In the Linnaean system of classification, the generic status of a species is part of its binomial name, and it is therefore important that the classification at the level of genus is consistent at least in related groups of organisms. Using maximum-likelihood phylogenetic trees constructed from a large number of complete or nearly complete mammalian cytochrome b sequences, I show that the distributions of intrageneric and intergeneric distances derived from these trees are clearly separated, which allows the limits for a more rational generic classification of mammals to be established. The analysis of genetic distances among hominids in this context provides strong support for the inclusion of humans and chimpanzees in the same genus. It is also of interest to decipher the main reasons for the possible biases in the mammalian classification. I found by correlation analysis that the classification of mammals of large body size tends to be oversplit, whereas that of small mammals has an excess of lumping, which may be a manifestation of the larger difficulty in finding diagnostic characters in the classification of small animals. In addition, and contrary to some previous observations, there is no correlation between body size and rate of cytochrome b evolution in mammals, which excludes the difference in evolutionary rates as the cause of the observed body size taxonomic bias.