A 250-GHz electron paramagnetic resonance (EPR) study of the slow rotational diffusion of two spin probes in toluene, viz., perdeuterated 2,2,6,6-tetramethyl-4-piperidone (PDT) and 3-doxylcholestane (CSL) is presented. EPR spectra were obtained in the slow-motional and near-rigid limit regions, which corresponds to rotational correlation times 10(10) > tau(R) > 10(-6)s. These two probes differ significantly in size and shape, permitting a detailed exploration of the sensitivity of 250-GHz EPR to different aspects of the molecular dynamics such as rotational anisotropy and non-Brownian diffusion. Nonlinear least-squares fitting based on full stochastic Liouville calculations provides a sensitive means for discriminating amongst motional models. PDT in toluene-d8 is found to be well described by an approximate free diffusion model, whereas the larger spin probe, CSL, is best described by Brownian diffusion. The slow-motional spectra at 250 GHz are most sensitive to the diffusional model, the (geometric) mean diffusional rate, and axial diffusional anisotropy but less sensitive to rhombic deviations from an axially symmetric diffusion tensor (i.e., to the general case R(x) not-equal R(y) not-equal R(z)). The slow-motional spectra of PDT were fit using anisotropic diffusion parameters determined from fast-motional spectra but are not very sensitive to such small anisotropies. For the case of Brownian diffusion, CSL was best fit with N(y) = R(y)/(R(z)R(x))1/2 = 9.0 (where the y axis is the long axis of the molecule and x and z are perpendicular axes), which differs appreciably from the fast-motional value of N(y) = 4.3 +/- 0.2 (and rho(x) = R(x)/R(z) = 0.5). However, a mixed model of free-diffusional motion about the y axis with Brownian motion of this axis yields an N(y) close to the fast-motional value with comparable overall quality in fit compared to full Brownian motion. An important feature of the 250-GHz studies is the ability to measure very accurately the magnetic tensors needed for the motional studies. The theoretical modifications needed for inclusion of a fully anisotropic rotational diffusion tensor in the slow-motional EPR simulations are also given.