2.1. Comparative behavioural approaches to thinking about the origins of bipedalism

The closest possible correlate of potential locomotor modes in the past is locomotion among extant species of apes and humans. As such, models of how and when hominin bipedalism emerged are frequently based on consideration of the characteristics and locomotor modes of extant apes. A long-standing and influential model sees the last common ancestor (LCA) of hominins and great apes as similar to an African ape, with a significant terrestrial knuckle-walking component to its locomotion (Richmond, Begun, & Strait, Reference Richmond, Begun and Strait2001; Washburn, Reference Washburn1967). When African apes move bipedally, they do so with a bent-hip–bent-knee posture that generates substantial torque about the flexed hip and knee that must be counteracted by sufficient muscle force to prevent the leg collapsing (Sockol, Raichlen, & Pontzer, Reference Sockol, Raichlen and Pontzer2007). Humans walk bipedally with an extended leg posture, reducing the muscle forces required and thus the associated energetic costs (Pontzer, Raichlen, & Sockol, Reference Pontzer, Raichlen and Sockol2009), providing a selection pressure that could theoretically drive the evolution of human-like bipedal walking from a knuckle-walking African-ape-like ancestor. The locomotion of Asian apes like hylobatids (gibbons and siamangs) and orangutans often involves arboreal bipedalism, or walking upright along the top of branches with an orthograde posture and an extended lower limb (Thorpe, Holder, & Crompton, Reference Thorpe, Holder and Crompton2007). Alternative models of the LCA that emphasise arboreal bipedalism, like those based on a hylobatian LCA (Tuttle, Reference Tuttle1974), see an easier transition to hominin bipedality if the LCA is already frequently using a bipedal posture and extended leg in the trees. The kinematics and locomotion of the LCA is thus a matter of keen debate: was it predominantly terrestrial or arboreal? When it stood upright, did it adopt the flexed with a bent-hip–bent-knee posture of an ape, or the more extended limb posture of a human?

Seeking answers to these questions from skeletal remains alone is not straightforward; as mentioned above, structures specialised for a variety of locomotor modes can function in multiple and overlapping ways that are not necessarily reflected in bone structure alone. In this Special Collection, Rosen and colleagues (2022) approach the question of the LCA's locomotor mode through the examination of bipedal frequency in extant zoo-housed primate taxa. The most arboreal among extant primates, hylobatids and orangutans, both utilise an orthograde posture and frequent arboreal bipedalism involving an extended leg posture. However, hylobatids have an additional trait key to human bipedalism: a long and flexible lumbar spine. It has previously been argued that these traits, shared by two taxa who now have distinct primary locomotor modes (bipedalism and brachiation), reflect the ancestral condition (e.g. hylobatian model; Tuttle, Reference Tuttle1974), but Rosen and colleagues explore whether or not it is arboreal individuals with these traits already who are most likely to move bipedally when on the ground. Among extant apes housed in zoos, they observed bipedal behaviour on the ground more often, and for longer per bout, among hylobatids than bonobos, chimpanzees, gorillas, orangutans or cercopithecines (Rosen et al., Reference Rosen, Jones and DeSilva2022). Of the 93 hylobatid individuals observed, bipedal bouts were documented among 83% of them, on average almost twice a day, and more than 35% of these bouts involved eight or more bipedal strides. Although strong selection could feasibly favour the evolution of obligate bipedalism from any of the observed taxa, as they all exhibited some bipedal behaviour, only the hylobatids already exhibited both an orthograde posture and a long and flexible lower back. An LCA with these traits as well may have been similarly predisposed to bipedal behaviour, in which case the orthograde posture and long lumbar spine of humans and hylobatids would be ancestral, and the short stiff lumbar spine and more pronograde posture of knuckle-walking apes would be the derived condition.

This behavioural work stimulates many interesting questions about the nature of the body form from which habitual bipedalism emerged, but also again highlights one of the main issues with the form–function paradigm that forms the central theme of this Special Collection: different skeletal structures can perform a given function reasonably well and can also simultaneously be used for a set of other independent functions. Even if a perfectly preserved fossil ape from exactly the right time was discovered, immediately before the divergence between hominin and great ape branches, and it had a long, flexible lumbar spine and seemingly an orthograde posture, this morphology can clearly be used in locomotor modes as diverse as habitual bipedalism and brachiation. The presence of one trait alone is not enough; approaches are necessary that can consider complexes of traits working together in flexible ways, as well as reconstruct soft tissue's role in both enabling functional diversity and generating structural variation in bone.

3.1. Mapping functional patterns of muscle activity

The strains experienced by a bone in the performance of a given behaviour are generated by a complex pattern of coordinated muscle activity within limbs and across joints. However, the direct in vivo measurement of bone strain produced during the performance of a particular task in humans is understandably uncommon (see review in Al Nazer, Lanovaz, Kawalilak, Johnston, & Kontulainen, Reference Al Nazer, Lanovaz, Kawalilak, Johnston and Kontulainen2012). Electromyography (EMG) has proven useful over the past decade in biological anthropology in assessing experimentally measured patterns of muscle activity during the performance of tasks important among past populations, to which patterns of bone functional adaptation thought to be generated by that task can be compared. Most work to this end thus far has focused on inferred skeletal signatures of key hunting and food-processing technologies, like querns, spears and projectile weaponry. Such research has provided support for behavioural interpretations linking changes in humeral asymmetry among European agricultural women to technological shifts in grain-grinding technology (Sládek, Hora, Farkasová, & Rocek, Reference Sládek, Hora, Farkasová and Rocek2016), but it has also challenged influential and longstanding hypotheses about the role of underhanded spear thrusting (Churchill et al., Reference Churchill, Weaver and Niewohner1996) in the pronounced humeral asymmetry among Neanderthals (Shaw, Hofmann, Petraglia, Stock, & Gottschall, Reference Shaw, Hofmann, Petraglia, Stock and Gottschall2012).

The development of projectile weaponry like the bow and arrow represented both a shift in cognitive complexity, being made and used only by Homo sapiens, and a key piece of the toolkit enabling human dispersal and adaptive success (Williams, Burke, & Lombard Reference Williams, Burke and Lombard2014; Shea, Reference Shea2006). Being predominantly made from organic materials, archaeological evidence of archery usually relies on finding stone, bone or metal arrowheads. However, the use of the bow and arrow requires substantial muscle force production at the elbow and shoulder in the draw (dominant) arm, force known as the draw weight that, for composite and longbows in the past, is in the range 90–130 lb or 40–60 kg (Hardy, Reference Hardy1992; Karpowicz, Reference Karpowicz2007; Soar, Reference Soar2006). The bow (non-dominant) arm must brace against these large forces to keep the arm straight, so archery may be detectable in asymmetrical humeral functional adaptation. Among the humeri of males thought to have performed archery, a pattern of relatively symmetrical responses to loading intensity (limb bone bending strength, external dimensions, total areas etc.) but asymmetrical responses to loading directionality (limb bone shape ratio) has been documented (Rhodes & Knüsel, Reference Rhodes and Knüsel2005; Stirland, Reference Stirland1993). Whether or not the performance of archery actually does involve symmetrical upper limb load intensity but asymmetrical load directionality is unclear, however. In this Special Collection, Tabitha Dorshorst and colleagues (2022) utilised EMG analysis to investigate the patterns of muscle activity at the shoulder and elbow in both upper limbs of participants throughout the drawing of a bow. Peak percentages of maximum voluntary contraction in the muscles of the bow and draw arms were similar, but were generated by the joint agonist in one arm and its antagonist in the other arm (Dorshorst, Weir, Hamill, & Holt, Reference Dorshorst, Weir, Hamill and Holt2022). This pattern of muscle activity about the shoulder and elbow during archery is consistent with signatures of symmetrical loading intensity overall but asymmetrical directionality inferred from skeletal remains.

3.2. Modelling activity-related external and internal forces

Although useful, EMG does not directly measure internal bone loading, as this is very difficult to measure in vivo. A variety of other parameters are also difficult to measure in vivo, like muscle–tendon lengths, moment arms and soft tissue deformation, all of which affect forces, kinematics and costs during locomotion. Musculoskeletal (MSk) modelling provides a computational means of estimating such parameters during the performance of a given movement (Arnold et al., Reference Arnold, Salinas, Hakawa and Delp2000; Wehner, Claes, & Simon, Reference Wehner, Claes and Simon2009; Baggaley et al., Reference Baggaley, Derrick, Vernillo, Millet and Edwards2022), so is a powerful tool in trying to better understand locomotor kinematics in the past. Using a combination of motion equations and muscle force prediction equations, MSk techniques can determine a range of kinematic and soft tissue parameters that would exist within a subject when conducting any specific activity that can be virtually defined or measured experimentally (e.g. walking). To do so, an MSk model typically requires detailed information on a subject's musculoskeletal anatomy, from cadaveric dissection, CT imaging, or magnetic resonance imaging. These anatomical data are then paired with experimental data on gait and muscle parameters to create a representative model of a subject moving; an MSk model climbing stairs is presented in Figure 2. The greatest accuracy is achieved by collecting all input data from the same individual (Charles, Grant, D'Août, & Bates, Reference Charles, Grant, D'Août and Bates2020), although this is not always possible. Even with a full subject-specific dataset, MSk models are sensitive to error in bony landmark identification, muscle tension and muscle geometries, all of which affect the calculated muscle and joint moments and forces (Broyde et al., Reference Broyde, Dempsey, Wang, Cox, Fagan and Bates2021; Martelli, Valente, Viceconti, & Taddei, Reference Martelli, Valente, Viceconti and Taddei2015; Valente et al., Reference Valente, Pitto, Testi, Seth, Delp, Stagni and Taddei2014). Even just the removal of soft tissue during dissection can change muscle/tendon paths enough to introduce large errors in muscle via points in an MSk model (Synek et al., Reference Synek, Lu, Vereecke, Nauwelaerts, Kivell and Pahr2019).

Figure 2. An example of a musculoskeletal model at different points in a stair-climbing simulation. Pink circles indicate the position of reflective motion capture markers placed at anatomical landmarks on the subject in life as they performed the movement.

Obviously, the sensitivity of even subject-specific models to estimation errors is a major challenge facing the effective use of MSk modelling among extinct fossil species, in whom bony preservation may be poor and muscle and gait variables are simply unknown. Existing efforts to simulate locomotor biomechanics in fossil hominins have faced difficulties in determining segment lengths, relative body segment inertial parameters, passive joint properties, muscle characteristics, muscle attachment locations and even the anatomy of entire limb segments that were missing (Nagano, Umberger, Marzke, & Gerritsen, Reference Nagano, Umberger, Marzke and Gerritsen2005; Wang et al., Reference Wang, Crompton, Carey, Günther, Li, Savage and Sellers2004). In order to overcome these issues, assumptions must be made that extant human or non-human primate data would have been similar enough to act as an appropriate substitute. Human or ape muscle, joint and relative body segment parameters can then be scaled to a given fossil's dimensions, and their experimentally collected gait data used to simulate a fossil walking (Nagano et al., Reference Nagano, Umberger, Marzke and Gerritsen2005; Wang et al., Reference Wang, Crompton, Carey, Günther, Li, Savage and Sellers2004). This practice can contribute to the already considerable disagreement that exists about hominin locomotion; for example, that australopithecines walked both with a flexed, bent-hip–bent-knee bipedal gait (Stern, Reference Stern2000; Stern & Susman, Reference Stern and Susman1983) and with an extended gait (Crompton, Yu, Weijie, Günther, & Savage, Reference Crompton, Yu, Weijie, Günther and Savage1998; DeSilva et al., Reference DeSilva, Holt, Churchill, Carlson, Walker, Zipfel and Berger2013; Raichlen, Gordon, Harcourt-Smith, Foster, & Haas, Reference Raichlen, Gordon, Harcourt-Smith, Foster and Haas2010), and that they walked bipedally both less efficiently than a human (Nagano et al., Reference Nagano, Umberger, Marzke and Gerritsen2005; Pontzer et al., Reference Pontzer, Raichlen and Sockol2009; Wang et al., Reference Wang, Crompton, Carey, Günther, Li, Savage and Sellers2004) and similarly or even more efficiently (Sellers, Cain, Wang, & Crompton, Reference Sellers, Cain, Wang and Crompton2005; Kramer, Reference Kramer1999; Kramer & Eck, Reference Kramer and Eck2000).

Considering reference data for something like muscle architecture from one species as representative of that of another is obviously not ideal, but is currently the only option available to us when considering extinct species. However, this is a problem even when modelling extant human kinematics: when collecting subject-specific datasets is not possible, most of the reference data for ‘generic’ human muscle parameters come from cadaveric dissections of elderly individuals (Klein Horsman, Koopman, van der Helm, Prosé, & Veeger Reference Klein Horsman, Koopman, van der Helm, Prosé and Veeger2007; Arnold, Ward, Lieber, & Delp, Reference Arnold, Ward, Lieber and Delp2010; Rajagopal et al., Reference Rajagopal, Dembia, DeMers, Delp, Hicks and Delp2016). These muscle parameters are probably not representative of most humans, as not only does muscle architecture (volume, length, cross-sectional area, fiber length, pennation angle, etc.) vary substantially, even among individual humans (Charles, Suntaxi, & Anderst, Reference Charles, Suntaxi and Anderst2019), it changes with age (Moore et al., Reference Moore, Caturegli, Metter, Makrogiannis, Resnick, Harris and Ferrucci2014; Narici, Maffulli, & Maganaris, Reference Narici, Maffulli and Maganaris2008). Compared with using elderly generic muscle architecture data, averaged data collected from young, healthy individuals (Charles et al., Reference Charles, Suntaxi and Anderst2019) does predict experimentally measured subject-specific isokinetic muscle torques to a higher degree of accuracy (Charles et al., Reference Charles, Grant, D'Août and Bates2020). However, even generic data from young adults still introduced error relative to subject-specific muscle data for some movements at the ankle, knee and hip. In this Special Collection, Patricia Kramer, Feuerriegel, Lautzenheiser, and Sylvester (2022) test the extent to which variation in the muscle architecture parameters of functional groupings of lower limb muscles influenced joint reaction force and muscle force estimates in complex movements. Kramer and colleagues modified a previously existing baseline MSk model to the size and proportions of 10 healthy volunteers, to which subject-specific motion and force data, and cadaveric muscle architecture parameters, were applied. The resulting joint and muscle forces of these models were compared with the same models but with generic young adult muscle architecture parameters applied instead (Charles et al. Reference Charles, Suntaxi and Anderst2019). Both across models and among individuals, muscle force patterns from the large lower limb functional muscle groups remained consistent across the gait, despite variation in the muscle architectural data used. However, the magnitude of those forces was affected by muscle parameter choices, with muscle and joint reaction forces varying as much as 15% depending on which muscle architecture dataset was applied. For some purposes, such as exploring shape differences in joint morphology among groups, this variation that muscle parameter choices induced in the resulting joint reaction forces may pose problems. In these instances, it is particularly important that the external kinematics, kinetics and muscle parameters used when comparing models are kept consistent (Kramer et al., Reference Kramer, Feuerriegel, Lautzenheiser and Sylvester2022).

4.1. The impact of walking with others and carrying load

Not only does a broad pelvis not necessarily sacrifice locomotor efficiency, it in fact provides a distinct advantage in two particular conditions: when walking with others and when carrying load. Neither of these conditions were considered in traditional biomechanically modelled or experimentally measured costs of locomotion, yet in reality, humans often walk with other humans, some or all of whom may be carrying loads while doing so. Both walking with others and carrying loads can affect the cost of walking: carrying reproductive-sized loads (pregnancy, child) slows your optimal walking speed, makes it more costly and increases the energetic penalty for walking at a suboptimal pace (Wall-Scheffler & Myers, Reference Wall-Scheffler and Myers2013), something which you may need to do when walking with other individuals (Wagnild & Wall-Scheffler, Reference Wagnild and Wall-Scheffler2013). These challenges are cross-cultural, and several solutions seem to have been universally converged upon. All societies have some type of infant-carrying device to reduce the cost of load carrying, like slings, wraps, cradleboards, etc. (Dettwyler, Reference Dettwyler1988; Konner, Reference Konner, Lee and DeVore1976; Lewis-Williams & Bannister, Reference Lewis-Williams and Bannister1991; van Hout, Reference van Hout2015), which can reduce the cost of carrying a child by 16% compared with carrying in the arms (Wall-Scheffler, Geiger, and Steudel-Numbers, Reference Wall-Scheffler, Geiger and Steudel-Numbers2007). Without an assistive device, the cost of carrying loads in the arms can be reduced instead by a broader pelvis, with savings of up to 12% (Wall-Scheffler et al., Reference Wall-Scheffler, Geiger and Steudel-Numbers2007). A broad pelvis also seems to confer greater speed flexibility (Wall-Scheffler & Myers, Reference Wall-Scheffler and Myers2013), which is important in keeping locomotor costs down when walking in a group. When the conditions of load-carrying and group mobility are incorporated into models of the cost of bipedal locomotion, a new perspective on the notably broad pelves of both male and female fossil hominins emerges (Wall-Scheffler, Reference Wall-Scheffler2012a; Wall-Scheffler et al., Reference Wall-Scheffler, Geiger and Steudel-Numbers2007; Wall-Scheffler & Myers, Reference Wall-Scheffler and Myers2013).

In this Special Collection, Cara Wall-Scheffler (Reference Wall-Scheffler2022) explores the extent to which the elevated metabolic costs of carrying a child in a sling are affected by interactions between the terrain over which you are carrying and the position of the child on the body while doing so. Over a 1000 m section of natural undulating terrain, a dorsal-carrying position was the least expensive for both sexes relative to front- or side-loading, and was the only position that did not require the parent to slow down their overall walking speed (Wall-Scheffler, Reference Wall-Scheffler2022). When encountering inclined sections of trail, parents carrying their children altered their kinematics regardless of load position, spending a significantly greater proportion of each stride in the stance rather than swing phase. This change was most dramatic for dorsal carrying, but for all positions, spending more time in stance phase probably enables better stability on a variable surface when carrying a precious cargo. In all positions (dorsal, side, front), mothers were able to carry their child more economically and more efficiently than fathers were, reflecting the relationship between a relatively wider pelvic breadth and reduced metabolic costs when carrying. The strategy that individuals employ to maximise their locomotor economy and efficiency is clearly highly variable, depending on their particular morphology, what they are carrying, where they are carrying it on their body, the terrain over which they are doing so and the kinematics that they employ. This complexity and also flexibility is simply not well captured in traditional biomechanical models of unloaded walking in fixed laboratory conditions.

4.2. The impact of substrate and terrain

Typically, locomotor costs when walking also tend to be derived from a subject moving not just alone and unloaded, but across a hard flat surface. On this substrate, long legs are associated with a lower cost of transport (Steudel-Numbers & Tilkens, Reference Steudel-Numbers and Tilkens2004; Steudel-Numbers, Weaver, & Wall-Scheffler, Reference Steudel-Numbers, Weaver and Wall-Scheffler2007). This aspect of the relationship between morphology and energetics has been influential in interpreting some of the earliest and most distinctive phenotypic changes associated with our genus Homo: the elongation of the lower limb and increases in stature. Based on the costs of moving on flat, hard surfaces, very short stature and short legs would reduce stride length, and thus increase the cost of moving the body a given distance. However, this makes it very difficult to explain the convergent evolution of particularly small body size repeatedly in modern humans independently in multiple different populations around the globe (Perry & Dominy, Reference Perry and Dominy2009), as well as among other hominins too, for example SE Asian hominins like Homo floresiensis (Brown et al., Reference Brown, Sutikna, Morwood, Soejono, Jatmiko, Saptomo and Due2004) and the recently discovered Homo luzonensis (Détroit et al., Reference Détroit, Mijares, Corny, Daver, Zanolli, Dizon and Piper2019). The answer may lie in the fact that not all substrates on which hominins and humans move(d) are hard and flat; when that condition changes, so too do the kinematics and energetic costs of locomotion.

Extinct and extant small-bodied populations have something in common: they live(d) in tropical rainforest, where vegetation is dense and the substrate is irregular and often compliant. Venkataraman et al. (Reference Venkataraman, Yegian, Wallace, Holowka, Tacey, Gurven and Kraft2018) used experimental data from the Batek and the Tsimane, two small-bodied human populations who forage in rainforest, to demonstrate that step length is constrained in such dense vegetation, forcing taller individuals to walk slower than their optimum speed. The energetic costs associated with constrained step length in this environment were not directly measured, but in a laboratory setting, walking on a soft, rather than hard, substrate eliminates the significant relationship between leg length and a lower cost of transport (Charles, Grant, D'Août, & Bates, Reference Charles, Grant, D'Août and Bates2021). Locomotion on variable surfaces is also associated with a range of kinematic alterations that affect the work done by the muscles. On soft surfaces like sand, although less muscular effort is required to provide cushioning (Tung, Franz, & Kram, Reference Tung, Franz and Kram2014), greater muscular work is required to control unwanted movement of the foot during footstrike and toe-off (Lejeune, Willems, & Heglund, Reference Lejeune, Willems and Heglund1998). The elastic recoil mechanisms of muscle–tendon units, crucial to human locomotor efficiency, are also dampened in substrates like sand (Lejeune et al., Reference Lejeune, Willems and Heglund1998), further increasing locomotor costs. Stride lengths tend to become shorter or more variable on soft or uneven surfaces (Dhawale & Venkadesan, Reference Dhawale and Venkadesan2021; Pinnington, Lloyd, Besier, & Dawson, Reference Pinnington, Lloyd, Besier and Dawson2005; Voloshina & Ferris, Reference Voloshina and Ferris2015; Voloshina, Kuo, Daley, & Ferris, Reference Voloshina, Kuo, Daley and Ferris2013), and leg posture can become more flexed (Pinnington et al., Reference Pinnington, Lloyd, Besier and Dawson2005). These kinematic changes in relation to substrate typically increase the metabolic costs of locomotion (Davies & Mackinnon, Reference Davies and Mackinnon2006; Gast, Kram, & Riemer, Reference Gast, Kram and Riemer2019; Lejeune et al., Reference Lejeune, Willems and Heglund1998; Lussiana, Fabre, Hébert-Losier, & Mourot, Reference Lussiana, Fabre, Hébert-Losier and Mourot2013; Voloshina & Ferris, Reference Voloshina and Ferris2015; Voloshina et al., Reference Voloshina, Kuo, Daley and Ferris2013), although not always (Dhawale & Venkadesan, Reference Dhawale and Venkadesan2021). However, the human locomotor system exhibits adaptability in key parameters like leg stiffness to accommodate changes in substrate and keep locomotor costs more stable: on uneven surfaces, runners reduce their joint stiffness (Dhawale & Venkadesan, Reference Dhawale and Venkadesan2021), but on soft substrates they increase it (Ferris, Louie, & Farley, Reference Ferris, Louie and Farley1998).

None of this variation in substrate, kinematics and costs is typically accounted for in traditional biomechanical models, highlighting the importance of research in more natural settings and conditions more typical of most of human evolution. In this Special Collection, Nick Holowka and colleagues (2022) explored how kinematics vary in relation to substrate in a natural forested setting, by assessing walking kinematics among the Tsimane of the Bolivian Amazon. Importantly, they quantified lower limb kinematics when walking not just across unbroken forest understory and an open hard-packed dirt field, but on forest trail as well. Trails are made and used by many human populations in heavily forested areas, probably because they provide some benefit to the energetic costs of locomotion and/or travel time, but these advantages are not well understood. Substantial changes in lower limb kinematics were evident when walking in unbroken forest compared with a hard-packed flat and open fields: the hips, knees and ankles were more flexed, the trunk more inclined, the foot swing higher and the foot strike flatter. These are in sharp contrast to the kinematics, like an extended leg, thought to be so crucial to energetically efficient bipedalism. Although walking with more flexed lower limb joints, inclined torso, a high-stepping gait and flatter foot strike would probably increase the cost of locomotion, in unbroken forest these changes help reduce the chances of tripping, lower the centre of gravity to aid balance and provide greater tactile sensitivity to the substrate, which are clear benefits when walking unshod on unstable and uneven surfaces. Forest trails mitigated some of the need for suboptimal walking kinematics, yet their construction and maintenance add their own energy costs, so more research is needed to determine how foraging strategies balance these costs and benefits in natural environments. However, for hominins moving on variable natural surfaces, some of them irregular, uneven, steep and/or slippery, it may be that the energetic advantages of bipedalism are more conservative, or at least more variable, than typical estimates of locomotor costs based only on walking over a flat open and hard substrate (Holowka, Kraft, Wallace, Gurven, & Venkataraman, Reference Holowka, Kraft, Wallace, Gurven and Venkataraman2022).

The importance of considering the environment in which a species moved when estimating the cost of locomotion is particularly crucial when interpreting the selective factors shaping diversity among hominins with wide geographic ranges, or between hominin species of African and Eurasian origins. For example, Homo neanderthalensis, a Eurasian species, had larger, broader bodies and relatively short legs and distal limb segments relative to Homo sapiens, a species of African origins (Holliday, Reference Holliday1997). As long distal limb segments, and relatively long limbs, are an advantage over flat terrain (Steudel-Numbers & Tilkens, Reference Steudel-Numbers and Tilkens2004; Steudel-Numbers et al., Reference Steudel-Numbers, Weaver and Wall-Scheffler2007), these differences are estimated to have increased locomotor costs for Neanderthals by up to 30% (Steudel-Numbers & Tilkens, Reference Steudel-Numbers and Tilkens2004). In contrast, short limbs, short distal limb segments and a large broad body provide a large volume of heat-generating tissue and minimal surface area through which to lose it, conferring a low surface-area-to-volume ratio that is a major advantage in cool conditions. Thus, the relative impact of thermoregulatory and locomotor selection pressures on the characteristic morphological differences between Neanderthals and humans is a matter of keen interest (Ocobock, Lacy, & Niclou, Reference Ocobock, Lacy and Niclou2021; Tilkens, Wall-Scheffler, Weaver, & Steudel-Numbers, Reference Tilkens, Wall-Scheffler, Weaver and Steudel-Numbers2007; Weaver & Steudel-Numbers, Reference Weaver and Steudel-Numbers2005). The interpretation of Neanderthal morphology must also take into account that some of the Neanderthals’ Eurasian range was not open and flat, but mountainous. Biomechanical modelling of lower limb forces when walking over uneven and steep terrain suggests that, in these conditions, having a shorter distal limb segment can be advantageous (Higgins & Ruff, Reference Higgins and Ruff2011). It is thus likely that Neanderthal limb and body proportions reflect optimisation to interrelating climatic and locomotion-related selection pressures. Among humans, this inter-relationship is probably even more complex, as we are a species of African origin now inhabiting every environment on the globe, and one for whom foraging strategies are extremely diverse and probably included a unique strategy: sustained running in the heat.

4.3. The interrelationship between locomotion and thermoregulation

Adaptations among humans enabling heat dissipation during sustained running are thought to have lifted one of the main constraints typically limiting the running abilities of other species (Carrier, Reference Carrier1984). Humans differ from most quadrupeds in the mechanisms through which we dissipate heat when running, including a prodigious ability to sweat that, unlike panting, has allowed evaporative heat loss over a large surface area independent of breathing (Taylor & Rowntree, Reference Taylor and Rowntree1973). The rate at which heat is lost is also maximised by our reduced body hair, enhancing both evaporative and convective cooling (Carrier, Reference Carrier1984). Ecogeographic patterning in surface-area-to-volume ratios of the body is also evident among humans globally, conforming broadly to Bergmann's and Allen's rules (Allen, Reference Allen1877; Bergmann, Reference Bergmann1847) and enhancing passive thermoregulation. Human physiology as well as body size, shape and proportions matter not just for locomotor efficiency while walking or running, but for our ability to efficiently dissipate the metabolic heat generated during locomotion in a particular environment as well. How are these costs reflected in the morphological variation of a single species with such a broad geographic range and diversity of foraging behaviours?

One unusual interdisciplinary way of investigating these questions is highlighted by a pair of papers originally published in 2019 and 2021 of which I was a part, which are included in this Special Collection as well. Together with Daniel Longman, Jay Stock and colleagues, we explored relationships between running performance and thermally optimised variation in body size, shape and proportions among ultrarunners racing in two extreme conditions: heat (southern Spain, Peruvian Amazon) and cold (Finnish Lapland, Nepalese Himalayas). Female athletes who successfully finished the hot races had significantly lower body mass and weight-for-height than those who finished the cold races, but significantly longer legs for a given height. On average then, female finishers in hot conditions had less tissue generating heat and more surface area through which to dissipate it, and the converse was true of finishers in cold conditions. Male athletes demonstrated similar patterns in relative leg length. However, participants racing in a given climate tended to have morphology more optimised for that climate in the first place, so these relationships could simply be reflecting self-selection. We thus also compared morphological traits among finishers and non-finishers in two consecutive years of the Al Andalus Ultimate Trail race, a multi-day 230 km semi-supported race. This race took place in hot, dry and relatively flat southern Andalucia at the height of summer, where daytime temperatures were in excess of 40°C. In these conditions, relationships between successful performance and morphology optimised for minimal heat generation and maximal heat loss were evident: all finishers had significantly longer legs for a given height than athletes who dropped out of the race (Longman et al., Reference Longman, Macintosh Murray, Roberts, Oakley, Wells and Stock2019), while female finishers were also significantly lighter and leaner than females who did not complete the race (Longman et al., Reference Longman, Murray, Roberts, Oakley, Wells and Stock2021). Thus, the morphological trends demonstrated between finishers of hot and cold races are probably reflecting at least some influence of thermally adapted morphology on performance.

Physical exertion for 230 km in temperatures exceeding 40°C clearly places individuals at the extremes of their thermoregulatory tolerance, where any trait that passively dissipates heat with minimal cost is likely to be an advantage. Morphology may reflect the costs of locomotion, but not independently of the environment in which that locomotion is occurring. The body size and proportions that were associated with better performance in the hot, dry and flat race described above (only 7.1 km of ascent over 230+ km of running) cannot be assumed to be the case in other environmental contexts. One of the races from which we recruited participants in the ‘cold’ group, the Everest Trail Race, starts at 2370 m of elevation, and over the course of 153 km racers climb 13.5 km and descend 13 km (Longman et al., Reference Longman, Murray, Roberts, Oakley, Wells and Stock2021). Sample sizes from this race were small, with seven females and 19 males taking part in the study, so relationships between performance and morphology were not explored in this group on its own. However, a brief perusal of trends among these mountain ultramarathon racers reveals that, for a given height, finishers were on average 8–8.5% heavier with 2.5% shorter legs relative to non-finishers of the same sex (unpublished data). If larger sample sizes demonstrated this to be a significant difference, it would be quite in opposition to the morphology that was associated with completion in the hot and flat conditions of the Al Andalus Ultimate Trail race.

Interestingly, the relationship between performance and thermally optimal morphology in a given set of environmental conditions may also vary by sex. In hot, dry and flat conditions, female ultramarathon runners demonstrated more widespread and pronounced relationships between performance and body size and limb proportions than males (Longman et al., Reference Longman, Murray, Roberts, Oakley, Wells and Stock2021). Differential reproductive endocrinology and energetics between males and females is well known (Bogin, Reference Bogin2021; Wells, Reference Wells2006), leading to characteristic patterns of sexual dimorphism in body size (Cutler, Reference Cutler1997; Eveleth, Reference Eveleth1975) and body composition (Norgan, Reference Norgan1997; Wells, Reference Wells2007) that emerge around puberty (Bogin, Reference Bogin2021). Across cultures, females tend to be smaller than males on average while carrying relatively less lean mass and relatively higher levels of peripheral subcutaneous fat (Taylor, Grant, Williams & Goulding, Reference Taylor, Grant, Williams and Goulding2010; Eveleth, Reference Eveleth1975). In endurance activities in the heat then, the smaller body size and lower lean mass for a given height of females relative to males should minimise locomotor and thermoregulatory costs, while their large reproductive fat reserves should maximise the relevant energy supply (Wall-Scheffler, Reference Wall-Scheffler2012b; Schütz et al., Reference Schütz, Billich, König, Würslin, Wiedelbach, Brambs and Machann2013). Among the athletes racing in the hot-weather ultramarathon who participated in our study, 65% of the 20 females who started the race completed it, while only half of the 38 males who started the race successfully completed it (Longman et al., Reference Longman, Macintosh Murray, Roberts, Oakley, Wells and Stock2019, 2021).

Ultraendurance racing is still quite a male-dominated sport (Zingg, Knechtle, Rosemann, & Rüst, Reference Zingg, Knechtle, Rosemann and Rüst2015), with participation rates for females typically around ~10–20% (da Fonseca-Engelhardt et al., Reference da Fonseca-Engelhardt, Knechtle, Rüst, Knechtle, Lepers and Rosemann2013; Peter, Rust, Knechtle, Rosemann, & Lepers, Reference Peter, Rust, Knechtle, Rosemann and Lepers2014; Shoak et al., Reference Shoak, Knechtle, Knechtle, Rüst, Rosemann and Lepers2013). However, despite this much lower participation, the performance gap between male and female finishing times in ultraevents can be extremely low, from around 4–6% in ultrarunning and ultraswimming (Waldvogel, Nikolaidis, Gangi & Rosemann, Reference Waldvogel, Nikolaidis, Gangi and Rosemann2019; Peter et al., Reference Peter, Rust, Knechtle, Rosemann and Lepers2014; Zingg et al., Reference Zingg, Rüst, Rosemann, Lepers and Knechtle2014) to virtually non-existent in ultra-long-distance cycling races of 400+ miles (Baumgartner, Victor Sousa, Nikolaidis, & Knechtle, Reference Baumgartner, Victor Sousa, Nikolaidis and Knechtle2020). Females are particularly more likely to match or even outperform men among the oldest age groups (Baumgartner et al., Reference Baumgartner, Victor Sousa, Nikolaidis and Knechtle2020; Stöhr et al., Reference Stöhr, Nikolaidi, Villiger, Sousa, Scheer, Hill and Knechtle2021; Waldvogel et al., Reference Waldvogel, Nikolaidis, Gangi and Rosemann2019), the longest races (Waldvogel et al., Reference Waldvogel, Nikolaidis, Gangi and Rosemann2019) and the races with the highest overall female participation (Knechtle, Valeri, Nikolaidis, Zingg, & Rosemann, Reference Knechtle, Valeri, Nikolaidis, Zingg, Rosemann and Rüst2016; Senefeld, Smith, & Hunter, Reference Senefeld, Smith and Hunter2016). The reasons why are not well understood, although differences in muscle characteristics, pacing strategies, oxygen and substrate utilisation, gastrointestinal physiology and endocrine function have all been implicated (for example, Tiller et al., Reference Tiller, Elliott-Sale, Knechtle, Wilson, Roberts and Millet2021). Interestingly, when it comes to ultrarunning, nearly all semi- or self-supported events in any environment require participants to run while wearing a backpack containing all of their supplies, nutrition, water and safety equipment. This dorsal load-carrying can alter the kinematics of the trunk and lower limb and increase the cost of running (Liew, Netto, & Morris, Reference Liew, Netto and Morris2017; Brown, O'Donovan, Hasselquist, Corner, & Schiffman, Reference Brown, O'Donovan, Hasselquist, Corner and Schiffman2014), costs that could potentially be mitigated by the broader female pelvis, assuming similar relationships in loaded running as documented in loaded walking (Wall-Scheffler, Reference Wall-Scheffler2012b).

4.4. Culture's impact on locomotor kinematics

The development of ever-more dynamic biomechanical models and experimental designs is necessary to incorporate thermoregulatory costs into our understanding of variation in structure and what it means for kinematics and locomotion in a given set of environmental and social conditions. The latter of these is perhaps more difficult to reconstruct in the past than terrain or climate, but is no less important in shaping the locomotor kinematics of a given morphology. Whether or not a behaviour is important for survival, group cohesion or play and/or is otherwise considered culturally and socially important can have an enormous impact on functional ability that is not necessarily reflected in the skeleton alone. Conversely, just because humans as a species have musculoskeletal adaptations that confer a superior capacity for sustained running (Bramble & Lieberman, Reference Bramble and Lieberman2004; Eng, Arnold, Biewener, & Lieberman, Reference Eng, Arnold, Biewener and Lieberman2015; Holowka & Lieberman, Reference Holowka and Lieberman2018; Rolian, Lieberman, Hamill, Scott, & Werbel, Reference Rolian, Lieberman, Hamill, Scott and Werbel2009; Venkadesan et al., Reference Venkadesan, Yawar, Eng, Dias, Singh, Tommasini and Mandre2020), that does not mean that endurance running is an important part of the locomotion of all humans or that, when we do run, we are any good at it.

Both a rearfoot strike pattern, where the foot contacts the ground heel-first, and ‘overstriding’, where it does so too far out in front of the body, are considered aspects of poor running technique, as they have been shown to reduce running economy and increase the risk of injury (Daoud et al., Reference Daoud, Geissler, Wang, Saretsky, Daoud and Lieberman2012; Folland, Allen, Black, Handsaker, & Forrester, Reference Folland, Allen, Black, Handsaker and Forrester2017; Lieberman, Warrener, Wang, & Castillo, Reference Lieberman, Warrener, Wang and Castillo2015). Among non-industrialised societies for whom running is a famously important part of cultural and spiritual identity, such as the Kalenjin in Kenya and the Tarahumara in Mexico, a rearfoot strike pattern or overstriding are rare (Lieberman, Reference Lieberman2014; Lieberman, Castillo, et al., Reference Lieberman, Castillo, Otarola-Castillo, Sang, Sigei, Ojiambo and Pitsiladis2015; Lieberman et al., Reference Lieberman, Venkadesan, Werbel, Daoud, Dandrea, Davis and Pitsiladis2010, Reference Lieberman, Mahaffey, Quimare, Holowka, Wallace and Baggish2020). However, this good running technique is not necessarily reflecting an evolved and naturally gifted running ability common to the human species, but rather the result of substantial learning, effort and practice. The standard approach utilised by inexperienced runners, in both industrialised and non-industrialised societies alike, seems to be to run with their natural walking kinematics: hitting the ground heel first, with the foot out in front of the body. This is documented both among non-industrialised societies in which running is uncommon, like the Daasanach of Ethiopia and Kenya and the Hadza of Tanzania (Hatala, Dingwall, Wunderlich, & Richmond, Reference Hatala, Dingwall, Wunderlich and Richmond2013; Pontzer et al., Reference Pontzer, Suchman, Raichlen, Wood, Mabulla and Marlowe2014), and among recreational runners in industrialised societies relative to elite competitive runners (Hasegawa, Yamauchi, & Kraemer, Reference Hasegawa, Yamauchi and Kraemer2007; Larson et al., Reference Larson, Higgins, Kaminski, Decker, Preble, Lyons and Normile2011). In this Special Collection, Ian Wallace and colleagues (Reference Wallace, Kraft, Venkataraman, Davis, Holowka, Harris and Gurven2022) explored running prevalence and kinematics for the first time among another non-industrialised society for whom running is uncommon: the Tsimane. Their walking kinematics are explored by this team elsewhere in this Special Collection (Holowka et al., Reference Holowka, Kraft, Wallace, Gurven and Venkataraman2022). Despite their excellent lifelong cardiovascular and physical fitness, and high levels of physical activity at all ages (Gurven, Jaeggi, Kaplan, & Cummings, Reference Gurven, Jaeggi, Kaplan and Cummings2013; Kaplan et al., Reference Kaplan, Thompson, Trumble, Wann, Allam, Beheim and Thomas2017), running is extremely uncommon: under 2.5% of their active movement over the course of three consecutive days was done at running pace. As inexperienced runners, not surprisingly they exhibited poor kinematics in unshod running trials: a rearfoot strike pattern was observed in 99% of running trials and overstriding in 98% of trials. This technique is optimal when walking, but not running; optimal running technique clearly requires learning and practice. If running is not a necessary or valuable activity in your society, this learning and practice is unlikely to happen. On the limited occasions where running is required, neither minimising risk of repetitive use injury nor maximising performance are likely to be major concerns, and suboptimal kinematics get the job done. However, if endurance running is a crucial part of your foraging strategy, is essential for survival in the environment in which you live and/or is culturally important in some way, enhancing performance and minimising injury become important. Wallace and colleagues (2022) suggest that individuals probably develop better running kinematics through both a process of unconscious natural self-optimisation and through social learning and observation.

The fact that we all share these adaptations to running, and that so too did other members of the genus Homo, suggests that running had a shared importance in our survival strategy at some point in our evolutionary history. Yet to run frequently and efficiently for sustained periods of time without getting injured, skeletal morphology on its own is not enough; humans must adjust their walking kinematics through social learning and practice, aided by a cultural identity that values and promotes that learning and practice. While it is impossible to know if endurance running was part of the cultural identity of past hominin and early human groups, it is quite possible that, just as its importance varies among cultures today, this may also have been the case in the past.