DOI: 10.1126/Science.1175822 , 67 (2009); 326 Science

Macrovertebrate Paleontology and the Pliocene Habitat of Ardipithecus ramidus Tim D. White, et al. Science 326, 67 (2009); DOI: 10.1126/science.1175822

The following resources related to this article are available online at www.sciencemag.org (this information is current as of October 14, 2009 ):

Updated information and services, including high-resolution figures, can be found in the online version of this article at: http://www.sciencemag.org/cgi/content/full/326/5949/67

Supporting Online Material can be found at: http://www.sciencemag.org/cgi/content/full/326/5949/67/DC1 A list of selected additional articles on the Science Web sites related to this article can be found at: http://www.sciencemag.org/cgi/content/full/326/5949/67#related-content This article cites 25 articles, 2 of which can be accessed for free: http://www.sciencemag.org/cgi/content/full/326/5949/67#otherarticles

This article has been cited by 2 articles hosted by HighWire Press; see: http://www.sciencemag.org/cgi/content/full/326/5949/67#otherarticles

This article appears in the following subject collections: Anthropology http://www.sciencemag.org/cgi/collection/anthro on October 14, 2009 Information about obtaining reprints of this article or about obtaining permission to reproduce this article in whole or in part can be found at: http://www.sciencemag.org/about/permissions.dtl www.sciencemag.org Downloaded from

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright 2009 by the American Association for the Advancement of Science; all rights reserved. The title Science is a registered trademark of AAAS. AUTHORS’SUMMARIES

Macrovertebrate Paleontology and the Pliocene Habitat of Ardipithecus ramidus

Tim D. White, Stanley H. Ambrose, Gen Suwa, Denise F. Su, David DeGusta, Raymond L. Bernor, Jean-Renaud Boisserie, Michel Brunet, Eric Delson, Stephen Frost, Nuria Garcia, Ioannis X. Giaourtsakis, Yohannes Haile-Selassie, F. Clark Howell, Thomas Lehmann, Andossa Likius, Cesur Pehlevan, Haruo Saegusa, Gina Semprebon, Mark Teaford, Elisabeth Vrba

ver since Darwin, scholars have (red crosses in figure) account for speculated about the role that nearly a third of the entire large mam- Eenvironment may have played in mal collection. Leaf-eating colobines human origins, evolution, and adapta- today exhibit strong preferences for tion. Given that all living great apes live arboreal habitats, and the carbon iso- and feed in trees, it has been assumed tope compositions of the fossil teeth are that the last common ancestor we consistent with dense to open forest shared with these forms was also a for- arboreal feeding (see figure). est dweller. In 1925, Raymond Dart The other dominant large mammal described the first Australopithecus, a associated with Ar. ramidus is the child’s skull, at Taung, South Africa. spiral-horned antelope, Tragelaphus Its occurrence among other fossils (the kudu, green circle). Today, these indicative of a grassland environment antelopes are browsers (eating mostly prompted speculation that the open leaves), and they prefer bushy to on October 14, 2009 grasslands of Africa were exploited by wooded habitats. The dental morphol- early hominids and were therefore ogy, wear, and enamel isotopic com- somehow integrally involved with the position of the Aramis kudu species origins of upright walking. are all consistent with such place- The Ardipithecus-bearing sediments ment. In contrast, grazing antelopes at Aramis now provide fresh evidence (which eat mostly grass) are rare in that Ar. ramidus lived in a predomi- the Aramis assemblage.

nantly woodland setting. This and cor- The large-mammal assemblage www.sciencemag.org roborative evidence from fossil assem- shows a preponderance of browsers blages of avian and small mammals and fruit eaters. This evidence is con- imply that a grassland environment was sistent with indications from birds, not a major force driving evolution of small mammals, soil isotopes, plants, the earliest hominids. A diverse assem- and invertebrate remains. The emer- blage of large mammals (>5 kg body gent picture of the Aramis landscape weight) collected alongside Ardipithe- Carbon and oxygen isotope analyses of teeth from the Ar. ramidus during Ar. ramidus times is one of a cus provides further support for this localities. Species are listed in order of abundance, and isotopic woodland setting with small forest Downloaded from conclusion. Carbon isotopes from data separate species by what they ate and their environment. patches. This woodland graded into tooth enamel yield dietary information nearby habitats that were more open because different isotope signatures reflect different photosynthetic and are devoid of fossils of Ardipithecus and other forest-to-woodland– pathways of plants consumed during enamel development. Therefore, community mammals. Finally, the carbon isotopic composition of animals that feed on tropical open-environment grasses (or on grass-eat- Ar. ramidus teeth is similar to that of the predominantly arboreal, small, ing animals) have different isotopic compositions from those feeding on baboon-like Pliopapio and the woodland browser Tragelaphus, indi- browse, seeds, or fruit from shrubs or trees. Moreover, oxygen isotopes cating little dietary intake of grass or grass-eating animals. It is there- help deduce relative humidity and evaporation in the environment. fore unlikely that Ar. ramidus was feeding much in open grasslands. The larger-mammal assemblage associated with Ardipithecus was These data suggest that the anatomy and behavior of early systematically collected across a ~9 km transect of eroding sediments hominids did not evolve in response to open savanna or mosaic set- sandwiched between two volcanic horizons each dated to 4.4 million tings. Rather, hominids appear to have originated and persisted years ago. It consists of ~4000 cataloged specimens assigned to within more closed, wooded habitats until the emergence of more ~40 species in 34 genera of 16 families. ecologically aggressive Australopithecus. There are only three primates in this assemblage, and the rarest is Ardipithecus, represented by 110 specimens (a minimum of 36 individuals). Conversely, colobine monkeys and a small baboon-like monkey When citing, please refer to the full paper, available at DOI 10.1126/science.1175822.

www.sciencemag.org SCIENCE VOL 326 2 OCTOBER 2009 67 RESEARCH ARTICLES

Consequently, most early hominid-bearing open-air fossil assemblages conflate multiple biotopes (7). Under such circumstances, it is not Macrovertebrate Paleontology surprising that many Pliocene hominid habitats have been referred to as a “mosaic” or “achanging mosaic of habitats” (8). Such characterizations risk and the Pliocene Habitat confusing noise for signal and local for regional environment, particularly for collection-biased as- of Ardipithecus ramidus semblages lacking temporal and spatial resolution. Initial assessment of the fauna associated with 1 2 3 4 5 Tim D. White, * Stanley H. Ambrose, Gen Suwa, Denise F. Su, David DeGusta, Ar. ramidus indicated “a closed, wooded” environ- 6,7 8,9 10 11,12 Raymond L. Bernor, Jean-Renaud Boisserie, Michel Brunet, Eric Delson, ment (9), an inference subsequently misquoted as 13 14 15 16 Stephen Frost, Nuria Garcia, Ioannis X. Giaourtsakis, Yohannes Haile-Selassie, “forest” (10). This interpretation was criticized on 17 18 19 20 21 F. Clark Howell, † Thomas Lehmann, Andossa Likius, Cesur Pehlevan, Haruo Saegusa, the basis that colobine monkeys and tragelaphine 22 23 24 Gina Semprebon, Mark Teaford, Elisabeth Vrba bovids might be unreliable indicators (11, 12). Taxonomic abundance. Several aspects of A diverse assemblage of large mammals is spatially and stratigraphically associated with Ardipithecus LowerAramisMemberlargermammalassem- ramidus at Aramis. The most common species are tragelaphine antelope and colobine monkeys. blage abundance data constitute strong indica- Analyses of their postcranial remains situate them in a closed habitat. Assessment of dental mesowear, tors of ancient biofacies and biotope (13). The microwear, and stable isotopes from these and a wider range of abundant associated larger mammals locality-specific subassemblages are remarkably indicates that the local habitat at Aramis was predominantly woodland. The Ar. ramidus enamel isotope 1 values indicate a minimal C vegetation component in its diet (plants using the C photosynthetic Human Evolution Research Center and Department of In- 4 4 tegrative Biology, 3101 Valley Life Sciences Building, Uni- pathway), which is consistent with predominantly forest/woodland feeding. Although the Early Pliocene versity of California, Berkeley, CA 94720, USA. 2Department Afar included a range of environments, and the local environment at Aramis and its vicinity ranged of Anthropology, University of Illinois, 607 South Matthews from forests to wooded grasslands, the integration of available physical and biological evidence Avenue, Urbana, IL 61801, USA. 3The University Museum, Uni- Ar. ramidus versity of Tokyo, Hongo, Bunkyo-ku, Tokyo 113-0033, Japan. establishes as a denizen of the closed habitats along this continuum. on October 14, 2009 4Department of Anthropology, Bryn Mawr College, Bryn Mawr, PA 19010–2889, USA. 5Department of Anthropology, Stanford ircumscribing the ecological habitat of We consider ecological habitat to mean the University, Stanford, CA 94305–2034. 6National Science the earliest hominids is crucial for un- biological and physical setting normally and Foundation, Sedimentary Geology and Paleobiology Program, 7 derstanding their origins, evolution, and regularly inhabited by a particular species. Our Arlington, VA 22230, USA. College of Medicine, Department of C Anatomy, Laboratory of Evolutionary Biology, Howard University, adaptations. Evidence integrated from a variety floral definitions follow the United Nations Edu- 520 W Street, Washington, DC 20059, USA. 8Paléobiodiversité of independent geological and paleontological cational, Scientific, and Cultural Organization et Paléoenvironnements, UMR CNRS 5143, USM 0203, Muséum sources (1–3) help to place Ardipithecus ramidus (UNESCO) classification of African vegetation national d’histoire naturelle, 8 rue Buffon, CP 38, 75231 Paris in its regional and local Pliocene environmental (6). Forests have continuous stands of trees with cedex 05, France. 9Institut de paléoprimatologie et paléonto- logie humaine, évolution et paléoenvironnements, UMR CNRS

settings. Here, we assess fossils of the larger overlapping crowns, forming a closed, often www.sciencemag.org 6046, Université de Poitiers, 40 avenue du Recteur-Pineau, vertebrates (mammalian families in which most multistory canopy 10 to 50 m high; the sparse 86022 Poitiers cedex, France. 10Collège de France, Chaire de species exceed 5 kg adult body weight) to reveal ground layer usually lacks grasses. Forests grade Paléontologie humaine, 3 Rue d’Ulm, F-75231 Paris cedex 05, characteristics of their diets, water use, and hab- into closed woodlands, which have less contin- France. 11Department of Anthropology, Lehman College, City 12 itat preferences. uous canopies and poorly developed grass layers. University of New York, Bronx, NY 10468, USA. Department of Vertebrate Paleontology, American Museum of Natural At Aramis 4.4 Ma (million years ago), pre- Woodlands have trees with canopy heights of 8 History, New York, NY 10024, USA. 13Department of dominantly terrestrial plants, invertebrates, and to 20 m; their crowns cover at least 40% of the Anthropology, University of Oregon, Eugene, OR, 97403– vertebrates were buried relatively rapidly on a land surface but do not overlap extensively. 1218, USA. 14Departamento Paleontología, Universidad Complutense de Madrid y Centro de Evolución y Comporta- low-relief aggrading floodplain, away from pe- Woodland ground layer always includes heli- Downloaded from miento Humanos, ISCIII, C/ Sinesio Delgado 4, Pabellón 14, rennially moving water capable of displacing ophilous (sun-loving, C4) grasses, herbs/forbs, 28029 Madrid, Spain. 15Ludwig-Maximilians-University of most remains (2, 3). Collection bias was avoided and incomplete small tree and shrub under- Munich, Department of Geo- and Environmental Sciences, by a systematic 100% collection strategy (1). stories. Scrub woodland has a canopy height Section of Paleontology, Richard-Wagner-Strasse 10, D-80333 16 Therefore, the large mammal assemblage spa- less than 8 m, intermediate between woodland Munich, Germany. Department of Physical Anthropology, Cleveland Museum of Natural History, 1 Wade Oval Drive, tially associated with Ardipithecus in the Lower and bushland. As proportions of bushes, shrubs, Cleveland, OH 44106, USA. 17Human Evolution Research Aramis Member allows for relatively robust and and grasses increase, woodlands grade into Center and Department of Anthropology, 3101 Valley Life precise environmental inference compared with bushland/thickets or wooded grasslands. Sciences Building, University of California, Berkeley, CA many other hominid-bearing occurrences. Reconstructing the Aramis biotope. Recon- 94720, USA. 18Senckenberg Forschungsinstitut, Senckenber- ganlage 25, D-60325 Frankfurt am Main, Germany. 19Départe- The assemblage was carnivore-ravaged and structing an ancient environment based on ver- ’ 7 ment de Paléontologie, Université de N Djamena, BP 1117, is consequently dominated by bone and dental tebrate macrofossils is often imprecise ( ). Even N’Djamena, Chad. 20University of Yuzuncu Yil, Department of fragments (3). It represents an attritionally de- assemblages from a single stratigraphic interval Anthropology, Faculty of Science and Letters, Zeve Yerlesimi rived fauna collected between two widespread may sample thousands of years and thus repre- 65080 Van, Turkey. 21Institute of Natural and Environmental marker tuffs that are today exposed along an sent artificial amalgamations of different biotopes Sciences, University of Hyogo, Yayoigaoka, Sanda 669-1546, Japan. 22Science and Mathematics, Bay Path College, 588 extended erosional arc (2, 3). The larger mam- shifting on the landscape through time. Even in a Longmeadow Street, Longmeadow, MA 01106, USA. 23Center mal fossil assemblage (4) comprises 3837 in- geologically isochronous assemblage, animals for Functional Anatomy and Evolution, Johns Hopkins dividually cataloged specimens assigned to 42 from different habitats may be mixed by moving University School of Medicine, 1830 East Monument Street, 24 species (6 of them newly discovered), in 34 water or by a moving lake or river margin. Eco- Room 303, Baltimore, MD 21205, USA. Department of Geology and Geophysics, Yale University, New Haven, CT genera of 16 families (1, 5), across a wide body logical fidelity can be further biased through 06520, USA. size range (Fig. 1A). Many of the sampled taxa unsystematic paleontological recovery, for ex- *To whom correspondence should be addressed. E-mail: provide evidence for the evolution of African ample, when only more complete, identifiable, [email protected] vertebrates. and/or rare specimens are collected. †Deceased

www.sciencemag.org SCIENCE VOL 326 2 OCTOBER 2009 87 Ardipithecus ramidus consistent in their taphonomy and taxonomy Relative and absolute abundance measures morphological and functional properties of the across the ~7 km distance from the easternmost for the large mammals in our collections from recovered Diceros sp. molars are similar to those (SAG-VP-7) to westernmost (KUS-VP-2) Ar. the Ardipithecus-bearing Lower Aramis Member of the extant browsing Diceros bicornis. ramidus localities (3). localities were assessed by the number of iden- Equids are rare. One, Eurygnathohippus sp. Contemporaneous localities between the two tified specimens (NISP) (n = 1930) and the mini- nov., is distinguished by its distal limb, which is tuffs farther south of the modern Sagantole drain- mum number of individuals (MNI) based on adapted to open-country running. Its elongate- age (SAG-VP-1 and -3, at the southeastern paleo- teeth (n = 330). Proboscideans, giraffids, and narrow snout with parabolic symphysis suggests transect pole) are relatively impoverished. They hippopotamids are rare (Fig. 1, B and C). The adaptation to selective feeding. The teeth of this lack this diverse and abundant mammal assem- rhinos Ceratotherium efficax and Diceros are rep- equid bear a low-blunt cusp morphology reflect- blage and contain no tragelaphines, no monkeys, resented by few specimens (NISP 6 and 1, MNI 4 ing habitual grazing. Large carnivores and no fossil wood or seeds, no birds, no micromam- and 1, respectively). Unlike most other waterside aardvarks are rare, in keeping with their trophic mals, and no Ardipithecus (table S1). Comple- Plio-Pleistocene assemblages, rhinos are more level (as in most other eastern African Plio- mentary structural, taphonomic, and isotopic data abundant than hippos at Aramis. The dental meso- Pleistocene assemblages). from localities on this pole of the paleotransect wear pattern and occlusal morphology of Pliocene Ardipithecus ramidus is represented at Aramis suggest a more open landscape that supported Ceratotherium efficax suggest that it was predomi- and environs by >110 cataloged specimens repre- more crocodilians, turtles, and hippopotamids, nantly a grazer but ate less abrasive forage with senting a minimum number of 36 individuals presumably associated with water-marginal set- respect to its highly specialized Pleistocene and [14 by upper second molar (M2) count] in the tings more axial in the drainage basin (2, 3). extant descendant Ceratotherium simum.The Lower Aramis Member. These numbers are rel-

Fig. 1. Aramis large mammals. (A)Sizerange illustrated by astragali. The Lower Aramis Mem- ber contains a wide range of mammalian taxa, illustrated by this image. Top left, Rhinocerotidae; on October 14, 2009 middle left, Ardipithecus ramidus (ARA-VP-6/500); lower left, small bovid. Included are other artio- dactyls, carnivores, and rodents. (B)Relativeabun- dance of larger mammal taxa at Aramis based on

dental MNI. (C)Dental www.sciencemag.org NISP based on dental individuals whose tooth crowns are more than half complete. The NISP value reflects all collected specimens identified to the taxon and excludes

bulk specimens (tooth Downloaded from crowns less than half complete). Associated dental specimens are counted as one. The MNI values use permanent molars segregated into upper and lower first, second, and third molars, respectively. Numbers for each taxon vary between NISP and MNI, but the relative proportions hold similar. Tragelaphine bovids and cercopithecid monkeys dominate, accounting for more than half of the assemblage, however counted.

88 2 OCTOBER 2009 VOL 326 SCIENCE www.sciencemag.org RESEARCH ARTICLES atively low compared with many of the other overall by a ratio of 1.4 to Pliopapio (colobinae assemblage based on taxonomically diagnostic macrovertebrate fossil species we collected. This NISP:cercopithecinae NISP). It is slightly larger craniodental elements (Fig. 1). Together, these rarity is consistent with that observed for hom- (12 kg female, 18 kg male) than this papionin taxa account for more than half of the larger inids in other well-known vertebrate assem- (8.5 kg, 12 kg) based on dental regressions (14). mammal specimens, whether counted by NISP blages (7). Kuseracolobus aramisi and Pliopapio Extant colobines exhibit strong preferences for or dental MNI. Both cercopithecid and bovid as- alemui are ubiquitous in the assemblage, ac- arboreal habitats; extinct African taxa range from semblages appear to be attritional and were rav- counting for 30% of both the larger mammal fully arboreal to highly terrestrial (15). aged heavily by carnivores after death (3). NISP and MNI. The colobine is numerically Bovids and primates, particularly tragelaphines Bovids help illuminate the local Aramis en- dominant within nearly all of the localities, and and cercopithecids, dominate the larger mammal vironment of the Ardipithecus-bearing localities. One useful index is the relative abundance of grazing versus browsing taxa, which can indicate the presence of open or closed conditions, 100% – A respectively (16 19). The most ecologically sensitive of these taxa include grazing, open-habitat 90% tribes such as Alcelaphini and Hippotragini versus the primarily browsing Tragelaphini or the 80% riparian-associated Reduncini. Reduncine bovids commonly dominate in African Plio-Pleistocene 70% faunal assemblages (Fig. 2), in keeping with fluviatile, swampy, or lake marginal depositional 60% Hippotragini conditions. Antilopini Alcelaphini Whether counted by NISP or dental MNI, 50% Reduncini Tragelaphus (whose modern congeners are as- Bovini sociated with wooded habitats) (20) is the nu- Neotragini 40% merically dominant Aramis bovid, comprising Aepycerotini 85% (NISP) of the bovid assemblage (Fig. 1), Tragelaphini

30% followed by Aepyceros (whose modern form on October 14, 2009 favors grassy woodland to wooded grassland environments). In contrast, alcelaphine and reduncine 20% bovids that are plentiful at other Plio-Pleistocene sites are rare at Aramis, accounting for a mere 10% 1% (NISP) and 4% (MNI) of all bovids. Aramis is unlike any other known African fossil assem- 0% blages in that Tragelaphus dominates the un- Aramis Basal Sidi Denen Shungura B Apak Kaiyumung Kanapoi Laetoli (879) Member Hakoma Dora (171) (64) (57) (125) (1061) gulates. (20–23) (Fig. 2). (66) (129) (429)

Alcelaphines and reduncines were found at www.sciencemag.org 100% slightly higher frequencies at locality SAG-VP-7 B at the eastern end of the Ardipithecus distribution 90% (although tragelaphines and aepycerotines still dominate there). This subtle difference between 80% SAG-VP-7 and other more westerly hominid- bearing localities is also indicated by cercopithecid 70% Hippotragini abundance. SAG-VP-7 has relatively fewer cer-

Antilopini copithecids and more alcelaphine and reduncine Downloaded from 60% Alcelaphini bovids (Fig. 2), potentially signaling that this east- Reduncini ernmost Ardipithecus locality was a transition zone 50% Bovini Neotragini between two biotopes. Aepycerotini 40% Functional morphology. Taxon-based ap- Tragelaphini proaches to the inference of paleohabitats are Kuseracolobus aramisi usually restricted to using identifiable cranioden- 30% Pliopapio alemui tal remains and assume that habitat preference 20% persists through evolutionary time. Another approach is to evaluate the anatomy of fossils with 10% respect to its implications for functional adaptations. These methods presume that mammals 0% exhibit skeletal and dental adaptations for lo- ALL ARAMIS ARA-VP-1 ARA-VP-6 KUS-VP-2 SAG-VP-7 (1441) (1073) (218) (53) (79) comotion and feeding that correlate with their preferred environment (24). Samples of extant Fig. 2. Aramis taxonomic abundance. (A) Comparison between the relative abundance (dental) of taxa are used to quantify the relations between bovid taxa at Aramis and other Plio-Pleistocene sites (21, 23, 45). The bovid fauna at Aramis is skeletal/dental traits and environmental variables, markedly different due to the dominance of tragelaphines. All frequencies are based on NISP, with the results then applied to fossil forms (25). except for Hadar, which is based on MNI. (B) Within-site comparison of the relative abundances of Here, we evaluate the “ecomorphology” of bovids and cerocopithecids. Among Lower Aramis Member localities, SAG-VP-7 has relatively lower the most common large mammals at Aramis, abundances of cercopithecids and higher abundances of alcelaphines and reduncines, potentially the bovids and cercopithecid monkeys. For the indicative of ecotonal conditions at this easternmost locality of the Ardipithecus distribution. Aramis bovids, we evaluated the astragali and

www.sciencemag.org SCIENCE VOL 326 2 OCTOBER 2009 89 Ardipithecus ramidus phalanges (25, 26) because other elements that arboreal. There was no clear evidence of terres- relative contributions of grass, trees, and shrubs can be revealing (metapodials and femora) were trial adaptation in 18 proximal radii. Hence, to its diets. Oxygen isotopes can reveal the de- not preserved in sufficient numbers. We used a based on current criteria, there is clear evidence gree that a species lives in, or consumes, water four-habitat grouping scheme (26)(SOMtext of arboreal locomotor adaptations, and a paucity from different sources (34). We sampled tooth S1). Of the 11 available intact bovid astragali of terrestrial indicators, in the overwhelming enamel bioapatite from 177 specimens encom- with statistically significant habitat predictions majority of the Aramis cercopithecoid postcranial passing a wide range of mammalian taxa within (accuracy >95%), 10 are classified as “forest” and sample (SOM text S2). the Ar. ramidus–bearing unit (Fig. 4, SOM text one as “heavy cover.” This is a clear signal, since Dental wear. The morphology, occlusal wear, S3, and table S4). These were analyzed blind to these methods typically produce more varied and stable isotope composition of dental re- taxon. Carbon isotopic ratios for grazers are high, habitat predictions when applied to fossil sam- mains also reveal the diet—and, indirectly, hab- whereas those for mixed feeders, browsers, and ples (27, 28). To lessen possible biases introduced itat preferences—of some Aramis mammals. forest floor feeders decrease systematically by confining the analysis to specimens sufficient- Differences in mesowear can distinguish among (SOM text S4). Oxygen isotope vales are low ly complete for measurement, we also examined extant browsers, grazers, and mixed feeders (32). for water-dependent species such as carnivores nonmetric traits of the phalanges and classified The Aramis neotragines, Giraffa,andTragelaphus and hippos in wet riparian habitats and higher the entire astragali/phalangeal sample by mor- cluster with extant browsers (Fig. 3 and table for water-independent browsers and open dry- photype (SOM text S1, tables S2 and S3, and S3), whereas Aepyceros falls between extant mixed habitat species. fig. S1). These results independently support feeders and nonextreme grazers. Rare Aramis In the Ardipithecus-bearing Lower Aramis the conclusion from metric prediction that these alcelaphines cluster with nonextreme grazers, Member assemblage, the aquatic carnivore animals inhabited a “forest” (in the analytical, whereas the rare bovine and equid fossils are Enhydriodon (an otter) has the lowest d18Oof not floral, sense). closest to extant coarser grass grazers. all species. Conversely, the ursid Agriotherium As with bovids, cercopithecid postcranial fea- The high cusps and colobine-like morphology (a bear) has the highest carnivore d18O, consistent tures are routinely posited to indicate locomo- of Pliopapio alemui (tall molars with high relief with anatomical evidence for an omnivorous tion (29–31). However, systematic studies of large and little basal flare) suggest that the two Aramis diet (35). Among herbivores, giraffids (Giraffa samples of extant taxa are generally lacking. We monkey taxa had similar diets. We sampled a and Sivatherium) have the highest d18O and therefore consider most proposed correlations mixed set of colobine and cercopithecine molars lowest d13C values, whereas grazing equids between cercopithecoid anatomy and locomotor for a blind test of microwear. No significant dif- (Eurygnathohippus), alcelaphines, bovines, hip-

mode to be of unknown reliability, pending ad- ferences were found between the two taxa. Micro- potragines, and rhinocerotids show the converse. on October 14, 2009 ditional study. Even so, the elbow is clearly a wear on the Aramis monkey molars is consistent Among primates, Kuseracolobus has higher key joint for distinguishing between arboreal and with both frugivory and folivory but not hard d18Oandlowerd13CthanPliopapio,whichre- terrestrial primate locomotion. Of 10 available object feeding. A diet of soft (but perhaps tough) sembles the difference between modern folivo- Aramis cercopithecoid distal humeri, 9 are clearly foods would be typical of colobines, and the same rous Colobini and more omnivorous Papionini consistent with “arboreal” substrate, whereas only may have been the case for the papionin (33). (36, 37). one is consistent with “terrestrial” substrate based Enamel isotopes. The carbon isotopic com- The carbon isotopic composition of four of on current criteria. Of 9 proximal ulnae, all are position of a mammal’s tooth enamel reflects the five Ardipithecus ramidus individuals is close to www.sciencemag.org Fig. 3. Mesowear analysis results for the 100 second molar paracone 80 apex of fossil ungulates. Cusp shape was scored 60 qualitatively as sharp, 40 rounded, or blunt. The 20 relative difference in Downloaded from height between tooth %sharp %round %blunt %high %low %sharp %round %blunt %high %low %sharp %round %blunt %high %low cusp apices and inter- Aepyceros Alcelaphini Bovini cusp valleys (occlusal relief) was qualitatively scored as either high or low (large or small Mesowear Results distance between cusp Cusp Shape apex and intercusp valley, respectively). Histo- Sharp Rounded Blunt grams show the results on the mesowear vari- %sharp %round %blunt %high %low cusp Eurygnathohippus ables measured (i.e., the relief percentages of sharp versus rounded versus blunt cusp shapes and the percentages of high versus low occlusal relief).

%sharp %round %blunt %high %low %sharp %round %blunt %high %low %sharp %round %blunt %high %low Giraffa Neotragini Tragelaphini

90 2 OCTOBER 2009 VOL 326 SCIENCE www.sciencemag.org RESEARCH ARTICLES that of Pliopapio, reflecting diets that included prey contributed more to their diet compared to (41). Consistently lower oxygen isotope ratios sup- small amounts of 13C-enriched plants and/or ani- their modern congeners in grazer-dominated open port geological evidence that Gona was close to malsthatfedonsuchplants.Ardipithecus con- savanna environments (37). This is congruent permanent water (43), but higher carbon isotope sumed slightly more of these resources than with the numerical dominance of browsing tra- ratios for all Gona browsers are inconsistent modern savanna woodland chimpanzees (38) gelaphines and accords with other evidence for with greater water availability (SOM text S3). but substantially less than later Plio-Pleistocene the dominance of woodlands in the 4.4 Ma local Other ecological approaches. An approach hominids (39, 40). The fifth individual has a environment occupied by Ardipithecus (2, 3). A to deducing ancient environment is to first as- d13C value of –8.5 per mil (‰), which is closer small number of rare grazing species—mainly sign each mammal taxon in a fossil assemblage to, though still lower than, the means for Aus- equids, alcelaphines, hippotragines, and some to an ecological category (usually based on diet tralopithecus africanus, Au. robustus, and early impala, rhino, and bovines—have high d13C and locomotion) and then compare the propor- Homo (39, 40). Slightly lower d18Ocompared and d18O, indicating that they fed on water- tions of these categories in the fossil sample to a with Pliopapio and Kuseracolobus suggests that stressed C4 plants in drier, open environments range of similarly categorized extant communi- Ardipithecus obtained more water from fruits, (41). These taxa comprise a small portion of ties (44, 45). This approach uses only the presence bulbs, tubers, animals, and/or surface sources. the overall assemblage. or absence of taxa, so it is subject to taxonomic The isotopic composition of the Aramis mam- Thelargerangeofd18O, particularly the large and taphonomic biases involving small samples mals between the two tuffs (Fig. 4 and table S4) difference (9.6‰) between water-independent and mixing. Furthermore, the results are often of conforms broadly to patterns expected for their (evaporation-sensitive) Giraffidae (Giraffa and low resolution because biased local fossil as- modern congeners across the forest-woodland- Sivatherium) and water-dependent (evaporation- semblages are compared to variably recorded savanna spectrum (37, 38) in the East African insensitive) Hippopotamidae, suggests a mean modern communities that pool multiple habitats rift and is consistent with other early Pliocene annual evaporative water deficit of ~1500 mm (21). Ardipithecus ramidus was previously inter- assemblages (39, 40). Relatively low primate, (41). Therefore, Aramis was a generally dry wood- preted as inhabiting a woodland or dry forest giraffid, tragelaphine, and Deinotherium d13C land setting far from riparian environments. based on a preliminary Aramis faunal list (about values indicate that small patches of closed can- Enamel isotopes of these taxa from nearby pene- 10% of the sample now available) (46). Al- opy forests were present, although woodlands to contemporary sites at Gona (42) (SOM text S3 though the full faunal list produces results con- wooded grasslands probably dominated. Low and fig. S2) have a d18O difference of only 4.6‰, sistent with this finding, these results are not d13C values for hyaenids suggest that browsing reflecting an annual water deficit of ~500 mm highly robust because the data broadly overlap

among distinct environments (e.g., open, riparian, on October 14, 2009 medium-density, and closed woodland) (47). closed C3 C4 closed C3 C4 Other measures of abundance also provide forest browse graze forest browse graze information on the trophic structure of mamma- ABO- lian community represented by the Ardipithecus- 18 bearing Lower Aramis Member. Although there are many grazing and carnivorous species (Fig. 5),

35 open/dry,

enriched leaf water these taxa are rare (48), so a strict presence/ absence evaluation distorts the ecological signal. O-

When measures of relative abundance (NISP and www.sciencemag.org 18

SMOW 30 MNI) are included, along with direct information on trophic levels from the stable isotope and O‰

18 mesowear results, a different picture emerges. δ depleted leaf water Closed/humid, These combined data show that the large 25 mammal biomass at Aramis was dominated by browsers and frugivores (including frugivorous animals that consume leaves as a substantial part -10 -5 0

-10 -5 0 drinking water

Downloaded from Meteoric/surface of their diet). It is unlikely that a plethora of mam- δ13 δ13C‰ C‰ PDB PDB mals dependent on browse and fruit would have Giraffa (b, wi) Aepyceros (m/g, wi/wd) Anancus (g, wi) been able to subsist in an environment without Sivatherium (b, wi) Hippotragini (g, wd) Deinotherium (b, wd) abundant trees, the presence of which is witnessed Tragelaphini (b, wi) Nyanzachoerus kanamensis (g, wd) Elephantidae (g, wd) Ugandax (m/g, wd) Nyanzachoerus jaegeri (g, wd) Hyaenidae (c, wd) by fossil pollen as well as abundant seeds, wood, Ugandax cf. gautieri (g, wi?) Kolpochoerus deheinzelini (b/m, wd) Agriotherium (c/o, wd) phytoliths, and rhizoliths (2). Simatherium (g, wd?) Hippopotamidae (g, wd) Enhydriodon (c, wd) Neotragini (b, wi) Kuseracolobus aramisi (b, wi) Hominid habitat. Establishing habitat (as Alcelaphini (g, wi) Rhinocerotidae (g/m, wd) Pliopapio alemui (b/o, wd?) opposed to general environment) is crucial for Damalops (g, wd?) Eurygnathohippus (g, wd) Ardipithecus ramidus (b/o, wd?) illuminating the paleobiology of any fossil species, including hominids. On the basis of mixed Fig. 4. Carbon and oxygen isotopic composition of mammal tooth enamel from the Lower Aramis d13 d18 fossil faunas, it has been previously proposed that Member of the Sagantole Formation in the Middle Awash Valley. (A)Individual Cand Ovalues “ T early hominids were apparently not restricted to plotted by taxon. (B) Bivariate means 1 SD. See SOM text S3 for methods and interpretations, table S3 ” for raw data and statistics, and fig. S1 for comparison with species also occurring in roughly con- a narrow range of habitats. [(8), p. 571]. However, temporaneous deposits at Gona (42). Food and drinking habits are inferred from closest living relatives this raises the question of whether the hominids actually occupied a wide range of habitats or and from carbon and oxygen isotope ratios. b, browser (C3 feeder); m, mixed grazer/browser (C3 and C4 whether taphonomic processes and sampling feeder); g, grazer (mainly C4 grasses); wi, water-independent (evaporation-sensitive) (41) or obtaining substantial amounts of water from green leaves; wd, water-dependent (evaporation-insensitive) (41), biases have mixed hominid remains with those relying on drinking water when plant leaves are dry; c, carnivore; o, omnivore, including diets with leaves of species from biotopes that hominids rarely, if fruit, tubers, roots, flowers (all predominantly C3), seeds, fungi, and vertebrate and invertebrate animal ever, frequented. Many fossil assemblages sim- matter. Diets, water use, and habitat preferences of species of extinct genera and families are indicated ply do not preserve the necessary temporal and in italics because they are more intrinsically uncertain. Interpretations are described and justified in spatial resolution needed to determine whether detail in SOM text S3. hominids preferred the riverine forest, lake margin,

www.sciencemag.org SCIENCE VOL 326 2 OCTOBER 2009 91 Ardipithecus ramidus

4. This assemblage is the one co-occurring with Ardipithecus and excludes the small contemporary samples of fossils 100% from the more easterly localities of SAG-VP-1 and -3; see (2, 3) and table S1 for details. 90% 5. Included in the larger mammal subassemblage analyzed here are the following taxa: Artiodactyla, Perissodactyla, 80% Proboscidea, Primates, Carnivora (except Viverridae), and Tubulidentata. RT 6. F. White, The Vegetation of Africa, Natural Resources 70% O Research, Vol. 20. (United Nations Scientific and Cultural I Organization, Paris, 1983). 60% C 7. T. White, in Evolutionary History of the Robust Australopithecines, F. Grine, Ed. (Aldine de Gruyter, New Fg 50% York, 1988), pp. 449–483. FG 8. M. G. Leakey, C. S. Feibel, I. McDougall, A. Walker, 40% MF Nature 376, 565 (1995). G 9. G. WoldeGabriel et al., Nature 371, 330 (1994). 10. N. E. Sikes, B. A. Wood, Evol. Anthropol. 4, 155 (1995). 30% B 11. B. R. Benefit, in African Biogeography, Climate Change, and Human Evolution, T. Bromage, F. Schrenk, Eds. 20% (Oxford Univ. Press, New York, 1999), pp. 172–188. 12. M. G. Leakey, in African Biogeography, Climate Change, 10% and Human Evolution, T. Bromage, F. Schrenk, Eds. (Oxford Univ. Press, New York, 1999), pp. 271–275. 0% 13. Taphonomic and curatorial biases inevitably compromise Faunal List MNI NISP quantitative interpretations of any assemblage, including Aramis. For example, a single hominid canine may break into only a few identifiable fragments, whereas one Fig. 5. Trophic ecovariable distributions by faunal list, dental NISP, and dental MNI. Comparisons of elephantid’s tusk or molar can shatter into thousands of the Aramis trophic structure based on the faunal list versus specimen-level, dental relative abundance identifiable fragments. Simple comparisons of fragment – abundance can therefore be misleading. Our abundance data as measured by NISP and MNI. Grazing and carnivorous species are abundant in the faunal list data take these potential problems into account (see Fig. based trophic structure, whereas browsers and frugivores dominate when NISP and MNI data are

1 for details). on October 14, 2009 incorporated. B, browser; G, grazer; MF, mixed feeder; FG, fresh grass grazer; Fg, Frugivore (includes 14. S. R. Frost, Am. Mus. Novit. 3350, 1 (2001). fruit and leaves); C, carnivorous; I, insectivorous; O, omnivorous; RT, root and tuber. 15. J. F. Oates, A. G. Davies, E. Delson, in Colobine Monkeys: Their Ecology, Behaviour, and Evolution, A. G. Davies, J. F. Oates, Eds. (Cambridge Univ. Press, Cambridge, bushland, savanna, and/or woodland habitats has not been found in the apparently more open 1994) pp. 45–74. demonstrably available within a few kilometers settings to the southeast. There is no evidence 16. P. Shipman, J. Harris, in Evolutionary History of the of most depositional loci within rift valley settings. of any taphonomic bias related to Ardipithecus Robust Australopithecines, F. Grine, Ed. (Aldine de Gruyter, New York, 1988), pp. 343–381. For example, Ardipithecus ramidus has also that might produce this pattern (3) and no evi- 17. E. S. Vrba, in Fossils in the Making, A. Behrensmeyer, been found at Gona, about 70 km to the north of dence of any other spatial or stratigraphic clus- A. P. Hill, Eds. (Univ. of Chicago Press, Chicago, 1980), pp. 247–271.

Aramis, in a valley margin environment where tering within the 4.4 Ma Lower Aramis Member www.sciencemag.org lake deposits interfingered with small fluvial interval. 18. R. Bobe, G. G. Eck, Paleobiology 27, 1 (2001). 19. G. Suwa et al., J. Vert. Paleontol. 23, 901 (2003). channels or lapped onto basaltic cones and flows Based on a range of independent methods 20. The Aramis Tragelaphus cf. moroitu has a body size close (43). At Gona, the dominance of C3 plants indi- for inferring habitat-based large samples of con- to that of the living nyala (T. angasii) and is likely a cated by paleosol isotopes contrasts with the C4 silient spatial, geological, and biological evidence direct descendent of the T. moroitu recorded from the plant signal in many associated ungulate grazers generated from diverse sources, we therefore Mio-Pliocene of Asa Koma and Kuseralee. 21. D. Su, thesis, New York University (2005). (indicated by enamel isotopic data). Levin et al. conclude that at Aramis, Ar. ramidus resided and 22. R. Bobe, J. Arid Environ. 66, 564 (2006). thus concluded that Ardipithecus “...may have usually died in a wooded biotope that included 23. K. E. Reed, J. Hum. Evol. 54, 743 (2008).

inhabited a variety of landscapes and was not as closed through grassy woodlands and patches of 24. H. A. Hespenheide, Annu. Rev. Ecol. Syst. 4, 213 (1973). Downloaded from ecologically restricted as previous studies suggest” true forest [sensu (6)]. There is no evidence to 25. D. DeGusta, E. S. Vrba, J. Archaeol. Sci. 30,1009 (2003). [(42), p. 232]. The Gona paleontological and associate this hominid with more open wooded 26. D. DeGusta, E. S. Vrba, J. Archaeol. Sci. 32,1099 isotopic data show only that a range of habitats grasslands or grassland savanna. (2005). was present, and the attribution of Ardipithecus to Isotopic data indicate that the Ar. ramidus diet 27. Y. Haile-Selassie et al., Geobios 37, 536 (2004). any particular set of the available biotopes is was predominantly forest- to woodland-based. This 28. L. J. Hlusko, Y. Haile-Selassie, D. DeGusta, Kirtlandia 56, problematical in this mixed assemblage (49). Fish, interpretation is consistent with evidence of the 163 (2007). 29. J. G. Fleagle, Yearb. Phys. Anthropol. 20, 440 (1976). birds, browsers, horses, and hominids are all fre- dental and skeletal biology of this primate (1). The 30. H. B. Krentz, in Theropithecus, N. Jablonski, Ed. quently found in a single mixed fossil assem- ecological context of 4.4 Ma Aramis hominids, (Cambridge Univ. Press, Cambridge, 1993), blage in a fluviatile or near-shore deposit. This combined with their absence or extreme rarity at pp. 383–422. does not mean that the fish were arboreal or that Late Miocene and Early Pliocene sites, suggest that 31. S. Elton, Folia Primatol. (Basel) 73, 252 (2002). 32. M. Fortelius, N. Solounias, Am. Mus. Novit. 3301, horses were aquatic. Neither do such data mean the anatomy and behavior of the earliest hominids 1 (2000). that the hominids exploited all potentially avail- did not evolve in response to open savanna or 33. Crushing and shearing areas of 10 cercopithecoid molars able habitats. mosaic settings. Rather, this clade appears to have yielded surfaces that could be included in this analysis. The Lower Aramis Member deposits provide originated within more closed habitats favored Surface images were made using an SEM at ×500 magnification, and microwear features were collected fossil samples that evidence a range of environ- by these peculiar primates until the origin of using the “Microwear” software (v. 2.2, 1996). Microwear ments in the region at 4.4 Ma (2, 3). However, Australopithecus, and perhaps even beyond (50). features included relatively few pits, with narrow pits and the consistent association of Ar. ramidus with a scratches. The microwear on the molars of the Aramis particular fauna and flora in deposits between monkeys is consistent with both frugivory and folivory, SAG-VP-7 and KUS-VP-2 suggests its persistent References and Notes but they were not routinely feeding on hard objects. 1. T. D. White et al., Science 326, 64 (2009). A diet of soft, but perhaps tough, foods would be typical occupation of a woodland with patches of forest 2. G. WoldeGabriel et al., Science 326, 65 (2009). of colobines, and the same might be true for the across the paleolandscape (2, 3). Ardipithecus 3. A. Louchart et al., Science 326, 66 (2009). papionin, which has tall molars with a large amount of

92 2 OCTOBER 2009 VOL 326 SCIENCE www.sciencemag.org RESEARCH ARTICLES

relief and has a low level of basal flare in comparison 44. P. J. Andrews, J. M. Lord, E. M. Nesbit Evans, Biol. J. Linn. RHOI; and grant SBR 98-71480 for mass spectrometry with other papionins. Soc. Lond. 11, 177 (1979). instrumentation at the Environmental Isotope 34. T. E. Cerling, J. M. Harris, Oecologia 120, 347 (1999). 45. K. Kovarovic, P. Andrews, L. Aiello, J. Hum. Evol. 43, 395 (2002). Paleobiogeochemistry Laboratory) and the Japan Society 35. B. Sorkin, Hist. Biol. 18, 1 (2006). 46. P. Andrews, L. Humphrey, in African Biogeography, for the Promotion of Science (G.S. and H.S.). We thank 36. M. L. Carter, thesis, University of Chicago (2001). Climate Change, and Human Evolution, T. Bromage, L. Bach, H. Gilbert, and K. Brudvik for illustrations; 37. S. H. Ambrose, M. J. DeNiro, Oecologia 69, 395 (1986). F. Schrenk, Eds. (Oxford Univ. Press, New York, 1999), the Ministry of Tourism and Culture, the Authority for 38. M. Sponheimer et al., J. Hum. Evol. 51, 128 (2006). pp. 282–300. Research and Conservation of the Cultural Heritage, and 39. M. Sponheimer, J. A. Lee-Thorp, in Handbook of 47. Our analysis also raised numerous questions about the the National Museum of Ethiopia for permissions and Paleoanthropology, W. Henke, I. Tattersall, Eds. assumptions and procedures underlying such efforts. facilitation; and the Afar Regional Government, the Afar (Springer, Berlin, 2007) pp. 555–585. 48. For example, there are 12 “grazing” taxa compared to people of the Middle Awash, and many other field 40. N. J. Van Der Merwe, F. T. Masao, M. K. Bamford, S. Afr. only 5 “browsing” taxa, but the former are represented workers for contributing directly to the data. J. Sci. 104, 153 (2008). by only 152 specimens, whereas the latter are represented 41. N. E. Levin, T. E. Cerling, B. H. Passey, J. M. Harris, by 758 (NISP). J. R. Ehleringer, Proc. Natl. Acad. Sci. U.S.A. 103, 11201 49. It is evident that in most rift-valley depositional settings, Supporting Online Material (2006). a variety of environments would almost always have been www.sciencemag.org/cgi/content/full/326/5949/67/DC1 42. N. E. Levin, S. W. Simpson, J. Quade, T. E. Cerling, available to hominids. Of primary interest is determining SOM Text S. R. Frost, in The Geology of Early Humans in the Horn of whether any one of these environments was the preferred Figs. S1 and S2 Africa,J.Quade,J.Wynn,Eds.(GeologicalSocietyof habitat of these primates. Mixed assemblages cannot Tables S1 to S5 America Special Papers, 2008), vol. 446, pp. 215–234. usually do this. References 43. J. Quade et al., in The Geology of Early Humans in the 50. T. D. White et al., Nature 440, 883 (2006). Horn of Africa, J. Quade, J. Wynn, Eds. (Geological Society 51. Supported by NSF (grants SBR-82-10897, 93-18698, 95- 4 May 2009; accepted 14 August 2009 of America Special Papers, 2008), vol. 446, pp. 1–31. 12534, 96-32389, 99-10344, and 03-21893 HOMINID- 10.1126/science.1175822 on October 14, 2009 www.sciencemag.org Downloaded from