It has become possible to compare ages of first appearance of Asian, European, and North American earliest Eocene mammals only in the past 3 yr, with identification of the global PETM marked by the Paleocene–Eocene carbon isotope excursion (CIE) on all three northern continents (11–13). This CIE coincides with an episode of intense global warming lasting ≈100 thousand years (Kyr) (14, 15), and the starting point of the excursion defines the P/E boundary (16, 17). It was during the PETM that euprimates, perissodactyls, and artiodactyls first appeared across the Holarctic continents. Early in the CIE interval, δ¹³C values decreased to a minimum and then gradually increased. The spike corresponding to the minimum value is situated 8.5–15 Kyr after the P/E boundary (14).

We correlated the CIE on the three northern continents and compared the stratigraphic positions and ages of the earliest records of Teilhardina on each continent (Fig. 1). The recently discovered Teilhardina asiatica from the Upper Lingcha Formation of China is from a level just above the P/E boundary, situated in the negative shift of the δ¹³C excursion but before the minimum value of the excursion (1,12). The type species Teilhardina belgica is known from Dormaal, just above the base of the fluviolagoonal Tienen Formation of Belgium (11). This base also lies within the negative shift of the δ¹³C excursion (18) and is estimated to be in an interval 4–10.5 Kyr after the P/E boundary (14). The early Eocene Willwood Formation of Wyoming has yielded five temporally successive Teilhardina species (19), of which Teilhardina brandti is the oldest. T. brandti, from the earliest Eocene (Wasatchian-0), was, until now, known from only one tooth (5), but several new specimens reported here reveal its phylogenetic importance. T. brandti and several other modern mammals (including artiodactyls and perissodactyls) first occur in a level that is situated above the minimum value of the δ¹³C excursion and that has an estimated age of 19–25 Kyr above the P/E boundary (14, 20, 21).

Fig. 1. Paleogeographic map showing hypothetical migration routes of Teilhardina during the earliest Eocene. (A–C) Timing of migration is obtained by correlations of the δ13C excursion in North America (A), Europe (B), and Asia (C). T. brandti (more ...)

Thus T. asiatica, T. belgica, and T. brandti were almost contemporaneous, but the slight differences in their ages suggest that Teilhardina appeared first in Asia, dispersed from Asia to Europe at ≈5–12 Kyr after the P/E boundary, and reached North America no later than 25 Kyr after the P/E boundary. The entire dispersal of Teilhardina across the three northern continents, therefore, probably occurred within 15–25 Kyr. Such a high level of precision in correlating intercontinental biotic events is unprecedented for the Early Cenozoic and results from discovery of the global CIE. We cannot, as yet, constrain local first appearances statistically within the CIE interval, but the reported differences are corroborated by the character analyses below (Table 4, which is published as supporting information on the PNAS web site).

Dispersal rates of living mammals are in the range of 1–10 km/yr (22). The distance covered by Teilhardina would be ≈20,000 km. Thus, even if Teilhardina dispersed at the minimum rate documented in extant mammals, ≈1 km/yr, it could have covered this distance in 20 Kyr, which is consistent with the dispersal rate obtained by correlation of the CIE on the three northern continents.

We hypothesize that the primate Teilhardina, and probably basal genera of several other modern orders, dispersed east-to-west around the Northern Hemisphere at the beginning of the PETM. They migrated from South Asia to Europe, crossing the Turgai Straits (23–25), and then dispersed to North America via Greenland (26). Recent description of T. asiatica (1), together with new specimens of T. belgica (3) and T. brandti, reported here, make it possible to evaluate dental morphology of Teilhardina on all three continents in detail (Figs. 2–4) to determine whether their anatomical differences are consistent with this dispersal hypothesis. Our comparisons indicate that morphological evidence mirrors the stratigraphic sequence of species just described.

Fig. 2. Earliest Eocene Teilhardina species: T. asiatica (A), T. belgica (B), T. brandti (C), and T. americana (D): lower teeth in occlusal view. Shown are IVPP V12357 (A), IRSNB M64 (B), UM 111434 (C1) (reversed in Fig. 3), UM 99031 (C2), USNM 493913 (C3) (reversed (more ...)

Fig. 3. Earliest Eocene Teilhardina species, T. asiatica (A), T. belgica (B), T. brandti (C), and T. americana (D): lower teeth in lingual view. Arrows indicate progressive elevation of metaconid on P4. T. brandti shown are UM 111434 (reversed, C1) and USNM 493913 (more ...)

Fig. 4. Earliest Eocene Teilhardina species, T. asiatica (A), T. belgica (B), T. brandti (C), and T. americana (D): last lower premolar (P4) in posterior view. Arrows indicate metaconid of P4, bars show increasing width of P4.

T. asiatica is morphologically most similar to T. belgica (1). These species share the following primitive characters: presence of P₁, large canine, narrow cheek teeth, and weak labial cingulids, which, in combination, make them more primitive than any other omomyid. Neither species has any obvious autapomorphic features. But T. belgica is more derived and closer to American Teilhardina species in having stronger reduction of the first three premolars, a lower protoconid on P_3–4–M₁, a wider P₄, and a squarer M₂.

New specimens of T. brandti were collected in 2003 and 2004 from three separate areas of basal Eocene (Wasatchian-0) age in the Bighorn Basin, Wyoming (Figs. 2–4 and Table 1). The new specimens corroborate the validity of the species, and their broad geographic distribution across the basin shows that T. brandti is an index fossil of the Wasatchian-0 fauna (like the primitive artiodactyl Diacodexis ilicis and the perissodactyl Hyracotherium sandrae). T. brandti is morphologically intermediate between European T. belgica and North American T. americana. T. brandti is slightly larger than T. belgica and about the same size as T. americana (Table 5, which is published as supporting information on the PNAS web site). In several characters, T. brandti more closely resembles T. belgica: the lower metaconid on P₄, the smaller and narrower hypoconulid lobe on M₃, the tendency to have better defined hypoconulids on M_1–2, the lower entoconid on M₂, and the more open talonid notch between the entoconid and the trigonid (Fig. 3). At the same time, T. brandti resembles T. americana in having slightly wider cheek teeth because of basal inflation of the crown, a stronger ectocingulum, or both. In fact, labial cingulid development is intermediate between the two species: some specimens have little or no cingulum, as in T. belgica, whereas others have a moderately to well developed cingulum, as in T. americana.

Table 1. List of specimens of T. brandti from earliest Eocene [Wasatchian (Wa)-0] of the Bighorn Basin, Wyoming

Whereas T. brandti is intermediate between T. belgica and T. americana, T. belgica is intermediate between T. asiatica and T. brandti (Tables 2 and 3). This morphocline suggests a basal omomyid lineage T. asiatica–T. belgica–T. brandti–T. americana, with a clear evolutionary gradient in dental characters. The Holarctic Teilhardina lineage is characterized by progressive reduction of the first three premolars, increasing elevation of the metaconid relative to the protoconid on P₄, and widening of P₄ and the molars (Fig. 4). Concomitantly, the medial incisor enlarged while the primitive caniniform canine became premolariform. Such gradual evolution was documented in endemic American lineages of omomyids, including Teilhardina (19, 27). This kind of dental modification probably reflects a shift in diet to a regimen richer in fruit and gums (28). The Teilhardina lineage is supported by an equally weighted parsimony analysis of 17 dental characters (Fig. 5). Only the omomyid Steinius vespertinus could modify the morphocline because S. vespertinus is more primitive than T. americana. However, Steinius occurs >1 million years later than T. americana (29) and could derive directly from a different Eurasian euprimate stock.

Table 2. Characters used in phylogenetic analysis

Table 3. Character matrix used for phylogenetic analysis

Fig. 5. Single most-parsimonious tree recovered in PAUP 4.0 b10 (34), derived from an equally weighted parsimony analysis of a matrix containing eight taxa and 17 dental characters (see Tables 3 and 5). Tree length = 36, consistency index excluding uninformative (more ...)

Fossil evidence suggests that North American omomyids came from Europe via the Greenland bridge rather than directly from Asia across the Bering land bridge. This hypothesis is in agreement with other Euroamerican mammal lineages studied from the earliest Eocene. The best species correlations are between representatives of modern orders from the Wasatchian-0 fauna of the Willwood Formation of Wyoming and those from the Tienen Formation of Dormaal, Belgium. Indeed, the earliest North American artiodactyl D. ilicis is slightly larger and more derived than Diacodexis gigasei from Dormaal (8, 30). D. ilicis is considered to be at the base of the American dichobunoid radiation. In addition to the species D. gigasei–D. ilicis, other closely allied Euroamerican species pairs in which the European species is slightly more primitive include the proviverrine creodonts Arfia gingerichi–Arfia junnei and Prototomus minimus–Prototomus deimos (31) and several lipotyphlan species. These Euroamerican lineages do not exclude a migration route from Asia to North America via the Bering land bridge for other taxa. Because some typical Asian late Paleocene groups are present in the late Paleocene or early Eocene of North America and are unknown in Europe, they probably dispersed directly from Asia to North America, possibly during the late Paleocene. This scenario could be the case for representatives of some modern orders, such as rodents, some perissodactyls, and limnocyonine creodonts.

Dispersal of the earliest Eocene primate Teilhardina and other modern mammals took place during the onset of the PETM and near the beginning of a marine transgression. This transgression followed a major eustatic lowering of sea level (11), which we infer exposed land bridges between the northern continents. The exceptionally high temperature at the beginning of the Eocene, combined with the existence of temporary land bridges, evidently permitted small subtropical mammals to cross high-latitude land bridges such as Greenland and Beringia. Because omomyids were strictly arboreal mammals, their rapid dispersal argues for the existence of a continuous evergreen forest belt at high latitudes during the PETM (32). This hypothesis is reinforced by recent studies showing evidence for a shift in the state of the climate system during this time, characterized by large increases in tropospheric humidity and enhanced cycling of carbon through terrestrial ecosystems (33).

This study is based on anatomical comparisons of new specimens of North American T. brandti with specimens of other species of Teilhardina (see Tables 1 and 5). Relative age of fossils is based on recent high-resolution carbon isotope stratigraphy (11, 12, 14).

We thank X. Ni and C. Li (Institute of Vertebrate Paleontology and Paleoanthropology, Beijing, China) for providing a cast of T. asiatica and allowing its illustration; R. Smith (Royal Belgian Institute of Natural Sciences, Brussels, Belgium) for access to specimens from his collection; P. Missiaen for running the cladistic analysis; J. Cillis for producing scanning electron microscopy photographs at the Royal Belgian Institute of Natural Sciences; W. Sanders for fossil preparation; and G. Gunnell, P. Missiaen, E. Steurbaut, S. Zack, and two anonymous reviewers for helpful comments. New fossils of T. brandti were collected under permits from the U.S. Bureau of Land Management, with support from National Geographic Grant 7630-04 and National Science Foundation Grants EAR-0000941 (to K.D.R.) and EAR-0125502 (to P.D.G.). This paper is a contribution to Research Project MO/36/011, financially supported by the Belgian Federal Science Policy Office. K.D.R.’s collaboration was facilitated by a Forschungspreis from the Alexander von Humboldt Stiftung.