We investigated two alternative hypotheses for the origin of crop pathogen species: that human-mediated agricultural practices drove the divergence of many crop plant pathogen species or that coevolutionary processes in natural populations of the crops' ancestors drove divergence of pathogen species. We distinguished between these two hypotheses by constructing a robust multigene phylogeny and estimating the dates of divergence among four, monophyletic species of smut fungi (Ustilago maydis, U. scitaminea, Sporisorium reilianum, S. sorghi) known to specifically infect maize, sorghum, sugarcane, and their wild ancestors. Without a fossil record for smut fungi, we calibrated the pathogen species' divergence times to their plant host divergence times. Specifically, a calibration date of 10,000 years was employed to test the hypothesis that the fungal species originated at the time of domestication of their current hosts and a calibration date of 50 million years was employed to test the hypothesis that the fungal species originated on wild ancestors of their domesticated hosts. Substitution rates at five protein coding genes were calculated and rates obtained for the 10,000 year calibration date were orders of magnitude faster than those commonly reported for eukaryotes, thus rejecting the hypothesis that these smut pathogen species diverged at the time of domestication. In contrast, substitution rates obtained for the 50 million year calibration were comparable to eukaryotic substitution rates. We used the 50 million year calibration to estimate divergence times of taxa in two datasets, one comprised solely the focal species and one comprised the focal species and additional related taxa. Both datasets indicate that all taxa diverged millions of years ago, strongly supporting the hypothesis that smut species diverged before the time of domestication and modern agriculture. Thus, smut species diverged in the ecological context of natural host plant and fungal populations.