Recent studies show Xenarthra to be even more isolated systematically from other placental mammals than traditionally thought. The group not only represents 1 of 4 primary placental clades, but proposed links to other fossorial mammal taxa (e.g., Pholidota, Palaeanodonta) have been contradicted. No unambiguous Paleocene fossil xenarthran remains are known, and Eocene remains consist almost exclusively of isolated cingulate osteoderms and isolated postcrania of uncertain systematic provenance. Cingulate skulls are unknown until the late middle Eocene, and the oldest sloth and anteater skulls are early Oligocene and early Miocene age, respectively; there are no nearly complete xenarthran skeletons until the early Miocene. Ecological reconstructions of early xenarthrans based on extant species and the paleobiology of extinct Neogene taxa suggest the group's progenitors were myrmecophagous with digging and perhaps some climbing adaptations. The earliest cingulates were terrestrial diggers and likely myrmecophagous but soon diverged into numerous omnivorous lineages. Early sloths were herbivores with a preference for forested habitats, exhibiting both digging and climbing adaptations. We attribute the rarity of early xenarthran remains to low population densities associated with myrmecophagy, lack of durable, enamel-covered teeth, and general scarcity of fossil localities from tropical latitudes of South America. The derivation of numerous omnivorous and herbivorous lineages from a myrmecophagous ancestor is a curious and unique feature of xenarthran history and may be due to the peculiar ecology of the native South American mammal fauna. Further progress in understanding early xenarthran evolution may depend on locating new Paleogene fossil sites in northern South America.