Natural selection, genetic drift, and trait correlation shaped hominin midfoot evolution Authors: Klara Komza, Bence Viola, Lauren Schroeder Published: 2026 ABSTRACT AbstractThe foot has undergone significant morphological change throughout the hominin lineage as our extinct relatives evolved to walk upright. The midfoot displays key adaptations for terrestrial bipedal locomotion, and is therefore considered a target of natural selection. Prior work suggests that early selection for bipedalism targeted the lateral elements of the foot, but that later, among australopiths, these elements were less morphologically diverse, and less influenced by directional selection than medial elements. Here, we apply methods from quantitative genetics to test which evolutionary processes (directional vs. stabilizing selection) shaped hominin midfoot diversification under a null hypothesis of genetic drift. Results demonstrate that the lateral midfoot may have been the primary target of selection for bipedalism in Ardipithecus ramidus, but that later in the hominin lineage, there is no consistent evidence that the lateral elements were less influenced by directional selection. Instead, we find a complex interaction of evolutionary processes shaping midfoot diversification, characterized by genetic drift, directional selection, and indirect responses to selection based on correlation between different elements. In sum, we emphasize that narratives of the evolution of bipedalism should consider the role of genetic drift and correlation among skeletal elements, as both have been historically underestimated.Subject terms: Biological anthropology, Evolutionary theory Introduction The human foot is a unique structure among primates that is specialized for obligate terrestrial bipedalism. The morphology of the midfoot (Fig. 1), which consists of the cuneiforms, navicular, cuboid, and metatarsal (Mt) bases, influences a primate’s capacity for arboreal or terrestrial locomotion1 and has been extensively studied in living primates and fossil hominins to understand the evolution of bipedalism in modern humans1–6. Comparative analyses of humans and nonhuman primates reveal that humans have a long midfoot relative to body mass and total foot length, as well as differences in articular surface shape that may reduce the foot’s mobility4,5. Overall, human midfoot proportions are hypothesized to lengthen the foot’s lever arm, and its articular surface morphology to limit midfoot mobility, ultimately converting the foot to an effective lever for propulsion3–5,7,8. Midfoot morphology in humans and chimpanzees. The foot on the left displays all elements of the midfoot in blue. Regions denoted with letters A–C differ between humans and chimpanzees and have been linked with locomotor differences. A The midfoot is proximodistally longer in humans than in chimpanzees, lengthening the foot’s lever arm and converting it to an effective lever for propulsion3. B The hallux is adducted in humans, limiting its grasping abilities but contributing to propulsion, and abducted in chimpanzees, enabling pedal grasping and arboreal locomotion2,4. C The proximal articular surface of the Mt4 is flat or sinusoidal in humans, likely contributing to limited lateral midfoot mobility, and convex in chimpanzees, contributing to greater lateral midfoot mobility5. Although these morphological features have been extensively studied, to date, we do not have a clear understanding of patterns of midfoot variation in fossil hominins, or of the underlying evolutionary processes shaping midfoot variation. Several models for the evolution of bipedalism make different assumptions about the foot morphology of the hominin-panin last common ancestor (LCA)6,9–16, but all are inherently difficult to test because of large temporal gaps in the fossil record and because many are based on modern primates that may not reflect the morphology of the LCA17. However, increasing fossil evidence may lend support to Kidd and colleagues9,10, who proposed that the lateral side of the foot was the earliest target of selection for bipedalism. This hypothesis was originally proposed based on an African ape-like LCA before the discovery of many important fossils, but has since received some support based on the morphology of the Ardipithecus ramidus foot. Lovejoy and colleagues11 observed that its lateral metatarsals (Mts 2–5) and cuboid display adaptations for bipedalism, whereas the medially-located Mt1 retains adaptations for arboreality. This is further supported by evolutionary modeling analyses of the Ar. ramidus forefoot, which suggest that Mt heads 2–5 were the primary targets of selection for bipedalism18. However, it is unclear whether this same evolutionary pattern is found in the midfoot. Later in the hominin lineage, among australopiths, the medial side of the foot has been proposed to be more morphologically diverse than the lateral side, with some Pleistocene taxa displaying more ape-like features, and other contemporaneous taxa showing more human-like features in Mts 1–219–24. However, it is difficult to determine whether this morphological diversity is indicative of locomotor variation because we do not have a clear understanding of normal levels of intraspecies morphological variation in extinct hominins. In addition, there may be less evidence of morphological diversity in some lateral elements due to preservation bias, as there are few cuboids and Mt4s in the fossil record. In sum, morphological variation and preservation biases, as well as our limited knowledge of the LCA’s anatomy have prevented a full understanding of midfoot diversity and evolution among hominins. Because of the foot’s crucial role in locomotion, it is often expected that much of its evolution was driven by natural selection25–27. The relative influence of selection on foot dimensions has been frequently inferred through biomechanical models that investigate the effectiveness of skeletal features for bipedal walking and running21–24. While such experimental approaches may clarify the biomechanical functions of different morphological features and their relationship with locomotion, variation in foot morphology may have also been shaped by morphological integration and covariance across skeletal elements28,29. Analyses of morphological integration between elements of the foot highlight that the foot acts as a functional unit, and that traits within the medial elements may be more strongly integrated than in the lateral elements28,29. As a result, the medial elements of the foot may have a limited capacity to evolve independently29, or to respond to natural selection. Research from the field of evolutionary quantitative genetics highlights the role that non-selective evolutionary processes (e.g., genetic drift) may also play in producing evolutionary change (e.g., refs. 30–37). Quantitative genetic methods developed by Lande and colleagues38,39 and Cheverud40 provide a testable way of measuring the influence of natural selection on morphological diversification using genetic drift as a null hypothesis. Previous applications of quantitative genetic approaches to paleoanthropological contexts emphasize that while some morphological features associated with biomechanical functions may have been influenced by natural selection, many have not, and others likely diversified because of their correlation with other skeletal traits30,34–36. Quantitative genetic analyses of the foot are limited, but there is some evidence of strong directional selection contributing towards a more robust hallux and shorter toes between chimpanzees and modern humans, and evidence of correlated trait responses to selection in the hands41. No work to date has tested the impact of different evolutionary processes on the hominin midfoot. Here, we aim to determine which evolutionary processes shaped the morphological diversity seen in the Plio-Pleistocene hominin midfoot in a series of hypothetical ancestor – descendant relationships using evolutionary quantitative genetics (for a full list of comparisons, see Tables 1, 2, and for justifications, see Supplementary Table 1). We analyze each element independently to assess the influence of natural selection on individual bones, and where we have three or more elements from the same foot, we also analyze the relative influence of natural selection on foot proportions. Table 1Full list of hominin comparisons for individual elementsElementComparisons (ancestor—descendant)Mt1P. troglodytes—Ar. ramidus (ARA-VP-6/500-089)Ar. ramidus (ARA-VP-6/500-089)—Burtele (BRT-VP-2/73c)A. afarensis (A.L. 333-54)—StW 573A. afarensis (A.L. 333-54)—A. africanus (StW 562)A. afarensis (A.L. 333-54)—OH 8A. afarensis (A.L. 333-54)—Burtele (BRT-VP-2/73c)Burtele (BRT-VP-2/73c)—OH 8StW 573—A. africanus (StW 562)StW 573—OH 8StW 573—H. naledi (U.W. 101-1443)A. africanus (StW 562)—P. robustus (SKX 5017)A. africanus (StW 562)—H. naledi (U.W. 101-1443)P. robustus (SKX 5017)—H. naledi (U.W. 101-1443)H. naledi (U.W. 101-1443)—H. sapiensOH 8—H. sapiensMt2P. troglodytes—Ar. ramidus (ARA-VP-6/1000)Ar. ramidus (ARA-VP-6/1000)—Burtele (BRT-VP-2/73)Ar. ramidus (ARA-VP-6/1000)—StW 573Burtele (BRT-VP-2/73)—OH 8StW 573—A. africanus (StW 89, 377, 595c)StW 573—A. africanus (StW 89, 595c)StW 573—A. africanus (StW 377)StW 573—OH 8OH 8—H. sapiensMt3P. troglodytes—Ar. ramidus (ARA-VP-6/505)A. afarensis (A.L. 333-157)—A. africanus (StW 387, 435, 477, 496)A. afarensis (A.L. 333−157)—OH 8A. africanus (StW 387, 435, 477, 496)—H. naledi (U.W. 101-1457)OH—H. sapiensH. naledi (U.W. 101-1457)—H. sapiensMt4A. afarensis (A.L. 333-160)—A. africanus (StW 485)A. afarensis (A.L. 333-160)—OH 8A. africanus (StW 485)—A. sediba (U.W. 88-22)A. africanus (StW 485)—H. naledi (U.W. 101-1456)A. sediba (U.W. 88-22)—H. sapiensOH 8—H sapiensH. naledi (U.W. 101-1456)—H. sapiensMt5A. afarensis (A.L. 333-13; 333-78)—A. africanus (StW 114-115)A. africanus (StW 114-115)—A. sediba (U.W. 88-33)A. africanus (StW 114-115)—H. naledi (U.W. 101-518)A. sediba (U.W. 88-33)—H. naledi (U.W. 101-518)H. naledi (U.W. 101-518)—H. sapiensMedial cuneiformA. afarensis (A.L. 333-28)—StW 573A. afarensis (A.L. 333-28)—OH 8StW 573—H. naledi (U.W. 101-1535)StW 573—OH 8OH 8—H. sapiensH. naledi (U.W. 101-1535)—H. sapiensIntermediate cuneiformP. troglodytes—Ar. ramidus (ARA-VP-6/500-045; 075)Ar. ramidus (ARA-VP-6/500-045; 075)—StW 573StW 573—OH 8StW 573—H. naledi (U.W. 101-1534)H. naledi (U.W. 101-1534)—H. sapiensOH 8—H. sapiensLateral cuneiformA. afarensis (A.L. 333-79)—StW 573A. afarensis (A.L. 333-79)—OH 8StW 573—OH 8StW 573—H. naledi (U.W. 101-683; 1764)OH 8—H. sapiensH. naledi (U.W. 101-683; 1764)—H. sapiensNavicularA. afarensis (A.L. 333-36; 47)—StW 573A. afarensis (A.L. 333-36; 47)—OH 8StW 573—OH 8OH 8—H. sapiensCuboidP. troglodytes—Ar. ramidus (ARA-VP-6/500-081)Ar. ramidus (ARA-VP-6/500-081)—OH 8OH 8—H. sapiensShown are all proposed relationships for each element, as well as the accession IDs for all hominin fossils included in the study sample. Full list of hominin comparisons for individual elements Shown are all proposed relationships for each element, as well as the accession IDs for all hominin fossils included in the study sample. Table 2Full list of hypothetical ancestor—descendant comparisons for foot proportionsComparison (ancestor—descendant)ElementsP. troglodytes—Ar. ramidusMt1, intermediate cuneiform, cuboidStW 573—H. nalediMt1, Mt2, intermediate cuneiform, lateral cuneiformOH 8—H. sapiensMt1, Mt2, Mt3, Mt4, medial cuneiform, intermediate cuneiform, lateral cuneiform, navicular, cuboidH. naledi (Foot 1)—H. sapiensMt1, Mt2, Mt3, Mt4, Mt5, intermediate cuneiform, lateral cuneiformH. naledi (Foot 4)—H. sapiensMt1, Mt2, Mt3, medial cuneiform, intermediate cuneiform Full list of hypothetical ancestor—descendant comparisons for foot proportions First, we test a lateral-to-medial model of midfoot evolution by investigating the evolutionary processes shaping the Ar. ramidus foot. An approach rooted in quantitative genetics that incorporates fossil hominin foot bones presents an empirical way of testing this evolutionary model, which has not been done to date. Under a null hypothesis of genetic drift, we predict that between chimpanzees and Ar. ramidus, we will find more evidence of directional selection acting on the cuboid, intermediate cuneiform, and Mts 2–3, than on the Mt1. To fully support a lateral-to-medial model of foot evolution, we also expect to find evidence of directional selection on traits of these elements that have been linked to a rigid midfoot, including the proximodistal length of the cuboid and the dorsoplantar height of the metatarsals1,4,6,11. Second, we investigate the evolutionary processes shaping morphological variation in Australopithecus and small-bodied Homo foot bones. Based on the observation that throughout the hominin lineage, some medial elements of the foot may have been more morphologically diverse than laterally-located elements, we predict that Mts 1–2 and all cuneiforms were influenced by a broader range of evolutionary processes than Mts 3–5 and the cuboid. Later in the hominin lineage, among members of the genus Homo, we expect to find fewer rejections of drift, and less evidence of directional selection on foot proportions, as all members of this genus have feet adapted for obligate terrestrial bipedalism42. Results Testing the null hypothesis of genetic drift The null hypothesis of evolution by genetic drift in elements of the midfoot was evaluated using regression tests based on theory from evolutionary quantitative genetics30,38,43,44, which assess proportionality among between- and within-group variance estimated using principal component (PC) analysis. If significant deviations from proportionality are observed, genetic drift can be rejected. If the regression slope is significantly greater than one, this may be indicative of directional selection acting on the major PCs or stabilizing selection acting on minor PCs. If the slope is significantly lower than one, this may be indicative of stabilizing selection acting on major PCs or directional selection acting on minor PCs. We focus mainly on results generated from logged and sex-standardized data to reduce the influence of extreme body size differences on our results (for rationale, see Supplementary Note 1 and refs. 35,45,46). In summary, results from unlogged data seem to largely reflect selection on body size (Supplementary Tables 2–4). In comparisons where we can reject the null hypothesis of genetic drift (Supplementary Table 3), we find evidence of greater between- than within-group variance on major PCs between hominins with large differences in body size (e.g., between Homo naledi and Homo sapiens). Among small-bodied hominins, we instead find evidence of greater within- than between-group variance on major PCs, highlighting the role that body size plays in human evolution. These results indicate that body size may reflect the line of least resistance within the hominin lineage, and that strong selection (either diversifying or stabilizing) on body size (as captured by the major PCs) may influence responses to selection in minor PCs47,48. In our analyses of logged, sex-standardized data, the null hypothesis of genetic drift can be rejected in 11 comparisons, mainly in the medial elements of the midfoot (i.e., the Mt1, Mt2, medial and intermediate cuneiforms; Table 3). However, after conducting a Benjamini-Hochberg (BH) correction on the p-values in aggregate, drift could only be rejected in 1/11 comparisons (Table 3). Nevertheless, patterns of selection do vary across these elements, with slopes being both higher and lower than one, indicating a combination of greater and lesser than expected between-group variance among PCs, and supporting our prediction that the medial side of the midfoot was subjected to a broad range of evolutionary processes. For most laterally-located elements (and all lateral elements after the BH correction), we could not reject the null hypothesis of genetic drift. Prior to the BH correction, we could only reject drift in the Mt5 between Australopithecus africanus and H. naledi, where we find a pattern of greater between- than within- group variance on major PCs, and lower between- than within-group variance on minor PCs (Table 3). Table 3Results of regression analyses of logged between- vs within-group variance in individual elementsElementComparisonMatrixSlope (95% Confidence interval)R2t-statisticp-value (raw)p-value (BH)Mt1A. afarensis—StW 573Pooled2.92 (0.45–5.39)0.832.480.090.76A. afarensis—OH 8Pooled1.87 (0.92–2.82)0.932.900.060.64StW 573—A. africanusHuman0.62 (0.33–0.90)0.944.330.020.55A. africanus—H. naledi*Human0.35 (–0.38 to 1.08)0.182.190.070.64Mt2StW 573—OH 8Human1.91 (1.05–2.78)0.902.930.040.64Ar. ramidus—StW 573Human-0.54 (–2.11 to 1.03)0.293.120.050.64Mt5A. africanus—H. nalediHuman1.47 (0.97–1.97)0.872.240.060.64Medial cuneiformStW 573—H. nalediPooled–0.08 (–1.23 to 1.06)0.0032.180.060.64StW 573—OH 8Pooled–0.09 (–1.15 to 0.96)0.012.390.040.64Intermediate cuneiformH. naledi—H. sapiensPooled2.53 (0.79–4.27)0.882.790.070.64OH 8—H. sapiensHuman3.12 (2.38–3.86)0.977.950.0010.06*Comparison includes proximodistal length of the Mt1.Shown here are comparisons where drift was rejected (p < 0.05), or where drift may have been rejected (0.05