SEARCH

SEARCH BY CITATION

Keywords:

  • biomechanics;
  • bipedalism;
  • computer modelling;
  • KNM-WT 15000;
  • load-carrying

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The first unquestionably bipedal early human ancestors, the species Australopithecus afarensis, were markedly different to ourselves in body proportions, having a long trunk and short legs. Some have argued that ′chimpanzee-like′ features such as these suggest a ‘bent-hip, bent-knee’ (BHBK) posture would have been adopted during gait. Computer modelling studies, however, indicate that this early human ancestor could have walked in a reasonably efficient upright posture, whereas BHBK posture would have nearly doubled the mechanical energy cost of locomotion, as it does the physiological cost of locomotion in ourselves. More modern body proportions first appear at around 1.8–1.5 Ma, with Homo ergaster (early African Homo erectus), represented by the Nariokotome skeleton KNM-WT 15000, in which the legs were considerably longer in relation to the trunk than they are in human adults, although this skeleton represents an adolescent. Several authors have suggested that this morphology would have allowed faster, more endurant walking. But during the same period, the archaeological record indicates a sharp rise in distances over which stone tools or raw materials are transported. Is this coincidental, or can load-carrying also be implicated in selection for a more modern morphology? Computer simulations of loaded walking, verified against kinetic data for humans, show that BHBK gait is even more ineffective while load-carrying. However, walking erect, the Nariokotome individual could have carried loads of 10–15% body mass for less cost, relative to body size, than AL 288-1 walking erect but unloaded. In fact, to the extent that our sample of humans is typical, KNM-WT 15000 would have had better mechanical effectiveness in bearing light loads on the back than modern human adults. Thus, selection for effectiveness in load-carrying, as well as in endurant walking, is indeed likely to have been implicated in the evolution of modern body proportions.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Bipedal locomotion is certainly the oldest, and arguably the most fundamental distinguishing characteristic of the human lineage. However, the origins, early evolution and adaptive context of human bipedality remain poorly understood. To a considerable extent, our lack of knowledge is related to the uniqueness of this adaptation among placental mammals. We seem to lack living analogues for the formative stages of human bipedalism.

It was Lovejoy et al. (1973) who first demonstrated that modern human bipedalism was not unique. In an analysis of the mechanics of abduction at the hip joint, they showed that australopithecines had evolved a pattern which, while unlike that of modern humans, was not functionally inferior. Further discoveries, particularly that of the ‘Lucy’ skeleton of Australopithecus afarensis, AL 288-1, at Hadar, have suggested that acquisition of the morphological characteristics of human bipedalism was not a sudden event, in which a ‘chimpanzee-like’ postcranial morphology was suddenly replaced by the long-legged, short-trunked configuration of modern humans. Stern & Susman (1983) suggest that Lucy's skeleton is rather a mosaic of features. Some features, such as the grasping toes, long arms and funnel-shaped thorax, they hold to be ‘chimpanzee-like’ and to reflect a climbing, forelimb-powered arboreal lifestyle; others, like the permanently adducted posture of the knees, are distinctly ‘human’, and others, such as the arrangement of the tarsal bones, are unique. Berge (1994, p. 271), in a biomechanical comparison of the hip and thigh of Australopithecus and Homo, suggests that australopithecine bipedalism ‘likely required a greater energetic cost than does human bipedalism’ because of a shorter stride length and frequency. Stern & Susman (1983) interpret Lucy's morphology as a compromise between the demands of terrestrial bipedalism and arboreal climbing, and argue that her gait would have been ‘bent-hip, bent-knee’ (BHBK), as is the voluntary bipedalism of untrained chimpanzees. The chimpanzee model is taken further by Hunt (1994). Based on field studies, Hunt argued that bipedalism in chimpanzees occurs predominantly in the form of ‘postural’ bipedalism, which chimpanzees adopt while reaching to feed from trees. In the chimpanzees, Hunt observed, functional bipedal locomotion was a rare activity that consisted of relatively ungainly ‘shuffling between trees’ (Wood, 1994, p. 588). Hunt has proposed that the ‘poor bipedal mechanics’ (Hunt, 1994, p. 198) of A. afarensis may be explained by an analogy to the flexed-knee ‘postural’ bipedalism of chimpanzees. He identifies several functional features of the postcranial anatomy of A. afarensis that he believes can be linked with a ‘postural’ rather than locomotor bipedalism, and goes on to suggest that locomotor bipedalism (the ‘pre-eminent bipedalism’ of Prost, 1980, p. 186), did not appear until the emergence of Homo ergaster (early African H. erectus) between 1.9 and 1.5 million years ago.

Recently, we have presented evidence from field studies and kiniesiological studies of captive animals that suggest that the common chimpanzee is not necessarily the best model for early hominid gait. Despite their quite distinctive skeletal morphology, orang-utans have long been observed in captivity to utilize erect body posture during voluntary bipedalism. Field studies now confirm that although unassisted bipedalism may not be much commoner than it is in common chimpanzees or bonobos, much more extended hip and knee posture is as typical of the (almost entirely arboreal) behaviour of wild orang-utans in their natural habitat as it is of captive (and hence much more ‘terrestrial’) orang-utans (Crompton et al. 2003; Thorpe & Crompton, 2004). Thus, there is no good reason to suppose that A. afarensis was any more anatomically incapable of erect posture during bipedality than are living orang-utans, and the more human-like knee morphology of A. afarensis suggests that bipedality is likely to have been better sustained. Some time ago, Crompton et al. (1998) used a computer simulation technique, inverse dynamic modelling, to predict the mechanical response of a model, representing the body proportions and build of the ‘Lucy’ skeleton, A. afarensis AL 288-1, to joint motion from bipedal walking in the erect, human manner; in the manner of the occasional voluntary bipedalism of common chimpanzees; and in the manner of humans simulating BHBK gait. They showed that although chimpanzee-like bipedalism did not appear to be compatible with Lucy's proportions, joint motion from both erect and BHBK human walking could drive a stable simulation. However, BHBK gaits nearly doubled the rate of work (mechanical joint power requirements) in erect walking, and imbalance in joint power suggested inefficency and the likelihood of increased heat load. Stern (1999) dismissed these findings as exaggerated, and suggested that any inefficiency would have been offset by possible reduction in stress across the sacroiliac joint due to the lower peak vertical ground reaction forces typical of BHBK, or more properly compliant, gaits, which might thus have facilitated an incipient bipedalism. Kramer & Eck (2000), however, using similar inverse modelling techniques to Crompton et al. (1998), also found that AL 288-1 could have been an efficient erect biped despite this early hominid's short legs [indeed, they seem to imply (Ward, 2002) that efficiency may have been greater in this early hominid than in modern female humans, perhaps as a consequence of the greater power needed to accelerate a long leg].

Over the last decade, there has been accumulating evidence, for example from Java, Georgia and Boxgrove, that H. erectus had a much earlier dispersal in the Old World than had been conventionally accepted. In particular, work by Swisher et al. (1994) suggests that this species is as old in Java as in it is in Africa. This early dispersal coincides with a period, presumably postdating A. afarensis but prior to 1.5 million years ago, in which there was a major reorganization in the anatomy and function of the postcranial skeleton of early hominids, as witnessed by the contrast between the short, rather conical, trunk, short legs and long arms of the skeleton of A. afarensis AL 288-1 ‘Lucy’ and the short, wide trunk, long legs but short arms of the skeleton of the Nariokotome skeleton of H. ergaster, KNM-WT 15000. The contrast in leg length immediately suggests longer stride length (McHenry, 1991) and therefore an ability of H. ergaster/erectus to walk faster and thus range more widely, but the work of Kramer & Eck (2000) suggests that increased leg length does not necessarily bring increased mechanical efficiency. However, Preuschoft & Witte (1991) suggest that AL 288-1′s short legs and long and funnel-shaped trunk favour static stability, and short legs may permit rapid changes of direction and rapid acceleration over short distances. Such characteristics are clearly relevant to the rather closed, woodland environments proposed by Bonnefille et al. (1987), for example, for Laetoli, a site with A. afarensis jaws as well as footprint trails, and the earlier deposits of Hadar. By contrast, Preuschoft & Witte (1991) argue that the dynamic response of a body with (taken together) long legs, a short but broad trunk and short arms suggests that the physique of KNM-WT 15000 was finely tuned for efficient ballistic walking, in other words endurant bipedalism. Comparative physiological studies by Wheeler (1992) suggest that that greater stature and more gracile physique also bring thermoregulatory benefits (see also Ruff, 1991) which, in the more xeric and open environments thought to be associated with H. ergaster, may have made it easier to range more widely.

Direct evidence of early hominid behaviour is hard to come by, but the Laetoli trackways, dated at 3.6 Ma, are a potential source of data on variables such as stride length, cadence, speed and the development of pressure under the foot at the time of A. afarensis. Unfortunately, we are uncertain of their authorship, so important parameters for calculation of these variables, such as stature and body proportions are uncertain. Most authors (e.g. White & Suwa, 1987, but see Tuttle et al. 1991) regard the footprints as both compatible with the morphology of A. afarensis and modern in functional aspect, and regard them as showing an adducted big toe and medial arch (see Deloison, 1991 and Clarke, 2003 for a contrary assessment of the latter features). Alexander (1984) showed from spacing of the Laetoli footprints that given the likely stature of the maker, the trails represent the equivalent of comfortable small-town walking speeds in modern humans.

A second source of ‘fossilized behaviour’ is of course the archaeological record, which for the period of interest is largely made up of lithics: stone tools and the raw material and debitage of their fabrication and use. In particular, the distribution of lithics across the landscape might be expected to inform us about ranging behaviour. A simple and robust measure of ranging behaviour can be derived from studying the distance between a given assemblage of stone tools and the source of the material of which the tools were made. Caution, however, must be employed in interpreting such evidence. McGrew (1993), for example, noted that chimpanzees, despite their small day- and home-ranges, can move material across a landscape in additive small journeys, and he concludes that early hominid transport may reflect a similar phenomenon. But the manufacturers of both the Oldowan and Acheulean industries – associated with Australopithecus or early Homo (H. habilis’, ‘H. rudolfensis’), and H. erectus (including H. ergaster), respectively – do appear to have transported raw materials. The best documented cases of Oldowan raw material transport are from Olduvai Bed I, where distances of 3–12 km have been established (Leakey, 1971; Hay, 1976). East Turkana also provides instances of the importation of raw material on to floodplains of the ancient lake, over distances of up to 20 km (Harris & Herbich, 1978). However, in Acheulean sites, evidence suggests that transport occurs more often – and over much greater distances. At Olorgesailie, Isaac (1977) notes occurrences of quartz brought over 40 km. At Kilombe, similarly, two obsidian bifaces appear among many hundreds made from local lavas, and the implication is again that long-distance transport occurred (Gowlett, 1982). At Gadeb, in eastern Ethiopia, dated at about 1.5 Ma, several obsidian bifaces apparently document a transport distance of over 100 km (Clark, 1980). Thus, the archaeological record suggests that transport both became more common and occurred over much greater distances, during the period in which Homo acquired its modern human-like postcranial skeleton. We may also look for patterns in the morphology of the tools themselves. Crompton & Gowlett (1993) and Gowlett & Crompton (1994) used a variety of multivariate morphometric techniques to analyse size-related variation in the African Acheulean. They showed that a common rule-set existed, which governed the adjustments in shape and dimensions that were made at increasing sizes. In nearly all cases, these adjustments had the effect of controlling for the disproportionate increase in weight that necessarily accompanies an increase in linear dimensions. The weight range of Acheulean bifaces is remarkably restricted: the great majority weigh less than 0.5 kg. If early African H. erectus was indeed the tool-maker, it does then seem to have taken weight into consideration when manufacturing tools, although the reasons might be associated with tool utilization rather than tool transport. Nevertheless, we might expect the weight of carried material to have been an effective selective influence on limb morphology of early Homo.

There is already good evidence to suggest that the morphology of H. ergaster is indeed adapted for efficient endurant walking, as well as fast walking. The present study, using similar, inverse dynamics modelling approaches as in our earlier study of bipedalism in AL 288-1 (Crompton et al. 1998), asks rather to what extent the morphology of A. afarensis AL 288-1 and H. ergaster KNM-WT 15000 are adapted for mechanical effectiveness in load-carrying.

Method

  1. Top of page
  2. Abstract
  3. Introduction
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

A suitable model of AL 288-1 had previously been constructed for the previous modelling study by Crompton et al. (1998), in which we had followed the quite consistent estimates of segment lengths etc. in Johanson et al. (1982), Jungers (1988a,b) and McHenry (1991). Using ADAMS (Mechanical Dynamics Inc., 1992) and the same techniques as described in Crompton et al. (1998), we proceeded to build a similar computer model of the body proportions and mass distribution characteristics of the Nariokotome skeleton, KNM-WT 15000, based originally on the Ruff & Walker (1993) reconstruction, which was available as a Kenya National Museum cast. All skeletal elements were laser scanned in our laboratory using a Cyberware 3030MM scanner, and we then proceeded to reconstruct the skeleton from the resulting individual solid models. The original stature estimate of 160 cm for KNM-WT 15000 was drawn from Lundy & Feldesman (1987) and Ruff & Walker (1993). However, alternative, shorter stature estimates (147 and 141 cm are provided in Ohman et al. 2003) based on new reconstructions of the trunk [original Ruff & Walker (1993) trunk estimate 61 cm; the ‘conservative’ model of Ohman et al. (2003) stature 147 cm, trunk 48.5 cm in length; and the shorter Ohman et al. (2003) reconstruction, stature 141 and trunk 42.8 cm)] were also modelled, holding limb length constant. However, these studies (Ohman et al. 2003) also indicated that the individual suffered from degenerative disease of the vertebral column. Nevertheless, as this is the only known specimen of this species, we currently have no way of knowing to what extent the shortness of the trunk in both the old (referred to below as Model A) and especially the new reconstruction (Model D) of KNM-WT 15000 is typical of the species, or an artefact of disease. Femur, tibia, humerus and ulna lengths of KNM-WT 15000 were taken from Ruff & Walker (1993); head and foot dimensions from Ohman et al. (2003); other dimensions were obtained by measurement from the cast. Segment and body weight was estimated from segment lengths using modern human data in Chandler et al. (1975a,b) and Jensen (1989). After body proportions and total mass were determined, mass distribution was estimated from the relative mass and relative moments of rotation of segments using the method of Chandler et al. (1975a,b) and Jensen (1989).

Inverse dynamic modelling of erect walking in KNM-WT 15000

We ran, through these androids, segment kinematics, taken from recorded performances of normal human walking (Wang et al. 2002a). All inertial models (‘androids’) were able to sustain stable erect walking. We then applied an optimization technique developed by one of us (Wang et al. 2002b), which modified the details of joint motion (kinematics) sufficiently to optimize the dynamic response of each simulation, to minimize the ratio of joint power expended to forward motion of the centre of mass. Resulting simulated ground reaction forces (the inverse of the forces applied by the body to the ground to drive walking), both sagittal and vertical, were very similar to curves we had measured for living adult humans using a Kistler forceplate (Wang et al. 2002a) confirming that our simulations of erect walking in KNM-WT 15000 were reliable. However, the ratio of forward displacement of the body centre of mass to the sum of lower limb joint power expended, i.e. the mechanical effectiveness of walking, was (by a small amount) best in the shorter Ohman et al. (2003) reconstruction. This would seem to offer biomechanical support to the contention of Ohman et al. (2003) that this reconstruction is more internally consistent than the alternatives, including the original Ruff & Walker (1993) model.

All three models of fossil species were then compared with two models of two modern male adults, one a Chinese of normal fitness (male_w), the other a European athlete (male_jt), and to a model of a 12-year-old modern human Chinese male (male_di). The last was included as a modern analogue of the proportions of KNM-WT 15000, generally regarded as an adolescent. The AL 288-1 model was driven with both the kinematics of human normal erect walking and of humans walking BHBK (see Crompton & Gowlett, 1993).

A simple environment was selected, a horizontal plane, to load the androids’ walking and produce ground reaction forces (GRFs).

Inverse dynamic models of loaded walking in KNM-WT 15000

Using (a) both inertial models based on both the original (Model A) and the new (Model D) reconstructions of the Nariokotome youth (and assuming a human-like pattern of size/mass distribution), (b) the existing models of AL 288-1, and (c) inertial models representing the proportions of living human adults and children, we proceeded to simulate bipedal walking both unloaded and with loads of different weights. Loads equivalent to 0–30% body weight were attached to the mid-back of each model, and motion files drawn from our earlier studies of erect and BHBK loaded walking in humans were applied to the models. Optimization routines were again applied to the simulations to modify the details of kinematics sufficiently to optimize the dynamic response of each simulation, to minimize the ratio of joint power expended to forward motion of the centre of mass. We then calculated simulated ground reaction forces for each simulation again (see for instance Fig. 1), to check that their pattern was similar to those we had recorded in equivalent performances of humans (Wang et al. 2002a).

image

Figure 1. Ground reaction forces (GRFs) in loaded walking. In each case, the horizontal axis shows phase of stance from heel strike (HS) to toe off (TO). Left: vertical (z) forces; right: sagittal (y) forces, where positive forces are decelerative, negative accelerative. Top pair, measured forces in newtons per kilogram (N kg−1); middle pair, force curves simulated for erect walking by AL 288-1 for loads 0–25% of body weight, y axis in newtons/body weight; bottom pair, force curves simulated for erect walking by KNM-WT 15000 (Model D).

Download figure to PowerPoint

Calculated results were normalized against the unloaded weight of the subject, not the loaded weight. It can be seen from Fig. 1 that the form of the vertical and sagittal GRFs is reasonably similar to that in real performances, so the simulations can be regarded as successful.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Unloaded erect walking simulations

We adopted joint power as our criterion of mechanical effectiveness. Joint power expresses the rate of doing work, and according to Winter (1990) is a good measure of whether a form of gait is or is not energy-effective. Tables 1 and 2 show that although identical motion functions were used to drive all models, different values of joint moments and powers resulted.

Table 1.  Comparison of average absolute powers (W/BV) in erect unloaded walking by various models
 al_288wt15kdmale_jtdimale_w
  1. W/BV: watts/(per body weight and per velocity). al_288 = AL 288-1; wt15kd = KNM-WT 15000, Model D, new reconstruction by Ohman et al. (2003); male_jt = a European adult male athlete; di = a 12-year-old Chinese male adolescent; male_w = an adult male Chinese.

Hip0.38090.25580.40240.35110.5019
Knee0.31050.25610.28670.35140.4453
Ankle0.60070.43000.60540.31670.8626
Sum1.29210.94191.29451.01921.8098
Table 2.  Comparison of peak positive and negative joint moments (Nm/BL) in erect unloaded walking by different models
 al_288wt15kdmale_jtdimale_w
  1. BL: (per body weight and per leg length). al_288 = AL 288-1; wt15kd = KNM-WT 15000, Model D, new reconstruction by Ohman et al. (2003); male_jt = a European adult male athlete; di = a 12-year-old Chinese male adolescent; male_w = an adult male Chinese.

Hip1.21–0.26 0.68–0.15 0.99–0.24 1.05–0.12 1.14–0.56
Knee0.41–0.70 0.11–0.70 0.32–0.74 0.13–1.05 0.60–0.71
Ankle1.73–0.01 1.14–0.02 1.60–0.02 1.03–0.01 1.93–0.04
Sum3.35–0.97 1.93–0.87 2.91–1.00 2.21–1.18 3.67–1.31

In general, KNM-WT 15000 requires lower moments and powers than does AL 288-1, particularly in the case of positive moments at the hip and knee, which are half those in AL 288-1. Although a typical value for the weight of a modern male adult is much higher than any likely values for AL-288-1 and KNM-WT 15000, both the joint moments and the powers for the male human adults are relatively small.

Loaded walking simulations

Figures 2 and 3 show that although timings do not differ extensively in erect walking for the two models, for proportionally equivalent loads, peak powers and moments required at the hip, knee and ankle joints are substantially greater for loaded walking in AL 288-1 than in KNM-WT 15000. To facilitate further comparison of the different models, we normalized joint powers to create dimensionless parameters. Various approaches to normalization exist:

image

Figure 2. Simulated joint moments (joint torques or turning forces) in loaded walking of AL 288-1 (left) and KNM-WT 15000 (right). Loads in AL 288-1, left 0–25% of body weight, in KNM-WT 15000 Model D, left 0–30% of body weight. Top: hip; middle: knee; bottom: ankle. Horizontal axis shows phase of stance from heel strike (HS) to toe off (TO). Vertical axes show moments in newton metres (Nm)/body weight × leg length (BL).

Download figure to PowerPoint

image

Figure 3. Simulated joint powers (i.e. rate of doing work) in loaded walking of AL 288-1 (left) and KNM-WT 15000 Model D (right). Loads in AL 288-1, left 0–25% of body weight, in KNM WT 15000, right 0–30% of body weight. Top: hip; middle: knee; bottom: ankle. Horizontal axis shows phase of stance from heel strike (HS) to toe off (TO). Vertical axes show power in watts (W)/body weight × leg length (BL). Positive values indicate power output (positive work), negative values imply negative work and probable energy storage.

Download figure to PowerPoint

  • 1
    Normalization by body mass (kg), speed (m s−1) of centre of mass (CM) and forward acceleration (m s−2) (see Fig. 4).
  • 2
    Normalization by body weight (kg and gravity) and speed of CM (m s−1) (Alexander & Goldspink, 1977) (see Fig. 5). This is a classical method, but it uses the gravity constant, g, to replace horizontal acceleration, and so is less intuitively related to the effectiveness of walking performance.
  • 3
    Normalization by body mass (kg) and displacement (m) of the body CM (Cavagna et al. 1977; Taylor et al. 1980; see Fig. 6). This method is easily understood and widely used, but it does not provide a truly dimensionless measure of power, and thus does not eliminate the effects of body size and mass from the comparisons to be made.
image

Figure 4. Comparison of joint power in various simulations of loaded walking, normalized by body mass (kg), speed (m s−1) of centre of mass and forward acceleration (m s−2). Horizontal axis, load as percentage of (unloaded) body weight; vertical axis, dimensionless absolute joint power. AL 288-1 BHBK = AL 288-1 driven by kinematics of human walking ‘bent-hip, bent knee’; AL 288-1 ERECT = AL 288-1 driven by kinematics of human erect walking; WT-15000 A = KNM WT 15000 original Ruff & Walker (1993) reconstruction, driven by kinematics of human erect walking; WT 15000 D = KNM-WT 15000 Ohman et al. (2003) model with shorter trunk length, kinematics from human erect walking; Ad Male W = adult Chinese human model, kinematics from human erect walking; Ad Male JT = adult European athlete, kinematics from human erect walking; Juv Male di =  12-year-old male Chinese adolescent, kinematics from human erect walking.

Download figure to PowerPoint

image

Figure 5. Comparison of joint power in various simulations of loaded walking, normalized by body weight (kg and gravity) and speed (m s−1) of centre of mass (CM). Horizontal axis, load as percentage of (unloaded) body weight; vertical axis, joint power (W/BV = watts/body weight and average velocity of CM). AL 288-1 BHBK = AL 288-1 driven by kinematics of human walking ‘bent-hip, bent knee’; AL 288-1 ERECT = AL 288-1 driven by kinematics of human erect walking; WT-15000 A = KNM WT 15000 original Ruff & Walker (1993) reconstruction, driven by kinematics of human erect walking; WT 15000 D = KNM-WT 15000 Ohman et al. (2003) model with shorter trunk length, kinematics from human erect walking; Ad Male W = adult Chinese human model, kinematics from human erect walking; Ad Male JT = adult European athlete, kinematics from human erect walking; Juv Male di = 12-year-old male Chinese adolescent, kinematics from human erect walking.

Download figure to PowerPoint

image

Figure 6. Comparison of joint power in various simulations of loaded walking, normalized by body mass (kg) and displacement (m) of centre of mass. Horizontal axis, load as percentage of (unloaded) body weight; vertical axis, joint power (W/BL = watts/body weight and leg length). AL 288-1 BHBK = AL 288-1 driven by kinematics of human walking ‘bent-hip, bent knee’; AL 288-1 ERECT = AL 288-1 driven by kinematics of human erect walking; WT-15000 A = KNM WT 15000 original Ruff & Walker (1993) reconstruction, driven by kinematics of human erect walking; WT 15000 D = KNM-WT 15000 Ohman et al. (2003) model with shorter trunk length, kinematics from human erect walking; Ad Male W = adult Chinese human model, kinematics from human erect walking; Ad Male JT = adult European athlete, kinematics from human erect walking; Juv Male di =  12-year-old male Chinese adolescent, kinematics from human erect walking.

Download figure to PowerPoint

As the acceleration due to gravity (used in the normalization method given in Alexander & Goldspink, 1977) acts only in a vertical direction, it is not ideal for comparisons of the effectiveness of forward progression in hominoids of different sizes. Method 1 replaces the gravity constant by the horizontal acceleration of CM, and does provide a dimensionless measure of power, and so is our favoured method.

The figures show that the three methods give different results, although all three show low costs of carrying for the model representing the proportions and mass distribution of the European male athlete. Figure 6, using normalization by body mass (kg) and displacement (m) of CM, gives relatively high power requirements at all loads for the juvenile Chinese model, which in the other two normalization methods uses nearly the least power for medium and large loads. This method also gives relatively low power requirements for loaded BHBK walking by AL 288-1, which at low loads (below 25% of body weight) are below the costs for the two models of Chinese humans. Nevertheless, as with the other two methods, BHBK loaded walking by AL 288-1 is more costly than erect loaded walking by the same hominid (except for loads less than 4% of body weight), and the other two power requirements of KNM-WT 15000 (both models) are low for loads less than 40% of body weight. As in Fig. 4, costs for KNM-WT 15000 model D are always well below those for AL 288-1 walking loaded, whether erect or BHBK. Figure 5, showing the results of normalization by body weight (kg and gravity) and speed of CM (m s−1), agrees with Fig. 4 in showing low costs for the model of the juvenile Chinese, and consistently higher costs for BHBK loaded walking by AL 288-1 than for any other model. However, it also shows high costs for both KNM-WT 15000 model D (with the shortest trunk) and the Chinese adult male (with a relatively long trunk) than for AL 288-1 for loads over 25% of body weight.

Figure 4 shows more consistency between loads for the different models than either Figs 5 or 6, which tends to support our preference for this method of normalization. Figure 4 shows very dramatically that BHBK walking would have been particularly disadvantageous to A. afarensis if this species had carried loads in this posture, however small. Models based on both the original (WT 15000 A) and the new (WT 15000 D) reconstructions of KNM-WT 15000 have substantially lower relative power requirements for normal walking than A. afarensis whatever the relative size of the load. The KNM-WT 15000 models can carry a load of, conservatively, 10–15% of body weight at no greater relative cost than AL 288-1 incurs walking erect and unloaded. Thus, the shorter trunk but longer legs of KNM-WT 15000 do indeed serve to allow it to walk more effectively, loaded or unloaded, but are perhaps of particular significance in offering ‘free’ loading. The results would not be very sensitive to error in trunk length estimates of KNM-WT 15000, as models A and D show parallel responses to loads and their curves lie close together in our two plots of dimensionless power (Figs 4 and 5).

In fact, the mechanical effectiveness of walking by the KNM-WT 15000 android is somewhat greater, carrying small loads, than for our modern human male adults (male_jt, male_w), although the advantage is reversed for loads above 40% of body mass. However, the inclusion of a Chinese adult male may have biased our sample towards a high value for trunk length, as a consequence of the relatively long trunk of Chinese individuals: inclusion of Nilotic individuals, had this been practical, would of course have tended to produce the opposite effect. Note that the model of a Chinese juvenile/adolescent male (di) also performed well, at low loads, compared with either human adult; Chinese populations born since the 1970s, with a higher quality diet, have tended to greater stature and this has been achieved primarily by increases in lower limb length, while trunk length has remained rather static. It is important to emphasize that although we have found intrageneric differences between models, which may be meaningful within the limitations of the accuracy of estimates of fossil proportions, and the representativeness of our samples, our aim in this study is to compare Homo to A. afarensis as represented by AL 288-1, not to make comparisons within Homo, for which latter purpose our sample is not well suited.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Why are there such differences among the models? The most distinctive simulation was undoubtedly loaded BHBK walking in A. afarensis.Wang et al. (2002a) compared a range of kinematic and kinetic parameters in humans carrying loads of up to 20 kg over the shoulder or on the back while walking upright or ‘BHBK’. The most marked differences were in joint moments, powers and energy transformation. In ‘BHBK’ walking, absolute moments at the knee were not only 2.5 times larger than in normal walking, but moments were negative, or extensor, for nearly the whole of the stance phase (60% of the cycle, as opposed to 30% of the cycle in erect loaded walking) to support the large flexor moment of the upper body and load. Nearly all forms of required powers were greater in BHBK loaded walking, but this effect was greatest at the ankle and knee. Most of the difference in powers reflected negative joint power, particularly at the knee and ankle, so that power output to drive motion was very reduced in BHBK walking. As a consequence, total joint power (work done) for equal displacement of the centre of mass was almost 1.5 times greater in BHBK than in erect loaded walking, amounting to a 40% increase in mechanical energy costs, and energy transformation (assessed using a particle mechanics approach, see Wang et al. 2003) was half that in loaded normal walking. These differences are sufficient to explain the particularly poor performance of the AL 288-1 BHBK loaded walking simulation. However, recall that our joint torque and power figures for loaded erect walking (Figs 2 and 3) do not show much difference in timing. Rather, AL 288-1 consistently requires substantially higher torques (moments) and greater power at the hip, knee and ankle, particularly for heavier loads.

Given that Figs 4 and 5 show dimensionless power, and that only the AL-288-1 BHBK walking uses a different set of kinematics, we know that only body proportions and mass distributions can account for the other differences in power requirements between models at equivalent loads. Therefore, one reason for the difference in powers may be that an increase in trunk length results in an increase of the segment's principal moments of inertia. According to the theorem of the moment of momentum (see for instance Meriam & Kraige, 1993) and if lower limb lengths remain constant while the trunk increases in length, moments at the hip and more distal joints have to increase to maintain the stability of the upper body. For the trunk to move forwards, moments are required at these joints to move the trunk as well as the legs. For the hip joint, the required moment will be determined by the moment of inertia, I, of the trunk, multiplied by ω, the angular velocity at the hip: I.ω. The moment of inertia of the trunk is proportional to the square of its length. The sum of moments required at the hip and more distal joints required for the trunk to move forwards will be (d(I · ω))/dt, and the energy required to move the trunk, E, then varies as 1/2(I · ω2). As the length, and hence the inertia, of the trunk increases, so will the energy required to move it. Thus, a relatively small proportion of the upper body (head and trunk) to lower limb length may be beneficial for bipedalism. This observation is in agreement with arguments made by Jungers & Stern (1983) and Preuschoft & Witte (1991).

Furthermore, however, and again holding the length of the lower limb constant, the addition of a back-carried load to a short trunk will increase the moment of inertia of the trunk, I1, by an amount, I2, which is again proportional to the dimensions of the load. Whether the net inertia (I1 + I2) is less than or greater than that of the trunk of a long-trunked subject such as AL 288-1 will depend on the dimensions and magnitude of the load. However, for a load equivalent to less than 10% of body weight, the inertia of the upper body, and hence the relative cost of loaded walking, in the short-trunked KNM-WT 15000 will be less than the relative cost of unloaded walking in the long-trunked AL 288-1.

Thus the relatively long trunk and head unit of AL 288-1 not only results in a high moment of inertia, but also imply a greater cost for load-carrying than applies for the shorter trunk of humans, and even more the very short trunk of KNM-WT 15000 model D. Although Figs 4–6 suggest that greater absolute body size and the relatively larger upper body of modern humans is not associated with a marked deficit in comparison with KNM-WT 15000, it must be borne in mind that the kinematics used to drive all models were those of adult modern human bipedalism, to which of course modern adult limb proportions and inertial properties will be best adjusted.

Witte et al. (1991) argue that the body form of A. afarensis serves the requirements of static stability. Their proposition was confirmed mathematically by Wang & Crompton (2003), who showed also, in contrast to Kramer & Eck (2000), that the greater stature and lower robusticity of H. erectus and later hominids does of itself reduce the power required to move a given mass a given distance. More specifically, however, Witte et al. (1991) show that several characteristics of the Nariokotome H. ergaster skeleton, such as the flat thorax and broad shoulders, ‘fine-tune’ dynamic response in endurant bipedal walking. However, the proportions of the upper and lower limbs also play a particular role. The efficiency of human walking is well known to be greatly enhanced by the exchange of kinetic and potential energies, which fluctuate half a cycle out of phase, owing to the inverted-pendulum-like motion of the body centre of gravity. Over long distances, the upper and lower limbs also tend to swing half a cycle out of phase, so that torques created by leg-swing are counteracted by opposite torques applied by the swinging arms, promoting mechanical efficiency by reducing motion other than in the desired, sagittal plane. Witte et al. (1991) suggest that the shorter arms of Homo, by better matching the natural pendular frequencies of the upper and lower limbs, therefore improve the efficiency of walking as compared with the long-armed, short-legged AL 288-1.

The archaeological record, by indicating an increase of ranging distances between the period of the Olduwan and Acheulian lithic industries, seems to implicate load-carrying in the evolution of modern body proportions, and the present study shows that shoulder- or back-borne loads can indeed be carried far more effectively by early African H. erectus and modern H. sapiens than by A. afarensis. However, there is also evidence to suggest that Homo species are also better designed for hand-carrying of loads. Wang et al. (2002c) demonstrated mathematically that coordination of upper and lower limb swing frequencies, which we have noted above is known to be a requisite of maximum efficiency in modern human bipedal walking, will be affected by the intermembral index (IMI, the ratio of upper to lower limb lengths) and also by any loads carried in the hand. There will be a relationship:

  • Mp = (Lp − 1)/2 − 3Lp

where Mp is the ratio of the mass held by the hand to the mass of the upper limb, and Lp the proportion of the length of the upper limb to that of the lower limb, IMI/100. Because Mp must be greater than or equal to zero, Lp must be equal to or larger than 2/3 (0.6667) and equal to or smaller than 1, so that 2/3 < Lp < 1. The resulting plot for Lp from 0.67 to 1 is given in Fig. 7 (redrawn from Wang et al., 2002c), which shows that chimpanzees, with long arms and an IMI of nearly 100, could not hand-carry any load without detriment to the match in swing frequencies. With an index of 80–90, AL 288-1 could have hand-carried loads some 15–50% of upper limb mass; KNM-WT 15000 at IMI 70, loads of 2–3 times upper limb mass; and modern humans, with IMI 68–70, up to 8 times upper limb mass. If we consider swing frequencies alone, the IMI of Homo is thus optimized for hand-carrying of loads. IMI values for WT 15000 and modern humans used here place them either side of a cusp in the curve, where small differences in IMI begin to make very large differences in the tolerable load. Our very small skeletal dataset may have led to error, but the theoretical consequences of alternative IMIs can readily be read off the graph. Note also that looking at ground reaction forces, Li et al. (2001) found that, in humans, turning forces (vertical free moments) under the feet are affected to the detriment of the effectiveness of walking by loads carried in the hand, so the effective optimum load in hand-carrying will be somewhat less than that predicted by swing frequencies alone.

image

Figure 7. Plot of the relationship of the load that may be carried in the hand without interfering with swing symmetry to intermembral index, IMI. Horizontal axis, intermembral index (IMI); vertical axis, tolerable load as proportion of upper limb mass. Pt = Pan troglodytes, Hss = modern humans. Redrawn from Wang et al. (2002c).

Download figure to PowerPoint

Implications for ranging behaviour of A. afarensis and early Homo

Aiello & Wheeler (1995) propose that the apparent ecological shift from partly closed environments associated with A. afarensis to the more open habitats associated with H. ergaster may have coincided with a redirection of the body's metabolic budget in favour of a rapidly expanding brain, itself made possible by a shift in trophic level, in which meat, and hunting, played a very important part. Aiello & Wells (2002) suggest that a narrower trunk indicates a shift to a higher-quality diet with more animal protein, and that the increased body size and change in tropic level is likely to have been accompanied by the use of fat as an energy buffer in a more unstable, xeric environment. In mammals, gut volume decreases with increase in food quality, and we might expect changes to the relative proportions of the trunk as a whole: in fact, the length of extremities relative to the trunk appears to have increased with increasing dietary protein in living human populations such as modern Chinese, and the reverse effect is reputed to occur in impoverished social groups in Europe.

Just as we found that energy transformation is more than halved in BHBK walking (Wang et al. 2003) we found that BHBK gait leads to doubling of physiological costs and double the increase in core temperature that occurs over the same period in normal walking. Recovery of temperature after BHBK walking may be estimated to require rest periods of 150% of activity time (Carey & Crompton, 2004). Thus, if A. afarensis walked ‘BHBK’ it is likely that activity time and hence ranging distances would have been small, and available food resources therefore limited. Erect walking, by contrast, would have imposed lower costs, permitted longer activity periods and larger ranges – and hence, by analogy to living primates as a whole, more selective/high-quality diet and possibly, by analogy to carnivores, have permitted a more predatory ecology. Load-carrying would only have accentuated the problems of BHBK walking. Acquisition of a more modern, long-legged, short-trunked morphology would, by contrast, have further enhanced the efficiency of endurant erect walking and reduced the costs of carrying. Our hypothesis that there is a direct relationship between the acquisition of modern postcranial proportions and increased ranging/transport distances at around 1.8–1.5 Ma appears to be borne out, although other selective factors, such as thermoregulatory influences (see Ruff, 1991; Wheeler, 1992) and adaptations for throwing (see Dunsworth et al. 2003), are likely to have played an important (although probably interdependent) role.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

R.H.C. thanks Professor John Gowlett for help with the archaeological literature and Professor Leslie Aiello for helpful comments on the manuscript. This research was funded by grants from the Natural Environment Research Council, The Engineering and Physical Sciences Research Council, The Biotechnology and Biological Sciences Research Council and The Leverhulme Trust.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Aiello LC, Wheeler P (1995) The expensive-tissue hypothesis. Current Anthropol. 36, 199221.
  • Aiello LC, Wells JCK (2002) Energetics and the evolution of the genus Homo. Annu. Rev. Anthropol. 31, 323338.
  • Alexander RMcN, Goldspink G (1977) Mechanics and Energetics of Animal Locomotion. London: Chapman & Hall.
  • Alexander RMcN (1984) Stride length and speed of adults, children and fossil hominids. Am. J. Phys. Anthropol. 63, 2327.
  • Berge C (1994) How did the australopithecines walk? A biomechanical study of the hip and thigh of Australopithecus afarensis. J. Hum. Evol. 26, 259274.
  • Bonnefille R, Vincens A, Buchet G (1987) Palynology stratigraphy and palaeoenvironment of a Pliocene hominid site (2.9–3.3 M.Y.) at Hadar, Ethiopia. Palaeogeogr. Palaeoclim. Palaeoecol. 60, 249281.
  • Carey TS, Crompton RH (2004) The metabolic costs of ‘Bent-Hip, Bent-knee’ walking in humans. J. Hum. Evol. in press.
  • Cavagna G, Heglund NC, Taylor CR (1977) Mechanical work in terrestrial locomotion: two basic mechanisms for minimizing energy expenditure. Am. J. Physiol. 233, R243R261.
  • Chandler RF, Clauser CE, McConville JT, Reynolds HM, Young JW (1975a) Investigation of the Inertial Properties of the Human Body. Technical Report DOT HS_801, 430, Wright Patterson Air Force Base, OH.
  • Chandler RF, Clauser CE, McConville JT, Reynolds HM, Young JW (1975b) Investigation of the Inertial Properties of the Human Body. AMRL Technical Report 74–137, Wright Patterson Air Force Base, OH.
  • Clark JD (1980) The Plio-Pleistocene environmental and cultural sequence at Gadeb, northern Bale, Ethiopia. In Proceedings of the 7th Panafrican Congress of Prehistory And. Quaternary Studies (eds LeakeyRE, OgotB), pp. 189193. Nairobi: TILLMIAP.
  • Clarke RJ (2003) Bipedalism and arboreality in Australopithecus. Cour. Forsch.-Inst Senkenberg 243, 7984.
  • Crompton RH, Gowlett JAJ (1993) Allometry and multidimensional form in Acheulean bifaces from Kilombe, Kenya. J. Hum. Evol. 25, 175199.
  • Crompton RH, Li Y, Wang W, Günther M, Savage R (1998) The mechanical effectiveness of erect and ‘bent-hip, bent-knee’ bipedal walking in. Australopithecus afarensis. J. Hum. Evol. 35, 5574.
  • Crompton RH, Li Y, Thorpe SK, Wang WJ, Savage R, Payne R, et al. (2003) The biomechanical evolution of erect bipedality. Cour. Forsch.-Inst Senckenberg 243, 115126.
  • Deloison Y (1991) Les Australopitheques: marchaient-ils comme nous? In Origine(s) de la Bipedie Chez les Hominides (eds CoppensY, SenutB), pp. 177186. Paris: Cahiers de Paleoanthropologie editions du CNRS.
  • Dunsworth H, Challis JH, Walker A (2003) Throwing and bipedalism: a new look at an old idea. Cour. Forsch.-Inst Senckenberg 243, 105110.
  • Gowlett JAJ (1982) Procedure and form in a Lower Palaeolithic industry: stoneworking at Kilombe, Kenya. Studia Praehistorica Belgica 2, 101109.
  • Gowlett JAJ, Crompton RH (1994) Kariandusi: Acheulean morphology and the question of allometry. Afr. Archaeol. Rev. 12, 140.
  • Harriss JWK, Herbich L (1978) Aspects of early Pleistocene hominid behaviour east of Lake Turkana, Kenya. In Geological Background to Fossil Man (BishopWW, ed.), pp. 529548. Edinburgh: Geol. Soc. London/Scottish Academic Press.
  • Hay RL (1976) Geology of the Olduvai Gorge. California: University of California Press.
  • Hunt KD (1994) The evolution of human bipedality: ecology and functional morphology. J. Hum. Evol. 26, 183202.
  • Isaac GLI (1977) Olorgesailie. Chicago, IL: University of Chicago Press.
  • Jensen RK (1989) Changes in segment inertial proportions between 4 and 20 years. J. Biomech. 22, 529536.
  • Johanson DC, Lovejoy CO, Kimbel WH, White TD, Ward SC, Bush ME, et al. (1982) Morphology of the Pliocene partial hominid skeleton (AL 288–1) from the Hadar formation, Ethiopia. Am. J. Phys. Anthrop. 57, 403452.
  • Jungers WL, Stern JT (1983) Body proportions, skeletal allometry and locomotion in the Hadar hominids: a reply to Wolpoff. J. Hum. Evol. 12, 3684.
  • Jungers WL (1988a) New estimates of body size in australopithecines. In Evolutionary History of the Robust Australopithecines (ed. GrineF), pp. 15126. New York: Aldine de Gruyter.
  • Jungers WL (1988b) Lucy's length. stature reconstruction in Australopithecus afarensis (AL 288–1) with implications for other small bodied hominids. Am. J. Phys. Anthrop. 76, 227231.
  • Kramer PA, Eck GG (2000) Locomotor energetics and leg length in hominid bipedality. J. Hum. Evol. 38, 651666.
  • Leakey MD (1971) Olduvai Gorge, Vol. III: Excavations in Beds I and 11, pp. 19601963. Cambridge: Cambridge University Press.
  • Li Y, Crompton RH, Günther MM, Alexander R McN, Wang WJ (1996) Characteristics of ground reaction forces in normal and chimpanzee-like bipedal walking by humans. Folia Primatol. 66, 13159.
  • Li Y, Wang WJ, Crompton RH, Gunther MM (2001) Free vertical movements in human walking and their relation to arm-swing. J. Exp. Biol. 204, 4758.
  • Lovejoy CO, Heiple KG, Burstein AH (1973) The gait of Australopithecus. Am. J. Phys. Anthrop. 38, 757780.
  • Lundy JK, Feldesman R (1987) Revised equations for estimating living stature from limb bones in the South African Negro. South African J. Sci. 83, 5455.
  • McGrew WR (1993) Chimpanzee Material Culture: Implications for Human Evolution. Cambridge: Cambridge University Press.
  • McHenry HM (1991) First steps? Analyses of the postcranium of early hominids. In Origine(s) de la Bipedie Chez les Hominides (eds CoppensY, SenutB), pp. 133142. Paris: Cahiers de Paleoanthropologie Editions du CRNS.
  • Mechanical Dynamics Inc. (1992) ADAMS, Version 6.1. Ann Arbor, MI: Mechanical Dynamics Inc.
  • Meriam JL, Kraige LG (1993) Engineering Mechanics, Vol. 2: Dynamics. New York: John Wiley.
  • Ohman JC, Wood C, Wood B, Crompton RH, Günther MM, Li Y, et al. (2003) Stature-at-death of KNM-WT 15000. Human Evol 17, 129114.
  • Preuschoft H, Witte H (1991) Biomechanical reasons for the evolution of hominid body shape. In Origine(s) de la Bipedie Chez les Hominides (eds CoppensY, SenutB), pp. 5979 Paris: Cahiers de Paleoanthropologie Editions du CNRS.
  • Prost JH (1980) Origin of bipedalism. Am. J. Phys. Anthrop. 52, 175190.
  • Ruff CB (1991) Climate and body shape in human evolution. J. Hum. Evol. 21, 81105.
  • Ruff CB, Walker A (1993) Body size and body shape. In The Nariokotome Homo Erectus Skeleton (eds WalkerA, LeakeyR), pp. 234265. Cambridge, MA: Harvard University Press.
  • Stern JT, Susman RL (1983) The locomotor anatomy of Australopithecus afarensis. Am. J. Phys. Anthrop. 60, 279317.
  • Stern JT (1999) The cost of bent-knee, bent-hip bipedal gait. A reply to Crompton et al. J. Hum. Evol. 36, 567570.
  • Swisher CC, Curtis GH, Jacob T, Getty AG, Suprijo A (1994) Widaasmoro. Age of the earliest known hominids in Java, Indonesia. Science 263,11181121.
  • Taylor CR, Heglund NC, McMahon TA, Lonney TR (1980) Energetic cost of generating muscular force during running: a comparison of large and small animals. J. Exp. Biol. 86, 918.
  • Thorpe SK, Crompton RH (2004) The locomotor ecology of wild orang-utans (Pongo pygmaeus abelii) in the Gunung Leuser Ecosystem, Sumatra, Indonesia: A multivariate analysis using log-linear modelling. Am. J. Phys. Anthrop. in press.
  • Tuttle RH, Webb DM, Baksh M (1991) Did Australopithecus afarensis make the Laetoli G footprint trails? Am. J. Phys. Anthrop. 12 (Suppl.), 175.
  • Wang W, Crompton RH, Wood CG, Li Y, Günther MM (2002a) Kinematics and kinetics of the joints of the lower limb during normal and bent-knee loaded walking. In The Engineering of Sport 4 (eds UjihashiS, HakkeS), pp. 739747. London: Blackwell.
  • Wang WJ, Crompton RH, Gunther MM, Savage R (2002b) A method to determine the optimum solution in multi-variable dynamic simulation, and a preliminary application to human walking. In The Engineering of Sport 4 (eds UjihashiS, HakkeS), pp. 339345. London: Blackwell.
  • Wang WJ, Crompton RH, Li Y, Gunther MM (2002c) Optimum ratio of upper to lower limb lengths under the assumptions of frequency coordination and hand-carrying of a load. J. Biomech. 36, 249252.
  • Wang WJ, Crompton RH (2003) Size and power required for motion with implications for the evolution of early hominids. J. Biomech. 36, 12371246.
  • Wang WJ, Crompton RH, Li Y, Gunther MM (2003) Energy transformation during erect and bent-hip-bent-knee walking by humans with implications for the evolution of bipedalism. J. Hum. Evol. 44, 563580.
  • Ward CV (2002) Interpreting the posture and locomotion of Australopithecus afarensis: where do we stand? Am. J. Phys. Anthrop. 45 (Suppl.), 185215.
  • Wheeler PE (1992) The thermoregulatory advantages of larger body size for hominids foraging in savannah environments. J. Hum. Evol. 23, 351362.
  • White TD, Suwa , G (1987) Hominid footprints at Laetoli – facts and Interpretations. Am. J. Phys. Anthrop. 72, 485514.
  • Winter DA (1990) Biomechanics and Motor Control of Human Movement, 2nd edn. New York: Wiley.
  • Witte H, Preuschoft H, Recknagel S (1991) Human body proportions explained on the basis of biomechanical principles. Z. Morph. Anthropol. 78, 407423.
  • Wood BA (1994) Four legs good, two legs better. Nature 363, 587588.