Proportional limb strengths signal an adaptive shift in arboreality in early human evolution Authors: Kristian J. Carlson, Christopher B. Ruff, M. Loring Burgess, Adam D. Sylvester, Tea Jashashvili, Jason L. Heaton, Travis R. Pickering, Lauren Sarringhaus, Timothy M. Ryan, Amélie Beaudet, Robin H. Crompton, A. J. Heile, Dominic Stratford, Kathleen Kuman, David Lordkipanidze, Ronald J. Clarke Published: 2026 ABSTRACT How much arboreality characterized the Australopithecus locomotor repertoire and how this differed from early Homo are vigorously debated. Most previous studies rely on traits of uncertain functional significance or are constrained by sampling concerns. Bone shaft strength is plastic and reflects in vivo mechanical loading. Because interlimb bone strength proportions among living apes correlate with arboreality, they are ideal contributors to this debate. We present the most comprehensive analysis yet of relative limb shaft strengths in Australopithecus and early Homo by adding two individuals [StW 573, ∼3.67 million years ago (Ma); and D3901/D4167/D4507, ∼1.77 to 1.81 Ma]. We show that Australopithecus individuals of varying body sizes spanning at least 0.5 Ma exhibit African ape–like interlimb strength proportions, indicating frequent arboreal behavior, and that the Australopithecus lower limb exhibits human-like intralimb strength proportions. In contrast, early Homo individuals are human-like in both strength proportions, reflecting fundamentally different selective pressures favoring terrestrial bipedalism and an unambiguous departure from arboreality by ∼1.8 Ma. INTRODUCTION Since the time of Darwin and Huxley, terrestrial bipedalism has been posed as a central tenet of the human ecological niche (1). Considerable subsequent work has been devoted to determining the changing role of this locomotor behavior in the human evolutionary record, beginning with the last common ancestor at the divergence of human and chimpanzee lineages ∼9.3 to 6.5 million years ago (Ma) (2–4). This cumulative body of evidence portrays a gradation of the selective importance of terrestrial bipedalism: from facultative to habitual to obligate behavior (5). Recent investigations of the earliest fossils attributed to the human ancestral pool of species (e.g., Sahelanthropus tchadensis) (6) have invigorated debate over whether terrestrial bipedalism is a fundamental adaptation early in human evolutionary history (7–9). Arboreality was undoubtedly a critical influence on the locomotor repertoire of S. tchadensis and other hominin taxa predating 4 Ma, such as Orrorin tugenensis (10) and Ardipithecus ramidus (11). The adaptive role of arboreality in later hominins, such as Australopithecus, remains very much a “deadlocked debate” [see page 50 in (4)] despite Australopithecus being the most well-documented early hominin genus in the human fossil record (12). Operating as a fundamental constraint in this debate is the issue of whether relevant morphological features are functionally significant primitive retentions undergoing stabilizing selection or selectively neutral traits (13). For example, morphological traits exhibited in the postcranial skeleton of Australopithecus that are considered adaptive for arborealism (e.g., relatively long forelimbs with larger articular surface areas) could alternatively be either holdover traits potentially signaling continuity in locomotor behavior with their earlier ancestors, or traits that are present but that are no longer functionally relevant. One potential solution to this dilemma is to use morphological traits that are known to exhibit developmental plasticity by responding to the mechanical demands of locomotor and postural behavior during life (5, 13–15). Internal structural characteristics of long bones are prime examples, because they have been demonstrated to be developmentally labile and responsive to applied mechanical loadings in experimental and comparative studies (16–21). Of particular relevance when characterizing behavior in early hominins, changes in locomotor repertoire and degrees of arboreality versus terrestriality during development have been shown to correlate with relative strength proportions of forelimb versus hindlimb diaphyses in both humans (22, 23) and our closest phylogenetic relatives: chimpanzees (24) and gorillas (19). Thus, these morphological traits reflect activity characterizing the realized niche of an organism performed over its lifetime (15) and are not saddled with the interpretive entanglements of phylogenetic inertia and neutral selection. Trends in these traits also would be robust to potential uncertainty resulting from lag between behavioral shifts and morphological change, i.e., behavioral drive (25). Relying on relative bone strengths as the basis for reconstructing early hominin ecological niches is constrained, however, by the rarity of hominin skeletons with sufficiently well-preserved and associated upper and lower limb remains. Within Australopithecus, only a single 3.2-Ma individual has been investigated to date (26). The partial skeleton (A.L. 288-1) known as “Lucy” and attributed to Australopithecus afarensis has femoral/humeral diaphyseal strength proportions intermediate between those of modern humans and chimpanzees, but closer to chimpanzees, indicating increased forelimb weight-bearing loads relative to modern humans. This implies a substantially higher degree of arboreal locomotor behavior in this small-bodied hominin compared with modern humans. Three additional partial skeletons predating 1.5 Ma have also been investigated in this manner (27, 28): OH 62, attributed to Homo habilis (29) or to Australopithecus cf. africanus (30); and KNM-ER 1808 and KNM-WT 15000, both attributed to Homo erectus. While both H. erectus individuals resemble modern humans in their femoral/humeral diaphyseal strength proportions, the OH 62 individual exhibits even more nonhuman ape-like relative limb strength than the A.L. 288-1 individual. Of these four partial skeletons, only the nonadult KNM-WT 15000 individual preserves material that can be used to assess relative bone strength within the hindlimb (i.e., femoral-tibial proportions). Thus, relative diaphyseal loading between Australopithecus hindlimb segments and how it relates to that of early Homo are unknown. The StW 573 hominin skeleton from Sterkfontein Cave, South Africa, known as “Little Foot” and attributed to Australopithecus prometheus (15, 31, 32) offers critical insights into the realized niche of an Australopithecus species. At ∼3.67 Ma, the Little Foot individual is among the earliest representatives of the genus Australopithecus (33, 34) and substantially expands the time frame available for investigation of the australopith grade to nearly 0.5 Ma earlier than Lucy. The previously analyzed A.L. 288-1 individual is small-bodied with an estimated body mass of 28.9 kg, placing it near the low extreme of Australopithecus individuals with estimated body masses [i.e., 35th out of the 37 individuals; (35)]. By comparison, the larger-bodied Little Foot individual, with an estimated body mass of 40.2 kg (see below), considerably extends the Australopithecus body size range derived from well-preserved partial skeletons. This offers crucial opportunities for insights, given the inverse relationship between body size and arboreality within many primate taxonomic groups [(21) and references therein], and uncertainties in statistical (36) and behavioral (37) extrapolation from the small-bodied Lucy to larger Australopithecus individuals. Moreover, the StW 573 individual also offers an opportunity to evaluate intra(hind)limb relative strength in Australopithecus for comparison to trends in interlimb relative strength. The Dmanisi hominins from Georgia are dated to ∼1.85 to 1.77 Ma and attributed to early Homo (38, 39). These hominins are represented by associated postcranial elements, which strategically expand the handful of early Homo individuals available for assessing relative limb strengths. The larger Dmanisi individual (D3901/D4167/D4507) stands out as unique within this select group in that it is not encumbered with complicating factors, such as subadult age [e.g., KNM-WT 15000; (40)] or pathological state [e.g., KNM-ER 1808; (27)]. Even the OH 62 partial skeleton is associated with some uncertainty in its characterization of relative limb strengths because the incomplete states of its femoral and humeral diaphyses complicate precise placement of targeted diaphyseal sections (28). Thus, the Dmanisi individual offers a secure opportunity to record relative interlimb and intralimb strength patterns for early Homo and evaluate them in comparisons with Australopithecus individuals. Our primary goal is to determine the degree to which members of Australopithecus, including larger-bodied individuals, are ape-like or intermediate between humans and extant nonhuman apes in their relative limb bone strengths (Fig. 1) and, by extension, their locomotor/positional behavior. We also assess these results against comparisons of articular proportions in the same limb bones to provide additional insights into locomotor performance. Our results clarify whether inferred changes in selective pressures on the hominin forelimb and hindlimb and their relative timing may have accompanied the transition from a forelimb primarily adapted to weight-bearing during locomotor/positional behavior to a forelimb potentially with enhanced manual manipulative capabilities. These changes may be associated with increased selection for specialized functions, including skilled lithic tool manufacture (41–43) and enhanced (high speed) throwing associated with hunting ability (44). Our results also clarify whether inferred changes in selective pressures on the hominin hindlimb may have concurrently affected its intralimb relative strengths. By using morphological traits in both limbs that are relatively unencumbered by the constraints of phylogenetic inertia across a broader size range of Australopithecus and an expanded sample of early Homo, we test whether an inferred fundamental adaptive shift in habitat use coincided with the transition from the australopith grade to the Homo clade. Femoral, humeral, and tibial diaphyseal sections and articular dimensions used in analyses. See fig. S3 (StW 573) and fig. S4 (Dmanisi) for all humeral sections, fig. S5 (StW 573) and fig. S6 (Dmanisi) for all femoral sections, and fig. S7 (StW 573) and fig. S8 (Dmanisi) for all tibial sections. Renderings shown with a 100-mm scale. Cross sections shown with a 10-mm scale. All sections are depicted as analyzed, after editing (see Materials and Methods), with the anterior surfaces toward the top and medial surfaces toward the left. RESULTS Extant hominoids Comparisons of interlimb diaphyseal strength and articular breadth proportions between extant humans, chimpanzees, Virunga mountain and other (Western lowland and grauer) gorillas, and orangutans are shown using plots of lower (hind)limb versus upper (fore) limb dimensions, with reduced major axis (RMA) regressions and 95% prediction intervals (PIs) for each group (Fig. 2). Summary regression statistics are reported (Table 1), along with average percentage of time spent in arboreal environments (for nonhuman taxa) based on behavioral observations (see also the Supplementary Text). Equivalence between RMA line elevations is used to assess proportional differences between groups (all slopes are nonsignificantly different). Log/log plots of structural variables and articular dimensions. Bivariate comparisons include (A) femoral 50% Zp to humeral 35% Zp, (B) tibial 50% Zp to humeral 35% Zp, (C) femoral head SI breadth to distal humeral ML breadth, and (D) tibial plateau ML breadth to distal humeral ML breadth for fossils and comparative samples, with reduced major axis (RMA) regression lines and 95% prediction intervals (PIs). Zp, polar section modulus; S-I, superoinferior; M-L, mediolateral; bd., breadth. Table 1.Extant taxa comparisons of interlimb proportions.The table reports RMA regressions of the first variable on the second variable. Arboreal percentage is calculated as the percentage of observation time spent in arboreal environments (see table S4 and the Supplementary Text for derivations). Elevation equivalency refers to equivalent line elevations in pairwise comparisons (P > 0.05). When the same letter appears in two or more rows for a variable, it indicates equivalent line elevations in pairwise comparisons. Slope SE, standard error of the slope; SEE, standard error of estimate; F50Zp, femoral midshaft torsional/average bending strength (cubic millimeters); T50Zp, tibial midshaft torsional/average bending strength (cubic millimeters); H35Zp, humeral mid-distal torsional/average bending strength (cubic millimeters); FHSI, femoral head superoinferior breadth (millimeters); TPML, tibial plateau mediolateral breadth (millimeters); HDML, humeral distal mediolateral articular breadth (millimeters).RegressionTaxon n % Arboreal r InterceptSlopeSlope SESEEElev. equiv.F50Zp/H35Zp Homo 12790%0.8172.5010.777*0.0120.127a Pan 9651%0.8981.3720.851*0.0370.072bGorilla (mtn.)267%0.8741.0720.9350.0830.060cGorilla (other)3727%0.8541.1070.9050.0730.073d Pongo 2098%0.9700.5440.9010.0480.072eT50Zp/H35Zp Homo 12030.7861.1570.909*0.0160.148a Pan 970.739−0.2190.9710.0630.122bGorilla (mtn.)260.8490.3540.8900.0850.065bGorilla (other)350.868−1.3031.0800.0850.077c Pongo 200.947−0.8250.9930.0700.101dFHSI/HDML Homo 12110.8120.3100.930*0.0150.038a Pan 950.7400.5420.787*0.0510.037b, cGorilla (mtn.)250.6890.4760.8170.1050.025dGorilla (other)370.676−0.2991.0010.1090.029b, e Pongo 200.940−0.3301.0210.0750.032c, d, eTPML/HDML Homo 8380.7881.0600.851*0.0180.039a Pan 930.7660.9970.814*0.0510.035b, cGorilla (mtn.)240.8331.0030.8160.0850.018b, d, eGorilla (other)340.634−0.6181.2190.1440.034c, d, f Pongo 200.9390.1541.0250.0760.032b, e, f*Slope significantly different (P < 0.05) from isometry (1.0). Extant taxa comparisons of interlimb proportions. The table reports RMA regressions of the first variable on the second variable. Arboreal percentage is calculated as the percentage of observation time spent in arboreal environments (see table S4 and the Supplementary Text for derivations). Elevation equivalency refers to equivalent line elevations in pairwise comparisons (P > 0.05). When the same letter appears in two or more rows for a variable, it indicates equivalent line elevations in pairwise comparisons. Slope SE, standard error of the slope; SEE, standard error of estimate; F50Zp, femoral midshaft torsional/average bending strength (cubic millimeters); T50Zp, tibial midshaft torsional/average bending strength (cubic millimeters); H35Zp, humeral mid-distal torsional/average bending strength (cubic millimeters); FHSI, femoral head superoinferior breadth (millimeters); TPML, tibial plateau mediolateral breadth (millimeters); HDML, humeral distal mediolateral articular breadth (millimeters). Slope significantly different (P < 0.05) from isometry (1.0). In every pairwise comparison, modern humans have significantly larger lower limb dimensions relative to upper limb dimensions (higher elevations) than any nonhuman group (Table 1). There is a consistent reduction in average femoral/humeral diaphyseal strength proportions from humans to mountain gorillas (which overlap considerably with humans), to other (nonmountain) gorillas, to chimpanzees, and lastly to orangutans (Fig. 2A). This sequential pattern corresponds inversely with the percentage of time spent in arboreal environments (Table 1). For tibial/humeral diaphyseal strength proportions, the distinction between humans and nonhuman groups is even greater (Fig. 2B), but there is more overlap among nonhuman groups, making an inversely sequential pattern between arboreality and strength proportions less consistent. For example, chimpanzees are indistinguishable from mountain gorillas and fall above other (nonmountain) gorillas in relative tibial to humeral strength, despite chimpanzees being more arboreal than either of these groups (Fig. 2B and Table 1). Interlimb articular comparisons again show modern humans to have relatively large lower limb dimensions, particularly femoral head breadth relative to distal humeral breadth (Fig. 2C), with somewhat less distinction between humans and nonhuman groups in proximal tibial breadth to distal humeral breadth proportions (Fig. 2D). However, there is considerable overlap among nonhuman groups in these articular proportions, with little correspondence in sequential patterns between hindlimb/forelimb proportions and degree of arboreality (Fig. 2 and Table 1). Modern humans exhibit stronger tibiae relative to femora when compared with all nonhuman hominoids, whereas both gorilla groups exhibit the weakest tibiae relative to femora (fig. S1). In contrast, chimpanzees and orangutans exhibit intermediate levels of tibial strength relative to femoral strength, although both are generally closer to modern humans (i.e., their regression lines fall within the modern human 95% PI) than to gorilla groups (fig. S1). Fossil hominins Seven fossil hominins with associated upper and lower limb long bones, dating from 3.67 to 1.47 Ma, were evaluated within the context of extant hominoid interlimb and intralimb proportions: StW 573 (A. prometheus); KSD-VP-1/1 and A.L. 288-1 (A. afarensis); OH 62 assigned to H. habilis by Johanson and colleagues (29) but to A. cf. africanus by Clarke (30); the large individual from Dmanisi (D3901/D4167/D4507), here assigned to “early Homo”; and KNM-ER 1808 and KNM-WT 15000 (H. erectus) (Fig. 2 and Table 2). Relative deviations in RMA standard error of estimate (SEE) units of each fossil from each extant group are reported (Table 3) and plotted (Fig. 3). Table 2.Fossil specimens included in the present study.Definitions for structural properties are provided in Table 1. Body mass is estimated in kilograms. See the Supplementary Text for details on estimation procedure. Sources for fossil data include: 1, present study; 2 (54); 3, (45); 4, (55); 5, (26); 6, (109); 7 (28); and 8 (30).PropertyTaxonDate (Ma)Body massF50ZpT50ZpH35ZpFHSITPMLHDMLSourcesStW 573 A. prometheus 3.6740.21504118394532.557.140.41, 2KSD-VP-1/1 A. afarensis 3.5559.5–––39.369.047.03, 4A.L. 288-1 A. afarensis 3.1828.91350–74128.650.230.13, 5, 6OH 62*H. habilis/A. cf. africanus1.8–1090–668–––1, 7, 8D3901/4167/4507Early Homo1.855.12531170376040.268.236.81KNM-ER 1808 H. erectus 1.671.63216–884–––5KNM-WT 15000 H. erectus 1.4751.0†1828161076344.966.838.21, 3, 5KNM-ER 1472Early Homo2.0451.9–––40.064.5–3KNM-ER 1481aEarly Homo1.9860.3–––44.069.3–3*See the Supplementary Text for derivation of cross-sectional properties.†Body mass at time of death; estimated body mass in adulthood is 76.7 kg (see also the Supplementary Text). Fossil specimens included in the present study. Definitions for structural properties are provided in Table 1. Body mass is estimated in kilograms. See the Supplementary Text for details on estimation procedure. Sources for fossil data include: 1, present study; 2 (54); 3, (45); 4, (55); 5, (26); 6, (109); 7 (28); and 8 (30). See the Supplementary Text for derivation of cross-sectional properties. Body mass at time of death; estimated body mass in adulthood is 76.7 kg (see also the Supplementary Text). Table 3.Deviations in interlimb proportions of fossils from the extant groups.The table reports deviations using RMA regressions of the first variable on the second variable (see Table 1 for definitions of variables). Deviations are reported in RMA SEE units, adjusted for the x position, from each extant group. Bold font in a row indicates the smallest deviation.Relative deviation of fossil specimen from the extant groupRegressionSpecimen Homo Pan Gorilla (mtn.)Gorilla (other) Pongo F50Zp/H35ZpStW 573−3.2061.223−1.314 0.057 5.820A.L. 288-1−2.6952.212 −0.289 0.8466.630OH 62−3.5370.908−1.070 −0.006 5.396D3901/4167/4507 1.153 8.4774.1225.19812.339KNM-ER 1808 1.926 9.7545.2676.31513.648KNM-WT 15000 −0.923 5.0891.7242.8439.262T50Zp/H35ZpStW 573 −1.564 3.7384.9526.6247.200D3901/4167/4507 1.253 6.9438.2819.74610.635KNM-WT 15000 0.954 6.6057.8829.38910.262FHSI/HDMLStW 573−5.1570.607 −0.379 1.2970.751KSD-VP-1/1−4.2112.092 1.344 2.4511.500A.L. 288-1−2.3332.624 1.172 2.6403.693D3901/4167/4507 0.515 6.5694.7085.5156.907KNM-WT 15000 2.073 8.4196.6007.0108.625TPML/HDMLStW 573−3.1350.865 0.725 1.9692.063KSD-VP-1/1 −1.962 2.3283.6262.5882.781A.L. 288-1 −0.770 3.0982.2903.1144.767D3901/4167/4507 1.759 6.4216.6824.8337.409KNM-WT 15000 0.832 5.4075.9204.4016.420 Deviations in interlimb proportions of fossils from the extant groups. The table reports deviations using RMA regressions of the first variable on the second variable (see Table 1 for definitions of variables). Deviations are reported in RMA SEE units, adjusted for the x position, from each extant group. Bold font in a row indicates the smallest deviation. Evaluating individual fossils relative to comparative samples. (A) femoral 50% Zp to humeral 35% Zp, (B) tibial 50% Zp to humeral 35% Zp, (C) femoral head SI breadth to distal humeral ML breadth, and (D) tibial plateau ML breadth to distal humeral ML breadth. Deviations, in RMA regression standard error of estimate (SEE) units, of fossils from the comparative sample RMA regressions illustrated in Fig. 2. Zp, polar section modulus; SI, superoinferior; ML, mediolateral. Both H. erectus individuals and the Dmanisi individual fall securely within and generally in the upper half of modern human distributions for all lower/upper (inter)limb strength and articular proportions and outside 95% PIs for other extant groups (Fig. 2). The comparatively lower position of KNM-WT 15000 in femoral/humeral strength, also falling just within the upper extent of the 95% PIs of mountain gorillas (Fig. 2A and Table 3), has been attributed to his juvenile age (27). In contrast, both the Australopithecus and OH 62 individuals fall below the 95% PIs of modern humans and are closest to either of the gorilla groups in femoral/humeral strength proportions (Figs. 2A and 3A and Table 3). This is also the case for femoral head to distal humeral articular proportions (Figs. 2C and 3C). In interlimb proportions involving the tibia, results indicate a more variable pattern. StW 573 falls closer to the diaphyseal strength proportions of modern humans than nonhuman groups (Figs. 2B and 3B) and KSD-VP-1/1 and A.L. 288-1 fall closer to modern humans in tibial plateau breadth to distal humeral articular proportions, but StW 573 falls closer to mountain gorillas and chimpanzees than modern humans in the latter proportion (Figs. 2D and 3D). However, in each of these cases where resemblance to modern human proportions is observed, the Australopithecus and OH 62 individuals fall in the lower half of the modern human distribution and well below the three H. erectus/early Homo individuals. In the proportions for which all three fossils can be compared (Figs. 2, C and D, and 3, C and D), A.L. 288-1 and KSD-VP-1/1 are more similar to modern humans than is StW 573. Comparing tibial/humeral diaphyseal strength versus femoral/humeral diaphyseal strength (Fig. 2, A and B), the relatively higher position of StW 573 in the former, i.e., greater similarity to modern humans, reflects the stronger tibia of StW 573 relative to its femur than in nonhuman hominoids. This is confirmed by evaluating intra(hind)limb relative strength and direct comparisons between tibial and femoral diaphyses, where StW 573 falls in the upper half of the modern human distribution for this proportional strength, along with Dmanisi and KNM-WT 15000 (fig. S1). Notably, all hominins, including modern humans, exhibit strong tibiae relative to femora when compared with nonhuman hominoids, which vary widely and exhibit overlap among one another in this intralimb proportion (fig. S1). We further explore relative femoral head size through direct comparisons of femoral head breadth to proximal tibial breadth (Fig. 4). Fossil specimens include two additional early Homo femora, KNM-ER 1472 and KNM-ER 1481a (Table 2). African apes have significantly smaller femoral heads relative to proximal tibial breadths compared with modern humans [P < 0.0001; (45)] and are nearly isometric in this relationship (RMA slope = 1.031 ± 0.015 SE), whereas modern humans exhibit clear positive allometry (RMA slope = 1.126 ± 0.012). The three Australopithecus individuals fall slightly below the African ape RMA regression line and, like African apes, are nearly isometrically scaled to one another [RMA slope = 0.9996 ± 0.0010, correlation coefficient (r) = 1.0]. Three Homo individuals (i.e., KNM-WT 15000, KNM-ER 1472, and KNM-ER 1481a) fall either near the modern human RMA regression line or above it, and at or above the upper limit of the African ape 95% PI. The Dmanisi individual plots slightly above the Australopithecus RMA regression line and well below the modern human RMA regression line, unlike any of these other Homo individuals. Thus, relative to knee joint size, hip joint size is small in African apes and Australopithecus, intermediate in the Dmanisi individual, and larger in modern humans and other early Homo or H. erectus individuals. Log/log plot of femoral head superoinferior breadth versus tibial plateau mediolateral breadth. Modern humans (black crosses), chimpanzees (open red circles), gorillas (closed red circles), and fossil individuals are shown with RMA regression lines and 95% PIs. Regression lines are provided for the human, nonhuman African ape, and Australopithecus samples, with 95% PIs provided only for the extant groups. Because some of the comparisons we conducted between extant groups and between fossils and extant groups involve extrapolations beyond the observed size range of the extant groups, where even relatively small, nonsignificant differences in regression slopes may distort spatial relationships, we also visually assessed these relationships using the common RMA slope for the entire pooled samples, plotted through each extant sample (figs. S1B and S2). These results do not alter any of our conclusions regarding relative strengths or articular proportions of the extant and fossil groups. DISCUSSION Relative limb strength, like certain other skeletal features such as femoral bicondylar angle (46, 47), reflects, in part, a developmentally plastic response to in vivo behavior (18). Greater relative strength of the forelimb in extant nonhuman hominoids compared with modern humans reflects greater involvement of the former’s forelimb in body weight support, especially during arborealism (19, 48). Additionally, relatively smaller hindlimb joint surfaces in extant nonhuman hominoids reflect lower joint loadings in this limb than typically accommodated by humans during their terrestrial bipedal walking or running dominated locomotor repertoire (49). In the present study, the forelimb to hindlimb relative strength and articular dimensions of the StW 573 individual are generally more African ape–like than modern human–like. Morphological traits observed in the StW 573 pectoral girdle (50) and inner ear (51) are also African ape–like, strengthening the inference of a substantial African ape–like degree of arborealism performed by this individual. In its relative limb strength and articular dimension patterns, StW 573 is even more African ape–like than the slightly younger and markedly smaller-bodied A. afarensis individual, A.L. 288-1 (26). This is consistent with other analyses of the external morphology of some other anatomical regions of A. afarensis, such as the foot, which have been shown to exhibit fewer African ape–like features compared with other Australopithecus species (12), even when including the younger taxa A. africanus and Australopithecus sediba (52, 53) in such comparisons. The present findings, using developmentally plastic traits, have critical implications for Australopithecus. The StW 573 individual, with an estimated body mass of 40.2 kg, is near the 50th percentile of 37 Australopithecus individuals with estimated body masses (35). This position demonstrates that an African ape–like functional signature in relative limb strength of Australopithecus is not exclusive to the smallest-bodied individuals such as A.L. 288-1. Relative femoral head to distal humeral articular dimensions in the even larger-bodied KSD-VP-1/1 A. afarensis individual are also African ape–like (Fig. 2C). Thus, results of the present study align different Australopithecus individuals spanning a broad spectrum of body sizes with African ape–like functional traits in the limbs. In contrast, StW 573 and A.L. 288-1 exhibit similar (isometric) interlimb length proportions (54), likely due, in part, to greater developmental constraints on such proportions, making them less sensitive indicators of in vivo behavior (19, 26). Variation in limb length proportions may also have less clear adaptive significance among early hominins (55, 56). Among diaphyseal strength proportions of extant hominoids, strength of the tibia to that of the humerus or femur shows less consistent relationships with degrees of arboreality than do femoral to humeral strength proportions. This difference may derive, in part, from shared load-bearing between the tibia and fibula in the leg, which, based on structural analyses of both bones, varies between extant hominoids as well as between early Homo taxa (57). Modern humans exhibit tibial strengths that more closely approach their larger femoral strengths compared with more disparate tibial/femoral strengths in nonhuman apes, especially gorillas (fig. S1). The modern human pattern appears to characterize all three early hominins examined here, including StW 573. Future analyses of the StW 573 fibulae will further investigate this issue. We suggest that an African ape–like degree of arborealism was a fundamental component of the niche of Australopithecus species on the basis of African ape–like functionally plastic features (e.g., relative limb strength) and articular proportions in the humerus and femur that are apparent across a broad range of Australopithecus body sizes. It is important to acknowledge, however, that terrestrial bipedalism also had an important role in the Australopithecus locomotor repertoire, because intra(hind)limb relative strength and some articular proportions of Australopithecus, such as proximal tibial to distal humeral articular size, are closer to those of modern humans than to African apes, consistent with increased load-bearing on the lower limb [see also (55, 58)]. Thus, there is no single, living analog for modeling how australopiths used both their upper and lower limbs to engage with their surrounding habitat. Articular proportions also show a more complex relationship with degrees of arboreality than some diaphyseal strength proportions, because more overlap exists between extant nonhuman groups. Articular size is related to joint reaction force, but it is also adaptive for joint excursion and mobility (59, 60). For example, despite having a relatively weak femoral diaphysis (Fig. 2A), Pongo has a femoral head that is proportionately as large as femoral heads in other nonhuman apes (Fig. 2C), which, combined with its relatively narrow neck, shallow acetabulum, and high neck-shaft angle (61), facilitates a substantial degree of hip abduction in orangutans (62, 63). Relative joint size can also be influenced by joint-specific mechanical factors, such as the hip abductor mechanism used by bipedal hominins. The relatively small femoral head of Australopithecus compared with that of Homo may be related to altered hip bipedal kinematics of the former that reduced its hip joint reaction forces (64). The estimate of femoral head breadth derived for StW 573 in the current study shows that this individual conformed to the same pattern as other Australopithecus individuals, implying a similar difference in hip biomechanics from that exhibited in early Homo. As previous studies (15, 20, 54) used a larger femoral head diameter for StW 573, future functional interpretations should incorporate the revised estimate reported here, which is based on reconstructed high-resolution computed tomography (CT) data (see the Supplementary Text). Interlimb strength and articular proportions of the Dmanisi individual are essentially modern human–like, similar to the pattern exhibited by two H. erectus individuals (KNM-WT 15000 and KNM-ER 1808). The Dmanisi individual also exhibits modern human–like intra(hind)limb relative strength. Other derived aspects of Dmanisi lower limb morphology corroborate this general functional equivalence to modern humans and concomitant commitment to terrestrial bipedalism (38, 65). Combining evidence from these Homo individuals effectively establishes the early Homo pattern in relative limb strengths and articular dimensions at ∼1.8 Ma as relatively modern human–like rather than African ape–like. The stark differences between these early Homo and Australopithecus patterns have critical implications for human evolution. Before ∼1.8 Ma, there was a behavioral shift resulting in subsequent members of the Homo clade emancipating their upper limb from weight-bearing loads. We attribute this change to a substantial reduction in the degree of arborealism concurrent with a full commitment to (i.e., habitual) human-like terrestrial bipedalism. We suggest that such a fundamental shift coincided with the transition from Australopithecus to Homo and was driven by one or more potential selective pressures. Stone tool knapping predates 3.0 Ma (66), suggesting that enhanced manual capabilities for lithic tool production were not among these. Moreover, selective pressures favoring dexterous manipulatory skills associated with tool use rather than explicit tool manufacture also seem unlikely because primates are fully capable of using fine instruments as tools, e.g., chimpanzee use of plant materials for termite fishing (67). Expanded ranging, possibly associated with increased meat consumption, merits consideration as an additional, alternative selective pressure favoring full commitment to terrestrial bipedality in the Dmanisi sample and early Homo (except OH 62, if it represents H. habilis) [see (56) and references therein]. In this respect, recent work (68) demonstrating a link between cognitive abilities associated with foraging and perceived effort during endurance walking bears consideration. Rates of encephalization in hominin species increase from Australopithecus to Homo (69). This shift generally coincides with the evolutionary timing that we propose for the shift toward femur-dominated interlimb strength ratios. Thus, we suggest that changes in foraging strategies that favor enhanced cognitive abilities and increased ranging (i.e., lower limb usage) might have intertwined both trends during this critical period of human evolutionary history. The modern human–like tibial/femoral relative strength proportion of StW 573 shows that this pattern was already established in early Australopithecus, whereas an African ape–like degree of arborealism indicated by the femoral/humeral strength proportion is observed in this same individual. A relatively strong tibia may be related to concurrent changes among australopiths in other aspects of the distal lower limb, i.e., the foot and ankle, that are adaptive for terrestrial bipedalism (70). In this respect, the overview signal of hindlimb loading reflected in femoral to humeral strength proportion may vary from discrete, localized loading of individual regions of the lower limb (71–74). Intralimb relative strengths from additional early hominins will help untangle the variable signals present among these indicators of hindlimb loading regimes and whether variable bipedal gait kinematics among early hominin taxa could be a contributing factor. Slight differences in bipedal gait mechanics may also distinguish Dmanisi individuals from earlier “erectus-like” and later H. erectus individuals. Hip relative to knee joint articular size in the Dmanisi individual is intermediate between that of modern humans and African apes and differs from the more modern human–like proportions exhibited in later H. erectus (e.g., KNM-WT 15000), as well as those in some earlier erectus-like Homo individuals (KNM-ER 1472 and KNM-ER 1481a). Considering that additional primitive features in the lower limb of the Dmanisi individuals have been noted, particularly in the foot (65, 75), this is suggestive of potential slight differences in bipedal gait mechanics of the Dmanisi individuals compared with those of earlier erectus-like and later H. erectus individuals within eastern Africa. This may have implications for the timing of the migration of Dmanisi ancestors and/or their source population from Africa. Given a temporal gap between ∼3.0 and 1.8 Ma in our new sample, we cannot pinpoint the timing of the proposed adaptive shift between the australopith grade and the Homo clade any more precisely. At ∼1.97 Ma, A. sediba is a late surviving member of the genus Australopithecus (76) with an australopith-like estimated cranial capacity of 420 cm3 in the MH1 individual (77). In their current state, partial skeletons of the MH1 and MH2 individuals attributed to A. sediba do not yet provide relative limb strengths for comparison to those of StW 573 and A.L. 288-1. Other morphological indicators of arborealism in the positional repertoire of A. sediba, however, have been proposed (78–81), suggesting that, when eventually available, A. sediba relative limb strengths may be less modern human–like than those of early Homo. The ∼1.8-Ma OH 62 partial skeleton represents a clear outlier compared with the early Homo pattern established in the present study. Its relative limb strength is as African ape–like as are those of StW 573 and A.L. 288-1. Whether OH 62 should be attributed to Homo or Australopithecus has been a subject of debate (30, 82–86). Based on our evaluations of developmentally plastic traits in the present study compared with potentially more conserved traits such as limb length proportions, we suggest that further consideration of the OH 62 individual as potentially representative of an australopith grade of arboreality is warranted. Last, while more recent hominins such as Homo floresiensis and Homo naledi do not fall within this temporal gap, they have some bearing on our efforts to characterize the relative limb strength pattern in the Homo clade. While these taxa are committed to terrestrial bipedalism, relative limb strengths cannot yet be evaluated in them (87, 88). When such traits become available in either taxon, we predict that any similarities to African ape–like proportions reflecting aspects of arboreality in their realized niches would represent convergence on the pattern characterizing Australopithecus. This would still be consistent with the fundamentally different realized niches of Australopithecus and early Homo species proposed here. MATERIALS AND METHODS Fossil specimens To acquire quantitative data from the StW 573 skeleton and Dmanisi hominin fossil assemblage to address study questions, we prioritized nondestructive acquisition of image data from humeri, femora, and tibiae (see the Supplementary Text). When necessary, we undertook minor in silico reconstruction of fossils before data collection (see below). To identify cross sections of interest on the fossils, we first aligned each long bone following criteria defined by Ruff (16) and Carlson (17) and adapted for in silico positioning by Carlson and colleagues (89). Next, targeted regions of interest along the shaft at 20, 35, 40 (humerus only), 50, 65, and 80% of biomechanical length (16) were located on the humerus (figs. S3 and S4), femur (figs. S5 and S6), and tibia (figs. S7 and S8) with 0% length represented by the distal end of the specimens. Evaluations of strength proportions use only 50% sections for the femur and tibia and the 35% section for the humerus, matching a previous comparative analysis of A.L. 288-1 (26). The average of right- and left-side values for the StW 573 tibiae was used in analyses. Properties for all sections of StW 573 and Dmanisi D3901/D4167/D4507 are reported in tables S1 and S2, respectively. Cross sections generated from rendered image data were extracted and manually edited to remove adherent matrix and areas of trabecular bone [for justification for removal of trabecular bone in mechanical analyses of diaphyseal sections, see (90)]. Periosteal and endosteal surfaces were well demarcated and preserved in all sections, with the exception of a small missing portion of the lateral surface of the humeral 20% section, which was reconstructed on the basis of adjacent preserved contours (fig. S1). Cross sections were rotated and flipped as necessary before analyses so that anterior is above and medial is to the left, regardless of whether elements are from the right or left side. Edited cross sections (Fig. 1) were analyzed using MomentMacro (https://fae.johnshopkins.edu/chris-ruff/), a plugin for ImageJ (https://imagej.net/ij/). MomentMacro calculates a number of cross-sectional properties, including bone areas, second moments of area, and section moduli (see tables S1 and S2). For bone strength analyses, the polar section modulus, Zp, was used. Zp is a good general index of overall diaphyseal strength, representing the geometric component of torsional and (twice) average bending strength (19), and has been used in a number of past studies of long bone strength proportions in early hominins as well as nonhuman primates (19, 26–28, 91–94). In these analyses, bone material properties are presumed to be constant, a reasonable assumption based on phylogenetic comparisons (95). Articular breadths and bone lengths for StW 573 and the Dmanisi specimens were obtained as described in the Supplementary Text. The estimate of femoral head breadth for StW 573 derived in the current study (32.5 mm) is substantially smaller than previously estimated [35.2 mm: table 5 in (54)], as we had the advantage of using high-resolution image data in its derivation. Sources for these dimensions and cross-sectional properties in other fossil specimens are reported in Table 2. Cross-sectional properties for OH 62 were derived as described in the Supplementary Text. T50Zp for KNM-WT 15000 was obtained from previously unpublished data and is reported as the average of right and left values. Estimated body masses were derived as described in the Supplementary Text. Here, we exclude the genus Paranthropus when using the term “australopith” in the manuscript, e.g., australopith grade. Information for all fossils included in the study is provided in table S3. Extant comparative samples The Pan sample included Pan paniscus (n = 19), Pan troglodytes schweinfurthii (n = 19), Pan troglodytes troglodytes (n = 34), and Pan troglodytes verus (n = 27). The Pongo sample included Pongo pygmaeus pygmaeus (n = 15) and Pongo pygmaeus abelii (n = 4), with one individual not identified at the species level. The modern human sample included anatomically modern Late Pleistocene and Holocene Europeans (n = 1279), ranging from 31,000 to 50 years before the present, composed mainly of archaeological skeletal material but also including some very recent cadaveric material, with a slight preponderance of males in the sample overall (92). Sample sizes for these groups varied somewhat between different variables (see Table 1). For Fig. 4, the modern human sample (n = 2760) includes several non-European samples [see details in (45)]. Gorillas were divided into two groups on the basis of marked locomotor and skeletal structural variation within the genus, as demonstrated in previous studies (19, 96): Gorilla beringei beringei (Virunga mountain gorillas, n = 17), and a second group comprising all other gorillas in our sample, including Gorilla beringei graueri, from both high (n = 6) and low (n = 13) elevations, and Gorilla gorilla gorilla (Western lowland gorillas) (n = 17). While there is a gradient in both behavior and limb bone strength proportions within the second group (19, 96), they were placed together here to somewhat simplify comparisons and because the former group, G. beringei beringei, is the most extreme among all gorilla taxa in both regards (Table 1) (19, 96). All individuals were adult and nonhuman individuals were all free-ranging. All groups contained both males and females, except for gorillas. Only female gorillas were included in the study to limit the body size range of the comparative sample and ensure more overlap between groups in quantitative analyses. Although there is some sex-related variation in locomotor behavior among gorillas (97, 98), the effects of this variation on structural proportions of the forelimbs and hindlimbs are relatively minor (19), and inclusion of male gorillas in the skeletal sample does not affect the positions of gorillas relative to other extant groups or fossils. For Fig. 4, however, following Ruff and colleagues (45), both male and female gorillas were included. Average percent arboreality (table S4), including locomotor and postural behaviors, was calculated for each group as described in the Supplementary Text. Morphological data for the extant Homo sapiens, Pan, Gorilla, and Pongo individuals were obtained from previous studies (16, 19, 45, 99, 100). Cross-sectional diaphyseal data for nonhuman taxa were derived through medical CT scanning and subsequent analysis using either SLICE (101) or MomentMacro. The modern human cross-sectional data were obtained from a study of Holocene European samples that used either medical CT or multiplanar radiography combined with external molding, followed by analysis with MomentMacro (92). In some previous studies, section moduli (including Zp) were calculated from second moments of area and external breadths. These values were converted here to true section moduli using relationships determined from sections that were analyzed in both ways (26). Similarly, in a subset of the nonhuman sample, data for the humeral 35% section were derived from a section at 40% of biomechanical length using specimens where both sections were analyzed (26). Articular dimensions were obtained through direct digital caliper measurements. Statistical analyses Statistical analyses were carried out in R (version 4.0.2) with graphs created using the ggplot2 package [version 3.3.2; (102, 103)]. Preliminary analyses indicated that several of the interlimb proportional relationships evaluated here were not isometric, i.e., regression slopes between variables departed significantly from 1.0 (Table 1). In this situation, ratios or log ratios between properties are potentially biased estimators of proportional differences (16). Therefore, following earlier studies (26, 28), interlimb proportions were evaluated using RMA regressions of natural log-transformed hindlimb against forelimb dimensions fit through the extant samples. No pairwise differences in slope between groups were significant, allowing comparisons of line elevations as indicators of proportional differences (104, 105). Individual fossil specimens were evaluated relative to each extant group using distances from RMA regression lines, expressed in SEE units (106). RMA analyses were carried out using the lmodel2 [version 1.7.3; (107)] and SMATR packages [version 3.4.8; (108)]. Graphs were created with 95% PIs for extant groups calculated as previously described (26). Although there were no significant differences in slopes between groups, because some of the extant groups, particularly female gorillas, were displaced in size relative to fossil specimens, there is still a chance that regression lines extrapolated into the size range of the fossils could provide a biased assessment of relative deviation in proportions from those groups. Therefore, we also plotted RMA lines through extant groups based on the common slope for all groups (104), along with fossil specimens, to allow another visual assessment of proportions (figs. S1B and S2). Last, we include a graph of relative deviations (in RMA SEE units) of each fossil from each extant group (Fig. 3).