids are similar. They are fluid and oriented in their local environment. This model does not require a second species immobilized by contact with the proteins. The labeled proteins are considered as very slowly reorienting macromolecular complexes such that the dynamic effects on the ESR spectrum arise mainly from faster internal processes. The importance of, and the potential inherent in, detailed spectral simulation based on well-conceived models are emphasized by illustrating the great range of spectral line shapes that these models can yield with suitable parameters. Simulations are compared with some recent experimental examples previously interpreted in terms of a two-site model. Reasonably good results are obtained for spin-probe-doped membranes when allowance is made for local ordering as well as for some distortion of the alkyl chain ends. Such effects are modeled by introducing an ordering potential λ and a diffusion tilt angle Ψ which describes the tilt of the nitroxide moiety relative to the rest of the alkyl chain. The effects of adding protein or of lowering the temperature are modeled by increasing the local ordering while decreasing somewhat the motional rates with some increase in alkyl chain distortions. In the case of spin-labeled membrane proteins, the model of very anisotropic rotation, in which the side chain containing the nitroxide is rotating more rapidly than the protein about an effective axis tilted relative to the N-O axis, is found to account for the extra splittings with just a single-site model.