Fossil data are ambiguous regarding the evolutionary origin of contemporary desert bighorn sheep (Ovis canadensis subspecies). To address this uncertainty, we conducted phylogeographic and population genetic analyses on bighorn sheep subspecies found in southwestern North America. We analyzed 515 base pairs of mtDNA control region sequence and 39 microsatellites in 804 individuals from 58 locations. Phylogenetic analyses revealed 2 highly divergent clades concordant with Sierra Nevada (O. c. sierrae) and Rocky Mountain (O. c. canadensis) bighorn and showed that these 2 subspecies both diverged from desert bighorn prior to or during the Illinoian glaciation (∼315–94 thousand years ago [kya]). Desert bighorn comprised several more recently diverged haplogroups concordant with the putative Nelson (O. c. nelsoni), Mexican (O. c. mexicana), and Peninsular (O. c. cremnobates) subspecies. Corresponding estimates of effective splitting times (∼17–3 kya), and haplogroup ages (∼85–72 kya) placed the most likely timeframe for divergence among desert bighorn subspecies somewhere within the last glacial maximum. Median-joining haplotype network and Bayesian skyline analyses both indicated that desert bighorn collectively comprised a historically large and haplotype-diverse population, which subsequently lost much of its diversity through demographic decline. Using microsatellite data, discriminant analysis of principle components (DAPC) and Bayesian clustering analyses both indicated genetic structure concordant with the geographic distribution of 3 desert subspecies. Likewise, microsatellite and mitochondrial-based F_ST comparisons revealed significant fixation indices among the desert bighorn genetic clusters. We conclude these desert subspecies represent ancient lineages likely descended from separate Pleistocene refugial populations and should therefore be managed as distinct taxa to preserve maximal biodiversity.