Data on limb bone lengths from 64 mammalian species were combined with data on 114 bovid species (Scott, '79) to assess the scaling of limb lengths and proportions in mammals ranging from 0.002 to 364 kg. We analyzed log-transformed data using both reduced major axis and least-squares regression to focus on the distribution across mammals of two key traits-limb length and metatarsal/femur ratio--associated with cursorial adaptation. The total lengths of both fore and hindlimbs scale in a manner very close to the M0.33 predicted by geometric similarity. Thus the relative limb lengths of large mammals, including bovids, generally regarded among the most cursorial of mammals, are very similar to those of the rodents and insectivores in this sample. Metatarsal/femur ratio also shows little change with changing mass, although bovids tend to have relatively longer metapodials than do other families in the sample. We argue that many of the remaining morphological traits associated with cursoriality (e.g., reduction in joint mobility and number of distal limb bone elements) promote cursoriality only at large body sizes. These results lead us to question the general perception that cursoriality is most widespread among large mammals. We also suggest that discussions of cursoriality should focus explicitly on the two partially independent aspects of performance that are otherwise confounded under this general term--speed and the ability to cover substantial distance.