Postcranial evidence of late Miocene hominin bipedalism in Chad

Bipedal locomotion is one of the key adaptations that define the hominin clade. Evidence of bipedalism is known from postcranial remains of late Miocene hominins as early as 6 million years ago (Ma) in eastern Africa1–4. Bipedality of Sahelanthropus tchadensis was hitherto inferred about 7 Ma in central Africa (Chad) based on cranial evidence5–7. Here we present postcranial evidence of the locomotor behaviour of S. tchadensis, with new insights into bipedalism at the early stage of hominin evolutionary history. The original material was discovered at locality TM 266 of the Toros-Ménalla fossiliferous area and consists of one left femur and two, right and left, ulnae. The morphology of the femur is most parsimonious with habitual bipedality, and the ulnae preserve evidence of substantial arboreal behaviour. Taken together, these findings suggest that hominins were already bipeds at around 7 Ma but also suggest that arboreal clambering was probably a significant part of their locomotor repertoire. Analyses of a thigh bone and a pair of elbow bones from Sahelanthropus tchadensis discovered in Chad suggest that the earliest hominin exhibited bipedalism with substantial arboreal clambering.

Bipedal locomotion is one of the key adaptations that define the hominin clade. Evidence of bipedalism is known from postcranial remains of late Miocene hominins as early as 6 million years ago (Ma) in eastern Africa [1][2][3][4] . Bipedality of Sahelanthropus tchadensis was hitherto inferred about 7 Ma in central Africa (Chad) based on cranial evidence [5][6][7] . Here we present postcranial evidence of the locomotor behaviour of S. tchadensis, with new insights into bipedalism at the early stage of hominin evolutionary history. The original material was discovered at locality TM 266 of the Toros-Ménalla fossiliferous area and consists of one left femur and two, right and left, ulnae. The morphology of the femur is most parsimonious with habitual bipedality, and the ulnae preserve evidence of substantial arboreal behaviour. Taken together, these findings suggest that hominins were already bipeds at around 7 Ma but also suggest that arboreal clambering was probably a significant part of their locomotor repertoire. Discoveries in Chad by the Mission Paleoanthropologique Franco-Tchadienne (MPFT) have substantially contributed to our understanding of early human evolution in Africa. The localities TM 247, TM 266 and TM 292 in the Toros-Ménalla fossiliferous area in the Lake Chad Basin have yielded, among hundreds of vertebrate remains, a nearly complete cranium (TM 266-01-60-1), three mandibles and several isolated teeth that represent a minimum of five adult individuals assigned to the hominin S. tchadensis 5,8 . The fossils were found in the TM Anthracotheriid Unit, and biochronological estimates and radiochronological age indicate that they are from about 7 Ma (refs. 7,9 ). Environmental indicators at Toros-Ménalla localities suggest a lacustrine fringe, in a desert vicinity, where open areas with dry and humid grasslands coexisted with arboreal cover 9 .
Three other hominin fossil remains were discovered at TM 266 in 2001 by the MPFT: one left femoral shaft (TM 266-01-063, unearthed in July 2001); and two right and left ulnae (respectively, TM 266-01-050, unearthed in July 2001, and TM 266-01-358, unearthed in November 2001; Supplementary Note 1). Although none of these limb bones can be reliably ascribed to any hominin craniodental specimen found at TM 266, the most parsimonious hypothesis is to assign these postcranial remains to the sole hominin species so far identified in this locality: S. tchadensis.

The femur
The hindlimb is represented by a left femoral shaft (TM 266-01-063) about 242 mm long ( Fig. 1 and Extended Data Fig. 1; descriptions also provided in Supplementary Note 2 and Supplementary Table 1) that lacks the distal epiphysis and most of the proximal one. The specimen is curved anteroposteriorly, similar to Australopithecus and to Orrorin tugenensis (BAR 1002′00 and BAR 1003′00 about 6 Ma from Lukeino, Kenya) 1,2 , and slightly more heavily built than the BAR 1003′00 specimen (Extended Data Figs. 2 and 3). The two major axes of the proximal and distal diaphyseal portions indicate that the neck was anteverted in a manner similar to fossil hominins 10 (Extended Data Fig. 4), which correlates with an antetorsion of the long axis of the shaft 11 . Femoral antetorsion is also reported in archaic fossil apes such as Ekembo, but is absent in the putative arboreal biped Danuvius guggenmosi 12,13 . The femur exhibits proximal platymeria immediately distal to the lesser trochanter ( Fig. 2a; accompanying comments are provided in Extended Data Fig. 5 and Supplementary Note 3, and see description in Supplementary Note 2), which is a trait observed in O. tugenensis and later hominins and suggests a correlation with neck elongation, an indicator of hominin bipedalism 3,4,11 . Similar to the Miocene ape D. guggenmosi 12 and to hominins [2][3][4] , the neck is anteroposteriorly compressed (but see ref. 14 for Dryopithecus fontani and Hispanopithecus laietanus; Fig. 1f,h,i).
A small but sharp relief, indicative of lateral and vertical gluteal tuberosity, continues distally into a rugose surface and blends with the lateral component of a broad 'proto-linea aspera' 3 (Fig. 1b,d,e,g; 12.2 mm in its narrowest width). In the posterior view, the lateral lip of the linea forms a well-marked sigmoid line. Similarly, the medial lip of the linea aspera is clearly marked. The spiral line for the insertion of the vastus medialis muscle is linear and merges distally with the medial lip of the linea aspera. Such a configuration is also observed in O. tugenensis and Ardipithecus ramidus 2,4,15 , albeit with a more salient proto-linea aspera in TM 266-01-063, but differs from modern humans in which a pilaster develops coincident with the posteromedial translation of the gluteus maximus muscle 3 ( Fig. 2b; accompanying comments are provided in Supplementary Note 3). Overall, the external morphology of TM 266-01-063 does not differ from O. tugenensis, and both are similar to the condition seen in later hominins.
Functionally, the anteroposteriorly compressed neck and the subtrochanteric platymeria reflect a hip subjected to high mediolateral bending movements. These movements are assumed to be counterbalanced by the gluteal muscles, which is consistent with the presence of distinct gluteal tuberosity in TM 266-01-063 (ref. 4 ). This array of features is compatible with habitual bipedalism in S. tchadensis 3,4,11 , considered here as a behavioural response repeated under comparable situations 16 and then potentially offering a selective advantage.
The cortical thickness distribution pattern of the femoral shaft is characterized by three relative reinforcement areas that correspond to the proximomedial, lateral and posterior portions of the diaphysis ( Fig. 3; accompanying comments are provided in Supplementary Note 3; see Extended Data Fig. 6 for TM 266-01-063 cortical thickness and bone cross-sectional geometries). Compared to extant hominoids (except gibbons), TM 266-01-063 most resembles a typical Homo sapiens (modern human), especially regarding its posterior and lateral aspects 17 (Fig. 3). Comparative data are, to our knowledge, not available for Ardipithecus, and high-resolution micro computed tomography (micro-CT) scan data are probably needed to precisely assess the cortical thickness in the BAR 1002′00 and BAR 1003′00 (O. tugenensis) fossils 18,19 . Available CT scan data show that S. tchadensis and O. tugenensis share an oblique (anteromedial to posterodistal) extension of bone thickening 20 . However, O. tugenensis does not show the posterior thickening or the expanded lateral thickening seen in TM 266-01-063 and H. sapiens.
The pattern and magnitude of bone distribution in organisms stem from a complex interplay between heredity and mechanical stimuli 21 , and early ontogenetic stages may display inherited patterns. Regarding locomotor autonomy, patterns of bone distribution record the individual activity level and the loadings experienced during ontogeny and adulthood. Hence, in the case of S. tchadensis, the hominin-like condition of the femoral bone distribution may retain an evolutionary signal, but more probably provides support for an individual locomotor activity pattern that includes bipedalism.
Cortical area and second moment of area (I x /I y and I max /I min ) are biomechanical parameters that are frequently used to infer habitual locomotor functions in primates, as they partially reflect how long bones resist loads, and have been used to infer how they grow in response to habitual loading 11 . The cortical area of the diaphyseal cross-section is a measure of resistance of the bone to axial compression or tension, whereas the second moment of area measures the resistance to bending and twisting loads.
In all cases, cross-sectional cortical area and second moment of area values for TM 266-01-063 fall within fossil hominin distributions (Supplementary Table 2   Many factors such as stature, body mass, muscle attachment sites, positional behaviour, ontogeny and sexual dimorphism may contribute to femoral bone mass distribution 22 . Considering size, values derived from the shape analysis of the femoral cross-section at 80% (centroid size, Extended Data Fig. 5) indicate that the S. tchadensis femur is in the range of hominin variation, apart from all other extant hominoids but H. sapiens. In addition, preliminary estimation of the body mass of S. tchadensis through univariate regression analyses 23 of anteroposterior and mediolateral subtrochanteric and midshaft femoral dimensions (Supplementary Table 1) produced values that ranged from 43.5 to 49.4 kg (but see ref. 24 ). These results are in the range derived from the same metrics reported for australopiths and early Homo specimens. Body estimates for Sahelanthropus fall between those estimated for O. tugenensis (from 39.9 to 45.7 kg; see supplementary information in ref. 25 ; but see ref. 26 for estimations of 47.7-50.1 kg) and Ardipithecus ramidus (ARA-VP-6/500, 51 kg (ref. 27 )). Body mass estimates for S. tchadensis remain close to the average body mass of Pan troglodytes (common chimpanzee), the earliest hominins and australopiths 28 . Functionally, the geometric properties of the femoral shaft of TM 266-01-063 indicate a greatest resistance to mediolateral bending stress. This condition, seen in A.L. 288-1 and in specimens of early Homo from Africa and Asia, is suggested to indicate a more lateral position of the body during the stance phase of gait in association with an increase in femoral neck length and biacetabular breadth 4,11 .
Functional interpretations remain difficult to formulate given the paucity of comparative data from early hominins and the state of preservation of the TM 266 femur. Nevertheless, in the proximal portion of TM 266-01-063, a structure formed by dense cancellous bone, consisting in an array of oblique trabeculae, is clearly identifiable in both transversal and parasagittal planes (Extended Data Fig. 8). In the parasagittal plane, this oblique structure originates posteriorly at the level of the distal base of the neck and ends anteriorly on the endosteal surface around the level of the distal portion of the lesser trochanter. In the transverse plane, the structure forms a spur that originates from the posteromedial endosteal surface and extends towards the anterolateral endosteal surface. This structure corresponds to the presence of a calcar femorale (CF) in TM 266-01-063. The CF is also documented in O. tugenensis (BAR 1003′00) 29 , in early hominins such as Australopithecus afarensis with MAK-VP-1/1 (ref. 30 ) and in modern humans 31,32 and is interpreted as a mechanical support for load distribution within the proximal femur in relation to habitual bipedal locomotion. In terrestrial bipeds, the CF facilitates the dispersal of compressive loads in the proximal femur by decreasing the stress in the posterior and medial aspects and increasing the stress in the anterior and lateral aspects 29,32 . TM 266 01-063 and BAR 1003′00 present both an oriented trabecular bundle, of columnar aspect, in cross-section set in the parasagittal plane, a morphology that recalls the human condition (Extended Data Fig. 9), and a similar proximodistal extension of the structure. Nevertheless, the CF is seemingly shorter and appears denser, with a tightened trabecular network, in the transverse cross-sections in BAR 1003′00. The observed differences might be due to different states of preservation and/or CT image acquisition settings between TM 266 01-063 and BAR 1003′00. A well-developed CF in TM 266-01-063 might represent a morphological adaptation for terrestrial bipedalism, which is strongly supported by the other lines of evidence derived from the external shape of the femur and structural analyses.

Ulnae
The forearm bones attributed to S. tchadensis consist of two partial left and right ulnae that lack the distal epiphyses (Supplementary Note 2). The similarity in size and shape of these ulnae may indicate that they are from the same individual, but no definitive evidence supports this inference. TM 266-01-050 is a left ulnar diaphysis 239 mm long (Fig. 4e-h  and Supplementary Table 1) and with an eroded proximal epiphysis. The right ulna (TM 266-01-358; Fig. 4a-d) is a proximal half shaft that is 155 mm long and with a partially preserved epiphysis. The shafts are curved in profile. Similar anteroposterior curvature is observed in A. kadabba 33 (ALA-VP 2/101) and later hominins (for example, L 40-19 and OH 36) as well as in apes from Africa 10,34-36 (see comparative views in Extended Data Fig. 10). This curvature contrasts with the straight ulnar shaft of D. guggenmosi, but this lack of curvature is probably due to pathological conditions 12 . A recent ulnar curvature analysis 34 places the ulna from Chad (TM 266-01-050) within the fossil hominin variation, close to apes from Africa in morphological space. It does not differ in this regard from OH 36 and L40-19 (presumably Paranthropus), U.W. 101-499 (Homo naledi) and, to a lesser extent, StW 573 (Australopithecus prometheus). In primates, such ulnar curvature is due to habitual loads exerted by the action of the brachialis muscle to maintain elbow flexion during arboreal climbing, which in turn involves the action of a powerful antagonistic forearm musculature, including the anconeus muscle and wrist and fingers extensors and flexors 36,37 . In this context, ulnar curvature is also suggested to be associated with arboreal behaviours in large-bodied primates, especially climbing and to some extent suspensory activities, for instance, in facilitating prono-supination and in countering habitual loads due to body weight and the action of digital and carpal flexors 36,38 .
The preservation of TM 266-01-050 allows partial assessment of the cross-sectional cortical bone distributions (Fig. 4  . The cortical bone is predominantly distributed anteroposteriorly, which is similar to orangutans and, to a lesser extent, to chimpanzees. This condition contrasts with that seen in gorillas, which tend to grow more bone mediolaterally. Relative mediolateral expansion is suggested to adjust for increased vertical and mediolateral forces that apply to the forelimb in terrestrial quadrupedal primates 39,40 . Compared to chimpanzees and orangutans, which tend to grow more bone anteroposteriorly 25 , the morphology of TM 266-01-50 more likely reflects bending loads associated with an array of arboreal locomotor modes. The geometry of the ulnar from Chad deviates from circularity, with a major axis oriented anteroposteriorly at the 80% level and anterolaterally at midshaft level, where maximum bone deposition occurs. Such a conformation is indicative of habitual loads exerted by the oblique cord and interosseous membrane along with the flexor digitorum profundus muscle anteriorly and along with the wrist and finger extensors posteriorly. Overall, the curvature and cross-sectional geometric properties of the ulna conform to greater anteroposterior resistance and optimized prono-supination abilities, which are indicative of habitual arboreal behaviours, including climbing and/ or 'cautious climbing' 41,42 , rather than terrestrial quadrupedalism 39,41,43 .
The preserved flat distal portion of the olecranon processes indicate that they were not projecting posteriorly as seen in apes from Africa 44 . In this regard, the distal portion of the olecranon most resembles the condition in Miocene apes 44,45 and in hominins 46 . The proximal epiphyses indicate an anteriorly facing trochlear notch, similar to that reported for fossils of hominins and Miocene apes 10,12,44,45,47 . Hence, the ulnae from Toros-Ménalla depart from the typical proximally oriented trochlear notch of the extant great apes 37,46 . In functional terms, a more anteriorly facing notch, which is associated with an olecranon aligned with the long axis of the forearm, favours leverage of the triceps at mid-flexion 37,45,46 . In A. ramidus, such a function has been linked to careful climbing and bridging 45 . Conversely, a more proximally facing notch reflects a habitually extended elbow, for example, during suspension 47 and quadrupedalism, whereas a posteriorly projecting olecranon favours leverage of the triceps in extension and is commonly associated with terrestrial quadrupedal locomotion in anthropoids 44 .
The ulnae from Toros-Ménalla display a keeled trochlear notch with a comparatively acute angle relative to later hominins and to apes from Africa. The distal keeling angle measured from TM 266-01-358 (distal keeling angle of 117°) is in the lower range of chimpanzees and Pongo pygmaeus (orangutan) and close to values reported for Oreopithecus bambolii and the left ulna A.L. 288-1t of A. afarensis 46 . Similarly, the proximal angle is acute (proximal keeling angle of 101°) and in the lower range of reported values for chimpanzees and gorillas and close to OH 36 and O. bambolii 46 . A pronounced trochlear keel probably augments mediolateral stability of the elbow in response to powerful superficial finger and wrist flexors and forearm pronators (flexor digitorum superficialis, flexor carpi radialis and ulnaris, and pronator teres muscles) 35,46 . Such a configuration was reported for D. guggenmosi 12 and is typical of the arboreal large apes that integrate climbing and suspension in their locomotor repertoire 46 and is unlikely to reflect habitual terrestrial quadrupedalism. Moreover, the two TM 266 ulnae trochlear notches present a morphology close to that of humans and chimpanzees, in which the middle portion is mediolaterally narrow relative to the distal half. This waisted configuration is an allometric consequence of size and unrelated to locomotor mode 46 . The distomedial quadrant of TM 266-01-358 is more developed than the concave distolateral one, an intermediate morphology between humans and chimpanzees, close to that of D. guggenmosi 12 , A. afarensis 46 and A. prometheus 10 . A developed medial portion of the trochlear notch is an adaptation for maximum joint compression medially, a configuration that could meet mechanical requirements in various non-terrestrial locomotor behaviours 12,46 .
The TM 266 ulnae lack prominent flexor apparatus enthesis, similar to orangutans, A. ramidus and later hominins. In this respect, it differs from the condition seen in Miocene apes and extant quadrupedal apes and monkeys 45,47 . A medially prominent flexor apparatus enthesis is assumed, through the activity and passive tension of the flexor muscles, to limit palmar dorsiflexion and to stabilize the forelimb during quadrupedal stance phase 48 . Its absence in S. tchadensis ulnae provides evidence to indicate that terrestrial quadrupedalism is not the primary locomotor behaviour of the Toros-Ménalla hominins.
In summary, the ulnae of S. tchadensis exhibit a combination of traits commonly seen in apes engaged in habitual arboreal activities. In particular, the ulnar morphology reflects habitual flexed forearm-arm postures, a stabilized elbow in flexion and extension, and powerful wrist and finger flexors along maintained capacities for prono-supination. Such a functional pattern is indicative of a form of arboreal locomotion compatible with pronograde and orthograde clambering (Supplementary Note 5), probably involving some degree of sure grasp and erratic limb excursion 41,42,49 , and compatible with the careful climbing described in previous studies of A. ramidus 45 .

Overall assessment
Given the combination of hominin-like traits identified in this study (compared to Miocene apes and extant non-human apes), the most parsimonious hypothesis remains that the postcranial morphology of Sahelanthropus is indicative of bipedality and that any other hypothesis would have less explanatory power for the set of features presented by the material from Chad. The multiplicity of attested and presumptive bipedalities currently proposed for several phylogenetically distinct hominoid taxa (for example, Orrorin, Ardipithecus, Australopithecus, Danuvius and Oreopithecus) strongly suggests that searching for a unique defining trait of bipedalism is hazardous ('magic trait' sensu 50 ). Rather, searching for specific functional complexes for inferring past postural and locomotor behaviours should be favoured.
The Toros-Ménalla femur exhibits several hallmarks of selection for bipedalism as a regular behaviour. Results from femoral cross-sectional contours, cross-sectional geometry properties and cortical bone distribution along a particularly transversally twisted shaft show that TM 266-01-063 presents distinctive hominin femoral characteristics (Table 1). In addition, a well-developed CF would facilitate the dissipation of compressive loads caused by bipedalism on terrestrial substrates 29,32 . Although present in some Miocene hominoids, a well-defined proto-linea aspera, the presence of lateral gluteal (3) Article tuberosity and associated subtrochanteric platymeria without hypotrochanteric or inferolateral fossae have been traditionally associated with enhanced hip flexion-extension 3,15 . Although they are part of a primitive functional complex [1][2][3][4]15 , they are in agreement with the overall functional pattern seen in habitual hominin bipeds 51 (Supplementary  Table 3 and references therein). Hence, on the basis of these multiple lines of evidence, we consider that the hindlimb bone discovered in Chad exhibited habitual bipedalism (contra opinion expressed in ref. 52 ), probably on terrestrial substrates. The ulnar material from Chad displays a suite of morphological features that are consistent with substantial arboreal behaviour, similar to in A. ramidus 15 and presumably O. tugenensis 1 and A. kadabba. All ulnar features, including the shaft curvature, the cross-sectional bone distribution and geometry properties, the trochlear morphology and orientation, converge to rule out terrestrial quadrupedalism (Supplementary Table 3). Instead, the functional pattern, inferred from the ulnae, points towards arboreal pronograde and orthograde clambering, probably involving some degree of sure grasp and erratic limb excursion 41,42,49 , without habitual suspensory activities such as forearm swing and/or suspension. Data from the femur also suggest orthograde clambering in the arboreal context, as part of a cautious climbing repertoire 42 , that involves weight-bearing functions for the hindlimbs 49 , as described in previous studies of A. ramidus 45 and possibly O. tugenensis 53 . Until now, the femur of Sahelanthropus supports early evidence of habitual bipedalism in the hominin clade 54,55 , thereby confirming previous interpretations of cranial material based on the relative orientation of the orbital plane and the foramen magnum, as well as the orientation and morphology of the nuchal plane 6,56-58 . In addition, our postcranial results favour a terrestrial component for the habitual bipedalism in S. tchadensis, as shown for O. tugenensis and late A. kadabba (5.2 Ma from Middle Awash, Ethiopia). Altogether, the analyses of the postcranial remains from Chad suggest that adaptation for bipedalism evolved soon after the human-chimp divergence 27 , in conjunction with the retention of osteological adaptations for arboreal positional behaviours.
The Toros-Ménalla postcranial material adds to previous interpretations of the environmental context of the Miocene hominins from the eastern African Miocene known so far. O. tugenensis is associated with open woodland areas with significant tree cover 59 , Ardipithecus kadabba (for which bipedalism is inferred from foot phalanx AME-VP-1/71 dated to 5.2 Ma (ref. 33 )) is associated with a mixture of woodland combined with wet grassland 60 , whereas A. ramidus (4.4 Ma from Aramis, Ethiopia) most probably inhabited a groundwater-fed grassy woodland, probably a palm grove 61,62 .
Taken as a whole, the earliest eastern African hominins shared an arboreal component in their environment. Regarding the fossiliferous area of Toros-Ménalla, paleoenvironmental proxies suggest a heterogeneous landscape that includes closed forest (probably riparian forests), palm grove and mixed grassland settings (from woodlands to savannahs and aquatic grasslands) 9,63 . Given the Toros-Ménalla surroundings and the inferred locomotor repertoire of S. tchadensis, the hominins from Chad were able to exploit both arboreal and terrestrial substrates to forage for food and access water resources. The association between a diverse locomotor repertoire (in trees and on the ground) and wooded habitats in mesic context for at least around 2.5 million years suggests that the ecological niche of these early hominins was not necessarily tied to the expansion of relatively dry, open areas. This niche could be depicted as opportunistic in its reliance on terrestrial and arboreal resources.
On the basis of molecular data, the chimpanzee-human last common ancestor was estimated to occur in Africa between 10 and 6 Ma (refs. 58,[64][65][66]. Fossil representatives of the panin clade are scarce 67,68 , but at least three hominine taxa have been described in this time interval in Africa: Samburupithecus, from Samburu Hills, around 9.5 Ma (ref. 69 ); Nakalipithecus from Nakali, around 9.8 Ma (ref. 70 ); and Chororapithecus from Chorora, about 8 Ma (refs. 71,72 ). These Miocene taxa are parsimoniously assigned to stem hominines 73 , even if Samburupithecus displays a particularly archaic morphology 73 . Chororapithecus displays derived dental affinities with Gorilla 71 . In light of this record and the lack of phylogenetic resolution, the ancestral condition of positional behaviour in apes from Africa and humans will remain elusive until new significant data become available. To date, the identification of the derived traits shared by hominins relies on the analysis of the earliest taxa of the clade. The early hominins Sahelanthropus, Orrorin and Ardipithecus share the same combination of non-honing C-P 3 complex and of features linked to bipedalism. This combination is parsimoniously interpreted as more similar to the condition observed in later hominins than in any other fossil and extant hominoids 74 . These are currently the only data available for formulating scenarios about the latest Miocene/ earliest Pliocene evolution of hominoids from Africa. In the absence of Mio-Pliocene fossils that display exclusive morphological affinities with Pan, cautionary tales about rampant homoplasy and character polarity in this evolutionary sequence 75 are untestable. Instead, our data support the hypothesis that the combination of a non-honing C-P 3 complex and habitual bipedalism is a synapomorphic signature of the hominin clade.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-022-04901-z.

Methods
The postcranial material from Chad is curated and conserved by the Centre National de Recherche pour le Développement (CNRD) in N'Djamena, Chad. Access to the palaeontological material collected by the MPFT is regulated by formal agreement between the Université de N'Djamena, the CNRD and the Université de Poitiers, and is available for study upon approval from Chad authorities. Access to the material for loan and/or study of the material, including original 3D microtomographic data, is available upon request to the CNRD, service de paléontologie, at nekoulnanc@yahoo.fr.
The original fossil specimens are measured to the nearest 0.1 mm using a Mitutoyo sliding digital calliper.

Comparative samples
The ulnae and femur from Chad were compared to extant and extinct hominoid specimens, including extant apes (humans, common and bonobo chimpanzees, gorillas and orangutans), Miocene apes and fossil hominins representing O. tugenensis, australopiths and early Homo. Priority was given to wild caught and non-pathological animals when gathering extant specimens. Comparative samples for fossil hominins and Miocene apes were gathered using published data, high-resolution images and high-resolution casts when available. We also acquired original data directly from CT scans, micro-CT scans and tridimensional meshes as far as possible. In this regard, only CT data for O. tugenensis (BAR 1002′00, BAR 1003′00 and BAR 1215′00) and A. prometheus (StW 573) were shared by the teams in charge of their study. The specimens used in the comparative analysis are listed in Supplementary Table 4.

CT imaging
High-resolution micro-CT images taken from the original femur and ulnae were used to assess the inner morphology of the bones. The material was scanned using an EasyTom XL Duo mCT (using a sealed Hamamatsu microfocus X-ray source (75 W, 150 kV) and an amorphous silicon-based detector (Varian PaxScan 2520DX, 1,536 × 1,920 pixel matrix; 127 mm pixel pitch, 16 bits, CsI conversion screen; RX-Solutions) at Plateforme PLATINA (A. Mazurier, IC2MP, University of Poitiers). For scanning procedures, the beam intensity was set at 90 kV and the tube current at 333 mA. Images of the TM 266-01-358 ulna were acquired with 3,584 projections, resulting in 3,036 slices of 730 × 825 pixels using a cone-beam reconstruction algorithm. The isovoxel size was set to 0.0525 mm. Images of the TM 266-01-050 ulna were acquired with 4,800 projections, resulting in 4,051 slices of 589 × 849 pixels. The isovoxel size was set to 0.0600 mm. Images of the TM 266-01-063 femur were acquired with 5,984 projections, resulting in 4,962 slices of 1,162 × 911 pixels. The isovoxel size was set to 0.0499 mm.

Processing of virtual models
Semi-automatic segmentation of the virtual fossil specimens and extraction of 3D surfaces were performed in Avizo Lite 2021 (Thermo Fisher). Cortical bone thickness distribution for the TM 266 femur was assessed in three dimensions using the Surface Thickness module in Avizo Lite from the outer surface of the femur to the outer surface of the segmented medullar cavity. All measurements based on 3D virtual models of the fossil specimen were done in Avizo Lite on 3D volumes and Fiji image software 76 on 2D slices. For comparative purposes, individual cortical bone thickness were divided by their maximal thickness.

CSGPs
The femur lacks the most part of the epiphyses, which complicates the assessment of its biomechanical length. However, CSGP values in Homo and Pan do not show significant differences between 45 and 55% of the femoral (biomechanical or maximum) length, which means that various midshaft estimates in this range provide comparable CSGPs 77,78 . In this regard, the nutrient foramen located at the level of the maximum femoral anteroposterior curvature was set to represent the 50% biomechanical level of the TM 266-01-063 femur. In addition, the 80% cross-sectional level was set at 10 mm below the distal border of the lesser trochanter, as previously recommended 79,80 . Midshaft and subtrochanteric cross-sectional levels were used to infer the total biomechanical length (about 284 mm) of TM 266-01-063, and the positions of the cross-sections were set at 65% and 35%, respectively. CSGP estimates were computed at midshaft; to obtain an assessment of the variation pattern of cortical bone distribution, positions were set at 80%, 65% and 35% of the biomechanical length. Surface alterations and fracture prevented assessment of the location of standard cross-sectional levels on TM 266-01-358. Consequently, CSGP variables were computed solely based on TM 266-01-050. Cross-section at 50% of the biomechanical length was located at the level of the nutrient foramen, whereas the one set at 80% was estimated using similar sized chimpanzee ulna (Pan paniscus) as an analog.
The percentage of cortical area and second moment of area were computed using Fiji image software 76 and EPMacroJ plug-ins 81 . Comparative data for CSGPs in extant and extinct hominoids are provided in Supplementary Table 2 and citations therein for the femur, and are from refs. 25,82 for the ulnae.
The total length of the femur (Supplementary Table 1) was estimated using the percentage difference between femoral biomechanical and total length in extant hominoids (Homo, Pan, Gorilla and Pongo; n = 64). Mean, minimum and maximum reported values are given using the pooled ape sample without a priori hypothesis for morphological affinities between TM 266-01-063 and any extant hominoid model.

Geometric morphometrics
Femoral anteroposterior curvature (femur anterior bowing) and the shape of the femoral external contours at 80% and 50% were analysed using 2D geometric morphometrics. Curvature was assessed from the anterior border of the femur, as the development of the pilaster in H. sapiens may affect its estimation. Two landmarks were digitized at the intersection between the anterior border of each aligned femur and the 80% and 35% cross-sectional levels. Between these two levels, 23 equally spaced semi-landmarks were digitized along the anterior contour. Landmark and semi-landmark coordinates were acquired using TPSDig2 (v.2.31) 83 . Landmark digitization was performed from the medial view of the oriented 3D virtual model of the femur for all extant and extinct specimens, except for one occurrence of O. tugenensis (BAR 1002′00) and A. prometheus (StW 573m).
Regarding the BAR 1002′00 specimen, the discrepancies between originally described anterior curvature and figured specimens 1,2,84 and the curvature observed from cast and original CT scan data (data used in refs. 20,52 ) led us to consider the two alternative reconstructions of the BAR 1002′00 femur in our analysis. The medial view of the reconstructed StW 573m femur was obtained from ref. 10 .
Shapes of the femoral external contours at 80% and 50% were assessed using 2 landmarks and 46 semi-landmarks. Landmark digitization was performed from the cross-section images issued from 3D virtual model of the femur for most extant and extinct specimens. Specimens, for which 3D data are not available or were not provided, were included based on their published cross-sectional images (Supplementary Table 3). Cross-sectional images were oriented so that the femoral anterior portion faces upward and the femoral medial portion toward the right. Two landmarks were digitized at the anterior and posterior ends of the anteroposterior midline of the cross-sections. Between these two landmarks, 46 equally spaced semi-landmarks were digitized on the femoral outer outline, 23 for each medial and lateral side of the femoral contour. To assess replicability, three blinded landmarking sessions were performed independently for TM 266-01-063 at 50% level by GD (n = 1) and FG (n = 2). Landmark and semi-landmark coordinates were acquired using TPSDig2 (v.2.31) 83 . In the same manner, two landmarking sessions were performed independently for O. tugenensis (BAR 1002′00) at 50% level by G.D. and F.G., which resulted in two close occurrences in the principal component 1-2 shape space.
Landmark and semi-landmark coordinates for femoral curvatures and femoral cross-sectional shapes were analysed using generalized Procrustes superimposition. The analyses were performed using the package Geomorph (v.4.0.0) 85,86 for R (v.4.0.3) 87 . Principal component analyses were performed on Procrustes residuals using Statistica software (Statsoft) with the specimens from Chad as a supplementary individuals. Principal component analysis was preferred over between-group principal component analysis 52 , as the latter may induce misinterpretations and introduce biases owing to small sample sizes 88,89 .

Femur diaphyseal torsion (antetorsion)
The diaphyseal antetorsion of the TM 266-01-063 femur was assessed between a proximal cross-section taken at the base of the lesser trochanter and a distal cross-section at 25% of the total biomechanical length. Each cross-section was oriented so that the femoral anterior portion faced upwards and the femoral medial portion towards the right, the mediolateral axis being horizontal. The longest axis of each femoral cross-section was assessed by the mean of Feret's diameter, as the longest distance between any two points along the cross-sectional contour, using Fiji image software 76 .

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Data availability
The postcranial material from Chad is curated and conserved by the CNRD in Chad. Access to the palaeontological material collected by the MPFT is regulated by formal agreement between the Université de N'Djamena, the CNRD and the Université de Poitiers and is available for study upon approval from Chad authorities. Access to the material for loan and/or study of the material, including original 3D microtomographic data, is available upon request to the CNRD, service de paléontologie, at nekoulnanc@yahoo.fr. Data supporting the findings of this study are available within the paper and its supplementary information files. Corresponding author(s): 2020-08-16114B FRANCK GUY Last updated by author(s): Apr 22, 2022 Reporting Summary Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.

Data
Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A description of any restrictions on data availability -For clinical datasets or third party data, please ensure that the statement adheres to our policy The datasets generated and/or analysed in this article are available from the corresponding author upon reasonable request.
2 nature portfolio | reporting summary

March 2021
Field-specific reporting Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Ecological, evolutionary & environmental sciences study design
All studies must disclose on these points even when the disclosure is negative.

Study description
The article presents a morphological and comparative study of three hominin fossil remains from Chad, including one femoral shaft and two ulnae. The anatomical description of the specimen in a comparative framework is completed by 1) three geometric morphometrics analyses of the femoral shaft curvature and diaphyseal cross-sectional contours; 2) an analysis of the femoral cortical bone thickness, 2D and 3D; 3) an analysis of the cross-sectional geometric properties for the femur and the ulna; 4) an analysis of the femoral trabecular structure; 5) an assessment of the morphological features of the fossil bones in a functional context.

Research sample
The comparative sample comprises specimens of extant (including humans), and extinct hominoids providing a phylogenetically close and functionally diverse assemblage for the analysis of the Chadian material. Selected specimens are all adults, males and females, and only wild caught specimens were selected in osteological collections. Datasets from the literature were referenced in the text and supplementary files.

Sampling strategy
No sample-size calculation was performed. Sample size for extinct hominoids is directly dependent on the available fossil remains (nature and preservation). Relevant post-cranial specimens, when available, were included in our sample. Adult and wild caught specimens were randomly collected in order to approach the fossil sample size. On average, ten specimens represent each taxa, which is relevant for an extant comparative sample considering the nature of the study.

Data collection
Data were collected using microCT scan of the extant and extinct specimen bones, by F.G. and G.D.
Timing and spatial scale Such information on our study design is not applicable, as some data were already available (microCT scans for some fossils and extant apes) while other have been collected during the 2017-2019 period (microCT scans and morphological observations), as later as 2020 for some specimens (eg Orrorin). As our study concerns osteological material from various institutions (not alive specimens), the temporality and the spatiality does not seem relevant (all apes are either from Africa or Asia from known geographic localities). Once the specimens were collected, all the dataset were collected in one time session per dataset over few weeks respectively.

Data exclusions
No data were excluded from the analyses.

Reproducibility
For landmarking and measurement data, authors (GD and FG) performed blinded sessions independently. All attempts to repeat the experiment were successful. For direct morphological observations, no need to verify the reproducibility as morphological evidences emerged from a consensus between the authors.