Postcranial studies of integration—or the coordinated variation of functionally or developmentally connected phenotypic traits (Olson and Miller, 1958; Wagner, 1984; Hallgrímsson et al., 2009; Klingenberg, 2014; Armbruster et al., 2014)—are uncommon in biological anthropology and evolutionary biology (Esteve-Altava, 2017). Moreover, such studies have generally focused on individual limb length measurements, rather than more detailed measurement systems that capture the three-dimensional (3D) shape in addition to length (though refer to the studies by Fabre et al., 2014; Conaway et al., 2018). Hallgrímsson et al. (2002), in a study on limb development in mice (Mus musculus) and rhesus macaques (Macaca mulatta), found stronger integration in both taxa between developmentally homologous limb elements (e.g., femur/humerus, tibia/radius, fibula/ulna, etc.) than in elements within limbs. Additionally, Schmidt and Fischer (2009) reported that limb proportions are largely conserved in a large sample of quadrupedal mammals, including several primate species, implying high levels of integration within limbs. However, they also reported slightly more variability in magnitudes of forelimb integration. They suggest that this may be due to the more diverse functional activities of the forelimb (foraging, feeding, infant carrying, etc.), particularly in primates. Slight increases in forelimb integration to accommodate arboreality along with nonlocomotor behaviors have been reported in other primate species as well (Rodman, 1979; Cant, 1988; Conaway et al., 2018). Additionally, Villmoare et al. (2014) found slightly reduced levels of within-limb integration compared with between-limb integration in vertical clinging and leaping strepsirrhines, relative to their more quadrupedal counterparts. Indeed, Fabre et al. (2014) found strong integration in the forelimbs of skunks, raccoons, and weasels, particularly in regions contributing to stability of the elbow joint (i.e., the distal humerus and proximal ulna), suggesting that even a small degree of functional independence of limbs may be detectable via the quantification of magnitudes of integration.

Since even small degrees of functional independence of limbs may be detectable in magnitudes of integration, organisms exhibiting more extreme functional divergence of forelimbs and hind limbs, i.e., brachiation in gibbons or bipedalism in humans, would be expected to exhibit magnitudes of within-limb integration comparable to or even stronger than those among developmentally homologous limb elements (Hallgrímsson et al., 2002). Most importantly, however, in order for a species to diverge away from a state of high integration among developmentally homologous limb elements, strong selection for reduced constraint would be required to overcome this canalizing hurdle (Hallgrímsson et al., 2002). In other words, we should expect to see lower overall integration in the postcranium of mammals whose limbs have diverged functionally to an extreme degree, particularly if the degree or nature of functional independence is variable.

In a study of primate limb proportions, Young et al. (2010) found a significantly reduced level of integration among developmental limb homologs in hominoids when compared to cercopithecoid and platyrrhine taxa. This finding is important because it suggests past selection among extinct hominoids for reduced constraint that may have allowed for the evolution of the diverse locomotor repertoires (including human bipedalism) and ensuing morphological diversity found in apes today (Young, 2006; Young et al., 2010; Pavlicev et al., 2010; Almécija et al., 2015). Indeed, Young et al. (2010) showed that partial correlations among adjacent hindlimb elements in humans (in terms of skeletal element proportions) were strong, whereas suspensory apes showed stronger correlations within forelimb elements. The relatively high integration of limb segments in quadrupedal primates may therefore be maintained to facilitate efficient locomotion on all four limbs, suggesting stabilizing selection for a likely ancestral mammalian condition (Lawler, 2008; Young et al., 2010; Rolian, 2014).

The expectation of low integration of the hominoid postcranium has not been universally corroborated, however. In a study of forelimb and hind limb shape covariance, Tallman (2013) found high covariation in hominoids and fossil hominins, particularly in the distal humerus and femur. Additionally, Marroig and Cheverud (2005) have suggested that decreased constraints on the evolution of limbs in hominoids may, somewhat paradoxically, slow evolutionary change rather than facilitate it. In other words, relaxed constraint may result in responses to selection in directions in morphospace that might otherwise have had little or no response due to higher integration of traits. In any case, the extent to which the findings of Young et al. (2010) are empirically supported when quantifying limb bone form using more complex higher dimension measurement systems (as opposed to measuring limb proportions) is currently unclear.

Indeed, there have been analyses of integration in more complex 3D structures such as the primate pelvis. In a study of crab-eating macaque (Macaca fascicularis) postcranial integration, Conaway et al. (2018) found that girdle elements (os coxa, scapula) had significantly lower integration than limb bones. The authors suggest that this may be due to lower redundancy of traits in their resampling protocol for girdle elements, as the sampled traits have a more complex 3D shape than limb bones, which have a simpler shape profile (Conaway et al., 2018). Additionally, previous work by Grabowski et al. (2011) and Grabowski (2013) has shown that the human pelvis is more evolvable, and therefore less integrated than other hominoid species Hansen (2003). Moreover, these results, as well as those presented by Mallard et al. (2017), showed that integration among different parts of the pelvis was low, perhaps as a reflection of constraints due to obstetrics and locomotion on its morphology (Grabowski et al., 2011, Grabowski, 2013).

Here we expand on previous work by quantifying postcranial morphological integration of four hominoid taxa, including anatomically modern humans (Homo sapiens), common chimpanzees (Pan troglodytes), western lowland gorillas (Gorilla gorilla), and lar gibbons (Hylobates lar), as well as two cercopithecoid species, crab-eating macaques (M. fascicularis) and vervet monkeys (Chlorocebus pygerythrus) as outgroups for comparison. We used these data to compare patterns of magnitudes of integration—or the magnitude of integration of a given trait relative to other traits—to address two hypotheses. First, we hypothesized that magnitudes of integration may be distinct in the postcranium of hominoids as compared to other catarrhine species (Hypothesis 1). To test this hypothesis, we made the following two predictions based on previous findings (e.g., Hallgrímsson et al., 2002; Young et al., 2010; Jung et al., 2021):

Prediction 1a: Hominoid taxa will have lower magnitudes of integration for anatomically defined modules relative to the two cercopithecoid taxa.

Prediction 1b: Hominoid taxa will have lower magnitudes of integration for developmentally homologous limb/girdle elements (e.g., femur/humerus etc.) than cercopithecoids.

Second, following Conaway et al. (2018), Grabowski et al. (2011) and Grabowski (2013), we hypothesize that magnitudes of integration of girdle elements (scapula, os coxa) will be lower, on average, than those for long bones in all taxa (Hypothesis 2). Inclusion of multiple taxa in the present study will allow us to test both the methodological suggestion of Conaway et al. (2018) that girdle element integration is affected less by trait redundancy and the biological conclusions of Grabowski et al. (2011), Grabowski (2013) and Mallard et al. (2017), namely, that integration of the human pelvis is low compared to other primates.

Recognizing morphological integration in the skeleton is challenging with simple analyses of trait covariance (Hallgrímsson et al., 2009). The sample analyzed here was selected and constructed to facilitate intertaxonomic analysis within a relatively small but variable clade. To address our hypotheses, similar methods to those utilized in the study by Conaway et al. (2018) were employed, wherein integration of modules, defined a priori based on anatomical and developmental criteria, was calculated using interlandmark distance data and the integration coefficient of variation (ICV; Shirai and Marroig, 2010).