Center of mass position does not drive energetic costs during climbing Free

The ability to climb vertical surfaces has evolved independently among most limbed animals (Cartmill, 1985; Preuschoft, 2002), conferring numerous ecological advantages, by providing access to a third dimension in which to forage, hunt or escape predators. However, apart from changes made to the autopodia for secure attachment to surfaces, few anatomical features distinguish specialized climbing animals (Cartmill, 1985; Labonte and Federle, 2015, 2016; Preuschoft, 2002; Spenko et al., 2006). As such, while there are arguments for a global biomechanical template (Goldman et al., 2006), there is no singular morphotype that defines a vertically climbing species (Granatosky and Young, 2023). This many-to-one relationship (Wainwright et al., 2005) makes it possible to test experimentally how various anatomical and behavioral characteristics affect the mechanics of climbing. Historically, climbing has been evaluated through the lens of metabolic efficiency (Bertuzzi et al., 2007; Booth et al., 1999; Hanna et al., 2008; Kozma and Pontzer, 2021), wherein it is either implied or explicitly stated that animals should alter their morphologies and behavior to minimize the energetic costs associated with locomotion (Bock and von Wahlert, 1965). However, despite the widespread manifestation of climbing, existing data on the associated energetic costs are limited (Bertuzzi et al., 2007; Booth et al., 1999; Hanna et al., 2008; Kozma and Pontzer, 2021).

Early research into the scaling costs of incline locomotion (Taylor et al., 1972) demonstrate that, from a mass-specific perspective, the additional costs associated with inclined locomotion are accrued independent of size. Increased body mass does not increase the mass-specific cost of moving uphill; rather, it costs the same amount of energy to lift a 1 kg mass by 1 m of height, no matter the body size of the organism (Elton et al., 1998; Hanna and Schmitt, 2011; Hanna et al., 2008; Taylor et al., 1972). Taylor et al. (1972) also report that the mass-specific cost of moving horizontally is much lower in large animals compared with small species, suggesting that mass does influence locomotor costs during horizontal locomotion. Therefore, the relative cost of moving uphill is greater for larger species than smaller ones.

Validated experimentally using insects (Full and Tullis, 1990; Lipp et al., 2005), it has been demonstrated that larger species incur a greater relative cost during incline locomotion compared with horizontal movement versus smaller taxa. Subsequently, data published by Hanna and colleagues (2008) and Hanna and Schmitt (2011) examined the energetics of climbing across five primate species, ranging in body mass from 0.17 to 1.40 kg. Their findings confirmed that, unlike in horizontal locomotion, the mass-specific cost of climbing does not consistently decrease with body mass at the same rate, supporting the notion that the energetics of climbing are determined by universal, mass-specific costs (Taylor et al., 1972; Elton et al., 1998). More recently, Kozma and Pontzer (2021) assessed the energetic efficiency of human climbing and correlated climbing efficacy with a range of morphometric characteristics (e.g. body mass, body height, appendage length and proportions). They observed that climbing velocity is the primary factor influencing energy expenditure, with faster climbing resulting in lower costs, while body size and body proportions did not significantly affect metabolic costs.

Owing in part to the myriad logistical limitations associated with collecting metabolic data directly from experimental subjects during climbing trials, many studies have instead sought to analyze the mechanics of climbing via mechanical modelling of center of mass (COM) mechanics (Autumn et al., 2006; Goldman et al., 2006; Young et al., 2023). During typical horizontal walking, the COM oscillates in a somewhat pendular fashion that allows for the conversion of potential energy to kinetic energy between strides (Cavagna and Margaria, 1966; Cavagna et al., 1977; Heglund et al., 1982; Kuo et al., 2005). During climbing, however, potential energy steadily increases, thus precluding this exchange mechanism (Autumn et al., 2006; Young et al., 2023). Instead, the mechanical efficiency of climbing is assessed by comparing total exerted mechanical against the theoretical minimum power required for ascent. Though comparative data are limited, it can be generally observed that most taxa (regardless of limb number, adhesive mechanisms or gait) adhere closely to the minimum power line, suggesting that the physical constraints of climbing compel animals to follow a singular global template for effectively ascending vertical surfaces (Autumn et al., 2006; Goldman et al., 2006; Young et al., 2023).

Several theoretical papers have outlined potentially determinative relationships between COM position and climbing efficacy (Autumn et al., 2006; Goldman et al., 2006; Young et al., 2023). Most simply, in the fore–aft plane, animals should rely on exclusively propulsive forces to maintain an upwards ascent. In the normal plane, meanwhile, limiting the distance between the COM and the substrate should theoretically reduce energetic costs (Cartmill, 1985; Jungers, 1977; Norberg, 1986). In other words, the forces (i.e. tensile forces produced by the limb positioned superior to the COM and compressive forces generated by limbs positioned inferior to the COM) necessary to counterbalance the backward gravitational torque are directly proportional to the offset distance between the COM and substrate (Young et al., 2024), and, in turn, to the mechanical work necessary for climbing (Fig. 1). This concept has been the subject of several biomechanical discussions (Bock and Winkler, 1978; Cartmill, 1985; DeSilva, 2009; Jungers, 1977; Norberg, 1986) and has been empirically evaluated in various animal species (Goldman et al., 2006; Isler, 2002, 2005; Young et al., 2023). Despite such theoretical considerations, in vivo COM movements in humans reveal somewhat unexpected findings. Zampagni et al. (2011) reported that expert climbers tend to maintain a COM position further from the wall than novices, which has been ascribed by the authors as a perceived trade-off between efficiency and stability – encroaching the COM up towards the wall reduces the horizontal offset between the COM and the ankle joints, a posture that is thought to be unstable (Pontzer and Wrangham, 2004; Zampagni et al., 2011). The explanation provided by Zampagni et al. (2011) aligns with previous descriptions of climbing strategies in chimpanzees, for which mechanical adaptations appear to similarly favor stability over efficiency (Pontzer and Wrangham, 2004). To evaluate this hypothesis, however, a much clearer relationship between COM kinematics and stability must be established, as well as a more direct assessment of the extent to which these supposedly competing optima are truly opposed (i.e. does increasing stability necessarily decrease efficiency? – and vice versa).

and consequently an increase in mediolateral kinetic energy:

and ultimately mediolateral work:

The sum of the moments is zero:

and the force in the normal plane (⁠⁠) is the product of normal distance from the substrate (⁠⁠) and the gravitational force (where is mass and is gravitational acceleration) divided by the distance between the upper and lower contacts (L):

The normal work (⁠⁠) is the integration of over the distance which the animal moved (Δdistance):

An increase in Pos_z would result in an increase in and ultimately total work (⁠⁠).

Finally, in the mediolateral plane, any movements of the COM may be considered a metabolic waste, as these fluctuations do not contribute to upwards progress (Autumn et al., 2006; Goldman et al., 2006). Theoretically, extraneous movements of the COM in the mediolateral plane should result in the accumulation of kinetic energy (through increased mediolateral velocities) and ultimately larger magnitudes of mechanical work (Fig. 1). Zampagni et al. (2011), however, demonstrated that expert climbers exhibit significant lateral oscillations of the COM, accompanied by a substantial redistribution of body weight between the two legs during the double support phase of climbing. These lateral oscillations are not random but follow a specific temporal pattern in all climbers, forming a typical ‘triangular-like’ path in the transversal plane (Zampagni et al., 2011). Novice climbers, in contrast, display fewer mediolateral COM oscillations. The widespread presence of this specific mechanism within expert climbers may reflect an energetic recovery system, wherein expert climbers utilize mediolateral undulations of their body mass to carry energy between strides. Such a mechanism for energy recovery may explain why experienced climbers tend to be more efficient, from an absolute metabolic perspective, compared with recreational climbers (Bertuzzi et al., 2007). However, such a theoretical mechanism for energy recovery has yet to be assessed and, moreover, the absence of a phasic relationship between kinetic and potential energies during climbing would preclude direct exchange.

Thus, as it stands, there is relatively little information and often conflicting interpretations about the relationship between COM movements and the metabolic costs associated with climbing. As such, we designed a study in which we experimentally manipulated climbing conditions to elicit changes in COM position and studied the associated metabolic costs. In so doing, we tested the following three hypotheses.

Hypothesis 1: mass-specific cost of transport will not vary between individuals. As previously observed (Full and Tullis, 1990; Hanna et al., 2008; Kozma and Pontzer, 2021), energy requirements of climbing are associated with mass (i.e. the cost for 1 kg of mass to gain 1 m of height is fixed). Thus, we predict that cost of transport between individuals will not vary as a product of body mass.

Hypothesis 2: the proximity of an individual's COM to the climbing wall will directly drive metabolic costs during climbing. Specifically, individuals who maintain a greater distance between their COM and the wall will experience higher metabolic costs, due to the increased muscular forces required to counteract the pitching moments generated by their body mass (Bock and Winkler, 1978; Cartmill, 1985; DeSilva, 2009; Jungers, 1977; Norberg, 1986).

Hypothesis 3: mediolateral movements of the COM will increase metabolic costs during climbing. Though expert climbers are reported to exhibit greater mediolateral oscillations of the COM than novice climbers (Zampagni et al., 2011), movements in the mediolateral plane do not directly contribute to upward movement, and are mechanically costly (Autumn et al., 2006; Goldman et al., 2006; Young et al., 2023). As such, we predict that high mediolateral COM oscillations will increase the metabolic costs of vertical climbing.

Climbing energetics data were collected from 14 individuals (body mass range: 56.10–98.90 kg, age range: 21.78–29.11 years old) at Inclusive Sports and Fitness (ISF; 5004 Veterans Memorial Hwy, Holbrook, NY 11741, USA) on a rock climbing treadwall mounted with rung handholds (see Fig. 2). All subjects gave their informed consent for inclusion before participation in the study and were free from any visible gait 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). All data collection protocols were approved by New York Institute of Technology's Institutional Review Board (IRB protocol: BHS-1731).

The COSMED K5 respirometry system (COSMED, Concord, CA, USA) was utilized to assess metabolic costs of climbing across three experimental conditions and to record the resting metabolic rate (RMR) of each climber. The COSMED system was calibrated following the manufacturer's protocols. Individuals were instructed not to consume any food at least 2 h prior to climbing trials (but were allowed to drink water); however, steps were not taken to validate this. After fitting all participants with appropriate mask sizes, participants were asked to sit or lie down (self-selected) in a dark, sensory stimuli-deficient room for 15–20 min to measure RMR. Individuals then each participated in three bouts of continuous climbing at a set, self-selected speed each lasting between 3 and 7 min and were allowed to rest for a minimum of 25 min between trials.

The three experimental conditions were designed to force different COM movements: a free choice condition (control), a sprawling condition (to force large mediolateral COM movements), and a ladder climbing condition (to force large normal COM movements; see Fig. 2). In the free choice condition, participants were asked to climb utilizing a combination of rungs of their choice. Rungs were cylindrical in shape and measured 0.30 m in length and 0.04 m in diameter. In the sprawling condition, the participants could only use the rungs mounted on the edges of the treadwall. These measured 0.92 m apart left to right (from center of rung to center of rung) and 0.60 m apart between the upper and lower rung. Lastly, in the ladder climbing condition, participants could only utilize the rungs mounted 0.60 m apart in a ladder fashion (Fig. 2).

All trials were captured at 120 Hz from a lateral and frontal view using synchronized GoPro video cameras (HERO10, GoPro, San Mateo, CA, USA). This footage was loaded into ImageJ (Schneider et al., 2012), calibrated with known distance, and used to calculate spatiotemporal variables including contact time, stride time, duty factor, stride length, stride frequency and velocity. A stride was defined from the touchdown of one hindlimb to the subsequent touchdown of the same hindlimb. A total of five strides were used per individual, per condition. The frontal video footage was also used to calculate mediolateral excursions through DeepLabCut (Mathis et al., 2018), a markerless pose estimation program using methods described by Young and colleagues (2022a,b, 2023). Briefly, 150 frames were randomly extracted and a total of three points (two of which demarcated a known distance used to calibrate the space, and one which marked the COM) were labeled in each frame. COM position was defined following Virmavirta and Isolehto (2014). It should be noted that these estimated COM positions were treated as static throughout the stride. While true COM position does not remain exactly static throughout a stride because of movements of the viscera (Daley and Usherwood, 2010) and limb positioning (Pavei et al., 2017), this variation is thought to be minimal and should not affect analyses of COM mechanics (see Farrell et al., 2015; Young et al., 2007). DeepLabCut (Mathis et al., 2018) was used to train the dataset and track the kinematic COM position. These positional data were cropped for a stride, and a maximum mediolateral excursion (maximum−minimum mediolateral COM position) was calculated using custom-written R code (http://www.R-project.org/). Maximum normal distance from the wall was measured using the lateral view footage; however, because of occasional obstruction of the climbing individual by a support beam attached to the treadwall apparatus, we opted for manual COM tracking using DLT Data Viewer in MATLAB rather than DeepLabCut (Mathis et al., 2018). A total of five strides per individual, per condition (free choice, sprawling, and ladder climbing) were extracted and an average maximum normal distance (per individual, per condition) was calculated.

All data were statistically assessed in R. We used the ‘lmer’ function of R package ‘lme4’ (Bates et al., 2015; https://CRAN.R-project.org/package=lme4) to run the linear mixed effect models. All models were assessed for normality and heteroscadecity using functions ‘shapiro.test’, ‘leveneTest’ (from the base R packages; http://www.R-project.org/) and the ‘check_model’ function of the ‘performance’ package to assess model diagnostics (Lüdecke et al., 2021). Individual was included as a random effect in all linear mixed effect models to account for possible individualistic differences. To assess whether the three conditions were successfully created (free choice, sprawling and ladder climbing), we created two linear mixed effect models with normal maximum distance and mediolateral excursion as response variables and climbing condition as the fixed effect. Then, to assess how energy levels (COT_net, COL_net and COL_net per cycle) were affected by climbing condition, velocity, body mass, normal maximum distance and mediolateral excursion, an additional three linear mixed effect models were created (one model per energy level). All models were also tested for the effects of sex, anthropometric variables (relative arm and leg lengths), strength metrics (grip strength and jump strength) and climbing experience; however, these were not significant in driving any changes in energy and were subsequently removed from all models (see Supplementary Materials and Methods). Lastly, all energy levels (COT_net. COL_net, and COL_net per cycle) were regressed against velocity, body mass, normal maximum distance and mediolateral distance using the ‘lm’ function from base R (http://www.R-project.org/) to run the linear models.

Throughout each trial, speed was relatively uniform across all three climbing conditions (mean±s.d. 0.18±0.05 m s⁻¹). Within free climbing, we observed the lowest duty factors (77.2±10.4%), stride lengths (0.65±0.14 m) and stride times (3.55±0.92 s), but the greatest stride frequency (0.30±0.08 Hz) and contact times (0.40±0.14 s). The spatiotemporal characteristics of the sprawling condition and the ladder climbing condition were nearly indistinguishable, with comparable duty factors (85.9±3.79% and 86.8±3.54%), contact times (0.31±0.12 and 0.30±0.11 s), stride times (4.28±1.53 and 4.37±2.77 s), stride frequencies (0.26±0.09 and 0.26±0.09 Hz) and stride lengths (0.77±0.25 and 0.75±0.22 m), respectively.

The energetic consumption was nearly indistinguishable when stratified for climbing condition (see Table 1), with an average for COL_net, COT_net and COL_net per cycle (over all three conditions) of 12.23±2.39 W kg⁻¹, 68.68±7.83 J kg⁻¹ m⁻¹ and 41.48± 8.96 J kg⁻¹ stride⁻¹, respectively (see Fig. 3). Contrary to our intended effects, the COM excursion in the mediolateral plane was greatest in the free choice condition (0.29±0.11 m, range 0.18–0.55 m), followed by ladder climbing (0.25±0.09 m, range 0.08–0.43 m) and sprawling (0.21±0.58 m, range 0.10–0.28 m), with the only statistically significant difference found between free choice and sprawling conditions (Figs 4 and 5; P=0.01; Table 2). In the normal plane, however, the ladder climbing condition did indeed yield the greatest COM to substrate distance (0.28±0.07 m, ranging 0.16 m–0.67 m), followed by sprawling (0.25±0.06 m, ranging 0.17–0.60 m) and free choice (0.13±0.07 m, ranging 0.19–0.60 m; all P<0.037; Table 2). All three conditions were statistically distinct from one another (all P<0.003; Figs 4 and 5; Table 2). The COM trajectories varied both from stride to stride within an individual and between individuals (Fig. 4). Differences in COT_net, COL_net and COL_net per stride were not driven by climbing condition (free choice, sprawling and ladder climbing), body mass or COM positional metrics (all P>0.128; Figs 5 and 6); however, velocity had a significant effect on COT_net and COL_net (P<0.001 for both; Table 2).

In this study, we experimentally manipulated climbing conditions to induce variation in COM positions within our subjects. We also selected a subject pool spanning a broad range of body mass (56.10–98.90 kg), comparable to levels observed in other extant hominoids (Smith and Jungers, 1997). Through these methods, we aimed to examine how both individual human morphotypes and specific COM conditions affected metabolic energy expenditure during climbing.

Contrary to our initial expectations (Autumn et al., 2006; Goldman et al., 2006; Jungers, 1977; Norberg, 1986; Zampagni et al., 2011), variation in COM position had minimal effects on mass-specific locomotor costs and overall metabolic costs per stride. Specifically, we observed that mediolateral COM excursions and normal COM excursions had no impact on any energetic measurement (COL_net, COT_net or COL_net per cycle). From a theoretical perspective, reducing the horizontal distance between the COM and the substrate should decrease metabolic energy costs, as the normal forces required to balance the body are directly proportional to this offset distance (see Fig. 1; Bock and Winkler, 1978; Jungers, 1977; Norberg, 1986). Across our experimental conditions, we noted substantial variation in this offset distance (∼0.2–0.7 m); however, this variable failed to predict mass-specific metabolic costs. This observation raises an intriguing question: why is metabolic energy expenditure during climbing seemingly indifferent to normal COM positioning?

To gain height, climbers must expend a minimum amount of energy equal to the product of the climber's mass, gravity and the total distance traveled upwards (i.e. potential energy; Autumn et al., 2006; Young et al., 2023). Any additional movements outside this propulsive plane are theoretically wasted energy, as they do not contribute to vertical ascent (Autumn et al., 2006; Goldman et al., 2006; Norberg, 1986). However, our data demonstrate that such out-of-plane movements represent only a tiny fraction of overall energy expenditure during climbing, challenging long-held assumptions that out-of-plane COM deviations contribute significantly to climbing costs. Instead, other costs – such as firing flexor musculature to maintain grip on the wall (Byrnes and Jayne, 2014) or the costs of swinging the limbs (Doke et al., 2005; Granatosky and McElroy, 2022) – may contribute more significantly to the metabolic costs of climbing. Indeed, the energetic expenditure of swinging the limbs has only become evident in recent years and accounts for approximately 8–33% of total locomotor costs (at least during horizontal locomotion) (Doke et al., 2005; Marsh et al., 2004; Pontzer, 2007). These unaccounted-for costs likely make substantial contributions to the overall metabolic expenditure, helping to explain the approximately 7-fold differences between the metabolic mass-specific cost of locomotion and the mechanical mass-specific cost of locomotion. Considering these factors, we argue that theoretical costs associated with mediolateral oscillations of the COM and horizontal distances from the substrate play a limited role in the metabolic cost of climbing.

Instead, we found that speed was the primary factor driving overall climbing costs, a finding consistent with previous research by Kozma and Pontzer (2021) and in line with trends observed in power requirements during climbing in other species (Autumn et al., 2006; Young et al., 2023). Some of the decreased costs associated with higher speeds may be attributable to the costs of muscular activation of maintaining a static position on the wall. By moving faster, climbers may be able to decrease these costs. However, once the mass-specific costs per stride are accounted for, any relationship with speed disappears, a topic thoroughly discussed by Heglund and Taylor (1988). Therefore, we choose to refrain from further discussions associating speed with metabolic cost during climbing.

Consistent with previous studies across various lineages (Full and Tullis, 1990; Hanna et al., 2008; Lipp et al., 2005; Taylor et al., 1972), we observed no differences in mass-specific metabolic costs across our sample (see Fig. 6). However, from a comparative and evolutionary perspective, climbing behavior is disproportionately observed in small-bodied taxa (Cant, 1992; Clemente and Dick, 2023; Granatosky, 2018; Labonte and Federle, 2015; Labonte et al., 2016). This misalignment raises a compelling question: if the mass-specific metabolic costs of climbing remain constant for a 1 kg and a 100 kg organism, why are so few large-bodied animals effective climbers?

Firstly, there are differential scaling relationships governing limb bone cross-sectional area (which increases to the second power with body length) and body mass (which increases to the third power with body length). Consequently, if two falling animals were to land on their limbs, a 100-fold increase in mass (e.g. from 1 kg to 100 kg) would amplify the relative pressure (force per unit area) experienced by the bones (assuming equal mechanical properties and falling velocity), as the geometric scaling of limb bones does not proportionally keep up with increases in mass. This problem is compounded further by the realization that, as body size increases, so does terminal velocity, a result of mass increasing cubically while surface area (and thus air resistance) increases to the second power (Hoppeler and Weibel, 2005; Schmidt-Nielsen, 1984). This effect is concisely summarized by J. B. S. Haldane, who vividly illustrates that when dropped down a thousand-yard mine shaft, ‘a mouse […] gets a slight shock and walks away; a rat is killed, a man is broken, a horse splashes’ (Haldane, 1926). Thus, attaining greater body mass substantially increases the risk of falls, with fitness risks scaling directly with body mass. While documented cases of falls in large-bodied arboreal taxa are infrequent, these events can be fatal (Arlet et al., 2009; Bwindi Impenetrable National Park, 2013).

Secondly, larger animals have an advantage when it comes to horizontal locomotion because they experience a lower mass-specific cost compared with smaller species. This means that, for a given horizontal distance traveled, larger animals expend less energy per unit of body mass (Hanna et al., 2008; Heglund and Taylor, 1988; Taylor et al., 1972, 1982). As such, the relative cost difference between climbing and horizontal walking is much greater for large-bodied species than for smaller taxa (Hanna et al., 2008; Taylor et al., 1972). Nevertheless, larger animals experience much greater absolute costs, reflecting the increased mass that must gain height during climbing. Thus, larger animals may find climbing to be more physically challenging compared with smaller species, making climbing a less efficient mode of locomotion for larger animals.

Given that the energetic costs associated with climbing are largely predetermined by gaining height, anatomical and kinematic changes centered on COM modulation have limited impact on the efficiency of climbing. This aligns with previous observations by Pontzer and Wrangham (2004) suggesting that chimpanzee anatomy is not selected to minimize the locomotor costs of arboreal locomotion. Instead, maneuverability, predator avoidance and stability should be considered the primary drivers of their locomotor morphology (Pontzer and Wrangham, 2004). Similarly, recent work on spatiotemporal gait characteristics in climbing mammals broadly shows characteristics consistent with increasing stability rather than energy-saving mechanisms or means to increase speed (Granatosky et al., 2019, 2021; Karantanis et al., 2016). Indeed, such considerations between safety over efficiency extend to non-mammalian tetrapods (Byrnes and Jayne, 2014).

In light of these findings, we argue that anatomical features associated with climbing should primarily be characterized by autopodial adaptations to increase stability. Such mechanisms involve changes inherent to the autopodia themselves and encompass the evolution of intrinsic mechanisms of adhesion and suction (e.g. setae in climbing geckos, mucus adhesion in arboreal frogs, wet adhesion in sucker-footed bats; Autumn et al., 2002; Federle et al., 2006; Mengüç et al., 2012; Riskin and Racey, 2010), specialized claws to generate passive clinging forces (Turnbull et al., 2023), or volar adaptations to the palm and sole in primates that increase dermal friction of the autopodia (Cartmill, 1974, 1985). Indirect mechanisms include adaptations that allow animals to utilize the autopodia in a more stable manner. For example, increasing dorsiflexion capabilities within the ankle joint through modifications of ankle flexor musculature maximize volar contact of the foot against the substrate during climbing (Venkataraman et al., 2013). Similarly, the siamang wrist exhibits a dorsopalmarly expanded triquetrum to protect against hypermobility and provide a stable platform for weight transfer through the ulna during climbing (Vanhoof et al., 2021). Finally, as discussed earlier, decreased body mass represents a global adaptation to reduce the fitness costs associated with falling. However, this may not be a climbing-specific adaptation but rather a feature associated broadly with arboreality (Cant, 1992).

This experimental design is faced with several limitations that must be addressed with regards to the interpretation of these data. Firstly, participants climbed a non-motorized, mobile treadwall, meaning downward motion of the wall was driven by gravitational acceleration acting on the participant's body mass against a set resistance. Further, the treadwall presents the climber with substrate conditions of limited ecological relevance (Barrey et al., 1993; Semaan et al., 2022). Therefore, it should be noted these findings may have limited evolutionary implications. Despite these limitations, use of the treadwall circumvents the need to introduce periods of non-climbing to bring the climbers back to the ground and restart climbing in energetic trials (Kozma and Pontzer, 2021). Additionally, participants chose the speed at which they felt they could sustain climbing for a minimum of 3 min and we kept the same resistance across all three climbing conditions to minimize speed differences between trials (0.18±0.05 m s⁻¹). As a result of speed self-selection, some individuals may have longer contact times and duty factors than those climbing at faster speeds; however, we used linear mixed effect models to mitigate these effects.

Secondly, the experimental conditions were intended to alter COM position in the normal plane, with the ladder climbing condition meant to force participants' COM far from the treadwall, and the sprawling condition selected to force participants' COM to be close to the treadwall. The most likely explanation can be drawn from the COM positional trackings (see Fig. 6). There was a tremendous range of variation in COM positioning both from stride to stride and, most prominently, between individuals. During trials, active behavioral modulation (e.g. degree of joint bending, climbing strategy) was essential and allowed each individual to adjust for unexpected instability (e.g. placement of hand and foot on rungs that rendered the participant momentarily unstable) and their level of fatigue. We did successfully achieve the highest normal COM maximums in the ladder climbing condition (0.52±0.07 m, 0.16–0.67 m); however, the sprawling condition (0.48±0.06 m, 0.17–0.60 m) did not minimize normal COM movements, rather falling intermediary to ladder climbing and free choice (0.44±0.07 m, 0.19–0.6 m), though all three experimental conditions were statistically distinct (all P<0.003). Rather than achieving the extremes in COM positioning in the normal plane, we did manage to create variation in COM position (∼0.20–0.70 m) and assess its effect on metabolic costs. Future experiments should be directed at creating climbing conditions that capture the extreme of the desired perturbations in distance from the wall.

Lastly, our participants ranged in climbing experience between no climbing experience and IRCRA level 22, equivalent to V7 following the Vermin scale or 5.13a following the Yosemite Decimal System (Draper et al., 2015). We included climbing experience as a fixed effect in our linear mixed effect models; however, the response variables (COL_net, COT_net and COL_net) were not driven by expertise in any model. The results of including experience as a fixed effect in the models must be taken with caution, however, as recruitment of individuals to participate in this study was difficult and in our sample of 14 individuals, only four could be characterized as experienced. We hope to control for these disparities by including individual as a random effect in each linear mixed effect model.

In this study, we manipulated experimental climbing conditions to induce variation in COM position and assess the metabolic impact of these changes. We observed that variation in COM position had minimal effects on mass-specific locomotor costs and overall metabolic costs per stride. Instead, these were largely driven by velocity. We also observed a clear disconnect between mechanical energy costs and absolute metabolic costs, which we ascribe to the major difference in magnitude between these two profiles. Collectively, we argue that our findings support previous assertions that stability and safety – as opposed to energetic efficiency – are the primary factors selected for during climbing.

We would like to thank Alex Lopez for access to Inclusive Sports and Fitness and the treadwall used within this study. We would also like to thank New York Institute of Technology College of Osteopathic Medicine for making this study possible by purchasing the COSMED K5 respirometry system. Additionally, we would like to thank the participants who donated their time and efforts for this strenuous activity. We would also like to acknowledge Inclusive Sports and Fitness and the New York Institute of Technology Center for Biomedical Innovation for co-funding this study. Lastly, we thank the two anonymous reviewers, and handling editor Monica Daley for their insightful comments to improve the quality of the manuscript.