All known Pleistocene fossils of manatees from North America are conspecific with the living West Indian manatee, Trichechus manatus. However, those of late Pleistocene (late Rancholabrean) age differ from all modern manatees, and are here assigned to a new subspecies, T. m. bakerorum. In contrast, early Pleistocene manatees from the late early Irvingtonian Leisey Shell Pit Local Fauna, Florida, do not clearly differ from modern West Indian manatees and are here referred to T. manatus, subspecies indeterminate. I hypothesize that warm intervals during the Quaternary allowed manatees from the Caribbean (represented today by the Antillean subspecies, T. m. manatus) to disperse northward into the United States. Ecological barriers (cool winters on the northern Gulf of Mexico coast, deep water and strong currents in the Straits of Florida) then impeded genetic contact with Caribbean populations and permitted evolution of endemic North American forms (T. m. bakerorum and the living Florida manatee, T. m. latirostris). Conditions much warmer than today might have eliminated the northern Gulf Coast barrier and allowed gene flow from the Caribbean to swamp such endemic gene pools. Return of glacial conditions reversed these northward range extensions and brought about periodic extinction of manatees in North America. This mechanism might explain the morphological uniqueness of the late Rancholabrean manatees, as well as the fact that early Pleistocene manatees from Florida are more similar to modern T. m. manatus than the late Pleistocene ones are. This model also helps explain phylogeographic patterns detected in mitochondrial DNA data from living manatees.