Australopithecus garhi: A New Species of Early Hominid from Ethiopia

See allHide authors and affiliations

Science  23 Apr 1999:
Vol. 284, Issue 5414, pp. 629-635
DOI: 10.1126/science.284.5414.629


The lack of an adequate hominid fossil record in eastern Africa between 2 and 3 million years ago (Ma) has hampered investigations of early hominid phylogeny. Discovery of 2.5 Ma hominid cranial and dental remains from the Hata beds of Ethiopia's Middle Awash allows recognition of a new species of Australopithecus. This species is descended from Australopithecus afarensis and is a candidate ancestor for early Homo. Contemporary postcranial remains feature a derived humanlike humeral/femoral ratio and an apelike upper arm–to–lower arm ratio.

The succession of early hominid genera and species indicates diversification into at least two distinct adaptive patterns by ∼2.7 Ma. A meager east African hominid record between 2 and 3 Ma has caused the pattern and process of this diversification to remain obscure. The Australopithecus afarensis (3.6 to 3.0 Ma) to A. aethiopicus(2.6 Ma) to A. boisei (2.3 to 1.2 Ma) species lineage is well corroborated by craniodental remains. In contrast, a suggested relationship between A. afarensis and early Homo has previously been evidenced only by relatively uninformative isolated teeth (1), a palate (2), and a temporal fragment (3).

The recovery of hominid remains from the Hata (abbreviation of Hatayae) Member of the Bouri Formation adds substantially to the inventory of fossils bearing on these phylogenetic issues. These remains comprise craniodental and postcranial elements from several areas in the Middle Awash. The first of these was discovered in 1990 at Matabaietu and Gamedah. Biochronology and Ar/Ar dating place these remains at ∼2.5 Ma. They include a small left parietal fragment (GAM-VP-1/2) and an edentulous left mandible corpus from Gamedah (GAM-VP-1/1), as well as a distal left humerus from Matabaietu (MAT-VP-1/1). It is impossible to attribute the humerus and parietal fragments to a genus. However, the small, fluvially abraded Gamedah mandible retains tooth roots and corpus contours. These demonstrate that it is not a robust Australopithecus.

It was not until 1996–1998 that we recovered additional hominid remains of comparable antiquity west of the modern Awash, at Bouri. The proximal half of an adult hominid ulna (BOU-VP-11/1) was found on the surface of the Hata beds by T. Assebework on 17 November 1996. On 30 November, White found a proximal femur and associated forearm elements of a smaller individual, ∼100 m to the WNW (BOU-VP-12/1A-G). Sieving and excavation revealed additional portions of this individual's femur in situ, 1 m above a 2.496 Ma volcanic ash, in a horizon with abundant catfish remains and medium-sized bovid fossils, the latter bearing cut marks (4).

This partial hominid skeleton includes fairly complete shafts of a left femur and the right humerus, radius, and ulna. A partial fibular shaft, a proximal foot phalanx, and the base of the anterior portion of the mandible were also found. There is no evidence that these remains represent more than one individual. Except for the in situ distal femoral shaft segment, all were surface finds lying within 2 m of one another. All are similarly preserved. Lengths can be accurately estimated for the phalanx, the femur, and the three arm elements. The foot phalanx is similar to remains of A.afarensis in size, length, and curvature. The mandible does not retain diagnostic morphology. No associated hominid teeth were found on the surface or in a large excavation.

Further search of the same ∼2.5 Ma horizon led to the discovery, 278 m farther NNW, of a partial hominid cranium (BOU-VP-12/130) on 20 November 1997 by Y. Haile-Selassie (Fig. 1). Another individual's crested cranial vault fragment (BOU-VP-12/87) was found 50 m south of the BOU-VP-12/1 skeleton excavation. At a more northerly locality in the Esa Dibo area ∼9 km away, A. Defleur found a fairly complete mandible, with dentition, of another hominid individual (BOU-VP-17/1) on 17 November 1997. An additional hominid humeral shaft (BOU-VP-35/1) was found ∼1 km farther north of the location of the mandible, on 4 December 1998, by D. DeGusta. On biochronological grounds, these Esa Dibo specimens are about the same age as the more southerly cluster of hominid remains at Bouri localities 11 and 12 (4).

Figure 1

Cranial parts of BOU-VP-12/130. (Top) Superior view of the original fossil. Nonstandard orientation (rotated posteriorly ∼10° from Frankfurt horizontal) to show maximum anatomy. (Bottom) Lateral view of casts to show cranial and maxillary profiles. Note that neither Frankfurt horizontal nor placement of the maxilla relative to the vault can be accurately determined and that reconstructed portions (indicated by oblique lines) are speculative. Photos ©David L. Brill 1999∖Atlanta.

Great uncertainty has continued to confound the origin ofHomo because of a lack of evidence from the interval between 2 and 3 Ma (5). The 2.5 Ma Bouri Hata hominids bear directly on these issues. In addition, they are closely associated with behavioral evidence of lithic technology (4).Australopithecus africanus from South Africa is roughly contemporary with the Hata remains. In eastern Africa, A.aethiopicus and at least one other putative lineage ancestral to early Homo are contemporaries in the Turkana Basin. The BOU-VP-12/130 cranial remains represent no previously named species. Only the recovery of additional specimens with associated crania and dentitions may allow the Bouri postcrania to be positively attributed to this new taxon. Therefore, the new species described below is established strictly on the basis of craniodental remains.

The following is a description of Australopithecus garhi, based on the BOU-VP-12/130 specimen: order, Primates Linnaeus 1758; suborder, Anthropoidea Mivart 1864; genus, AustralopithecusDART 1925; and species, Australopithecus garhi.

Etymology. The word garhi means “surprise” in the Afar language.

Holotype. ARA-VP-12/130 is an associated set of cranial fragments comprising the frontal, parietals, and maxilla with dentition. It was found by Y. Haile-Selassie on 20 November 1997. The holotype is housed at the National Museum of Ethiopia, Addis Ababa.

Locality. Bouri Vertebrate Paleontology locality 12 (BOU-VP-12) is on the eastern side of the Bouri peninsula, west of the modern Awash River, in the Middle Awash paleoanthropological study area, Afar depression, Ethiopia. The BOU-VP-12/130 holotype was found at 10°15.6199′N, 40°33.8445′E, at ∼550 m elevation.

Horizon and associations. The holotype was recovered from silty clays within 2 m of the top of the Maoleem vitric tuff, which has been dated to 2.496 Ma by Ar/Ar. Vertebrate fossils, including additional hominids, were found at the same stratigraphic horizon on nearby outcrops (4).

Diagnosis. Australopithecus garhi is a species of Australopithecus distinguished from other hominid species by a combination of characters presented in Table 1. It is distinguished from A.afarensis by its absolutely larger postcanine dentition and an upper third premolar morphology with reduced mesiobuccal enamel line projection and less occlusal asymmetry. Australopithecus garhi lacks the suite of derived dental, facial, and cranial features shared by A. aethiopicus, A.robustus, and A. boisei.Australopithecus garhi is distinguished from A.africanus and other early Homo species by its primitive frontal, facial, palatal, and subnasal morphology.

Table 1

List of characters. Listed are characters widely used in consideration of hominid phylogenetics (11, 13) that are preserved on the BOU-VP-12/130 holotype cranium. Because of arbitrary boundaries of presence or absence criteria, variability within species, limited sample sizes, and possible correlation between features, we caution against a numerical cladistic application of these tabulated data. Rather, this character list is meant to demonstrate the phenetic status of the single knownA. garhi specimen with respect to features used to evaluate early hominid fossils. Note that despite the large postcanine dentition, no shared derived characters link A. garhi with A.robustus or A. boisei. The “earlyHomo” column comprises specimens assigned by various authors to both H. habilis and H.rudolfensis. Abbreviations are as follows: mod., moderate; asym., asymmetric; disp., disparate; sym., symmetric; rect., rectangular; para., parabolic; var., variable; conv., convergent; div., divergent; ant., anterior; prom., prominent; proc., procumbent; cont., continuous; interm., intermediate.

View this table:

Dental description. The postcanine dental size is remarkable, at or beyond the known nonrobust and A.robustus extremes (Fig. 2). The anterior dentition is also large, with I1 and canine breadths equivalent to or exceeding those of their largest knownAustralopithecus and early Homo homologs. Thus, despite exceptional postcanine size, dental proportions of the holotype deviate markedly from the robust Australopithecus condition. The canine-to-premolar/molar size ratios are comparable to those ofA. afarensis, A. africanus, and early Homo. Relative canine to incisor alveolar length is most similar to that of A. africanus. Postcanine wear, with developed angular facets and retention of buccal cusp saliency, differs distinctively from the robustAustralopithecus pattern. The upper P3 is more derived than that of A. afarensis and most A.africanus specimens in exhibiting a reduced mesiobuccal crown quadrant and a weak transverse crest. The buccolingual narrowing of premolars and first molars often seen in early Homo is absent.

Figure 2

Dental size ofA. garhi compared with other early hominid taxa and specimens. (A) Canine breadth for various taxa. (B) The square root of calculated (MD × BL) premolar area. (C) The square root of calculated (MD × BL) second molar area. (D) Canine breadth relative to postcanine tooth size for various taxa. (E) Anterior alveolar length (mesial I1 to distal C) relative to postcanine tooth size. In (A) through (C), taxon means, standard deviations, ranges, and sample sizes (in parentheses) are given. All measures were taken by T.W. and G.S. on originals except for A.L. 444-2 and A.L. 417-1 (A.afarensis) and A.L. 666-1 (Homo), which are from (2, 14). Dental metrics for the BOU-VP-12/130 specimen are as follows (XX broken; parentheses = estimate; mesiodistal measure reported first, followed by buccolingual): RI1 XX, (9.2); RI2 6.9, 6.8; RC 11.6, 12.9; RP3 (11.0), 16.0; RP4 XX, XX; RM1 XX, XX; RM2 (14.4), (17.7); RM3 (15.2), 16.9; LI2 6.7, 7.0; LC 11.7, 12.9; LP3 (11.4), 16.0; LP4 (11.4), 16.0; LM1 (14.4), (16.5). □, A. afarensis; ○, A. africanus; ▵, Homo; ⧫, A. robustus; ▴,A. boisei; •, BOU-VP-12/130.

Cranial description. The lower face is prognathic, with procumbent incisors. Canine roots are placed well lateral to the nasal aperture margin. The premaxillary surface is separated from the nasal floor by a blunt ridge and is transversely and sagittally convex. The palate is vertically thin (∼3 mm at M1/M2 midline). The zygomatic roots originate above P4/M1. The dental arcade is U-shaped, with slightly divergent dental rows (Fig. 3). The temporal lines encroach deeply on the frontal, past the midsupraorbital position, and probably met anterior to bregma. The postglabellar frontal squama is depressed in a frontal trigon. The localized frontal sinus is limited to the medial one-third of the supraorbital surface. Postorbital constriction is marked. The parietal bones have a well-formed, bipartite, anteriorly positioned sagittal crest that divides above lambda. An endocast was made from the aligned parietals and frontal and was completed by sculpting by R. Holloway. Cranial capacity was about 450 cm3, as measured by water displacement.

Figure 3

The most complete palates of A. afarensis (A.L. 200-1a; canine reset) (A) and A. boisei (OH-5) (B) compared with that of A. garhi (C and D). The photograph (©David L. Brill 1999∖Atlanta) was mirror-imaged on midline.Australopithecus garhi has relatively large canines likeA. afarensis and absolutely large but morphologically nonrobust premolars and molars. Drawings ©L. Gudz.

Taxonomic discussion. There is no current agreement about how many pre-erectus Homo species should be recognized or even on how the genus Homo should be defined. The traditional conservative definition emphasizes adaptive plateau. Ironically, by this definition, the early Homo speciesH. rudolfensis and H.habilis might be better placed inAustralopithecus, as this would affiliate the major adaptive breakthroughs in anatomy and behavior that characterize H.erectus (ergaster) with the earliest defined occurrence of Homo. If A. garhi proves to be the exclusive ancestor of the Homo clade (see discussion below), a cladistic classification might assign it to genusHomo. Here we provisionally adopt the conservative, grade-sensitive alternative, emphasizing its small brain and large postcanine dentition by assigning the new Bouri species toAustralopithecus. This attribution as well as our diagnosis and description may require emendation when additional individuals representing the species are recovered and firm postcranial associations are established (6).

Although the Bouri Hata postcrania cannot presently be assigned to the new species A. garhi, they illuminate aspects of hominid evolution. The past few years have witnessed a rash of attempts to estimate early hominid limb length proportions from fragmentary and unassociated specimens. These specimens have been used to generate a variety of functional and phylogenetic scenarios. Accurate estimates of the limb proportions of early hominids, however, must be confined to the very few specimens that actually preserve relevant elements, such as the A.L. 288-1 (“Lucy”) specimen and KNM-WT 15000. The new Bouri VP-12/1 specimen is only the third Plio-Pleistocene hominid to provide reasonably accurate limb length proportions. The Olduvai Hominid 62 specimen of Homo habilishas been erroneously argued to show humerus-to-femur proportions more primitive than those of “Lucy” (7, 8), but its femur length cannot be accurately estimated. Other studies of limb proportions in early Australopithecus species are based on unassociated joints and not on actual (or even estimated) limb lengths (7).

The postcranial remains recovered from BOU-VP-11, -12, and -35 cannot be conclusively allocated to taxon. The BOU-VP-12/1 specimen features a humanlike humeral/femoral ratio (Fig. 4). This ratio may be an important derivation relative to A.afarensis, because it marks the earliest known appearance of the relative femoral elongation that characterizes later hominids. However, as in A. afarensis, the specimen's brachial index is apelike. This suggests that upper arm–to–lower arm ratios persisted into the basal Pleistocene and that the first hominid with modern forearm proportions was probably Homo erectus(ergaster). Because the A. afarensisforearm was also long relative to both the humerus and femur, the femur must have elongated before forearm shortening in early hominids.

Figure 4

Probable stages in the progressive differentiation of hominid long bone proportions (all bones shown to the same scale). The humerus (top), antebrachium (middle), and femur (bottom) are of about equal length in chimpanzees (Pan). Modern humans differ in two primary ways. Although our humerus is virtually the same length, the femur is elongated and the antebrachium is shortened. These changes appear to have emerged fully by ∼1.5 Ma in H. erectus [all three limb segments are virtually complete in KNM-ER-15000 (15)]. On the basis of the other two partial skeletons in which long bone length can now be reliably estimated, the modern human pattern appears to have emerged in two stages: (i) elongation of the femur, which is intermediate in length relative to the humerus in A.L. 288-1 but exhibits modern proportions in BOU-VP-12/1; and (ii) shortening of the antebrachium, which retains primitive proportions in both specimens (16). Drawings ©L. Gudz.

The BOU-VP-11 ulna is from a larger individual, as is the BOU-VP-35/1 humeral shaft (estimated total humeral length = 310 to 325 mm). The latter is absolutely longer than the humerus of BOU-VP-12/1 but is less rugose and probably bore a smaller deltopectoral crest. These differences (in both size and rugosity) are well within the species ranges of extant hominoids. If both humeri represent the same taxon, they could reflect sexual dimorphism, which would be comparable to that currently seen in A. afarensis. However, it is perilous to speculate on differences between only two specimens as they may reflect only fluctuating intraspecific variation in morphology and body mass.

The few and fragmentary nonrobust Turkana Basin hominids that span the 2.7 to 2.3 Ma time range are similar to the Bouri specimens in both size and aspects of morphology. Postcanine dental arcade length of BOU-VP-12/130 is equivalent to that of Omo 75-14, whereas individual teeth of the smaller Middle Awash mandibles (GAM-VP-1/1 and BOU-VP-17/1) are comparable in size to the smaller specimens of the Omo nonrobust collection. It is also important that BOU-VP-17/1 exhibits a derived lower P3 morphology (1) most similar to the Omo nonrobust and early Homo conditions and a dental arcade shape concordant with that of the holotype of A.garhi.

On the basis of size, BOU-VP-12/130 is a male. The craniodental size dimorphism documented for the closely related A.afarensis and A. boisei therefore predicts smaller individuals in A. garhi. The biochronologically contemporary and morphologically compatible BOU-VP-17/1 and GAM-VP-1/1 specimens are considerably smaller and may be females. This would suggest a shift in either or both body and dentognathic sizes to averages greater than in A.afarensis. More specimens are needed to test this hypothesis.

The discovery of A. garhi provides a strong test of many phylogenetic hypotheses that have addressed the relationships among Plio-Pleistocene hominid taxa. The South African species A. africanus was once widely considered to be the most primitive hominid. Discoveries of A.afarensis at Hadar and Laetoli displaced A.africanus. This more primitive sister species (A.afarensis) was in turn supplanted when the increasingly older and more primitive sister taxa A. anamensis(9) and Ardipithecus ramidus (10) were identified. However, the geometry of post-afarensis hominid phylogeny continues to be the focus of debate.

The position of A. africanus relative to the emergence of the genus Homo has been particularly difficult to resolve, even in the face of unduly elaborate phylogenetic analyses (11). One reason for this difficulty is the fundamental disagreement on whether early Homo comprises one sexually dimorphic (H. habilis) or two (H. habilis and H.rudolfensis) species. Most phylogenetic efforts have placedA. africanus as the link between A.afarensis and early Homo. This hypothesis has been widely, but not universally, accepted. Most predicted that a population of A. africanus would be found in eastern Africa when the 2.5 Ma gap there was filled by fossil discoveries.

The 2.5 Ma A. garhi is derived toward megadontia from A. afarensis, but in cranial anatomy it is definitively not A.africanus. Neither is it a representative of the contemporary A. aethiopicus. It is in the right place, at the right time, to be the ancestor of early Homo, however defined. Nothing about its morphology would preclude it from occupying this position. The close spatial and temporal association between A. garhi and behaviors thought to characterize later Homo provide additional circumstantial support. The temporal and possible phylogenetic placements of various hominid taxa relative to the new species from Bouri are reviewed inFig. 5.

Figure 5

(A) A cladogram depicting relationships among widely recognized early hominid taxa, including the new species A. garhi. Note that an additional clade is required when two contemporary forms of early Homo are recognized. A variety of possible cladograms have been generated from the data available in the hominid fossil record, but none of these satisfactorily resolve the polychotomy illustrated here (11). This cladogram adds A.garhi to the unresolved node. (B) The chronological relationships of early hominid taxa. Age is given in Ma. (C to F) Alternative phylogenies depicting possible relationships among early hominid taxa. Note that these alternatives do not exhaust the possibilities and that not all are entirely consistent with the cladogram. It is not presently possible to choose among these alternatives.

Plio-Pleistocene hominid phylogenetics is bedeviled by atomization of functionally correlated character complexes that probably emanate from restricted genomic shifts as well as inadequate fossil samples (particularly for early Homo). Table 1compiles characters available for A. garhi and related taxa bearing on phylogenetic placement. The discovery of the KNM-WT 17000 specimen of A. aethiopicusdemonstrated the pervasiveness of homoplasy in hominid evolution (12). Specimens such as KNM-ER 1590, KNM-ER 1470, KNM-ER 1802, Malawi UR 501, and Omo 75-14 make it obvious that some earlyHomo specimens exhibit megadontia evolved in parallel with robust Australopithecus. Australopithecus garhiis certainly megadont, at least relative to craniofacial size. However, its lack of derived robust characters leaves it as a sister taxon toHomo but absent many derived Homo characters. A strictly cladistic analysis of available data has continually failed to resolve the issue of the position of A. africanus(11). The resulting currently unresolved polychotomy (Fig. 5) stems from the fact that those characters most widely used in early hominid phylogenetic systematics are predominantly related to masticatory adaptation and are known to be both interdependent and susceptible to parallel evolution. Other characters such as cranial base flexion and craniofacial hafting are even more poorly understood. The atomization of such morphological complexes has led to lengthy trait lists, but the valence of the individual “characters” is clearly compromised. Such exercises have been useful in establishing the extensive homoplasy present among early hominids, but such confirmation only accentuates the precarious nature of phylogenetic reconstructions based on an incomplete and highly fragmentary fossil record.

Even a combination of all available temporal, spatial, and (circumstantial) behavioral evidence fails to resolve whether the origin of Homo was from South African A.africanus or east African A. afarensis(or both). We now know that a nonrobust species derived fromA. afarensis persisted in eastern Africa until at least 2.5 Ma. Only additional fossils will confirm whether this form participated in a rapid evolutionary transition or transitions resulting in an early form or forms of Homo. Such rapid transition may be signaled by the recently described A.L. 266-1 palate from Hadar deposits that are claimed to be 2.33 Ma (2). This palate is more derived than that of A. garhi. IfA. garhi is the direct ancestor of earlyHomo, as represented by such younger specimens as KNM-ER 1590 and KNM-ER 1470, additional major craniofacial changes must have occurred after 2.5 Ma, many of them as direct consequences of brain enlargement. Novel behavioral shifts associated with meat and marrow procurement by means of lithic technology may have played instrumental selective roles during this critical and perhaps short period of evolution.

  • * To whom correspondence should be addressed. E-mail: timwhite{at}


View Abstract

Stay Connected to Science

Navigate This Article