A multivariate analysis of the postcranial skeleton of extant bovids reveals patterns of osteological features indicative of ecological adaptations for habitat use and body size. The morphological patterns that characterize the postcranial anatomy of bovid species from each habitat type were identified with stepwise canonical discriminant analysis and decision trees, a technique based on machine learning. The analyses were carried out using 43 measurements from 110 extant bovid species. The discriminant functions and decision trees obtained allow a perfect discrimination among bovids adapted to open plains, forests and mountainous areas (100% of correct reclassifications obtained in all comparisons), using sets of variables measured in all major limb bones as well as combinations of variables derived exclusively from single limb elements. Given that the adjusted algorithms involve small sets of postcranial measurements, they can also be applied to noncomplete specimens preserved in archaeological and paleontological assemblages, thus being useful for estimating the locomotor performances of ancient taxa. These algorithms, indicative of ecological adaptations for habitat use, combined with those adjusted with craniodental measurements for estimating the dietary preferences of bovid species, have the potential for characterizing the paleoautecology of extinct taxa and may be used in paleoenvironmental reconstruction. We also analyze if multiple regression equations show higher predictive ability for estimating body mass than simple regression equations, and propose the best algorithms obtained from postcranial morphological variables measured in each single major limb bone[Acta Zoologica Sinica 52(6):971 –987,2006].

Corrigendum：  The abbreviation "Bogr" in Figures 2-3 indicates the position in the morphospace of the yak, Bos grunniens, a species adapted to mountainous areas, and does not make reference to the nilgai Boselaphus tragocamelus, as incorrectly written in the legends of these figures.