Birds and humans are successful bipedal runners, who have individually evolved bipedalism, but the extent of the similarities and differences of their bipedal locomotion is unknown. In turn, the anatomical differences of their locomotor systems complicate direct comparisons. However, a simplifying mechanical model, such as the conservative spring–mass model, can be used to describe both avian and human running and thus, provides a way to compare the locomotor strategies that birds and humans use when running on level and uneven ground. Although humans run with significantly steeper leg angles at touchdown and stiffer legs when compared with cursorial ground birds, swing-leg adaptations (leg angle and leg length kinematics) used by birds and humans while running appear similar across all types of uneven ground. Nevertheless, owing to morphological restrictions, the crouched avian leg has a greater range of leg angle and leg length adaptations when coping with drops and downward steps than the straight human leg. On the other hand, the straight human leg seems to use leg stiffness adaptation when coping with obstacles and upward steps unlike the crouched avian leg posture.
Habitual bipedalism has evolved multiple times in several different animal orders. Avian bipedalism traces back 230 million years to their dinosaur ancestors [1,2], whereas hominins have a significantly shorter bipedal history of at least 3.6 million years [3,4]. Although both humans and birds move bipedally, there are a number of morphological disparities, which will affect their locomotion. Bird leg morphology consists of a near horizontal femur orientation and an elongated tarsometatarsal (proximal foot bones in humans) segment resulting in a digitigrade posture. In stark contrast, humans are plantigrade, and during quiet standing, we have a vertically orientated femur. These simple morphological differences result in dissimilarities during locomotion, including in leg posture and gait characteristics such as step length and gait transitions [5–7]. However, even with these anatomical differences between bipedal species, and although there may be stability disadvantages associated with bipedal locomotion , habitually bipedal species apparently negotiate uneven terrain as successfully as multi-legged species.
Despite different speeds and morphologies, both humans and ground birds are highly adapted to bipedal walking and running , even when moving over varied and sometimes unpredictable changes in ground level. These changes can be visible (e.g. roots, hills, stones) or camouflaged (e.g. puddle of unknown depth or high grass hiding terrain). In the case of minor disturbances, both humans and birds run across varied terrain with ease, having little problem dealing with irregularities in the ground. However, owing to differences in anatomy, e.g. body size and leg geometry [5,7], the same absolute perturbation in ground level results in a different perturbation relative to their leg length or standing hip height. Therefore, when only a single route is available for running over a perturbation, small birds (e.g. quail) and children have to deal with relatively bigger perturbations than large birds (e.g. ostrich) and adult humans.
To directly compare birds of different sizes, a ‘posture index’, defined as the hip height divided by the sum of leg segment lengths (measured at mid-stance during running), was established . Four species of cursorial ground birds (quail, pheasant, guinea fowl, turkey) that have been investigated in a study by Birn-Jeffery et al. , can successfully negotiate visible obstacles up to 50% of their standing hip height, whereas ostriches only manage visible obstacles up to 10% standing hip height . Guinea fowl have also been shown to successfully negotiate visible and camouflaged drops of 40% standing hip height [10,11] and visible downward steps of 6 cm (approx. 30% of standing hip height; ). Similar to ostriches, human studies suggest successful negotiation while running over uneven ground with visible and camouflaged drop heights of 10 cm (approx. 10% of standing hip height; [13,14]), visible obstacles up to 15 cm (approx. 15% of standing hip height; [15,16]) and visible steps (down and up) of 10 cm . However, none of these experiments were designed to test performance limits, but rather to understand the control strategies used. As a result, these studies demonstrate that both humans and birds use simple feed-forward control strategies (e.g. swing the leg backward just before it touches the ground) to negotiate small perturbations during running, which can be described by a simplifying mechanical model [10,11,15,17,18].
The aim of this review is to compare the adaptation strategies used by birds and humans when running on level and uneven ground in order to derive general control strategies. The anatomical differences between their locomotor systems complicates direct comparisons, but a simplifying mechanical model, such as the conservative spring–mass model (figure 1), can facilitate a direct comparison of the locomotor strategies across morphologically disparate species. We are not reviewing the spring–mass model, and its derivatives, itself, but merely using the model as a comparative tool. Using this minimalist model, rather than other derived models, provides a simple method where results are easier to interpret, therefore allowing us to determine whether underlying control principles differ in bipedal species during uneven terrain locomotion.
2. Spring–mass model for running
To describe avian and human running, the simplest model that serves as a template is the conservative spring–mass model (figure 2; [19,20,24]). According to global dynamics, spring–mass modelling simplifies the segmented avian and human leg to a single physical (massless) spring, and the body to a point mass (figure 1). Because the model is conservative, meaning that the total mechanical energy of the centre of mass does not change over a step, it can be completely described by only three parameters: angle of attack, spring stiffness and resting length of the spring (figure 1; [19,20,25]). In modelled spring–mass running, the simplest strategy is running with a fixed angle of attack and constant spring stiffness . However, only a very small subset of these parameter values (figure 3, stability plot) are viable for successful locomotion without additional control .
The stability range of the spring–mass model can be further enhanced by using a pure feed-forward control strategy. One option is to swing the leg backward with a constant retraction speed [27–29]. This backward rotational behaviour of the swing leg increases the angle of attack proportionally to flight duration, without any sensory feedback in the contact phase [10,11,30]. As the flight time during running on uneven ground depends on ground-level height (the higher the next ground contact, the shorter the flight time, and the lower the next ground contact, the longer the flight time ), a drop or downward step in ground level results in a steeper angle of attack, and thus, a redistribution of potential energy to kinetic energy, resulting in greater forward speed. On the other hand, an upward step in ground-level height leads to a flatter angle of attack resulting in lower forward speed.
Although swing-leg retraction is an important stabilizing mechanism, it cannot increase stability simply by ever increasing the rotational velocity [29,31]. On the one hand, low leg retraction velocities improve the robustness against variations in terrain height, whereas high leg retraction velocities minimize peak force and improve ground speed matching [32,33]. Karssen et al.  investigated whether an optimal swing-leg retraction rate that maximizes disturbance rejection, and minimizes impact losses and foot slipping exists; and concluded that no unique retraction rate exists that optimizes all goals mentioned at the same time. Thus, a potential trade-off has emerged between optimal swing-leg trajectory to regulate leg loading for injury avoidance, and facilitating a steady gait through disturbance rejection.
To facilitate a stable steady gait (constant speed and bounce height), a strategy during running on uneven ground could be to maintain the centre of mass trajectory. Simulations revealed that this goal can be achieved by using the flight phase to set the correct parameters of the system (angle of attack and spring stiffness) at the instant of touchdown [28,35,36]. Without feedback, this feed-forward control automatically adjusts the angle of attack and spring stiffness with the falling time and thus, to the falling height . This means that during running on uneven ground the massless spring rotates continuously at varying rates from the apex (start of falling time) resulting in a steeper angle of attack and a stiffer spring when stepping down, or in a flatter angle of attack and a compliant spring when stepping up. However, based on the conservative spring–mass model, every running pattern can be stabilized by tuning the adaptation rates of angle of attack and spring stiffness, but also by adapting the resting spring length .
The spring–mass model is a conservative system that simplifies both the human and avian leg to a single physical spring (figure 2). It does not consider leg mass, trunk orientation (e.g. upright trunk in humans versus near horizontal trunk in birds), muscle reflexes and/or anticipatory control (e.g. feed-forward versus pre-reflexive behaviour). To account for anticipatory control, a non-conservative variant of the spring–mass model becomes necessary [9,37]. By adding a damper, such a system simulates realistic energy losses, and by adding an actuator in series with the spring and damper, the system replaces the lost energy. Simulations reveal that a trajectory optimization strategy (e.g. optimal ground reaction force) performed much better under disturbances than an optimal level-ground strategy . To account for trunk movement, the simple conservative spring–mass model can be extended by adding a freely rotating trunk . Suspending the freely rotating trunk with springs to a point above the centre of mass (termed virtual pivot point ) retains the descriptive power of the spring–mass model. However, its ability to cope with perturbations in ground level has not been evaluated so far. Although extended spring–mass models can provide better descriptions, on the specifics, of the biomechanics of walking and running, they are not necessary to explain general locomotor strategies. It is for this latter reason that in this review we use the simple spring–mass model to determine the similarities in neuromechanical control strategies across bipeds in uneven terrain.
3. Avian and human running
3.1. Running on level ground
During level avian running, the leg angle at touchdown is usually between 40° and 68° (table 1). When normalized to body size, relative step length of small avian bipeds was much greater than that of large bipeds, indicating a more horizontally aligned touchdown leg angle in smaller birds . However, in all measured avian bipeds, step length increased and thus, leg length at touchdown decreased with running speed . In comparison with birds, the leg angle at touchdown during level human running is steeper at around 60–80° (table 1), and decreases (as observed in avian bipeds) with running speed from 78° at 2 m s−1 to 71° at 4 m s−1 . However, from simulations, we know that higher running velocities require one of two changes to leg parameters; either a flatter leg angle at touchdown for constant leg stiffness or a higher leg stiffness assuming a constant leg angle at touchdown .
Leg stiffness cannot be calculated directly, because the avian and human leg is not a physical spring (figure 2). However, because the leg does behave in a spring-like manner during running, effective leg stiffness can be estimated under the assumption that the entire body behaves like a spring–mass system. Compared with humans, the flatter leg angle at touchdown in birds is accompanied by more compliant legs. In level avian running, leg stiffness is between six body weights per initial leg length (bw/l0) and 15 bw/l0 (table 1). Humans, on the other hand, run with stiffer legs, with leg stiffness between 17 and 40 bw/l0 (table 1). However, for both human and birds, leg stiffness does not change significantly with speed .
Besides differences in leg stiffness and leg angle at touchdown, swing-leg retraction occurs in both avian and human running [10,15,17,29,46,47], and has been shown to reduce foot-velocity relative to the ground and therefore, landing impact . In level running, humans and pheasants use similar dimensionless leg retraction velocities . In addition, leg retraction velocity increases with increasing running speed in both . In pheasants, the changing rate of the leg angle increased from 94° s−1 at 2.2 m s−1 to 132° s−1 at 2.8 m s−1 and in humans, the changing rate of the leg angle increased from 36° s−1 at 2 m s−1 to 82° s−1 at 4 m s−1 . From simulations, we know that low leg retraction velocities reduce fall risk, whereas high leg retraction velocities reduce injury risk . An optimization for one of these factors inherently limits the other; however, compliant legs relax this trade-off as for the same leg retraction velocity compliant gaits are more robust to a sudden terrain drop .
3.2. Coping with drops and downward steps
The leg angle at touchdown depends on flight time and swing-leg retraction [10,11,30]. As a result of longer flight time and continuous swing-leg retraction during running over a camouflaged drop, guinea fowl leg angle increased at touchdown . Here, two different strategies could be observed: strategy 1, the leg contacts the ground with a steeper angle (with leg angle at touchdown increasing by about 35° from 40° to 75°; figure 5a), resulting in smaller ground reaction forces. The distal joints therefore behave as springs, converting potential energy into kinetic energy, and the animal stabilizes its running pattern by speeding up during the drop. This strategy was observed in 63% of the trials , and is consistent with the passive dynamics of a simple spring–mass model (no net change in centre of mass energy is required). In strategy 2, the leg contacts the ground with a more extended knee (leg lengthening; figure 5b) and a shallower angle (with leg angle at touchdown increasing by 15° from 40° to 55°; figure 4). Now, the distal joints behave as dampers, absorbing energy, and the animal stabilizes its running pattern by decreasing total mechanical energy. This strategy was observed in 37% of the trials. When the substrate drop was visible, the leg angle changes in guinea fowl were similar to the camouflaged drop trials . Even when running a visible step down (a permanent drop in the ground-level height; figure 5, dashed horizontal line), the leg angle changes in guinea fowl remained similar to those observed during an unexpected drop . As the leg angle at touchdown in birds does not differ between visible and camouflaged drop trials , it is possible that guinea fowl (or ground-dwelling birds in general) do not perceive the single drop down and therefore do not change their motor control as is observed in humans [13,14]. However, guinea fowl when running down a visible step still have similar leg angle changes to those of an unexpected drop step, but do also show significant locomotor differences in the pre-perturbation step suggestive of tuning occurring dependent on the environment .
Similar to the guinea fowl observed strategy 2 (increased leg angle at touchdown and leg lengthening; figure 5b), humans run across a camouflaged drop of 10% standing hip height (10 cm) with constant speed but with leg angle at touchdown increasing by 9° from 57° to 66° . However, in contrast to the unexpected ground-level changes in birds, it seems that human runners adapt their motor control prior to the camouflaged contact and anticipate a drop of approximately 0–10 cm (strategy 3; figure 5c; ). Thus, in humans, the camouflaged ground-level changes were not unexpected . When humans can use visual feedback of the perturbation, the leg adjustment is smaller and the leg angle at touchdown increases by only 3° from 57° to 60° (figure 4; ). The visual perception of the perturbation allows an adaptation of the motor control prior to the perturbation (figure 5c). When human runners become aware of the drop (or step down) in ground level, they lower their centre of mass (by reducing leg stiffness) by about 40% of drop (or step) height in preparation of the perturbation . Thus, flight time decreases when compared with the camouflaged drop , which results in smaller leg angle changes.
However, in the human experiments, a constant running speed was demanded , therefore, it is not known whether speeding up would be a natural stabilizing reaction in humans similar to strategy 1 observed in guinea fowl . Unlike the humans who were forced to steady speed, birds could adjust speed, which may be part of their control strategy in those situations. On the other hand, humans generally use a steeper leg angle at touchdown, thus, a leg angle adjustment of about 35°, as used by guinea fowl in strategy 1, would result in a fall. Nevertheless, the leg angle adjustment observed during human running across (visible and camouflaged) drops and (visible) steps are similar albeit smaller than those observed in birds (figure 4).
In birds, leg length at touchdown did not significantly differ among sequential hidden drop trials. However, during the observed strategy 1, the leg contacts the ground with a shorter relative length, whereas during strategy 2, the leg contacts the ground with an extended knee, resulting in longer initial leg length . Similar to strategy 2, during human running on surfaces with visible and camouflaged drops, runners lengthen their leg at touchdown . In humans, leg length at touchdown increases by about 3% from 1.00 to 1.03 m at a visible drop of 10% standing hip height and from 1.01 to 1.04 m at a camouflaged drop of 10% standing hip height . Unlike birds, the longer leg at touchdown in humans was achieved by plantar flexion of the ankle joint and not by an extension of the knee joint [13,17]. Because humans already run with a relatively straight leg posture, their leg architecture does not allow for much knee extension. Thus, leg extension in humans is only possible with ankle plantar flexion and therefore, its contribution is small. In humans, the increased plantar flexion of the ankle joint at touchdown can be attributed to an increased muscle pre-activation of the M. gastrocnemius medialis . The M. gastrocnemius is experimentally accessible and exhibits broadly similar activation and strain patterns for running across many animals . In turkeys and guinea fowl, the gastrocnemius muscle is likely to be capable of rapidly switching between economic force development and high work output, dependent on the conditions [50,51]. However, in contrast to human runners, in birds pre-activation of the M. gastrocnemius does not change prior to the drop step .
Because large leg lengthening is only feasible with crouched avian leg geometry, but not with the straight human leg structure, leg stiffness adaptation is—theoretically—an option to stabilize running. Nevertheless, humans did not adjust leg stiffness during a downward step neither for a single drop of different height nor for a step down [13,17]. In guinea fowl, leg stiffness varied considerably and did not differ significantly when stepping into a visible or camouflaged drop . However, when running down a visible step effective leg stiffness increased across all steps, which suggest that guinea fowl tune gait dynamics dependent upon context including the anticipated step down .
3.3. Coping with obstacles and upward steps
Owing to shorter flight time and swing-leg retraction, on visible upward steps or obstacles (to camouflage a step up is difficult), the leg angle at touchdown decreases for both birds and humans [9,15,17,22]; adjusting to the vertical height of the step [9,15,22]. At the highest vertical step (+15 cm, approx. 15% standing hip height) in humans the leg angle at touchdown decreased (in accordance with the predictions of the spring–mass model; figure 3) by about 6° from 68° to 62° (figure 4; ). An adjustment to the number of subsequent elevated contacts was not observed . Similar to humans, in pheasants, the leg angle at touchdown decreased by about 7° from 55° to 48° on the highest obstacle (obstacle heights: 50% leg length; figure 4; ). Similar to the pheasants, leg angle adjustment on the obstacle step did not differ significantly across several species of avian bipeds spanning a 500-fold size range from quail to ostriches .
In situations where the human runner is well aware of an upward step, initial leg length at touchdown (l0) on the obstacle step is also shortened in proportion to the height of the obstacle and decreases by about 5%, from 0.94 l0 at ground-level height, to 0.9 l0 at the highest vertical step (15% standing hip height; ). In humans, the shorter leg at elevated contact was achieved by increased bending of the ankle and knee joint [15,17]. The increased dorsiflexion of the ankle joint at touchdown was attributed to a decreased muscle pre-activation of the M. gastrocnemius medialis [16,43]. On the obstacle step, birds across a 200-fold size range also consistently adopt a shorter leg length at touchdown relative to level ground, but with no significant linear trend with increasing obstacle height [9,22]. In contrast to humans, at slow speeds (when given more time to visually assess oncoming terrain when compared with higher speeds) guinea fowl increase (and not decrease) anticipatory muscle activation before obstacle contact . Furthermore, the results reveal synergistic co-activation of muscles across the limb, rather than independent control of individual muscles .
In addition to the leg length and leg angle adjustment, human runners also adjust leg stiffness to the vertical height of a disturbance . At the highest vertical step (15% standing hip height), leg stiffness was reduced by about 26% . In a multi-joint system, the global leg stiffness depends on a combination of the geometry of the system and the local stiffness of the joints [18,54]. Because joint stiffness also depends on the level of activation of the muscles acting on the joint [43,55], the decreased global leg stiffness can be (in part) attributed to an increased bending in the joints and decreased muscle (pre-) activation of the M. gastrocnemius medialis [16,43]. Leg stiffness, though, was adjusted not only to the vertical height, but also to the number of subsequent elevated contacts . When running a step upward, leg stiffness decreased by about 20% on a single obstacle of 10% standing hip height and decreased by about 9% on a permanently elevated step of 10% standing hip height . Thus, the adjustment of leg stiffness in humans seems to be actively achieved, and context-dependent . Nevertheless, humans seem to exploit this possibility, whereas birds do not. In birds, leg stiffness varied greatly even within a single step category or obstacle height . The lack of active leg stiffness adjustments seen in birds may purely be a result of the huge variety of leg geometries available by adjusting both leg angle and leg length. This difference in use of leg stiffness adaptations to obstacles may indicate an underlying difference in control targets or purely present owing to restrictions in leg geometries between the species.
Both birds and humans are successful bipedal runners, which have individually evolved bipedalism. Despite differences in anatomy, both avian and human running can be described by the conservative spring–mass model. Thus, the model facilitates direct comparison of the locomotor strategies that birds and humans use when running. Using this comparative method, it then becomes obvious that humans use significantly steeper leg angles at touchdown and stiffer legs while running compared with birds.
Despite these disparities, the swing-leg adaptations (leg angle and leg length kinematics) used by birds and humans while running on uneven ground appear (in general) similar when coping with drops and downward steps (leg angle at touchdown and initial leg length increased) and with obstacles and upward steps (leg angle at touchdown and initial leg length decreased). Nevertheless, owing to their leg morphology, birds have a greater range of leg angle and leg length adaptations when coping with drops and downward steps than the restrictive straight leg of humans. If humans need a greater range of leg lengthening options, they have to move with a less upright posture, which in turn results in certain disadvantages. Some authors have suggested that early hominins walked bipedally in a way that involved more flexed hips and knees, likely resulting in higher energy costs of locomotion [56,57]. Similar to birds, such a crouched leg posture seems to be better adapted for leg lengthening and results in more compliant legs. However, when compared with birds, the straight human leg does not allow for larger leg extension but seems to be better adapted for leg stiffness adaptation when running a step upward. Because the adjustment of leg stiffness in humans seems to be actively achieved, we assume that birds adjust their running pattern more passively than humans.
Interestingly, the reduction in leg stiffness in humans and the leg lengthening adjustment observed in birds seems to be a general control strategy used for many different ground-level perturbations. More precisely, humans reduce leg stiffness on the obstacle or upward step but also in preparation of a drop or downward step , whereas birds lengthen their legs on the drop or downward step, but also in preparation of an obstacle or upward step (vaulting strategy ).
We have reviewed and compared the adaptation strategies used by birds and humans when running on level and uneven ground using a simple conservative spring–mass model. The implications of using this model are that morphological differences (e.g. leg mass, muscle reflexes or trunk orientation) between human and birds are not considered, but may affect species locomotion. One example, the percentage of body mass located in the legs differs between species; in humans about 40% , in ostriches about 37%  and in guinea fowl about 23% . Compared with the massless spring in the spring–mass model, this will generate higher moments of inertia owing to leg oscillation (backward and forward swing) that can be expected to have a significant impact on the overall dynamics. However, to compare the adaptation strategies used by birds and humans with respect to leg mass, reflexes and/or trunk orientation further studies with more experimental data and related models are required. Therefore, as used here, the simplifying assumptions of the conservative spring–mass model can reveal general locomotor control policy between two morphologically disparate groups of animals.
R.M. generated the main idea and coordinated the work. All authors wrote the manuscript. R.M. created the figures.
We declare we have no competing interests.
This research received no funding from any agency in the public, commercial, or non-profit organization.
- Received July 5, 2016.
- Accepted August 25, 2016.
- © 2016 The Author(s)
Published by the Royal Society. All rights reserved.