RESEARCH ARTICLE modern humans (on average this torsion ranges from neutral to 1-5 Implications for hominin pedal evolution degrees (Supplementary Fig.2)).In dorsal view (Fig.2c),the shaft Comparisons with the earlier Ar.ramidus and contemporaneous curves towards the medial side of the foot,a feature that in combination Au.afarensis provide a morphological and chronological context within with the aforementioned axial torsion acts to further direct the which to view BRT-VP-2/73.Several relevant pedal elements are also articular head towards the hallux.This complex is characteristic of represented in the South African samples from Sterkfontein and extant African apes and Ar.ramidus and is indicative of a grasping Swartkrans(see Supplementary Information for further discussion). great toe. The earlier Ar.ramidus pedal remains indicate a mosaic foot capable In contrast to the hallucal metatarsal,the superior surface morphology of terrestrial bipedal toeing-offon the lateral four metatarsophalangeal of the articular head of the second metatarsal does conform to the'typical joints (oblique metatarsal axis)while still maintaining a functionally morphological pattern shared by Ardipithecus and Australopithecus abductable,grasping hallux.By contrast,the foot of Au.afarensis In distal view,the head is roughly triangular in shape and its rounded possessed a longitudinal pedal archa permanently adducted great dorsal apex is domed above the epiphyseal junction.This dorsal toedorsal doming of its hallucal head,anteriorly canted bases doming creates the distinctive transverse gutter between the subchondral on its proximal phalanges(also shared by Ar.ramidus),and clearly margin and the diaphysis,indicating the passive hyperdorsiflexion at used a human-like transverse metatarsal axis during the latter stages of the metatarsophalangeal joint that occurs during bipedal heel-off toe-off. through toe-off6 Although BRT-VP-2/73 is contemporaneous with Au.afarensis at The articulating second proximal(BRT-VP-2/73e)and intermediate around 3.4 Myr ago(see Fig.5),it differs significantly from the known (BRT-VP-2/73h)phalanges also exhibit the derived anatomical feet of Australopithecus.Its hallux is short and the hallucal metatarsal features shared with Ar.ramidus and Au.afarensis.The base of head lacks dorsal doming.The bases of the second and fourth the proximal phalanx exhibits the dorsiflexive anterior cant (100) metatarsals of BRT-VP-2/73 do not have the expanded dorsoplantar conforming to the dorsiflexion dome of the associated metatarsal head dimensions seen in Ardipithecus and Australopithecus,features Like Ar.ramidus and Au.afarensis,the shaft exhibits strong curvature, that along with the associated rugose ligamentous attachments would although not as much as in chimpanzees(Fig.2e and Supplementary resist midtarsal and tarsometatarsal dorsiflexion and midfoot Fig.4b;see also Supplementary Information for angle measurement breaking9-23.However,its lateral metatarsophalangeal joints (MTs methods and discussions).The inclination of the proximal articular 2,3 and 4)do conform morphologically to the Ardipithecus and surface in combination with the bone's longitudinal curvature results in Australopithecus pattern,in having dorsally domed heads and an the characteristic sulcus on the dorsal surface where the articular sur- anterior cant to the phalangeal bases. facejoins the shaft.The second intermediate phalanx(BRT-VP-2/73h) BRT-VP-2/73 also resembles Ar.ramidus in combining an abducent is relatively long compared to the associated proximal phalanx(Sup- hallux and the medially directed torsion of the second metatarsal.It is plementary Table 3). also similar to Ar.ramidus and Au.afarensis in the metatarsophalangeal The third ray (BRT-VP-2/73f)is represented only by an isolated joints of the other rays,indicating that these adaptations in the lateral metatarsal head (see Fig.2a).It too conforms to the pattern seen foot are among the earliest anatomical modifications to hominin in Ardipithecus"and Australopithecus(Supplementary Fig.5)in terrestrial bipedality.The height of the hallucal metatarsal base exhibiting dorsal doming.The dorsoplantar height of the third suggests that a well-developed transverse pedal arch preceded the metatarsal head exceeds that of the second metatarsal,a relationship development of a permanent longitudinal arch.However,the lack of more common in Pan,Gorilla and Australopithecus than in anatom- dorsoplantar expansion of the metatarsal bases(MTs2 and 4)suggests ically modern humans wherein the second metatarsal head is usually that this midtarsal stabilizing feature seen in both Ar.ramidus"and taller (Supplementary Table 2). Au.afarensis was absent in this specimen. The fourth ray is represented by a complete metatarsal(BRT-VP The most surprising feature observed in the BRT-VP-2/73 forefoot 2/73g)and its associated proximal phalanx(BRT-VP 2/73d;Fig.2e,f). is the length of the fourth metatarsal relative to the first and second A ratio of the estimated dorsoplantar height of the metatarsal base and metatarsals.The currently available Ardipithecus and Australopithecus the bone's length indicates that the fourth metatarsal does not have (eastern and South African)fossil record is not adequate to assess the expanded,stabilizing base morphology seen in Au.afarensisand accurately the significance of this particular feature.However,in light Homo but,rather,is similar to Pan and some Old World monkeys of its occurrence in some Miocene apes(for example,KNM-RU 2036) (Supplementary Fig.6). it may represent the primitive state in early hominins.Nonetheless,it is The most unexpected feature seen in the fourth metatarsal is its clear that the BRT-VP-2/73 foot skeleton represents a hominin that, relative length when compared to the associated first and second unlike the contemporaneous Au.afarensis,retained a grasping capacity metatarsals.The fourth metatarsal is absolutely longer than is the that would allow it to exploit arboreal settings more effectively.Yet, second metatarsal,a condition not previously encountered in extant judging from its lateral metatarsophalangeal complex,when on the apes or hominins.The fourth metatarsal is also much longer than is ground it was at least facultatively bipedal,although it may have the hallucal metatarsal and in this ratio,the fossil specimen again fails practiced bipedality in a novel fashion probably similar to Ar.ramidus. to align with extant apes or hominins and is most similar to Old Unlike Au.afarensis,it did not have a longitudinal pedal arch,nor was World monkeys (Fig.3d,f).At present,no associated fossil elements it capable of efficiently using the transverse metatarsal axis. allow a similar comparison in Ardipithecus or Australopithecus and,as Although the taxonomic affinity of BRT-VP-2/73 is currently a consequence,no judgment can be reliably made regarding the polarity indeterminate,there is adequate morphological evidence that it does of this character.A relatively longer fourth metatarsal is the usual not belong to the contemporaneous species Au.afarensis.Regardless condition in Old World monkeys and it also occurs in some Miocene of its taxonomic affinity,however,this specimen is the first strong apes(KNM-RU 2036;ref.17),indicating that it probably represents the evidence indicating multiple hominin lineages,adaptively separated primitive condition. (at least in the foot skeleton),in the 3-4-Myr-ago time interval.A The proximal phalanx of the fourth ray(BRT-VP-2/73d;Fig.2e)is final,but important,note for the metatarsal ratios used in the PCA well preserved and similar to those observed in Ardipithecus"and performed in this study,anatomically modern humans and gorillas Australopithecus.It has the shallow transverse sulcus where the overlap substantially and BRT-VP-2/73 falls in the gorilla cluster.It proximal articular surface cants anteriorly into the curvature of the is unclear at this point what the functional implications of this overlap shaft.It presents a higher degree of dorsal canting than does the might mean;it requires further investigation as it has important con- phalanx of the second ray (104,see Supplementary Fig.4a and sequences for the interpretation of locomotor behaviour in early Supplementary Information for discussions). hominins. 568 NATUREI VOL 483 29 MARCH 2012 2012 Macmillan Publishers Limited.All rights reservedmodern humans (on average this torsion ranges from neutral to 1–5 degrees (Supplementary Fig. 2)). In dorsal view (Fig. 2c), the shaft curves towards the medial side of the foot, a feature that in combination with the aforementioned axial torsion acts to further direct the articular head towards the hallux. This complex is characteristic of extant African apes and Ar. ramidus and is indicative of a grasping great toe. In contrast to the hallucal metatarsal, the superior surface morphology of the articular head of the secondmetatarsal does conform to the ‘typical’ morphological pattern shared byArdipithecus and Australopithecus6,12,16. In distal view, the head is roughly triangular in shape and its rounded dorsal apex is domed above the epiphyseal junction. This dorsal doming creates the distinctive transverse gutter between the subchondral margin and the diaphysis, indicating the passive hyperdorsiflexion at the metatarsophalangeal joint that occurs during bipedal heel-off through toe-off11,16. The articulating second proximal (BRT-VP-2/73e) and intermediate (BRT-VP-2/73h) phalanges also exhibit the derived anatomical features shared with Ar. ramidus11 and Au. afarensis6,12. The base of the proximal phalanx exhibits the dorsiflexive anterior cant (100u) conforming to the dorsiflexion dome of the associated metatarsal head. Like Ar. ramidus and Au. afarensis, the shaft exhibits strong curvature, although not as much as in chimpanzees (Fig. 2e and Supplementary Fig. 4b; see also Supplementary Information for angle measurement methods and discussions). The inclination of the proximal articular surface in combination with the bone’s longitudinal curvature results in the characteristic sulcus on the dorsal surface where the articular sur￾face joins the shaft. The second intermediate phalanx (BRT-VP-2/73h) is relatively long compared to the associated proximal phalanx (Sup￾plementary Table 3). The third ray (BRT-VP-2/73f) is represented only by an isolated metatarsal head (see Fig. 2a). It too conforms to the pattern seen in Ardipithecus11 and Australopithecus6 (Supplementary Fig. 5) in exhibiting dorsal doming. The dorsoplantar height of the third metatarsal head exceeds that of the second metatarsal, a relationship more common in Pan, Gorilla and Australopithecus than in anatom￾ically modern humans wherein the second metatarsal head is usually taller (Supplementary Table 2). The fourth ray is represented by a complete metatarsal (BRT-VP 2/73g) and its associated proximal phalanx (BRT-VP 2/73d; Fig. 2e, f). A ratio of the estimated dorsoplantar height of the metatarsal base and the bone’s length indicates that the fourth metatarsal does not have the expanded, stabilizing base morphology seen in Au. afarensis12 and Homo but, rather, is similar to Pan and some Old World monkeys (Supplementary Fig. 6). The most unexpected feature seen in the fourth metatarsal is its relative length when compared to the associated first and second metatarsals. The fourth metatarsal is absolutely longer than is the second metatarsal, a condition not previously encountered in extant apes or hominins. The fourth metatarsal is also much longer than is the hallucal metatarsal and in this ratio, the fossil specimen again fails to align with extant apes or hominins and is most similar to Old World monkeys (Fig. 3d, f). At present, no associated fossil elements allow a similar comparison inArdipithecus orAustralopithecus and, as a consequence, no judgment can be reliably made regarding the polarity of this character. A relatively longer fourth metatarsal is the usual condition in Old World monkeys and it also occurs in some Miocene apes (KNM-RU 2036; ref. 17), indicating that it probably represents the primitive condition. The proximal phalanx of the fourth ray (BRT-VP-2/73d; Fig. 2e) is well preserved and similar to those observed in Ardipithecus11 and Australopithecus6,16. It has the shallow transverse sulcus where the proximal articular surface cants anteriorly into the curvature of the shaft. It presents a higher degree of dorsal canting than does the phalanx of the second ray (104u, see Supplementary Fig. 4a and Supplementary Information for discussions). Implications for hominin pedal evolution Comparisons with the earlier Ar. ramidus and contemporaneous Au. afarensis provide a morphological and chronological context within which to view BRT-VP-2/73. Several relevant pedal elements are also represented in the South African samples from Sterkfontein and Swartkrans10 (see Supplementary Information for further discussion). The earlierAr. ramidus pedal remains indicate a mosaicfoot capable of terrestrial bipedal toeing-off on the lateral four metatarsophalangeal joints (oblique metatarsal axis11) while still maintaining a functionally abductable, grasping hallux. By contrast, the foot of Au. afarensis possessed a longitudinal pedal arch6,12,18, a permanently adducted great toe12,14,16, dorsal doming of its hallucal head6,14, anteriorly canted bases on its proximal phalanges6,16 (also shared byAr. ramidus11), and clearly used a human-like transverse metatarsal axis during the latter stages of toe-off. Although BRT-VP-2/73 is contemporaneous with Au. afarensis at around 3.4 Myr ago (see Fig. 5), it differs significantly from the known feet of Australopithecus. Its hallux is short and the hallucal metatarsal head lacks dorsal doming. The bases of the second and fourth metatarsals of BRT-VP-2/73 do not have the expanded dorsoplantar dimensions seen in Ardipithecus11 and Australopithecus12,14, features that along with the associated rugose ligamentous attachments would resist midtarsal and tarsometatarsal dorsiflexion and midfoot breaking19–23. However, its lateral metatarsophalangeal joints (MTs 2, 3 and 4) do conform morphologically to the Ardipithecus and Australopithecus pattern, in having dorsally domed heads and an anterior cant to the phalangeal bases. BRT-VP-2/73 also resembles Ar. ramidus in combining an abducent hallux and the medially directed torsion of the second metatarsal. It is also similar toAr. ramidus andAu. afarensisin the metatarsophalangeal joints of the other rays, indicating that these adaptations in the lateral foot are among the earliest anatomical modifications to hominin terrestrial bipedality. The height of the hallucal metatarsal base suggests that a well-developed transverse pedal arch preceded the development of a permanent longitudinal arch. However, the lack of dorsoplantar expansion of the metatarsal bases (MTs 2 and 4) suggests that this midtarsal stabilizing feature seen in both Ar. ramidus11 and Au. afarensis12 was absent in this specimen. The most surprising feature observed in the BRT-VP-2/73 forefoot is the length of the fourth metatarsal relative to the first and second metatarsals. The currently available Ardipithecus and Australopithecus (eastern and South African) fossil record is not adequate to assess accurately the significance of this particular feature. However, in light of its occurrence in some Miocene apes (for example, KNM-RU 2036) it may represent the primitive state in early hominins. Nonetheless, it is clear that the BRT-VP-2/73 foot skeleton represents a hominin that, unlike the contemporaneous Au. afarensis, retained a grasping capacity that would allow it to exploit arboreal settings more effectively. Yet, judging from its lateral metatarsophalangeal complex, when on the ground it was at least facultatively bipedal, although it may have practiced bipedality in a novel fashion probably similar to Ar. ramidus. Unlike Au. afarensis, it did not have a longitudinal pedal arch, nor was it capable of efficiently using the transverse metatarsal axis. Although the taxonomic affinity of BRT-VP-2/73 is currently indeterminate, there is adequate morphological evidence that it does not belong to the contemporaneous species Au. afarensis. Regardless of its taxonomic affinity, however, this specimen is the first strong evidence indicating multiple hominin lineages, adaptively separated (at least in the foot skeleton), in the 3–4-Myr-ago time interval. A final, but important, note for the metatarsal ratios used in the PCA performed in this study, anatomically modern humans and gorillas overlap substantially and BRT-VP-2/73 falls in the gorilla cluster. It is unclear at this point what the functional implications of this overlap might mean; it requires further investigation as it has important con￾sequences for the interpretation of locomotor behaviour in early hominins. RESEARCH ARTICLE 568 | NATURE | VOL 483 | 29 MARCH 2012 ©2012 Macmillan Publishers Limited. All rights reserved