Comparative kinetics of humans and non-human primates during vertical climbing Free

Climbing within an arboreal milieu is recognized as a key adaptive trait of the primate order (Cartmill, 1974a, 1985; Fleagle et al., 1981; Granatosky, 2018; Preuschoft, 2002). Indeed, the origins of the primate lineage have been traced to fine-branch arboreal experts specializing in the navigation of these complex three-dimensional environments (Cartmill, 1972; Schmitt and Lemelin, 2002; Toussaint et al., 2020). This ecological niche necessitates the ability to scale vertical supports of unpredictable sizes, a biomechanically challenging task that imposes numerous demands including requiring the individual to cling to a vertically oriented surface while actively propelling their body weight upwards against the force of gravity. This task is made more difficult by the generally large body sizes of primates compared with those of most specialized arboreal vertebrates (Cant, 1992; Mittermeier et al., 2013), and by the absence of claws (a mechanism that enables many arborealists to passively cling while supporting their body weight) in almost all primate families (Cartmill, 1974b, 1979).

Several previous attempts have been made to understand vertical climbing in non-human primates (e.g. Druelle et al., 2024; Granatosky et al., 2019; Hanna, 2006; Hanna and Schmitt, 2011; Hirasaki et al., 1993; Isler, 2002, 2004, 2005; Neufuss et al., 2018; Samuel et al., 2018; Schoonaert et al., 2016; Wunderlich and Ischinger, 2017). However, fewer studies have quantified the force profiles associated with vertical climbing in primates. Black-handed spider monkeys (Ateles geoffroyi) were observed to employ a hindlimb-driven climbing gait in which hindlimb propulsive forces were significantly greater than those generated by the forelimb (Hirasaki et al., 1993). By contrast, Japanese macaques (Macaca fuscata) were reported to balance propulsive forces equally between the hindlimb and forelimb (Hirasaki et al., 1993), though this trend is not ubiquitous among Macaca as long-tailed macaques (Macaca fascicularis) demonstrate a hindlimb-driven gait similar to Ateles (Hanna and Schmitt, 2011), likely attributable to longer tails shifting the center of mass (COM) of the long-tailed macaques caudally (Lemelin and Schmitt, 2004). The most comprehensive study to date was presented by Hanna et al. (2017) and analyzed kinetic profiles from eight taxa, ranging in size from ∼200 g (Loris tardigradus) to ∼9 kg (M. fascicularis) and spanning strepsirrhines, platyrrhines and catarrhines. Of these, five species (Cheirogaleus medius, Saimiri sciureus, Aotus nancymaae, Eulemur mongoz and M. fascicularis) exhibited hindlimb-driven propulsion, one species (Daubentonia madagascariensis) demonstrated a balanced hindlimb:forelimb ratio similar to M. fuscata, and two lorisid taxa (L. tardigradus and Nycticebus pygmaeus) expressed a novel kinetic profile in which the forelimb generated the majority of propulsive forces during climbing (Hanna et al., 2017). In the normal plane, all species are reported to exhibit pulling and pushing forces in both the forelimb and hindlimb – a finding that is not seen in the forelimbs of geckos (Autumn et al., 2006), frogs (Young et al., 2023b) or sloths (Young et al., 2023a), or in the beaks of tripedal parrots (Young et al., 2022).

Unsurprisingly, very little is also known about the kinetics of human climbing; yet, as committed terrestrial bipeds with a long evolutionary history of movement in the trees, humans represent a fascinating case study in the context of arboreal ability. Certainly, humans are capable of effectively climbing in an arboreal context, as evidenced by numerous ethnographical accounts of forest foraging communities (Demps et al., 2012; Dutta et al., 1985; Endicott, 1979; Frisbie, 1971; Ichikawa, 1981; Kaplan et al., 1985; Kitanishi, 1996; Marlowe, 2004, 2010; Valli and Summers, 1988). Moreover, the rising frequency of free-solo ascents of cliff faces >1 km further proves humans to be similarly adept climbers even in non-arboreal contexts (Dybiec et al., 2021; Sparks, 2016).

Despite these abilities, humans are generally considered to be anatomically maladapted for climbing. Indeed, paleoanthropological reconstructions of climbing abilities – particularly in the context of australopithecines and other early hominins – typically utilize a ‘human-like’ ankle and foot as evidence that a hominin could not climb trees, or else did so poorly (Kraft et al., 2014; Latimer, 1991). As an activity, human climbing has also been related to high incidences of injuries, particularly to the forelimbs (Backe et al., 2009). This observation may be attributed to the propensity of novice climbers to overly rely upon their forelimbs to pull themselves upwards, as opposed to driving themselves upwards from their legs – a behavior which may increase the risk of forelimb injuries (Horst, 2019; Rooks, 1997). Additionally, as obligate bipeds, humans regularly load their hindlimbs with their full body weight, yet the forelimb experiences such loads far less regularly and thus may be maladapted to handle high loading magnitudes.

Biomechanically, several traits have been hypothesized to typify proficient climbing. Firstly, propulsive forces should be hindlimb driven, with a greater proportion of generated forces coming from the hindlimb as opposed to the forelimb (Preuschoft, 2002). This trend is observed in most, but not all, climbing primates (Hanna and Schmitt, 2011; Hanna et al., 2017; Hirasaki et al., 1993). While independent studies have measured total force magnitudes generated by the human forelimb (Fuss and Niegl, 2008) and hindlimb (Baláš et al., 2014) during climbing, to date no study has assessed the relative contributions of each limb during vertical climbing. As such, it remains unclear to what extent humans follow the typical primate pattern of force expression when scaling vertical substrates.

where the normal force is directly proportional to the distance between the climber's COM and the wall. Proficient climbers should, then, position their COM as close to the wall as possible to reduce these limb loading demands (Fig. 1). However, such data have yet been collected either in humans or in non-human primates.

Lastly, it has been speculated that proficient climbers may make use of pendular mechanics in the mediolateral plane, oscillating their COM between strides to conserve energy from one stride to the next (Zampagni et al., 2011). When ascending a rock wall, experienced human climbers exhibit greater mediolateral oscillations of their COM during ascent than novices by shifting their body weight onto a single supporting leg when moving the contralateral arm upwards, followed by a reallocation of body weight onto the opposing leg at the end of their double support phase (Zampagni et al., 2011). However, it is unknown to what extent this behavior is utilized by non-human primates during climbing.

In this study, we present data on single limb forces and COM kinematics during human climbing across a broad spectrum of expertise. We then compare these data with forces measured during vertical pole climbing in a series of non-human primates to test the overarching hypothesis that force generation profiles during climbing are consistent between human and non-human primates. These data are then collectively compared with biomechanical predictions of ‘optimal’ climbing strategies to assess the overall efficacy of primate climbing from a mechanistic standpoint. We assess the following three explicit predictions.

The hindlimb is loaded in compression during climbing, whereas the forelimb experiences tensile loads that are less readily resisted by bone. Moreover, unlike quadrupedal primates, the human hindlimb is loaded more habitually with body weight than the forelimb. Thus, we predict that humans will exhibit a greater bias towards hindlimb-powered climbing than other primate species to protect their forelimbs against injuries, resulting in a higher contribution of hindlimb forces in the propulsive plane.

Following biomechanical principles of vertical climbing, normal forces experienced by the hindlimb and forelimb must be equal in magnitude and opposite in direction to balance the body and prevent toppling (Bock and Winkler, 1978). The absolute magnitudes of these forces are proportional to the distance between the substrate and the individual's COM. As humans are considered less proficient climbers than other primates, we predict that normal force magnitudes in both forelimbs and hindlimbs (and, by extension, the distance between the COM and the substrate) will be greater in humans than in non-human primates during climbing.

Within humans, we predict that climbing expertise will modulate force profiles and COM movements. Specifically, we predict that experienced climbers will exhibit: (a) greater relative loading of the hindlimb and decreased relative loading of the forelimb than more novice participants; (b) a reduction in the overall magnitude of normal forces experienced by the limbs; and (c) a lower wall-to-COM distance, contributing to reduced normal force magnitudes.

Data were collected from 44 individuals (body mass range: 26.39–96.16 kg, age range: 9–38 years old) at Inclusive Sports and Fitness (ISF; 5004 Veterans Memorial Hwy, Holbrook, NY 11741, USA). All subjects gave their informed consent for inclusion before participation in the study. All subjects were healthy without musculoskeletal pathologies. Individuals ranged between no climbing experience and International Rock Climbing Research Association (IRCRA) level 22, equivalent to V7 following the Vermin scale or 5.13a following the Yosemite Decimal System (Draper et al., 2015). As only eight participants had meaningful climbing experience (i.e. ranked on the IRCRA scale), for the purposes of statistical comparison (see below) all individuals were classified as either ‘novice’ (n=36) or ‘experienced’ (n=8). All data collection protocols were approved by New York Institute of Technology College of Osteopathic Medicine Institutional Review Board (IRB protocol: BHS-1731).

Single limb forces (sampled at 1200 Hz) were collected by instrumenting a single cylindrical hand hold (0.30 m in length and 0.04 m in diameter) upon a rock climbing treadwall (Brewer Fitness, Randolph, MA, USA) with a high load triaxial force plate [model OR6-7-2K-26403 (464×508×82.5 mm); Advanced Mechanical Technology, Inc., Watertown, MA, USA; see Fig. 1]. The rungs were arranged such that there was 0.41 m between each subsequent rung (top to bottom) and offset at 108 deg in a zig-zag formation (see Fig. 2). All force data were collected using NetForce (Advanced Mechanical Technology, Inc.) and began by calibrating the force plate to remove any drift or offsets from previous recordings. Participants were allowed two practice runs to acclimate to the moving platform before performing between five and 10 climbs with ample rest between each ascent (see Fig. 2 and Movie 1). Only trials with no visual acceleration or deceleration, and that included a clear handfall and/or footfall on the force plate, were considered successful. All trials were recorded using two GoPro cameras (HERO10, GoPro, San Mateo, CA, USA) at 120 Hz from lateral and posterior views. Speed was calculated within ImageJ (Schneider et al., 2012) using the approximate position of the individual's COM as it moved across a known distance to calibrate the space. A stride was based on a reference limb touchdown to subsequent touchdown of the same reference limb.

Single limb forces were processed using a custom-written MATLAB script (see Young et al., 2023a,b) to correct for direction of travel, orientation and contacting limb (right or left). Forces were standardized such that fore–aft forces were split into positive propulsive and negative braking forces, normal forces were divided into positive pushing and negative pulling forces, and mediolateral forces were divided into positive laterally directed forces and negative medially directed forces. All forces were filtered through a low-pass Fourier filter at 15 Hz and normalized to a percentage of the participant's body weight (% BW) for peak forces and body weight seconds (% BWS) for impulses to allow statistical comparison between individuals.

To correlate the maximum wall-to-COM distance with single limb forces, we used the machine-learning pose-estimation software DeepLabCut (Mathis et al., 2018) and tracked the positional distance of climbing individuals from the laterally positioned videos following previously established methods (Young et al., 2022). Briefly, from all the lateral footage, 150–200 frames (total) were chosen at random. We chose to label five points – the predetermined COM of the individual, two points on the edge of the wall itself (used to draw a vertical line which we subtracted from maximum COM position to get wall–COM distance), and two additional points on the upper and lower borders of the force plate used as the known dimensions to calibrate the pixel-space post hoc (see Fig. 1). The COM position was based on previously published studies (Lee and Farley, 1998; Virmavirta and Isolehto, 2014). These frames were labeled by a single individual to prevent inter-user error and used to train the deep neural network. Following training, subsequent videos were fed through the program, and positional values (in pixel space) of each labeled point (COM, wall points and force plate points) were output as a function of time. These positional outputs were cropped for the contact phase of both the forelimb and hindlimb (e.g. from touchdown of the forelimb to liftoff of the same forelimb or from the touchdown of the hindlimb to the liftoff of the same hindlimb), and the maximum COM position for each single limb hit was determined. To output maximum wall­-to-COM distance, we subtracted the position of the wall using a custom-written MATLAB script.

Climbing data for eight non-human primates (L. tardigradus, N. pygmaeus, C. medius, E. mongoz, D. madagascariensis, S. sciureus, M. fascicularis and A. nancymaae) were aggregated from previous trials collected and published by Hanna et al. (2017). We acknowledge differences between the non-human primate single limb forces published by Hanna et al. (2017) and those reported in this paper. Namely, while pulling and pushing normal forces were reported in both the forelimb and the hindlimb for all species by Hanna and colleagues (2017), we report exclusively pulling forelimb forces and occasional pulling forces in the hindlimbs of strepsirrhines, though hindlimbs assumed predominantly pushing forces. The inconsistency in force profiles is likely due to a processing error that was corrected for in the reanalysis of these forces (see Hanna et al., 2024). Briefly, a consequence of utilizing poles in the experimental set up of climbing trials introduces a degree of freedom in that the climbing primates can freely rotate around the pole. While trials wherein the animal is directly above the force plate registered accurate force readings in biologically relevant planes (i.e. fore–aft forces reflect propulsive and braking forces, normal forces reflect pulling and pushing forces, and mediolateral forces reflect inwardly directed and outwardly directed forces), these planes, namely the mediolateral and normal forces, no longer reflect the behavior of the animal as they shift around the pole. Attention was taken to adjust the directionality and planes of force readings depending on the plane in which the animal climbed the pole (either directly on top with normal forces perpendicular to the substrate or to the left or right of the substrate wherein the normal forces would instead be registered in the mediolateral plane and vice versa). All trials in which the animals were in between planes (i.e. diagonally oriented with respect to the substrate) were excluded from analyses. We believe the pulling (negative) and pushing (positive) normal force profiles reported in Hanna et al. (2017) may reflect some mediolateral forces which often fluctuate between medial (negative) and laterally (positive) oriented forces within a single stride.

The full details of the experimental setup can be found in Hanna et al. (2017) but, in brief, each primate was encouraged to climb a vertically oriented pole of diameter 1.27–3.81 cm, varying on the basis of the animal's size [see Table 1 in Hanna et al. (2017) for full summary]. A central section of the pole was instrumented using a triaxial force transducer to collect ground reaction forces in the propulsive, normal and mediolateral planes. All trials were recorded using high-speed cameras at either 60 or 120 Hz.

To maximize methodological comparability between human and non-human primate datasets, the raw trials collected by Hanna et al. (2017) were reanalyzed here using the original video files and raw force traces. Hanna et al. (2017) presented the absolute forces (that is, force orientation specific to the force transducer). Here, we analyzed force orientation relative to the animal's position on the pole. As such, some deviation between the findings presented here and those originally published by Hanna et al. (2017), particularly with respect to the normal forces, are expected. Raw force traces were passed through a custom-written MATLAB code [see Young et al. (2023a,b) for code] to correct for sidedness (left or right substrate contacting limb). As in the original study, only trials in which the animal exhibited steady-state locomotion (i.e. was not accelerating or decelerating) were considered for analysis, and trials were only used wherein clear differentiation between limb contacts on the force plate could be observed. Forces were passed through a low-pass filter ranging from 100 to 1000 Hz (see Table S6) and standardized such that fore–aft forces were split into positive propulsive and negative braking forces, normal forces were divided into positive pushing and negative pulling forces, and mediolateral forces were divided into positive laterally directed and negative medially directed forces. Similarly, force magnitudes were normalized to a percentage of the participant's body weight (% BW) for peak forces and body weight seconds (% BWS) for impulses (although not statistically analyzed, see Fig. S1). Speed for each trial was calculated and provided by Hanna et al. (2017).

All data were analyzed using linear mixed effect models in R using the packages ‘lmerTest’ and ‘lme4’. First, the human data were analyzed independently. Three linear mixed effect models (one per plane of movement: fore–aft, normal and mediolateral) were constructed for both absolute peak forces and impulses in each plane. The fixed effects of interest were climbing experience (novice versus experienced) and limb (hindlimb versus forelimb); however, speed [well known to affect ground reaction forces, see Granatosky et al. (2020) for statistical discussion], and anthropometric and demographic effects [e.g. sex, age, ape index and height] were also included in the models to account for these effects. Ape index is calculated as a ratio between arm span and height and is a metric for relative arm length (Kozma and Pontzer, 2021). Age, ape index and height were included as an interaction effect (e.g. age×ape index×height) to account for potential collinearity. Individual was included as a random effect in all models to account for behavioral idiosyncrasies between participants following Bates et al. (2015). Directionality [e.g. braking versus propulsion (fore–aft), pulling versus pushing (normal), and medially directed versus laterally directed (mediolateral)] was included as a fixed effect only in the mediolateral models as fore–aft forces were exclusively propulsive, and normal forces were entirely differentiated (e.g. the forelimb was purely tensile and the hindlimb was purely compressive), and therefore accounted for in the fixed effect ‘limb’.

As the fixed effect ‘limb’ had a significant effect on fore–aft and normal forces, the data were subsequently split into a forelimb and a hindlimb data-frame and two secondary linear mixed effect models were constructed (one per limb, per plane of movement) to further assess the remaining fixed effects (see above). The mediolateral data showed significant differences in directionality (e.g. medial versus lateral forces); therefore, these data were split into a medial and a lateral data-frame and two secondary linear mixed effect models were constructed to further assess the additional fixed effects.

In order to directly assess the effect of the COM-to-wall distance on peak normal forces (see Introduction), we also constructed a linear mixed effect model with COM-to-wall distance and climbing experience (novice versus experienced) as the fixed effects of interest, whilst controlling for differences in speed, limb (forelimb versus hindlimb), and demographic and anthropometric variables (age, ape index, height, sex). Again, age, ape index and sex were included as an interaction term (e.g. age×ape index×height) to account for potential collinearity. Individual was included as a random effect to account for repeated measures and individual idiosyncrasies. As the fixed effect ‘limb’ significantly drove differences in peak normal force, an additional two linear mixed effect models were created (one for the forelimb, one for the hindlimb), to discern which of the remaining fixed effects were driving these differences.

Subsequently, three comparative models (again, one per plane of movement) were constructed to compare each non-human primate taxon with our human data. These models again used absolute peak forces and impulses as response variables, and included species, limb, speed and directionality as fixed effects and individual as a random effect (anthropometric and demographic effects used in human models were excluded from comparative analyses). As humans exhibited purely pushing forces in the hindlimb, hindlimb pulling forces generated by other species could not be compared (i.e. many non-human primate species exhibited both pulling and pushing forces in the hindlimb).

Lastly, after highlighting trials in which a clean consecutive forelimb and hindlimb contact could be isolated (e.g. within the same climbing trial, the climber successfully made contact with the force plate in the forelimb and the hindlimb), we calculated a limb force ratio by dividing peak forelimb forces by peak hindlimb forces. One final linear mixed effect model was then conducted using this ratio as the response variable, species and speed as fixed effects, and individual as a random effect.

Human climbers exhibited a symmetrical gait pattern with an average speed of 0.23±0.09 m s⁻¹ and a contact time of 2.57±0.92 s per stride (Table 1). Propulsive peaks and impulses were significantly greater (P<0.001) in the hindlimb (88.88±16.80% BW, 135.17±64.03% BWS) than in the forelimb (28.78±10.29% BW, 32.00±17.91% BWS), regardless of experience level, or anthropometric factors (Tables 1 and 2, Fig. 3). Male climbers (76.02±65.73% BWS), however, had significantly lower impulses (P=0.031) than female climbers (89.67±73.53% BWS). When stratified by limb (e.g. forelimb and hindlimb), these fore–aft force magnitudes (peak forces and impulses) were not driven by experience, demographic or anthropometric factors (all P>0.074; Table S1) contrary to Prediction 3.

While demographic and anthropometric factors had little effect on peak forces and impulses within the fore–aft plane, their significance was revealed in the normal plane. Force profiles differed between the forelimb (−27.65±6.86% BW, −34.77±13.30% BWS) and the hindlimb (25.53±7.09% BW, 27.85±10.03% BWS) but did not vary within each limb as a product of climbing experience or anthropometric factors (all P<0.001; Fig. 3 and Table 2; Fig. S2), again contrary to Prediction 3. Male climbers, however, generated greater magnitudes of peak normal forces (27.66±7.71% BW) than their female counterparts (25.22±5.62% BW; P=0.013; see Table S1). Additionally, within both the forelimb and hindlimb datasets, we failed to detect any appreciable differences in peak normal forces and impulses driven by experience (all P>0.782; Table S1), though the hindlimb was significantly affected by myriad anthropometric and demographic factors (all P<0.055; Table S1).

Finally, male climbers (8.76±4.92% BW) exhibited greater magnitudes of mediolateral peak forces than female climbers (8.22±4.48% BW; P<0.001; Table 2), regardless of experience level or anthropometric factors, though this difference was not significant in impulses (P=0.927; Fig. 3, Table 2). Further separating the data by directionality (e.g. medial versus lateral), peak forces and impulses registered at ∼5–10% BW (Tables 1 and 2) in both the forelimb and the hindlimb, with the hindlimb generating greater magnitudes (all P<0.019) of mediolateral forces irrespective of anthropometric (all P>0.092) or demographic factor (all P>0.058), or experience (all P>0.736; Table S1). The only exception existed in medial peak forces, wherein the male climbers (8.13±4.68) generated greater magnitudes of medial forces than female climbers (7.37±4.36; P=0.006; Table S1).

Overall, we observed a direct positive relationship between COM-to-wall distance and peak normal force magnitude (P<0.001); however, experience level did not drive any of these differences (P=0.802; Fig. 4, Table 3). This difference was significantly pronounced in the forelimbs of climbers (P=0.022; Table S2), but marginal in the hindlimbs (P=0.054; Table S2). The hindlimb peak forces and impulses were, however, significantly affected by demographic and anthropometric factors (all P>0.021; see Table S2).

In the fore–aft plane, humans and non-human primates exhibit exclusively propulsive forces (see Fig. 5; note M. fuscata and A. geoffroyi are plotted for graphical purposes only but were not analyzed statistically because of low sample sizes). As our initial model revealed significant differences between the forelimb and hindlimb (P<0.001), data were subsequently divided and reanalyzed within each limb. In the forelimb, humans demonstrated significantly lower propulsive forces than all non-human primate species (all P<0.008; Table S3). Similarly, humans exhibited significantly greater hindlimb propulsive forces than six of the eight non-human primates (barring C. medius or S. sciureus; Table S3). Expressing these force proportions as a ratio (following Prediction 1), humans demonstrated significantly greater hindlimb dominance compared with the rest of the non-human primate sample (all P<0.001; Fig. 6; Table S4). Thus, Prediction 1 was supported. Saimiri sciureus, C. medius, M. fascicularis, A. nancymaae, E. mongoz and Daubentonia madagascariensis also exhibited hindlimb dominance, but at a lower magnitude than in humans (Fig. 5). Conversely, lorisids (L. tardigradus and N. pygmaeus) exhibited forelimb dominance (Fig. 6).

Humans demonstrated a complete functional differentiation between the forelimb (pulling, tensile) and hindlimb (pushing, compressive). This matched the limb loading patterns observed in our other anthropoids (M. fascicularis, S. sciureus and A. nancymaae). By contrast, all strepsirrhines exhibited pull–push transitions (Figs 5 and 7) in which an initial forelimb pull was followed by a push approximately twofold in magnitude (Figs 5 and 7).

As this initial model revealed a significant difference in normal force magnitudes between the forelimb and hindlimb (P<0.001; Table S3), data were subsequently divided and reanalyzed within each limb. Normal forelimb forces were significantly lower in humans than in all non-human primates (all P<0.008; Fig. 5) barring C. medius (P=0.087; Table S3), for which comparisons did not yield statistical significance. In the hindlimb, normal forces were also significantly lower in humans than in six of eight species, but comparisons to C. medius (P=0.151) and D. madagascariensis (P=0.164) did not reach statistical significance (Table S3). Thus, Prediction 2 was not supported.

As significant differences between medially and laterally oriented forces were observed, data were analyzed independently between orientation. In the medial orientation, force magnitudes were lower in Homo sapiens than in any non-human primate in the hindlimb (Fig. 5; Table S5; all P<0.001). Forelimb medial forces were also lower in humans than in any species barring N. pygmaeus (P=0.221).

For laterally oriented forces, humans exhibited greater force magnitudes than any non-human primate in both the forelimb and hindlimb (Table S5; all P<0.001). It is unclear, however, whether these statistics reflect a difference in climbing mechanics between humans and non-human primates or are product of differing experimental conditions (see limitations below).

In this study, we conducted a comparison of the kinetic profiles of vertical climbing between humans and a sample of non-human primates. Humans provided a unique opportunity for studying climbing kinetics as they have a long evolutionary history associated with arboreal locomotion, but anatomically modern humans have primarily adopted terrestrial bipedalism and abandoned vertical climbing. We hypothesized that certain adaptations related to bipedalism, such as exclusively hindlimb-loaded weight support, might have an impact on climbing abilities. The main objective of this study was to empirically evaluate climbing kinetics in a comparative context, comparing humans with non-human climbing primates. Additionally, we examined how climbing kinetics might vary among humans with different levels of expertise, encompassing a diverse range of participants.

Contrary to our initial expectations, we did not observe any kinetic differences among our human participants based on their climbing experience. We observed highly consistent force profiles, with significantly lower peak propulsive forces in the forelimb compared with the hindlimb. Furthermore, there were no substantial differences observed in the fore–aft, normal or mediolateral planes as a product of experience, even after controlling for anthropometric and demographic factors (as shown in Table 2).

We also predicted that experienced climbers would maintain a COM position closer to the substrate while climbing, as this posture would reduce the demand for normal plane forces on both the hindlimb and forelimb. However, our analysis showed that the mean COM distances were virtually identical between novice and expert climbers (mean 43.7 cm versus 44.7 cm), and the differences between groups were not statistically significant (P=0.204; Fig. 3). Nonetheless, we did observe a consistent relationship between COM position and peak normal force magnitude in both experienced and novice climbers. COM position from the wall most strongly predicted greater peak normal forces in the forelimb, and this relationship was less pronounced in the hindlimb. The ability to predict the magnitude of single limb forces, at least in the normal and fore–aft plane, is largely dependent on the limb's position relative to and/or distance from the COM (Dickinson et al., 2022; Granatosky et al., 2018; Larson and Demes, 2011; Raichlen et al., 2009; Bock and Winkler, 1978; Norberg, 1986). As such, these data provide direct evidence for long-held assumptions based on models of force distribution during climbing.

These data present an intriguing natural experiment, offering indirect evidence on how changes in body proportions and stature throughout human evolution might affect climbing performance. Over the past two million years of human evolution, there has been a general increase in stature and a decrease in relative arm length (Hunt, 1994; Ruff, 2000; Stulp and Barrett, 2016; Styne and McHenry, 2008). These anatomical changes have been linked to reduced reliance on climbing abilities, often directly associated with the stresses imposed on the limbs (Drapeau and Ward, 2007; Stern, 2000). In our sample, we had a diverse range of individuals with considerable variation in body proportions. However, our data show that neither stature nor arm span has any influence on the magnitudes of forces acting upon the limbs. While these findings are admittedly indirect, they suggest that climbing performance, at least from a limb loading perspective, may not have been negatively impacted by changes in body proportions during early primate evolution. These findings, in addition those reported in previously published experimental work (Kozma and Pontzer, 2021), collectively raise questions about the relative biological importance of changing limb proportions throughout human evolution. They also suggest that the proposed consequences for climbing frequency in the literature may be overstated.

Unlike non-human primates, modern humans utilize an exclusively hindlimb-powered gait during habitual terrestrial locomotion. Thus, it was predicted that during vertical climbing, humans would exhibit the strongest hindlimb bias in propulsive force generation (i.e. have the highest proportion of propulsive forces generated by the hindlimb, and the lowest proportion of propulsive forces generated by the forelimb). This hypothesis (Prediction 1) was strongly supported. It was interesting to note that both lorisid species broke from the trend observed in humans and other non-human primates in exhibiting forelimb bias in propulsive force generation when climbing (Fig. 6) – a pattern also seen during vertical climbing in sloths (Young et al., 2023a). This shared trait between lorisids and sloths may thus be reflective of adaptations towards below-branch, inverted quadrupedal locomotion practiced frequently by both species. Indeed, this notion is bolstered by recent observations that non-human primates adopt a forelimb-biased weight support strategy during inverted quadrupedal walking (Dickinson et al., 2022). Thus, it appears that despite the reorientation of the substrate, limb-usage patterns of primates are strongly conserved between typical bipedal/quadrupedal walking and vertical climbing.

One of the most surprising findings of our study was the phylogenetic distinction associated with normal forces. Namely, strepsirrhines (E. mongoz, C. medius, D. madagascariensis, L. tardigradus and N. pygmaeus) all demonstrated a pull–push transition in their hindlimbs (i.e. experiencing both tensile and compressive forces) while there was a complete absence of this trait in anthropoid primates (M. fascicularis, A. nancymaae, S. sciureus and H. sapiens; see Fig. 7). Contrary to data published by Hanna et al. (2017) that report pulling and pushing normal forces in the forelimb of anthropoids, we report exclusively pulling forelimb forces. The inconsistency in force profiles is likely due to animals freely rotating about the pole during climbing, altering their respective orientation relative to the force plate. We accounted for this difference in our re-analysis.

A pull–push transition refers to an initial pulling (i.e. tensile) peak in the normal plane at the beginning of the stance phase, followed by a pushing peak (as shown in Fig. 7). This force pattern has been observed in the hindlimbs of climbing species such as Australian tree frogs (Young et al., 2023b), geckos (Wang et al., 2015), parrots (Young et al., 2022) and sloths (Young et al., 2023a), and is, upon visual inspection, the result of the limb being positioned cranial to the COM at touchdown and subsequently moving to a caudal position relative to the COM by the end of the stance phase. The presence of this phenomenon may be ascribed to the behavior of hindlimb overstriding (i.e. placing the hindlimb above the COM at touchdown). The point at which the hindlimbs undergo this transition in force profiles falls approximately 10–30% into the stance phase (mean±s.d. 18.92±6.92%), occurring earliest in E. mongoz (9.80%) and latest in L. tardigradus (28.56%). While dedicated kinematic analyses did not form a part of this study, we did note that L. tardigradus exhibited a hyper-protracted hindlimb in trials where a transition was observed (a trait previously observed and described by Hanna, 2006). However, further research is needed to quantify this relationship. We further observed a single instance of a forelimb push–pull transition in the aye-aye, which may be explained by a similar mechanism wherein the forelimb was positioned caudal to the COM at touchdown and eventually crossed the COM during the stride, resulting in a shift from pushing to pulling forces.

Given the significant limb excursion potential in both the primate forelimb (Larson et al., 2000) and hindlimb (Larson et al., 2001), one would expect to observe pull–push across all primate taxa. However, the observation that anthropoids break from the typical tetrapod pattern of a hindlimb pull–push transition during vertical climbing is intriguing, and further investigation into kinetic patterns of other anthropoids – particularly among hominoids – would be of great value for interpreting the potential role of this functional differentiation in driving the biomechanical similarities between vertical climbing and bipedal walking previously described in humans (Fleagle et al., 1981; Gebo, 1996; Lovejoy et al., 2009; Stern and Susman, 1981; Wunderlich and Ischinger, 2017).

Previous attempts to quantify mediolateral forces during vertical climbing have often revealed significant variation, making it difficult to draw clear functional interpretations (Carlson et al., 2005; Schmitt, 2003; Young et al., 2023a,b). Within our human sample, we observed consistent oscillations of both medially and laterally directed forces. These oscillations do not differ between the forelimbs and hindlimbs, and both limbs exhibit a slight bias towards laterally directed forces. This directionality is likely a result of the sprawled posture adopted when scaling a mechanical treadwall. By contrast, non-human primates show a clear bias towards medially directed forces, likely attributable to the use of a pole wherein medially directed forces aid in ‘gripping’ during locomotion on thin branches (Carlson et al., 2005; Granatosky and Schmitt, 2017; Schmitt, 2003). This difference in experimental setup represents a major practical limitation of this study (see below); thus, future studies should endeavor to describe the kinetic profiles of vertical pole climbing within humans. However, recent work on tree frogs revealed no differences in mediolateral forces between flat and round substrates (Young et al., 2023b), though the adhesive quality of frog autopodia may muddy these comparisons.

A study of this nature is of course faced with several limitations that must be disclosed for the reader to draw their own conclusions of the data. First, to accommodate climbers of all experience levels (including true novices), the experimental design was relatively simplistic in terms of its mechanical demands. While such a design was necessary to allow direct comparisons between novice and expert climbers, this simplicity likely attributed to the similar kinetic profiles generated across subjects. It is possible that more complex wall designs that place more taxing demands on the climber may yield more significant differences in force profiles, a hypothesis that will be explored in future work. Related to this, our simplistic setup essentially elicited ladder climbing. Such movements are not biologically relevant with regards to human evolution and do not replicate true rock climbing or scaling vertical large-diameter substrates such as tree trunks (Demps et al., 2012; Dutta et al., 1985; Endicott, 1979; Frisbie, 1971; Ichikawa, 1981; Kaplan et al., 1985; Kitanishi, 1996; Marlowe, 2004, 2010; Valli and Summers, 1988). The significant disparity between our experimental conditions and the natural environment calls for additional research. Nevertheless, the fact that no researchers have gathered three-dimensional force data from a non-human great ape climbing a vertical pole underscores the challenges involved in adapting such equipment for human participants to use. However, we contend that the difference in substrate conditions is unlikely to significantly affect limb-loading patterns during climbing. In recent years, numerous studies have investigated the loading patterns of tetrapods during climbing (Hanna et al., 2017; Hirasaki et al., 1993; Wang et al., 2015; Young et al., 2022, 2023a,b). Remarkably, there is very little variation in normal and fore–aft forces across these species, suggesting that the forces exerted by a single limb in these two planes are governed by common mechanical requirements (Bock and Winkler, 1978; Cartmill, 1985; Goldman et al., 2006; Preuschoft, 2002). The only direction where differences between climbing a flat wall and a pole might be detected is in the mediolateral plane. Indeed, when it comes to vertical climbing in humans, there are consistent oscillations of both medially and laterally directed forces. These force patterns remain consistent between the forelimbs and hindlimbs, with both limbs experiencing a slight bias towards lateral forces, likely due to the spread-out posture typical for climbing vertical walls. In contrast, non-human primates exhibit a clear bias towards medially directed forces, which is likely attributed to their use of poles, where medially directed forces assist in gripping (Carlson et al., 2005; Granatosky and Schmitt, 2017; Schmitt, 2003) while moving on thin branches. This variation in experimental setup represents a significant practical limitation of our study, preventing meaningful comparisons in the mediolateral plane between vertically climbing humans and non-human primates. We acknowledge such a limitation and note to the reader that our inferences are based on overall patterns, rather than magnitude or specific species differences.

Finally, because of the complexities of subject recruitment, the number of experienced climbers (n=8) was substantially smaller than our novice subject pool (n=36). It is possible that a broader group of individuals that spanned higher grades of climbing skill may have induced greater variance. Related to this, we observed significant effects of age and, consequently, the size of the participants for several of the kinetic parameters. Although we accounted for such potential variance in our statistical models, there was the unavoidable fact that the fixed inter-rung distance meant that relative stride length differed dramatically between our youngest and oldest individuals. Again, our statistical models found no such effect, but it is important to note this consideration when interpreting these data.

Humans, despite their evolutionary history of arboreal locomotion, have predominantly adopted terrestrial bipedalism and abandoned vertical climbing. Nevertheless, anatomically modern humans do exhibit the ability to climb (as seen in the wide variety of climbing styles), though the manner in which they scale vertical substrates does differ from that of non-human primates. Contrary to initial expectations, experience did not significantly impact the kinetic profiles of human climbers, suggesting that humans (unlike non-human primates) demonstrate a stereotyped – and possibly constrained – climbing style. Both novice and experienced climbers displayed consistent force profiles, with hindlimbs generating significantly higher propulsive forces compared with forelimbs. Both experienced and novice climbers showed a relationship between COM position and peak normal force magnitudes. Humans exhibited a clear hindlimb bias in propulsive force generation during climbing, a pattern not observed in non-human primates. This bias is likely associated with the exclusive hindlimb-powered gait used by modern humans during terrestrial locomotion. Phylogenetic distinctions were observed in normal forces, with strepsirrhines showing a pull–push transition in hindlimbs during climbing, while anthropoid primates, including humans, did not exhibit this pattern. The absence of this trait in anthropoids during vertical climbing is intriguing and warrants further investigation. Such functional differentiation may help to explain the evolution of bipedalism in ancestrally climbing hominoids. Mediolateral forces displayed differences between humans and non-human primates, likely influenced by the experimental setup. Overall, this study enhances our understanding of the biomechanics of climbing in humans and how they compare with the biomechanics of climbing in their non-human primate counterparts.

We would like to thank Jandy Hanna for providing all the raw data necessary to replicate force analyses for the comparative context, as well as for meaningful suggestions to the initial draft of the manuscript. We would also like to recognize Jon Gustafson for his contribution in data preparation as well as Andrea Sanseviro for her help in scheduling human participants. We also thank the Center for Biomedical Innovation at New York Institute of Technology and Inclusive Sports and Fitness for co-funding this study. Lastly, we would like to thank Evie Vereecke, one anonymous reviewer and Monica Daley, the handling editor for their comments and for helping to improve the quality of the manuscript. The authors report that some of the subjects used for data collection are also co-authors of the manuscript. This by no means influenced the interpretation or presentation of the data.