We tested the hypothesis that hind limb proportions may be used to predict locomotor performance in a sample of 49 species of primarily cursorial mamals. Data on maximal sprint running speeds taken from published sources were related to measurements of hind limb lengths. To control for statistical complications due to the hierarchical nature of phylogenetic relationships, we used Felsenstein's (1985) independent contrasts method for analysing comparative data, and a composite phylogeny for all 49 species, based on a variety of published sources. The independent contrasts method indicates that maximal running speed does not covary significantly with body mass for this sample of mammals (mass range= 2.5–2,000 kg). Even though quality of the available speed data is highly variable, both metatarsal/femur ratio—the traditional index of ‘cursoriality’ in mammals—and hind limb length (corrected for body size) are significant predictors of maximal running speed. When only fully curorial species are included in the analyses (n = 32), hind limb length still significantly predicts speed (r2= 16%), but MT/F ratio does not. Although ungulates tend to have larger MT/F ratios than do Carnivora, they are not generally faster; relatonships between speed and limb proportions within the two clades show no significant differences. These and previous results suggest that hind limb proportions and maximal running speed may not have evolved in a tightly coupled fashion. Prediction of locomotor performance of extinct forms, based solely on their limb proportions, should be undertaken with caution.