1The positional repertoire of an individual consists of a combination of postural and locomotor modes whose frequency varies according to age, body mass and environmental conditions (Doran, 1992; 1997; Hunt, 1992; 1994; Wells and Turnquist, 2001; Druelle and Berillon, 2013; Sarringhaus et al., 2014). In palaeoanthropology, reconstructing palaeo-repertoires is a major issue, especially given the central role of bipedalism in defining the evolutionary trajectory of the hominin lineage (Lequin and Marchal, 2021). Comparative anatomical studies of extant primates, whose positional repertoires are well documented, suggest that fossil hominins – unlike extant humans – engaged in bipedalism in combination with other modes (for a recent review, see Stamos and Alemseged, 2023). These inferences are mainly based on the principle that a morphology resembling that of extant humans is typically interpreted as indicative of bipedalism, whereas more ape-like anatomical features are associated with alternative behaviours, such as climbing or quadrupedalism. The pelvis and the femur in particular provide some of the clearest examples of such functional interpretations, since the differences in their morphology in extant humans and non-human primates are very distinct. For instance, a wide, short ilium and laterally oriented ischium; or for the femur, a long neck, marked linea aspera and bicondylar angle, are associated with bipedalism.
2However, external morphology also reflects long-term adaptations, which are influenced by plesiomorphic characteristics in predominantly arboreal primates (Napier, 1967; Stanford, 2006; Thorpe et al., 2007; 2014; Senut, 2014). Furthermore, different morphologies can effectively accommodate multiple locomotor modes without sacrificing functionality (Murray, 2022): examples include the capacity for climbing of extant humans (Venkataraman et al., 2013) and the efficiency of some primates in actually performing bipedalism although they do not have the morphology which is presumed to be adapted to it. The absence of any perfect correspondence between external morphology and behaviour has already been underlined (Druelle and Berillon, 2014; Rosen et al., 2022). The approach via external morphology can be complemented by an alternative approach that would improve the ability to capture actual behaviours among individuals.
3The concept of bone functional adaptation, adapted from Wolff’s law (Roux, 1881; Wolff, 1892), holds that bone tissue is continuously reshaped by bone (re)modelling in response to mechanical stimuli that are primarily generated by the displacement of body mass and by muscular contractions (Currey, 2003; Ruff et al., 2006; Willie et al., 2020). The magnitude and distribution of cortical and trabecular bone can therefore provide indirect evidence of habitual activities, such as posture and locomotion. Among the most widely used parameters for assessing these structural adaptations are those derived from the cross-sectional geometry (CSG) of long bone diaphyses (e.g. Lieberman et al., 2004; Ruff, 2008). In vivo experimental studies have shown that the geometry of the cortical bone in a long bone diaphysis, taken perpendicularly to its longitudinal axis, responds to mechanical loading in a manner consistent with engineering beam theory (Alexander, 1981).
4The ratio of humeral to femoral bending strength and rigidity (calculated from cross-sectional properties) can be used, for instance, to identify whether an individual relies more on the fore limbs (humeral>femoral) or the hind limbs (humeral<femoral) (e.g. Schaffler et al., 1985; Kimura, 2003; Patel et al., 2013). This ratio has been used to put forward hypotheses on the positional repertoire of fossil hominins (Ruff, 2008; 2009; Ruff et al., 2016; Marchi et al., 2017).
5However, the cross-sectional properties of the long bone shaft may be considered as an imperfect proxy for inferring behaviours, since the orientation of the cross-sectional shape or cortical distribution don’t always align with the principal axis of loading (Lieberman et al., 2004). This limitation is especially relevant in palaeoanthropological contexts, where interpretations often rely on isolated bones due to the fragmentary nature of the fossil record, which makes it even more challenging to confidently reconstruct behaviour from structural properties alone. For example, concerning the femur, all extant hominoids except Homo sapiens display mediolaterally oriented cortical bone (e.g. Carlson, 2005; Puymerail, 2011; Sarringhaus et al., 2016; Nadell et al., 2021), despite their diverse positional repertoires, which include brachiation, suspension and knuckle-walking. In fossil hominins, early Homo sapiens, Neandertals and Homo erectus (s.l) all display substantial differences in their femoral cross-sectional properties, despite most likely sharing a common (single?) mode of locomotion (Puymerail et al., 2012; Trinkaus and Ruff, 2012; Rodríguez et al., 2018; Chevalier and de Lumley, 2022). Thus, the relationship between isolated bones and the positional repertoire of an individual has remained unclear, resulting in diverging interpretations for isolated fossil remains (Macchiarelli et al., 2020; Daver et al., 2022; Cazenave et al., 2024).
6Here, we aim to address to what extent femoral structural properties, considered in interaction with femoral and pelvic morphology, reflect positional repertoires in extant primates, and how these insights can inform reconstructions of fossil hominin locomotion. To address this question, we analysed femoral structural properties in several datasets including extant humans and non-human primate groups, as well as in fossil samples of the genus Homo. Each sample was chosen to explore a specific aspect of the relationship between femoral structure and locomotor behaviour, by making either intra- or interspecific comparisons.
7Also, while external morphology and internal bone structure have both contributed valuable insights into primate locomotor adaptations, morphological traits often reflect long-term evolutionary trends but may lack sensitivity to individual or behavioural variations. In contrast, the structural properties of long bones – although responsive to mechanical loading – are shaped by multiple factors, including phylogeny, life history and plastic responses, making their interpretation context-dependent. To address these limitations, we also adopted an integrative approach that considers both the external morphology and the internal structure of postcranial elements.
8The structural properties of the femoral diaphysis were assessed following a common protocol, detailed below, which provides a consistent methodological framework across all the sections of the manuscript. The pelvic and femoral morphologies were analysed using landmark-based geometric morphometrics (GMM) to explore patterns of integration between the pelvis and the femur. Depending on the specific questions addressed in each section of this study, additional analytical approaches were applied. These complementary methods are presented in the respective Materials and methods of the following sections.
9The structural properties were calculated from 20% (distal) to 80% (proximal) of the femoral biomechanical length (or diaphyseal length for individuals with unfused epiphyses) using the morphomap R-package (Profico et al., 2020). These properties include cross-sectional geometry (CSG) parameters such as the relative cortical area (%CA=cortical area of the section / total area of the section), second moments of area (Ix, Iy, J, Imax and Imin) and section moduli (Zx, Zy and Zp). The description and biomechanical interpretation of these CSG parameters are provided in table 1. In addition to CSG parameters, overall cortical distribution was assessed by means of cortical thickness maps (see methods in Bondioli et al., 2010 and Profico et al., 2020). Cortical thickness maps were generated for each femur and represent an individual-specific standardised scale from 0 (minimum thickness, in blue) to 1 (maximum thickness, in red). This approach enabled inter-individual comparisons regardless of body size, and emphasises heterogeneity in cortical distribution rather than absolute thickness measurements.
Table 1
Cross-sectional properties and their usual biomechanical interpretation. "Rigidity" indicates resistance to deformation while "strength" indicates resistance against failure |
Paramètres de géométrie de section et leur interprétation biomécanique classique : le terme "rigidity" traduit une résistance à la déformation, tandis que le terme "strength" traduit une résistance à la rupture
10Structural properties were calculated with 3D models of reconstructed femora from (micro)CT-scans using Avizo software v8.1 (Thermo Fisher Scientific). Details on the sample used for each section are provided in the supplementary material of this article (Supplementary tables S1-S4).
11In each section, CSG values were compared between groups using Wilcoxon-Mann-Whitney tests at every 5% interval along the biomechanical length. Additionally, standardised thickness values from the individual cortical thickness maps were analysed using principal component analysis (PCA) to assess inter-individual variation in cortical distribution.
12In this section, we explore how the structural properties of the femoral diaphysis relate to positional repertoires, by testing whether these properties are sensitive to changes in behaviour within a single species, using the example of the olive baboon (Papio anubis). For detailed information, see the published version of this chapter: Do femoral biomechanical properties follow locomotor changes in primates? An ontogenetic study of olive baboons (Papio anubis) in the American Journal of Biological Anthropology (Cosnefroy et al., 2022).
13The olive baboon (Papio anubis) is described as a primarily terrestrial quadruped (Rose, 1977; Hunt, 2016). However, like other primates, their positional repertoire changes during ontogeny with, in the firsts months of life, individuals who are carried by their mother or grasp her around the waist (Boulinguez-Ambroise et al., 2021). Once they achieve locomotor autonomy, they develop a varied positional repertoire that changes significantly with age (Druelle and Berillon, 2013; Anvari et al., 2014; Druelle et al., 2016). At around 0.5 years, quadrupedalism accounts for 60% of locomotion, alongside climbing (27.5%), leaping (5.4%), and a small proportion of bipedalism (1.9%) (Druelle et al., 2016). By 1.75 years, quadrupedalism increases to 76.7%, while climbing and leaping decline sharply. In adults, the repertoire is almost exclusively quadrupedal (98.7%), with minimal climbing (1%) and no observed leaping or bipedalism (Rose, 1977; Druelle and Berillon, 2013; Hunt, 2016).
14Concerning the structural properties of the femoral diaphysis, previous work focusing on the midshaft section has shown equivalent strength in femoral vs. humeral robusticity corresponding to quadrupedal behaviour (Ruff, 2003), posterior reinforcement of the cortical bone in adults (Puymerail, 2011) and a significant influence of heritability on these properties (Hansen et al., 2009). However, the link between femoral structural properties and locomotor changes in baboons remains poorly documented. Finally, these changes in behaviour make the olive baboon an appropriate model for testing the relationship between femoral structural properties and positional repertoire. We tested this relationship by cross-analysing structural and behavioural data for a single social group of baboons for which substantial knowledge is available for different stages of their development (e.g. Berillon et al., 2010; Druelle et al., 2017; Boulinguez-Ambroise et al., 2021).
15We studied a sample of 50 femora from 50 individuals housed at the CNRS primatology station (UPS 846) and representing 5 developmental stages based on thigh segment growth rate and locomotor changes (see Leigh, 2009; Druelle et al., 2016). These individuals included newborns (N=4, younger than two days), younglings (N=2, younger than 6 months), infants (N=8, 0.5 to 1.8 year), juveniles (N=6, 1.8 to 4.75 years) and adults (N=30, older than 4.75 years). To avoid any bias due to sexual dimorphism, we studied the structural properties of the femur in females except for two newborn males.
16The cross-sectional properties studied include the relative cortical area (%CA), the ratio of second moments of area relative to anteroposterior and mediolateral axes (Ix/Iy) and overall cortical distribution via standardised cortical thickness maps.
17Our results show that newborns and younglings differ from all the other groups for each structural property by having a higher %CA, increased anteroposterior cortical reinforcement (Ix/Iy>1) and distinct cortical distribution (figure 1). Given that these distinctions are mainly observed at midshaft, they probably result from the primary ossification spreading from the centre of the diaphysis, but also increased bone robusticity to avoid fractures in immature skeletons (Young et al., 2010). In older groups, we observed a shift with age of cortical reinforcement toward the proximal diaphysis, probably related to the continuous increase in body mass from younger to older individuals. This shift may reflect the growing mechanical role of the proximal femur in supporting body mass during locomotion, itself linked to the proximal migration of the thigh’s centre of mass as baboons get older (Raichlen, 2005; Druelle et al., 2017). However, structural properties do not significantly differ between infants and adults, despite differences in their positional repertoires.
Figure 1
Structural properties of the femoral diaphysis in the Papio anubis ontogenetic sample. Mean cross-sectional properties (CSG) by group with (a) relative cortical area (%CA) and (b) the ratio of anteroposterior/mediolateral second moment of area (Ix/Iy); (c) Results of a Principal Component Analysis based on cortical thickness maps data. Black: adults (N=30), red: juveniles (N=6), green: infants (N=8), blue: younglings (N=2), yellow: newborns (N=4) (modified from Cosnefroy et al., 2022) |
Propriétés structurelles de la diaphyse fémorale dans l’échantillon ontogénétique de Papio anubis. Moyenne des paramètres de géométrie de section (CSG) par groupe avec (a) l’aire corticale relative (%CA) et (b) le ratio des seconds moments d’aire antéropostérieur/ médiolatéral (Ix/Iy) ; (c) Résultats d’une Analyse en Composantes Principales basée sur les valeurs des cartographies d’épaisseur corticale. Noir : adultes (N=30), rouge : juvéniles (N=6), vert : enfants (N=8), bleu : jeunes enfants (N=2), jaune : nouveau-nés (N=4) (modifié d’après Cosnefroy et al., 2022)
18Our results thus indicate a conservative pattern in femoral structural properties once locomotor autonomy is achieved, as also observed in humans (Swan et al., 2020), despite further changes in behaviour. We interpret this as the rapid and early establishment of a specialised biomechanical pattern adapted to the adult baboon’s environment and daily activity substrate (mostly open or semi-arboreal grasslands, with an arboreal component that is not necessarily visited during daily activities, see Rose, 1977; Kunz and Linsenmair, 2007; Palombit, 2017; Elton and Dunn, 2020) that probably presents a selective advantage. Our results also indicate that femoral cortical distribution in the early stages closely resembles that of other catarrhines (Morimoto et al., 2012; 2018), suggesting a common organisation of cortical thickening of the femoral diaphysis among this clade.
19We conclude on the fact that the ontogeny of the femoral diaphysis structural properties in olive baboons seems to be governed by a set of developmental factors, of which potentially a large part is genetically determined, and do not accurately reflect changes in their positional repertoire.
20In this section, we explore how the structural properties of the femoral diaphysis relate to positional repertoires, by testing whether these properties are specific to different locomotor categories across catarrhine taxa. For detailed information, see the published version of this chapter: Phylogenetic and biomechanical influences in the structural pattern of the femoral diaphysis among catarrhines in the Revue de Primatologie (Cosnefroy et al., 2024a).
21Catarrhine primates exhibit a wide diversity of locomotor behaviour in their repertoires that includes several forms of quadrupedalism, suspension, leaping, brachiation, climbing, and even bipedalism (for a review, see Hunt, 2016).
22Biomechanical analysis of the femur through cross-sectional geometry has shown locomotor-related patterns in some primate groups. For instance, cercopithecoid leapers exhibit stronger anteroposterior bending rigidity than climbers (Burr et al., 1981; 1989; Ruff, 2002; Ruff et al., 2019), while terrestrial quadrupeds tend to have more balanced properties (Ruff, 2003; Puymerail, 2011; Cosnefroy et al., 2022). In great apes, however, femoral bending rigidity consistently displays a mediolateral orientation, in both knuckle-walkers and suspensory taxa (e.g. Carlson et al., 2006; Sarringhaus et al., 2016; Nadell et al., 2021), which differs from the anteroposterior bending rigidity translated by the anteroposterior (AP) cortical reinforcement observed in extant humans. Therefore, the relationship between behaviour and femoral biomechanics does not seem to be straightforward in all primates. Other studies among a variety of primate taxa have also pointed to the influence of phylogeny in shaping the femoral structure at both the intra- (Morimoto et al., 2011; Cosnefroy et al., 2022) and inter-species levels (Morimoto et al., 2012; 2018).
23The hypothesis that femoral cross-sectional properties are directly linked to the primary locomotor mode of a species remains uncertain and requires to be further tested at a broader scale than previously done. Here, we examine differences in femoral cross-sectional properties across different catarrhine genera, some of which share similar primary locomotor modes (mainly among quadrupeds), while others differ substantially in their locomotion, such as bipeds and brachiators.
24We studied diaphyseal cross-sectional geometry and cortical distribution on the femora of 127 adult catarrhines. The sample encompasses specimens from extant hominoids, including bipeds with Homo sapiens (N=32, 17F/15M), knuckle-walkers with Pan genus (N=28, 10F/14M/10ND), Gorilla genus (N=10, 3F/7M) and brachiators with hylobatids (N=4, 1F/3M) and extant cercopithecoids including plantigrade quadrupeds with Papio anubis (N=38, 29F/9M) and Macaca mulatta (N=15, 9F/6M). The CSG include %CA, Ix/Iy, and Imax/Imin, while cortical distribution was assessed using standardised cortical thickness maps. Given their small sample size, hylobatids were excluded from CSG statistical comparison.
25Our results show that within locomotor groups, CSG generally indicates greater variation than expected, including in the diaphyseal section considered, which highlights the importance of considering the whole diaphysis rather than just the midshaft section.
26Among quadrupeds, CSG analyses show that African apes have significantly lower %CA along the entire diaphysis compared to cercopithecoids (figure 2). In contrast, Ix/Iy and Imax/Imin reveal similarities in the distal diaphysis between Pan and macaques, which engage in more arboreal behaviour, such as vertical climbing, than gorillas and baboons (see similar results for the humerus in Patel et al., 2013). Hylobatids, meanwhile, show intermediate values across all CSG parameters and do not clearly differ from other taxa, regardless of locomotor behaviour. This may reflect a less distinct internal structure due to lower biomechanical loading of the femur in their arboreal lifestyle. Finally, and as expected, Homo sapiens, the only biped in the sample, is clearly distinguished from all other groups, especially at midshaft, by an anteroposteriorly reinforced cortical distribution (Ix/Iy>1).
Figure 2
Cross-sectional properties of the femoral diaphysis in the catarrhines sample. Percentages indicate the location of the cross-section on the biomechanical length of the femur. (left) Relative cortical area (%CA), (middle) ratio of anteroposterior/mediolateral second moment of area (Ix/Iy), (right) ratio of maximal/minimal second moment of area (Imax/Imin) (modified from Cosnefroy et al., 2024a) |
Paramètres de géométrie de section de la diaphyse fémorale dans l’échantillon de catarrhiniens. Les pourcentages indiquent la position de la section transversale par rapport à la longueur biomécanique du fémur. (gauche) Aire corticale relative (%CA), (milieu) ratio des seconds moments d’aire antéropostérieur/médiolatéral (Ix/Iy), (droite) ratio des seconds moments d’aire maximal/minimal (Imax/Imin) (modifié d’après Cosnefroy et al., 2024a)
27In non-human primates, our results indicate that the locomotor signal in the femoral structure is weaker than expected. This suggests that the biomechanical impact of quadrupedalism or suspensory behaviours has limited influence on the shaping of femoral structural properties. The Principal Component Analysis (PCA) of cortical thickness distribution separates cercopithecoids from hominoids (figure 3), with the hominoid cluster including both knuckle-walker groups and brachiators. This pattern is therefore more likely to reflect phylogenetic connections than functional differences, highlighting a strong genetic influence that may obscure signals of locomotor adaptation (e.g. Hansen et al., 2009; Morimoto et al., 2012; Cosnefroy et al., 2022).
Figure 3
Results of a Principal Component Analysis based on cortical thickness maps data of the catarrhine sample (modified from Cosnefroy et al., 2024b) |
Résultats d’une Analyse en Composantes Principales basée sur les valeurs des cartographies d’épaisseur corticale (modifiée d’après Cosnefroy et al., 2024b)
28The absence of a direct relationship between femoral structural properties and positional behaviours in extant primates, both within and between species, has led us to conclude that diaphyseal structure alone does not fully capture adaptations to locomotor behaviour. Here, we investigate how these properties can be integrated with external morphological features of the femur and the pelvis, and how this integration can shed light on locomotor adaptation. For detailed information, see the published version of this chapter in the American Journal of Biological Anthropology (Cosnefroy et al., 2024b).
29Integration refers to the coordinated variation of anatomical traits, and provides key insights into evolutionary and functional relationships within the skeleton, which are often limited when based on a single trait (Olson and Miller, 1958). This approach is based on semiautonomous units, or modules, that present a higher magnitude of integration among themselves than with other parts of the organism (Cheverud, 1996; Wagner, 1996).
30In primates, studies of morphological integration between limb elements have, for instance, shown stronger integration in cercopithecoids and quadrupeds compared to hominoids and climbers, with a low level of integration being interpreted as a consequence of functional dissociation between the upper and lower limbs during arboreal locomotion (Young et al., 2010; Agosto and Auerbach, 2022). More generally, a reduced integration is generally associated with greater phenotypic plasticity (Wagner et al., 2007; Hansen and Houle, 2008).
31Here, we measured morphological integration between the pelvis and the femur of catarrhine primates that exhibit distinct positional behaviours. Our aim was to quantify the magnitude of coordinated variation within the femur-pelvis module, which is a key feature in locomotor adaptation among hominins (Ruff, 1995). Additionally, we investigated what we called the morpho-structural integration between the external morphology of the femur or pelvis and the cross-sectional properties of the femoral diaphysis, in order to provide a more comprehensive understanding of skeletal adaptation by combining both external and internal traits.
32The sample consists of 55 femora and hemi-pelvises from Homo sapiens (N=19, 9F/10M), Pan genus (N=22, 8F/10F/4ND) and Papio anubis (N=14, 12F/2M).
33Morphological integration between external femoral and pelvic morphology was measured using the Covariance Ratio (CR) with the geomorph R-package (Adams, 2016). CR was performed in each group on Procrustes coordinates of 3D landmarks set for both the femur and the pelvis (figure 4). We used the CR effect size difference to assess differences in magnitude of morphological integration between pairs of groups. Morpho-structural integration between femoral or pelvic morphology and CSG of the femoral diaphysis was measured using 2-block partial least squares (2b-PLS) analyses (Rohlf and Corti, 2000). The morphological block was defined either by Procrustes coordinates directly, or by PC scores (typically PC1 and PC2) derived from a principal component analysis (PCA) based on these coordinates. The structural block consisted of CSG variables (either Ix/Iy or Imax/Imin) measured along the entire diaphysis (61 sections) or from selected regions (15 sections per shaft segment).
34All analyses of morphological and morpho-structural integration were repeated after removing allometry. This was done by calculating the centroid size of each landmark configuration (for both femur and pelvis) and extracting the residuals from multivariate regressions of Procrustes coordinates on the log-transformed centroid size (Monteiro, 1999).
35Our findings show distinct patterns of morphological and morpho-structural integration within the sample (figure 4): the group that presents the highest magnitude of morphological integration also shows the lowest morpho-structural integration – and conversely, the group with the strongest morpho-structural integration displays the weakest morphological integration.
Figure 4
Schematic representation of patterns of integration in the femur-pelvis module of the sample. White boxes indicate analyses of external morphology, while black boxes indicate analyses of the internal structural properties of the femoral diaphysis. Morphological integration, assessed by the Covariance Ratio (CR), is highest in Papio anubis, intermediate in the Pan genus and lowest in the Homo sapiens sample. Morpho-structural integration, assessed by two-blocks Partial Least Squares (PLS) analysis, is highest in Homo sapiens, intermediate in the Pan genus and lowest in the Papio anubis sample |
Représentation imagée des modèles d’intégration du module fémur-pelvis au sein de l’échantillon. Les encadrés blancs indiquent les analyses de la morphologie externe, tandis que les encadrés noirs correspondent aux analyses des propriétés structurelles internes de la diaphyse fémorale. L’intégration morphologique, évaluée par Covariance Ratio (CR), est la plus élevée chez Papio anubis, intermédiaire chez le genre Pan et la plus faible au sein de l’échantillon Homo sapiens. L’intégration morpho-structurelle, évaluée par une analyse aux moindres carrés (PLS) à deux blocs, est la plus élevée chez Homo sapiens, intermédiaire chez le genre Pan et la plus faible au sein de l’échantillon Papio anubis
36Morphological integration between the femur and pelvis was higher in baboons – primates with a stereotyped, quadrupedal locomotion – than in the Pan genus and extant humans, suggesting strong functional coordination between these elements. In contrast, extant humans showed the lowest morphological integration, probably reflecting different selective pressures on the femur (locomotion) and pelvis (locomotion and parturition) (Grabowski, 2013; Huseynov et al., 2016; Ruff, 2017). The Pan sample displayed intermediate values.
37Morpho-structural integration was higher in humans – particularly between femoral morphology and femoral CSG – than in Pan and baboons. This probably reflects complex adaptation to habitual bipedalism involving both external morphological and internal structural traits, while emphasising the central role of the femur in supporting these adaptations.
38Although baboons exhibited high morphological integration, they showed the lowest morpho-structural integration. This suggests that their adaptation to stereotyped quadrupedal locomotion relies primarily on external morphological features, which may also explain the low variability observed in their femoral CSG (Cosnefroy et al., 2022). In Pan, the morpho-structural integration patterns were more variable (depending on whether allometry was removed or not) and generally intermediate, potentially reflecting their broader and more diverse positional repertoire.
39Interestingly, morpho-structural integration based on PC scores and femoral CSG revealed that certain morphological traits, as captured by the PC scores, tend to co-vary with structural properties at specific locations along the diaphysis. For example in humans, the proximal 3/4 of the femoral shaft – where they differ markedly from Pan and baboons due to an anteroposterior cortical bone distribution (Cosnefroy et al., 2024a) – shows strong integration with morphological traits such as the neck-shaft angle, femoral head orientation and diaphyseal curvature.
40While caution is needed in interpreting the results for morpho-structural integration – given that this is the first study to assess such integration, and that a high variable-to-sample (p/N) ratio may weaken statistical power – our findings suggest complex integration patterns that probably reflect locomotor adaptations within the femur-pelvis module. These adaptations may rely either on external morphology alone or on coordinated morpho-structural relationships. We further suggest that assessing morpho-structural integration between PC scores and CSG is a promising approach for exploring complex anatomical variation, offering valuable insights into the form-function relationship for future research (Murray, 2022).
41While the other sections of this paper focus on identifying morphological and structural differences among extant primates with varying positional behaviours, here we approach the question from the opposite perspective. We investigate differences in femoral morphology and structure between two taxa – Homo sapiens and Neandertals – that share a common and singular positional mode: bipedalism.
42Neandertals and Homo sapiens are both obligate bipedal hominins (see locomotor grade criteria in Stamos and Alemseged, 2023). However, they present numerous differences in the morphological and structural features of their locomotor skeleton (e.g. Marchal, 2000; Weaver, 2009; Chapman et al., 2018). The femur in particular reflects these differences (e.g. Twiesselmann, 1961; Trinkaus and Ruff, 1999; 2012).
43One of the main differences in femoral anatomy between Neandertals and Homo sapiens lies in the orientation of their respective cortical bone distribution within the diaphysis. CSG in Neandertals typically show a mediolateral (ML) cortical distribution, which resembles the plesiomorphic condition seen in earlier Pleistocene hominins (e.g. Puymerail et al., 2013; Chevalier and de Lumley, 2022). In contrast, Homo sapiens exhibit a more anteroposterior (AP) cortical distribution, which represents an autapomorphic feature of the species (Trinkaus and Ruff, 2012). This distinction has been hypothesised to be the by-product of different proportions in the pelvis (Ruff, 1995; Trinkaus and Ruff, 2012), which is broader in Neandertals than in Homo sapiens (Marchal, 2000).
44Another main distinction between the two taxa lies in their respective femoral curvature in the sagittal plane, with greater curvature in Neandertals compared to Homo sapiens (Shackelford and Trinkaus, 2002; De Groote, 2011; Chapman et al., 2018). However, differences within Homo sapiens populations complicate this distinction, as variations in Pleistocene Homo sapiens and Neandertals tend to overlap, raising the question of whether this distinction reflects a functional difference between highly mobile hunter-gatherer and sedentary lifestyles (Shackelford and Trinkaus, 2002). Moreover, the biomechanical significance of long bone curvature remains in debate, although some authors, based on experimental studies, suggest that a sagittal curvature helps to convert bending stresses into biomechanically more favorable axial compression (Bertram and Biewener, 1988; Taylor et al., 1996). In this case, sagittal curvature can be seen as a mechanism to resist bending stresses, similar to the role of cortical bone distribution in the diaphysis (Ruff, 2008). By extension, these two features may be biomechanically interconnected.
45Here, we investigated the reasons behind differences in femoral diaphyseal curvature and cortical bone distribution between Neandertals and Homo sapiens, by exploring whether these traits represent alternative biomechanical strategies or integrated traits of a coordinated morphological response to manage bending loads during bipedal locomotion. To address this, we analysed curvature and cortical distribution and their relationship along the femoral shaft in Neandertals and in several Homo sapiens samples with varying mobility levels. We tested the correlation between the two traits within each group, under the hypothesis that a negative correlation would indicate biomechanical independence, while a positive correlation would suggest functional integration as part of a shared adaptive response. This approach allowed us to evaluate whether Neandertals and Pleistocene Homo sapiens relied on distinct or comparable long-term strategies for bipedalism, and to explore whether mobility-related variation within Homo sapiens reflects the emergence of specific adaptive trends in the species.
46The sample consists of 139 femoral diaphyses from 123 individuals, covering Neandertals (N=9), Pleistocene Homo sapiens (PHS, N=6), Mesolithic hunter-gatherers (N=9), Neolithic/Bronze Age/Mediaeval agro-pastoralists and farmers (NBMA, N=61) and extant humans (N=38).
47Cortical distribution was measured by using the following CSG: the ratio of mediolateral and anteroposterior second moments of area (Ix/Iy). In addition, we analysed %CA to evaluate not only the distribution but also the amount of cortical bone and its potential correlation with curvature. All CSG measurements were taken at each percentile from 20% to 80% of the biomechanical length when available. For incomplete femora, biomechanical length was either obtained from the literature or estimated using anatomical landmarks following the guidelines of Trinkaus and Ruff (2012) for fossil hominins.
48Sagittal curvature was assessed using a novel approach adapted from previous studies on primates and humans (Yamanaka et al., 2005; Dupej et al., 2017; Imamura et al., 2021). The morphomap R-package calculates CSG based on the centroid of the cortical area (CCA). The successive CCA coordinates, which define the neutral axis of the diaphysis, were used to measure sagittal curvature in the sagittal plane. At each point of interest (C), curvature was calculated as the orthogonal distance [CH] from C to the midpoint H of a segment defined by two equidistant CCA points (A and B), scaled by the segment length (figure 5; table S5). Therefore, the closer the CH distance is to 1, the greater the sagittal curvature of the femur. This method allows for an overall measurement of the diaphyseal curvature for complete femora, but also locally within a 20% diaphyseal segment (from -10% to +10% relative to C). This method has two advantages: it measures curvature at the same cross-sectional levels where CSG are calculated – allowing for local comparisons between curvature and CSG – while also enabling analyses on incomplete femora.
Figure 5
Measurement of the sagittal curvature defined by the successive (y) coordinates of C along the longitudinal axis. B and A are always equidistant from C and can be shifted towards the longitudinal axis; with [AC]=10% of length, [CB]=10% of length and [AB]=20% of length. As an example, the figure displays local sagittal curvature at 50% of length. The [CH] length is not scaled on the figure. Red contours of cross-sections indicate the periosteal surface; green contours indicate the endosteal surface |
Mesure de la courbure sagittale définie par les coordonnées successives (y) de C le long de l’axe longitudinal. B et A sont toujours équidistants de C et peuvent être déplacés vers l’axe longitudinal ; avec [AC]=10 % de la longueur, [CB]=10 % de la longueur et [AB]=20 % de la longueur. À titre d’exemple, la figure illustre la courbure sagittale locale à 50 % de la longueur. La longueur [CH] n’est pas mise à l’échelle sur la figure. Les contours rouges des sections transversales indiquent la surface périostéale, et les contours verts indiquent la surface endostéale
49The correlation between curvature and CSG was assessed at each diaphyseal level by using Pearson correlation tests with a Bonferroni correction, comparing local curvature and CSG values at the same cross-sectional locations.
50Our results confirm that Homo sapiens exhibit a more extensive anteroposterior (AP) cortical distribution than Neandertals, particularly in Pleistocene Homo sapiens (PHS), at nearly every location along the diaphysis (figure 6). No differences were found in %CA, indicating that while the total amount of cortical bone is similar between Pleistocene specimens of the two taxa, they differ in how this bone is distributed. Within Homo sapiens, a temporal decrease in AP cortical distribution was observed: PHS show higher AP distribution than Mesolithic individuals, who in turn exhibit higher values than NBMA and extant humans. This trend correlates with the decrease in mobility from late Upper Palaeolithic hunter-gatherers to chronologically more recent populations (e.g Ruff et al., 1984; Larsen, 1995; Stock and Pfeiffer, 2001; Holt, 2003; Sparacello and Marchi, 2008; Shaw and Stock, 2011; 2013; Davies and Stock, 2014; Ruff, 2018; Holt and Whittey, 2019; Sparacello et al., 2020). The %CA tend to be stable in Homo sapiens, with only the extant sample showing an increase when compared to NBMA. Considering that the %CA is not a perfectly reliable biomechanical proxy for bone loading (Ruff, 2008) and is influenced by multiple factors such as age, sex, hormonal regulation and nutrition (Cowgill et al., 2023) – none of which are controlled for in this study – this trend remains difficult to interpret. Nonetheless, we can hypothesise that the higher cortical bone proportion in extant populations compared to the archaeological sample reflects a reduced emphasis on energetic economy in bone structure, probably associated with fewer evolutionary and/or developmental constraints in modern humans.
Figure 6
Comparison of local sagittal curvature (left) with the anteroposterior to mediolateral second moment of area ratio (Ix/Iy) and the relative cortical area (%CA), measured at the same diaphyseal location across groups: Neandertals (N=9), Pleistocene Homo sapiens (PHS, N=6), Mesolithic (N=9), Neolithic/Bronze/Middle Ages (NBMA, N=61), and extant humans (N=38). * indicates a significant difference in the Mann-Whitney-Wilcoxon test with p<0.05 |
Comparaison de la courbure sagittale locale (à gauche) avec le ratio des seconds moments d’aire antéropostérieur/médiolatéral (Ix/Iy) et l’aire corticale relative (%CA), mesurés au même niveau diaphysaire pour chaque groupe : Néandertaliens (N=9), Homo sapiens pléistocènes (PHS, N=6), Mésolithique (N=9), Néolithique/âge du Bronze/Moyen Âge (NBMA, N=61) et humains actuels (N=38). * indique une différence significative du test Mann-Whitney-Wilcoxon avec p<0.05
51Concerning curvature, Neandertals exhibit a higher overall curvature, increased distal local curvature, and a more distal apex of curvature compared to all Homo sapiens samples. Among the latter, PHS, Mesolithic and NBMA individuals show similar femoral curvature patterns, all generally more curved than in extant humans (Cosnefroy, 2024; figure 6). Within Homo sapiens, the Extant sample exhibits a lower curvature locally (figure 6), which can also be hypothesised as a reduction in bending stress in modern populations. However, the absence of clear differences within archaeological and fossil Homo sapiens samples suggests that, although mobility behaviours (and therefore biomechanical stresses on the lower limb) tend to decrease over time, femoral curvature does not seem to be the trait that reflects these behavioural changes.
52The correlation tests indicate very few correlations between CSG and local curvature in the entire diaphysis at the intra-group level in all samples (table 2). Therefore, based on our hypothesis, our results do not statistically support the hypothesis of a direct equilibrium between cortical distribution and sagittal curvature, suggesting that local variations in curvature may not function as an analogous biomechanical response to bending loads, as established for cortical distribution. However, it should be noted that groups with greater AP cortical distribution tend to exhibit a lower curvature, indicating potential alternative strategies: AP reinforcement in Homo sapiens (especially PHS) and high sagittal curvature in Neandertals. The Amud 1 Neandertal, with its unusually high AP distribution and low curvature, also supports this pattern.
Table 2
Pearson’s correlation test with Bonferroni correction between local curvature and cross-sectional properties (CSG) along several locations of the femoral diaphysis. Significant correlations: adjusted p-value<0.05* |
Test de corrélation de Pearson avec correction de Bonferroni entre les valeurs de courbure locale et les paramètres de la géométrie de section (CSG) à différents niveaux de la diaphyse fémorale. Corrélations significatives : valeur-p ajustée<0,05*
53These findings show that, despite having obligate bipedal locomotion in common, late Pleistocene Homo groups certainly employed different biomechanical strategies to achieve it, relying on cortical distribution in Homo sapiens and increased femoral curvature in Neandertals. However, it remains unclear whether these differences represent adaptive responses to distinct loading regimes in the two taxa, driven by (1) variations in mobility and environmental factors, such as terrain and climate (e.g. Lieberman, 2002; Weaver, 2003; Marchi et al., 2006; Stock, 2006; Shaw and Stock, 2013; Cowgill, 2014; Holt and Whittey, 2019), (2) constraints imposed by external morphologies like the pelvis (Ruff, 1995) and knee (e.g. Trinkaus and Rhoads, 1999; Trinkaus, 2000; Rosas et al., 2020; Cazenave and Radovčić, 2023), or (3) the influence of thigh musculature, which largely acts on cortical distribution (Duda et al., 1998) and which has been suggested to contribute to increased curvature in Neandertals (Chapman et al., 2018).
54We can also hypothesise that the biomechanical distinctions between Pleistocene Homo sapiens and Neandertals reflect differences in loading regimes, potentially linked to distinct types of mobility behaviour between hunter-gatherer groups of the two species (Lieberman et al., 2004). Alternatively, as suggested by Kubicka et al. (2018), biomechanical differences may result from differences in skeletal plasticity between Neandertals and Homo sapiens, implying that both species could have experienced the same loading behaviours but responded differently at the physiological level.
55Finally, these distinctions may simply reflect the separate evolutionary histories of the two lineages, representing heritable features rather than functional adaptations.
56The structural properties of the femur have long been described as providing critical insights into how the hind limb adapts to mechanical loads during locomotion (Ruff et al., 2006). However, this work highlights the limitations of relying solely on the cross-sectional properties of a single bone to reflect the actual behaviours of individuals or primate groups, especially when their positional repertoire includes a variety of locomotor modes. Therefore, these properties should not be considered a perfectly accurate proxy for reconstructing the past behaviours of fossil hominins. In particular, measuring cross-sectional properties along the entire femoral diaphysis allowed us to identify several significant local differences in the diaphysis of primates that share similar positional repertoires.
57The analysis of cortical distribution using cortical thickness maps enabled the identification of group-specific features and suggested that overall cortical distribution patterns may be influenced more by phylogeny than by mechanical loading (Morimoto et al., 2011; 2012). This supports the idea that bone structure not only reflects functional adaptation but also carries long-term evolutionary signals inherited from ancestors. By extension, we believe that such structural patterns could be used for recognising phylogenetic relationships and for the taxonomic attribution of isolated fossil remains (Bleuze, 2022).
58With a view to further exploring the link between cortical distribution and locomotor adaptations, experimental studies could offer valuable insights into how strains produced by specific behaviours relate to the local distribution of cortical bone along the diaphysis (e.g. Swartz et al., 1989; Barak et al., 2011; Zani et al., 2015). To address this question, we turned to finite element (FE) modelling, a computational method originally developed in engineering that divides complex structures into small elements to simulate their mechanical response to a priori defined forces. This approach is particularly relevant in palaeoanthropology since it can be applied to test loading regimes on virtual bones, including fossils, without destructive manipulations. We explored this potential by simulating simplified mechanical constraints and comparing the resulting strain distribution with the cortical bone distribution observed in the femoral diaphysis of an olive baboon individual, using a specimen-specific FE model of its femur and comparing it to its cortical thickness distribution (figure 7). Although exploratory, this approach illustrates how FE modelling can bridge the gap between observed structural patterns and the mechanical demands of locomotion, and lays the groundwork for more refined analyses using realistic loading conditions.
Figure 7
Comparison of (a) cortical thickness distribution with standardised values ranging from 0 (thinnest point, in blue) to 1 (thickest point, in red), and (b) Von Mises strain distribution (with high strain in red and low strain in blue) in a finite element model. Both analyses were performed on the V935E Papio anubis right femur |
Comparaison entre (a) la distribution de l’épaisseur corticale avec de valeurs normalisées de 0 (point d’épaisseur le plus faible, en bleu) à 1 (point d’épaisseur le plus fort, en rouge) et (b) la distribution des contraintes Von Mises (fortes contraintes en rouge et faibles contraintes en bleu) dans un modèle par éléments finis. Les deux analyses ont été réalisées sur le fémur droit du spécimen Papio anubis V935E
59Our results were not directly interpretable as the simulated strains appeared to be overly simplified, in particular because they did not account for muscular contractions, which represent a significant portion of the constraints experienced during gait (Duda et al., 1998). A more accurate approach would be to simulate the strains associated with an actual gait cycle of an individual and apply these dynamic constraints in multiple phases of the cycle to the finite element model. This would make it theoretically possible to identify regions of the diaphysis where cortical reinforcement aligns with the strains experienced. Such simulations could be incorporated using a musculoskeletal model of the individual (of any species), following the "patient-specific" approach commonly used in biomechanical and clinical research (e.g. Wu et al., 2016; Song et al., 2019) as well as in studies of non-human primate locomotion (e.g. Ogihara et al., 2010; O’Neill et al., 2022). This approach could help to develop a precise definition of the relationship between cortical reinforcement and mechanical loading and identify the most informative diaphyseal regions.
60Finally, by demonstrating how external and internal traits integrate in order to provide a more comprehensive understanding of skeletal adaptations, this work highlights the need for an integrative framework to better assess form-function relationships (Murray, 2022). For example, the analysis of morpho-structural integration revealed that the relationship between external morphology and structural properties in the hind limb is not uniform across primates and may reflect distinct biomechanical adaptations. Similarly, when taken alone, the variability in the orientation of cortical distribution in the femoral diaphysis among fossil Homo is difficult to interpret in relation to locomotor behaviours (Puymerail et al., 2012; Trinkaus and Ruff, 2012; Chevalier and de Lumley, 2022). However, when analysed together with another feature such as the femoral curvature, it allowed us to put forward a scenario for the emergence of contemporaneous adaptive strategies in Late Pleistocene Homo, with the examples of Pleistocene Homo sapiens and Neandertals. This highlights the importance of considering multiple traits simultaneously and exploring how their connections can shed light on complex skeletal adaptations.
61Supplementary material Table S1 [https://journals.openedition.org/bmsap/16350?file=1]
62Detailed list of the specimens studied in the section entitled Variability of femoral structural properties at the intra-specific level
63Supplementary material Table S2 [https://journals.openedition.org/bmsap/16351?file=1]
64Detailed list of the specimens studied in the section entitled Diversity of femoral structural properties at the inter-specific level
65Supplementary material Table S3 [https://journals.openedition.org/bmsap/16358?file=1]
66Detailed list of the specimens studied in the section entitled The integration of (external) morphological and (internal) structural properties
67Supplementary material Table S4 [https://journals.openedition.org/bmsap/16364?file=1]
68Detailed list of the specimens studied in the section entitled Diversity of femoral morphology and structure in bipeds within one positional mode: a focus on Late Pleistocene hominins
69Supplementary material Table S5 [https://journals.openedition.org/bmsap/16365?file=1]
70This table presents a template for sagittal curvature analysis. To use it, please follow the following instructions:
- specify the femoral biomechanical length in the A2 case;
- fulfill the D column with the y coordinates of the center of area (C) point. This is obtained by using the morphomapCSG function of the morphomap R-package (Profico et al., 2020).
71The H column (CH_global) reports the measurement of overall sagittal curvature at each cross-sectional level. It is calculated along the diaphysis from 20% to 80% biomechanical length. In this case, point A is fixed at 20% and point B at 80% of the total length (see figure 5). Values are expressed in millimetres (mm).
72The L column (CH_local) reports the measurement of the local sagittal curvature at each cross-sectional level. It is calculated on 20% length segments. In this case, points A and B are positioned 10% of length away from point C (distally for A, proximally for B), and the [AB] segment moves along the shaft as the calculation is repeated across the length.
Acknowledgements: I would like to thank the Société d’Anthropologie de Paris for the honour of awarding me its prize for the year 2024. I am deeply grateful to my thesis supervisors, François Marchal and Gilles Berillon, for their confidence in me to carry this project forward. My heartfelt thanks go to all those who supported and contributed to this work – whether by co-authoring articles, providing access to data and osteological collections, offering technical assistance or simply by their encouragement and guidance. I am especially grateful to Isabelle de Groote, Kristian Carlson, Isabelle Crevecoeur, Amélie Beaudet, Lionel Tholon, Camille De Becdelièvre, Laurence Bellaiche, Pauline Brige, Laura Buck, Robert Carlier, Cyrille Cazeau, Tara Chapman, Kathia Chaumoître, Tony Chevalier, Guillaume Daver, François Druelle, Emmanuel Gilissen, Anna-Marria Kubicka, Franck Lamberton, Hila May, Antoine Perrier, Antonio Profico, Sofija Stefanovic, Jean-Christophe Theil, Adrien Thibeault, Nicole Torres-Tamayo, Christine Verna, and Sebastien Villotte. I also wish to thank the institutions and facilities that provided access to data, imaging and collections: the Centre Européen de Recherche en Imagerie Médicale (Marseille, France), Voxscan (Dommartin, France), the Hôpital Raymond-Pointcarré (Garches, France), the Cyceron platform (Caen), the Centre d’Imagerie Médicale Bachaumont (Paris, France), the Royal Museum of Central Africa (Tervuren, Belgium), the Pôle Imagerie Médicale de l’Assistance Publique des Hôpitaux de Marseille (Marseille, France), the University of Belgrade, the Neanderthal Museum, the Royal Belgian Institute for Natural Sciences, the Ulmer Museum, the Museum für Ur- und Frühgeschichte Thüringens (Weimar), the Muséum National d’Histoire Naturelle (Paris), the Museum of the First Piasts at Lednica (Poland), the Open Research Scan Archive of the Penn Museum, the French National Research Agency (ANR-15-CE33-0004 Gravett’os), the Tel Aviv University, the Jagiellonian University and the CERMEP in Tautavel. Finally, my thanks for their financial support to Aix-Marseille University’s Doctoral School 251, ANR HoBis and IRN Bipedal Equilibrium.