Chimpanzee locomotor energetics and the origin
of human bipedalism
Michael D. Sockol*, David A. Raichlen†, and Herman Pontzer‡§
*Department of Anthropology, University of California, Davis, CA 95616;†Department of Anthropology, University of Arizona, Tucson, AZ 85721;
and‡Department of Anthropology, Washington University, St. Louis, MO 63130
Edited by David Pilbeam, Harvard University, Cambridge, MA, and approved June 12, 2007 (received for review April 9, 2007)
Bipedal walking is evident in the earliest hominins [Zollikofer CPE,
Ponce de Leon MS, Lieberman DE, Guy F, Pilbeam D, et al. (2005)
Nature 434:755–759], but why our unique two-legged gait evolved
remains unknown. Here, we analyze walking energetics and bio-
mechanics for adult chimpanzees and humans to investigate the
long-standing hypothesis that bipedalism reduced the energy cost
of walking compared with our ape-like ancestors [Rodman PS,
McHenry HM (1980) Am J Phys Anthropol 52:103–106]. Consistent
with previous work on juvenile chimpanzees [Taylor CR, Rowntree
VJ (1973) Science 179:186–187], we find that bipedal and quadru-
pedal walking costs are not significantly different in our sample of
adult chimpanzees. However, a more detailed analysis reveals
significant differences in bipedal and quadrupedal cost in most
individuals, which are masked when subjects are examined as a
group. Furthermore, human walking is ?75% less costly than both
between bipedal and quadrupedal walking, as well as between
chimpanzees and humans, is well explained by biomechanical
differences in anatomy and gait, with the decreased cost of human
walking attributable to our more extended hip and a longer
hindlimb. Analyses of these features in early fossil hominins,
coupled with analyses of bipedal walking in chimpanzees, indicate
that bipedalism in early, ape-like hominins could indeed have been
less costly than quadrupedal knucklewalking.
biomechanics ? human evolution ? locomotion ? limb length ? inverse
divergence of the human lineage from the other apes. One
enduring hypothesis for this transition is that bipedalism evolved
to reduce locomotor costs in early hominins, relative to the
ape-like last common ancestor (LCA) of chimpanzees and
humans (3). Under this scenario, reducing the energy cost of
walking provided early hominins with an evolutionary advantage
over other apes by reducing the cost of foraging. Such an
advantage may have been especially important given the cooler,
drier climate that prevailed at the end of the Miocene (4, 5) and
that would have increased the distance between food patches.
Testing this hypothesis requires comparative data not only on
the cost of locomotion (COL) in humans and chimpanzees but
also on the biomechanical determinants of these costs. The only
previous study of chimpanzee locomotor cost used juvenile
chimpanzees and indicated that bipedalism and quadrupedalism
were equally costly in chimpanzees and that both were more
costly than human locomotion (6). Although this study has been
central to the debate over energetics and the evolution of
bipedalism (3, 7), the reliability of these data has been ques-
tioned because adult and juvenile locomotor mechanics and
costs can differ substantially (7) and because of recent evidence
that bipedalism is more costly than quadrupedalism in other
did not include a biomechanical analysis of the determinants of
chimpanzee locomotor costs, limiting the potential application
of the study to the hominin fossil record.
of the earliest hominins (2) and thus marks a critical
Here, we compare human and adult chimpanzee locomotor
energetics and biomechanics to determine links among anatomy,
gait, and cost. Our study focuses on two primary questions. First,
do adult chimpanzees follow the pattern of costs found previ-
ously for juveniles (6)? Second, do differences in anatomy and
gait between bipedal and quadrupedal walking, as well as
between chimpanzees and humans, explain observed differences
in cost? Using this biomechanical approach to link differences in
anatomy and gait to cost, we then examine what changes, if any,
would lower the cost of bipedalism for an early hominin, such
that bipedalism would be more economical than the ape-like
quadrupedalism of the last common ancestor.
We focused on walking speeds because walking is the gait
commonly used during terrestrial travel in wild chimpanzees (9).
We tested two sets of predictions; first, based on recent studies
of primate mechanics and energetics (8, 10), we predicted that
bipedal and quadrupedal (i.e., ‘‘knucklewalking’’) costs will
differ in adult chimpanzees and that both bipedal and quadru-
pedal walking in chimpanzees will be energetically more costly
relative to other quadrupeds and humans. Second, following
previous work (11, 12), we predicted that these differences in
cost would be explained by corresponding differences in (i) the
force required to support bodyweight during each step and (ii)
the volume of muscle activated to generate one unit of ground
To test these predictions, we collected metabolic, kinematic,
and kinetic data during walking from five chimpanzees, aged
6–33 years, and four adult humans (see Table 1 and Methods).
The magnitude of ground force was estimated as the inverse of
the duration of foot–ground contact time, tc, per step (11, 13),
and the volume of muscle activated per unit of ground force,
Vmusc, was estimated by using inverse dynamics (14) (see Meth-
ods). Following Roberts et al. (12), we predicted that the COL
(ml of O2kg?1s?1), varies as the ratio of Vmuscand tc, Vmusc/tc.
Thus, any difference in Vmusc/tc, either between species or
between gaits, should lead to a proportional difference in COL.
Results and Discussion
The mass-specific cost of transport (ml of O2 kg?1m?1) for
chimpanzees was greater than expected for their body size (15)
(Fig. 1). In contrast, human walking was less expensive than
Author contributions: M.D.S., D.A.R., and H.P. contributed equally to this work; M.D.S.,
D.A.R., and H.P. designed research; M.D.S., D.A.R., and H.P. performed research; M.D.S.,
D.A.R., and H.P. analyzed data; and M.D.S., D.A.R., and H.P. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Abbreviations: COL, cost of locomotion (measured as the mass-specific rate of oxygen
force during locomotion.
§To whom correspondence should be addressed at: Department of Anthropology,
Washington University, 119 McMillan Hall, St. Louis, MO 63130. E-mail: hpontzer@
This article contains supporting information online at www.pnas.org/cgi/content/full/
© 2007 by The National Academy of Sciences of the USA
July 24, 2007 ?
vol. 104 ?
no. 30 ?
expected for their body size and substantially (?75%) less
expensive than chimpanzee locomotion (Fig. 1). Within the
significantly, more costly (?10%) than quadrupedal walking
(Fig. 1). This finding is consistent with previous work on juvenile
chimpanzees (6), which indicated that bipedal and quadrupedal
locomotion were equally costly for chimpanzees. However,
differences in bipedal and quadrupedal cost varied among
individuals (Fig. 1), and in contrast to the study by Taylor and
Rowntree (6), most subjects exhibited significant differences
between gaits. For three chimpanzees (C1-C3) bipedalism was
32.2% more expensive (P ? 0.001, Student’s paired t test), but
for two other chimpanzees, bipedal costs were similar (P ? 0.39;
C5) or even less than quadrupedal costs (P ? 0.05; C4).
Given these results, we investigated potential biomechanical
sources of the observed differences in cost among chimpanzees
and between chimpanzees and humans. As predicted, differ-
ences in kinematics and estimated muscle activation explained
observed differences in cost between bipedal and quadrupedal
walking as well as between humans and chimpanzees. In the
three chimpanzees for which Vmusc were estimated (subjects
C1-C3; see Methods), an increase in Vmusc and shorter tc in-
creased Vmusc/tc by 35.2% (?5.2%) during bipedal walking
compared with quadrupedal walking. This difference corre-
sponds closely to the observed 32.2% (?3.2%) increase in COL
during bipedal walking for these subjects (Figs. 2 and 3). When
human walking was compared with chimpanzee bipedal walking,
humans activated smaller muscle volumes per unit of ground
differences caused a 79.4% (?1.6%) lower ratio of Vmusc/tc,
which corresponded closely to the observed 76.8% (?2.6%)
decrease in locomotor cost (Fig. 3). Similarly, although tcfor
quadrupedal chimpanzees were slightly longer than for humans,
Vmusc/tcwas 72.8% (?4.6%) lower for humans than for quadru-
pedal chimpanzees, matching the 68.5% (?4.3%) difference in
COL (Figs. 2 and 3).
Interspecific differences in tc and Vmusc point directly to
anatomical and kinematic sources for the observed differences
in cost between chimpanzees and humans. First, the shorter legs
of chimpanzees (Table 1) lead to shorter tcfor a given speed (16)
during bipedal walking (Fig. 2c), which increases the magnitude
of the ground reaction force (GRF) impulse for each step. That
is, because of their shorter hindlimbs, bipedal chimpanzees must
generate greater ground forces at a faster rate than humans,
thereby increasing bipedal costs (11, 13). Conversely, the long
forelimbs of chimpanzees increase tcand decrease ground force
impulses during quadrupedal walking. Second, the bent-hip,
bent-knee gait of chimpanzees positions their body’s center of
mass anterior to the hip joint and increases the moment arm of
the GRF. This posture generates large external flexion moments
(Fig. 2a) that, when combined with chimpanzees’ long muscle
fibers (17), must be opposed by activating a correspondingly
large volume of hip extensor muscle (Fig. 2b). Additionally, the
long muscle fibers (17) and crouched posture of chimpanzees
result in large Vmuscat the knee (Fig. 2b). In contrast, humans
decrease Vmuscby adopting an upright posture, which orients the
GRF vector nearer to the hip and knee joints and confines large
moments to the ankle, where muscle fibers are short (Fig. 2a).
Thus, although the long forelimbs of chimpanzees enable them
to knucklewalk using longer tc than humans at dynamically
similar speeds (Froude number ?0.2), walking costs are lower in
humans than in chimpanzees.
The influence of tcand joint angle on locomotor cost is further
supported by subject C4. Only this chimpanzee used longer tc
during bipedal walking and flexed her knee and hip to a similar
degree during bipedal and quadrupedal walking (Fig. 4). As
expected, C4 was also the only subject whose costs were lower
during bipedal versus quadrupedal walking, although not as low
as in humans (Fig. 1). These results highlight how slight kine-
matic changes can lead to large variations in locomotor cost and
in posture can affect cost (18–20). Note, however, that chim-
panzees cannot employ the full hip and knee extension typical of
humans because of their distally oriented ischia, which reduce
the ability of the hamstrings to produce an extensor moment
Table 1. Chimpanzee and human costs of transport
Cost of transport, ml of O2kg?1m?1
C1, 6-year-old male
C2, 9-year-old male
C3, 19-year-old female
C4, 33-year-old female
C5, 27-year-old female
Chimpanzees, n ? 5
Humans, n ? 4
Individual means (with SEs in parentheses) were calculated from four speeds in each gait for each subject. Species means were
calculated from individual means. Froude number was calculated from hip height by following Alexander and Jayes (34). P values are
one-tailed, for paired-samples Student’s t tests.
pedal walking (blue), chimpanzee bipedal walking (red), and human walking
(SI) Table 2].
Net cost of transport (ml of O2kg?1m?1) for chimpanzee quadru-
www.pnas.org?cgi?doi?10.1073?pnas.0703267104Sockol et al.
when the femur is extended relative to the pelvis (21, 22).
Chimpanzee pelvic anatomy thus requires them to walk with a
flexed hip and knee throughout their stride. In contrast, human
ischia are oriented dorsally, allowing large hamstrings extensor
moments when the femur is fully extended (21).
Results of our biomechanical analyses linking anatomy and
gait to cost generate two testable predictions for the hominin
fossil record. If locomotor economy was a selective force behind
hominin bipedalism, then early hominin lower limbs should be
longer than those of apes, and the ischia of early hominin pelves
allow us to test these predictions in the earliest hominins, but an
increase in leg length is apparent in partially complete specimens
of Australopithecus afarensis (AL-288) and Australopithecus af-
ricanus (22–24), consistent with selection to increase tc and
thereby lower locomotor cost. Furthermore, A. afarensis (AL
288-1) and A. africanus (STS-14) both have a more dorsally
oriented ischium compared with chimpanzees (21, 22, 25). These
the hamstrings when the hip was extended beyond that of
chimpanzees, thereby reducing Vmuscand thus lowering walking
However, previous morphological investigations have sug-
gested that australopithecines walked with greater knee and hip
indicate extension. Note the large hip flexion moments in chimpanzees relative to humans. Horizontal dashed lines indicate zero joint torque. (b) Vmuscper
chimpanzee hindlimbs during bipedalism (red), and human hindlimbs (yellow). (c) Mean tc(in seconds) during walking in chimpanzees and humans. Note that
Froude numbers are similar for all groups (Fr ? 0.2; Table 1), but absolute speeds are slightly higher for humans (Table 1).
Comparison of walking mechanics in chimpanzees and humans. (a) GRF vectors and joint torque for humans and chimpanzees. Figures show joint
Fig. 3. Comparison of differences in Vmusc/tc(white bars) and COL (gray bars) between gaits and species. Error bars indicate ?1 SEM percent difference.
Sockol et al.
July 24, 2007 ?
vol. 104 ?
no. 30 ?
flexion, and therefore greater cost, than modern humans (25,
26), although this issue remains intensely debated (27). None-
theless, even if early hominins used a bent-hip, bent-knee form
of bipedalism (25), our results suggest that early transitional
forms would have reaped some energy savings with minor
the last common ancestor of humans and chimpanzees was a
chimpanzee sample (Fig. 4) demonstrates that some members of
the last common ancestor population likely had the ability to
extend their hindlimb more fully and to use longer tc during
bipedal locomotion. These kinematics would have decreased the
cost of bipedal walking below that of quadrupedal knucklewalk-
ing in these individuals (Fig. 4, Table 1). Even small increases in
energy savings from slight increases in hindlimb extension or
length may have provided critical selection pressure for early
hominins (31, 32).
Our results, therefore, support the hypothesis that energetics
played an important role in the evolution of bipedalism. Unfor-
tunately, a lack of postcranial evidence from the earliest homi-
nins and their immediate forebears prevents us from testing the
hypothesis that locomotor economy provided the initial evolu-
tionary advantage for hominin bipedalism. However, regardless
of the context under which bipedalism evolved, our biomechani-
cal analysis of adult chimpanzee costs, coupled with previous
analyses of early hominin pelvic and hindlimb morphology,
suggests that improved locomotor economy may have accrued
very early within the hominin lineage. Future fossil discoveries
from the earliest hominins will resolve whether this energetic
advantage was in fact the key factor in the evolution of hominin
Five chimpanzees (two males and three females; mean age, 18.2
years; range, 6–33 years) were trained over the course of 4
months to walk quadrupedally (i.e., knucklewalk) and bipedally
on a treadmill (Smooth 9.15; Smooth Fitness, Sparks, NV).
Three of these subjects (C1-C3) were also trained to walk down
a force-plate-equipped track. All subjects were socially housed in
large, outdoor enclosures at a United States Department of
Agriculture registered and approved facility. Institutional Ani-
mal Care and Use Committee approval was obtained before the
beginning of the study, and institutional animal care guidelines
were followed throughout.
During treadmill trials, subjects wore loose-fitting masks that
collected expired air, and the mass-specific COL was measured
via established open-flow methods (15). COL was measured at
a range of speeds for each individual. Only trials lasting a
minimum of 3 min, and in which the oxygen-consumption rate
visibly plateaued, were included for analysis. Multiple COL
measurements were taken at each speed for each subject, and
means were used for subsequent analyses. For a subset of
treadmill trials, a set of kinematic variables, including tc(i.e.,
duration of stance for one foot or hand) was collected via
high-speed video (125 frames/s; Redlake, Tucson, AZ).
During force-plate trials, subjects walked down a 10-m track
equipped with an embedded force-plate (Kistler, Amherst, NY)
recording at 4 kHz, providing vertical and fore-aft GRFs.
Simultaneous kinematic data were collected via high-speed
video (125 frames/s; Redlake), with joint centers for forelimbs
and hindlimbs (shoulder, elbow, wrist, hip, knee, and ankle)
marked on each subject by using nontoxic water-based white
paint. Because flexion and extension of the limb joints occur
primarily in the sagittal plane during walking and because
mediolateral forces were smaller than anteroposterior ground
forces and generally ?10% of vertical ground forces, we re-
stricted our analyses to the sagittal plane. Force-plate trials were
cleanly and if fore-aft GRF traces indicated constant forward
Body mass and external measurements for each subject were
used to calculate segment inertial properties by following
moments by following Winter (14), by using force and kinematic
data. Joint moments were combined with published data on
chimpanzee muscle moment arms (17) to calculate the opposing
extensor muscle forces generated for each muscle group. The
volume of muscle activated for each step was then calculated by
following Roberts et al. (12), by using published muscle-fiber
Previous work (12) has shown that the mass-specific energy
activated to apply a unit of ground force, such that COL ? k ?
(Vmusc/tc), where k is a constant relating oxygen consumption and
force production (ml of O2N?1). This relationship holds because
the energy cost of terrestrial locomotion derives primarily from
muscle forces generated to support bodyweight.
For comparison with humans, a similar dataset of locomotor
cost, kinematics, and muscle activation was collected for a
segment and the horizontal (e.g., thigh angle is 90° when the thigh is perpendicular to the ground). Knee angle is measured as the angle between the thigh
and bipedal (red) and strides (P ? 0.05, Student’s paired t test).
Comparison of thigh angle, knee flexion, and tcfor C4 versus other chimpanzees (n ? 4). Thigh angles is measured as the angle between the thigh
www.pnas.org?cgi?doi?10.1073?pnas.0703267104Sockol et al.
sample of four humans (one female and three males). Subjects
were recreationally fit adults with no gait abnormalities and gave
informed consent for this study. Human subjects committee
approval was obtained before this study, and institutional guide-
lines were followed throughout. Methods for obtaining locomo-
tor cost and kinetic data were identical to those used for
chimpanzees, with the following exceptions: kinematics were
measured via a high-speed infrared motion analysis system
(Qualisys, Gothenburg, Sweden), data for muscle-fiber lengths
and joint mechanical advantage were calculated by following
Biewener et al. (20), and segment inertial properties were
calculated by following Winter (14).
We thank D. E. Lieberman, A. A. Biewener, P. S. Rodman, H. M.
McHenry, D. M. Bramble, and two anonymous reviewers for useful
comments on earlier versions of this manuscript. A. A. Biewener, D. E.
Lieberman, J. Jones, and P. Rodman generously provided necessary
equipment. This project was supported by grants from the National
Science Foundation (Grant BCS-0424092) and the L. S. B. Leakey
1. Darwin C (1871) The Decent of Man, and Selection in Relation to Sex (John
Murray, London); reprinted (1981) (Princeton Univ, Princeton).
2. Zollikofer CPE, Ponce de Leon MS, Lieberman DE, Guy F, Pilbeam D, Likius
A, Mackaye HT, Vignaud P, Brunet M (2005) Nature 434:755–759.
3. Rodman PS, McHenry HM (1980) Am J Phys Anthropol 52:103–106.
4. Cerling TE, Harris JM, MacFadden BJ, Leakey MG, Quade J, Eisenmann V,
Ehleringer JR (1997) Nature 389:153–158.
5. Cerling TE, Ehleringer JR, Harris JM (1998) Philos Trans R Soc London B
6. Taylor CR, Rowntree VJ (1973) Science 179:186–187.
7. Steudel-Numbers KL (2003) J Hum Evol 44:255–262.
8. Nakatsukasa M, Ogihara N, Hamada Y, Goto Y, Yamada M, Hirakawa T,
Hirakasi E (2004) Am J Phys Anthropol 124:248–256.
9. Hunt KD (1992) Am J Phys Anthropol 87:83–105.
10. Thorpe SKS, Crompton RH, Wang WJ (2004) Folia Primatol (Basel) 75:253–
11. Kram R, Taylor CR (1990) Nature 346:265–267.
12. Roberts TJ, Chen MS, Taylor CR (1998) J Exp Biol 201:2753–2762.
13. Pontzer H (2007) J Exp Biol 210:484–494.
14. Winter DA (1990) Biomechanics and Motor Control of Human Movement
(Wiley, New York).
15. Taylor CR, Heglund NC, Maloiy GMO (1982) J Exp Biol 97:1–21.
16. Hoyt DF, Wickler SJ, Cogger EA (2000) J Exp Biol 203:221–227.
17. Thorpe SKS, Crompton RH, Gunther MM, Ker RF, Alexander RM (1999)
Am J Phys Anthropol 110:179–199.
18. Carey TS, Crompton RH (2005) J Hum Evol 48:25–44.
19. Crompton RH, Li Y, Wang WJ, Gunther MM, Savage R (1998) J Hum Evol
20. Biewener AA, Farley CT, Roberts TJ, Temaner M (2004) J Appl Physiol
21. Robinson JT (1972) Early Hominid Posture and Locomotion (Chicago Univ
22. Fleagle J, Anapol FC (1992) J Hum Evol 22:285–305.
23. Jungers WL (1982) Nature 297:676–678.
24. McHenry HM (1991) Am J Phys Anthropol 85:149–158.
25. Stern JT, Susman RL (1983) Am J Phys Anthropol 60:279–317.
26. Susman RL, Stern JT, Jungers WL (1984) Folia Primatol (Basel) 43:113–156.
27. Ward CV (2002) Yearb Phys Anthropol 45:185–215.
28. Washburn SL (1967) Proc R Anthropol Inst Gr Br Ireland 3:21–27.
29. Pilbeam DR (1996) Mol Phylogenet Evol 5:155–168.
30. Richmond BG, Strait DS (2000) Nature 404:382–385.
31. Steudel KL (1994) Evol Anthropol 3:42–48.
32. Leonard WR, Robertson ML (1997) Curr Anthropol 38:304–309.
33. Raichlen DA (2004) J Hum Evol 46:719–738.
34. Alexander RM, Jayes AS (1983) J Zool 201:135–152
Sockol et al.
July 24, 2007 ?
vol. 104 ?
no. 30 ?