Hominin first metatarsals (SKX 5017 and SK 1813)
from Swartkrans: A morphometric analysis
B. Zipfela, R. Kidd.
A Bernard Price Institute for Palaeontological Research, University of the Witwatersrand,
PO Wits, 2050 Wits, South Africa
bSchool of Biomedical and Health Sciences, University of Western Sydney, Cambelltown,
NSW 2560, Australia
Received 29 June 2005; accepted 1 January 2006
Two hominin metatarsals from Swartkrans, SKX 5017 and SK 1813, have been reported by
Susman and Brain [1988. New first metatarsal (SKX 5017) from Swartkrans and the gait of
Paranthropus robustus. Am. J. Phys. Anthropol. 79, 451–454] and Susman and de Ruiter
[2004. New hominin first metatarsal (SK 1813) from Swartkrans. J. Hum. Evol. 47, 171–181].
They found these bones to have both primitive and derived traits indicating that, while being
bipedal, these hominines had a unique toe-off mechanism. We have undertaken additional
multivariate morphometric analyses, comparing the fossils to the first metatarsals of modern
humans and extant apes. The largest proportion of discrimination lies in the different
locomotor functions: apes on the one hand and the humans and fossils on the other. While the
fossils have the closest affinity to humans, they have a unique biomechanical pattern
suggesting a more facultative form of bipedalism. The implications of this are, while
morphometric analyses do not necessarily directly capture the described primitive and derived
traits, the associated functional pattern is held within the broader morphology of the bone.
r 2006 Elsevier GmbH. All rights reserved.
Pedal elements within the fossil record are extremely rare, in particular the
forefoot elements consisting of the metatarsals and phalanges. The metatarsus
contains functional features that clearly discriminate among the extant Hominoidea
(e.g. Day and Napier, 1964; Archibald et al., 1972; Susman et al., 1984; Aiello and
Dean, 1990), and compared to extinct hominins, contributed to an understanding of
the evolution of the human foot (Keith, 1929; Morton, 1935; Lewis, 1981; Susman,
1983). Within the fossil record there is as yet no complete pre-human metatarsus
available comprising all five bones. The most complete hominin foot is that from
East Africa, the OH8 foot from Olduvai, of which the metatarsal heads are missing
from all five bones. In contrast, the Hadar fossils A.L. 333-115 (c. 3.0–3.4Ma)
consist of only the metatarsal heads (Susman et al., 1984). The Stw573 fossil
assemblage from Sterkfontein, South Africa (possibly as old as 3.5 Ma), initially
reported by Clarke and Tobias (1995) and more recently by Kidd and Oxnard
(2004), contains only a proximal hallucal metatarsal fragment. Stw 562, also from
Sterkfontein, as yet not formally described, is an almost complete first metatarsal
bone (Susman and de Ruiter, 2004).
Susman and Brain (1988) described an isolated undistorted left hallucal metatarsal
(SKX 5017) recovered from the ‘‘lower bank’’ of Swartkrans Member 1. Member 1 is
estimated to be approximately 1.5–1.8Ma and has yielded more than 130 hominin
individuals. Of these, more than 95% are attributed to Paranthropus (Brain, 1981)
and SKX 5017 to Paranthropus robustus (Susman and Brain, 1988). SKX 5017 is
unique as it represents the only documented undamaged hominin first metatarsal
from the Plio-Pleistocene range. This specimen is from the same time range as the
OH8 hominin, roughly 1.8Ma (Day and Wood, 1968) and the possibility of it being
contemporaneous with the Stw573 hominin cannot be ruled out should the
revised estimated dates of the Sterkfontein formation by Berger et al. (2002) be
proven to be correct.
Recently, Susman and de Ruiter (2004) reported on another Swartkrans
metatarsal, that of the SK 1813 right hallucal metatarsal. The specimen was
recovered from samples labeled ‘‘channel fill’’. The channel fill itself consists of
rubble discarded by miners and thrown into backfill holes during limestone
operations (Brain, 2004a). This later became known as Member 2 (Butzer, 1976;
Brain, 2004b), yet the context of the fossil is not entirely clear as the channel fill
represents a highly disturbed surface (Susman and de Ruiter, 2004). The specimen
SK 1813 is complete although the shaft was broken just beyond the mid-point during
removal from the breccia and a portion of bone was dislodged from the medial
aspect of the proximal articular surface.
Susman and Brain (1988) based their study of SKX 5017 on observations,
measurements, and radiography. The non-metrical observations revealed that
the base, shaft and head suggested human-like foot posture and human-like
dorsiflexion of the first metatarsophalangeal joint, while the mediolateral diameters
of the distal articular surface indicates that the human-like toe-off mechanism was
absent in Paranthropus. On the lateral margin of the base, dorsal to the peroneal
tubercle, there is a small area that served as a contact point for the second
This contact point is similar to that found in humans originally described
by Singh (1960) as a ‘‘variable articular facet’’ and later classified by Romash et al.
(1990) as a facet that is either not present, transitional or well developed. This may
be a feature of a non-opposable hallux in humans (Day and Napier, 1964) and
suggests a smaller angle between the first and second metatarsals (Fritz and
Susman and de Ruiter (2004) in their report on SK 1813 found that it bears
a strong morphological resemblance to SKX 5017 and in addition to the
descriptive morphology, they carried out a discriminant functions analysis
including SK 1813, SKX 5017, Stw 562, samples of chimpanzees, bonobos, gorillas
and modern humans. The multivariate results suggest an affinity of the Swartkrans
specimens to humans. Susman and de Ruiter (2004) concluded that while
bipedal, these early hominins possessed a unique toe-off mechanism as a result
of the mosaic of primitive and derived traits. Whilst we are in agreement
with the evaluations of the Swartkrans metatarsals and specimens of this nature
are extremely rare, it was considered appropriate to obtain additional evidence
utilizing alternative methods. The development of obligate bipedalism is now
generally considered to be one of the most significant adaptations to occur within the
hominin lineage. To this end, therefore, a multivariate study utilizing principal
components and canonical variates analyses (CVAs) of the Swartkrans first
metatarsals were carried out.
Materials and methods
The materials used in the current study consist of the first metatarsals from SKX
5017 and SK 1813. These were made available courtesy of the Transvaal Museum,
South Africa. Osteometric data sets were collected by one of the authors (RSK) of
samples from four extant species, namely modern human (Homo sapiens),
chimpanzee (Pan troglodytes), gorilla (Gorilla gorilla), and orang-utan (Pongo
pygmaeus). Females and males of each species were treated as groups in their own
right and with the exception of the orang-utans, were more or less equally
represented. The human sample comprised 18 males and 18 females, the chimpanzee
and gorilla samples consisted of 20 males and 20 females from each group and the
orang-utans consisted of 11 males and 16 females. The human specimens were
Victorian British (The Spitalfields Collection), made available courtesy of the British
Museum of Natural History. The chimpanzee and gorilla samples by the Powell-
Cotton Museum, England, and the orang-utan sample by the Smithsonian
Institution, Washington DC. The specimens used in this study were restricted to
adults as these have completely fused epiphyses. Where possible, dimensions were
collected from the left-hand side, previous studies demonstrating no significant
difference in variation between sides (Steudal, 1984; Kidd, 1995).
All dimensions have been chosen so as to reflect the broad morphology of
the bone. For the purpose of this study, a minimum number of variables thought
to be representative of the general size and shape of the bones were utilized and
are defined below.
Measurements obtained from the first metatarsal were defined in relation to the
long dimension of the base, considered to be coincident with the sagittal plane and
the transverse plane to be at right angles to this. The measured dimensions were
based on those defined by Martin and Saller (1957).
Using these defined planes, the following linear variables are defined:
(1) The functional length is measured from the posterior articular surface to the
extreme of the anterior articular surface.
(2) The height of the base is the maximum height measured from the most superior
point on the base to the most inferior point on the base in the assumed sagittal
(3) The breadth of the base is measured at right angles to (2) above.
(4) The height of the head is the maximum height measured from the most superior
point on the distal articular surface to the most inferior point of the distal
(5) The breadth of the head is the maximum bone span measured at right angles to (4)
(6) The height of the shaft is measured at the mid-shaft in the sagittal plane.
(7) The breadth of the shaft is measured at right angles to (6) above.
All dimensions were obtained using standard digital sliding calipers. All readings
were taken in millimeters and recorded to 0.01mm with the bone held and orientated
by hand. While this level of accuracy is undoubtedly spurious, the number was
recorded without rounding in order to avoid the pitfall of false recording of data. All
seven dimensions were utilized as these could be measured on both SKX 5017 and
SK 1813 even though the latter bone was damaged.
Although this was a multivariate study, an initial univariate analysis was
undertaken. There are several reasons for this. First, it is useful in identifying
erroneously recorded data. Second, it is a useful method for obtaining a broad
comparison of the size and variance of each variable in different groups. Third, it is
essential in the interpretation of subsequent multivariate analyses.
The standard univariate descriptors of mean, standard deviation and coefficient of
variation were examined. Subsequently, Student ‘‘t’’ tests were undertaken to
investigate significance of differences of means within groups. Plots of means against
their standard deviations revealed a clear positive regression; as a consequence, all
data were subsequently transformed to their natural logarithms. The multivariate
objective of the study was to establish patterns of morphological discrimination
within and between the groups, initially using principal components analysis (PCA)
(Blackith and Reyment, 1971; Bryant and Yarnold, 2001) and subsequently using
CVA (Reyment et al., 1984; Albrecht, 1980, 1992). Computations for both analyses
were undertaken using PC SASs 8.2 (2003).
PCA does not make any a priori patterns of inter-relationship such as sex
differences or the identification of a particular group or groups. It thus shows the
distribution shape of the pooled group of organisms and can therefore be used as a
cluster finding tool including both fossils within the overall structure. In the current
study, the PCA served primarily as an exploratory exercise to validate the data for
subsequent CVA and to examine the relationship of the fossils to each other as well
as to the extant species. The PCA produces two standard outputs: eigenvalues
(indicating the proportion of the total information contained within each principal
component) and eigenvectors.
CVA defines the maximum discrimination between groups, relative to the
variation within the group (Reyment et al., 1984) and unlike PCA, requires
an a priori definition of the groups. CVA produces four standard outputs:
group means for each group on each canonical variate, eigenvalues of each canonical
variate (indicating the proportion of discrimination within the variate), canonical
coefficients, and Mahalanobis D2 distance matrix. In the CVA used in this
study, each fossil was entered directly as part of the overall canonical structure
as a sample size of unity, rather than by interpolation into the matrix of
extant species. The canonical component of this study was undertaken twice,
once for each fossil. The reason for this is that a weighted analysis was used. While
there is much debate with regards to the relative merits of weighted and unweighted
analyses (e.g. Albrecht, 1980, 1992), they do serve to maximize the amount of
discrimination held within early variates. However, the inclusion of both fossils (i.e.
two samples of size unity) is likely to produce distortion; the analysis was thus
As univariate values primarily reflect size, it is difficult to interpret these results as
minor shape differences tend to be swamped. All fossil values were ape-like but did
not consistently fall within the range of any particular species. All the values of SK
1813, with the exception of the breadth of the mid-shaft were smaller than SKX
5017. The length and proximal height and breadth values of both fossils fall within
or approach the mean values of the chimpanzees and orang-utans. The head
heights of the fossils fall within the range of the chimpanzee males and gorilla
females. The shaft height of SKX 5017 is comparable to both humans and gorillas,
whilst that of SK 1813 lies between the chimpanzees and gorillas. As there is great
variation within each dimension of each species group, and the fossils are isolated
without any indication of their context within their own taxa, multivariate analysis is
Principal components analysis
The majority of the variation lies within the first two principal components,
together accounting for just over 91% of the total variation. A plot of the first two
components, giving the position of each individual, is given in Fig. 1. The
eigenvectors from the first principal component are all of positive sign and would
tend to indicate that most of the variance contained within this principal component
is associated with size and size-related shape (Table 1). Both the fossils lie centrally
on the first component with SK 1813 lying within the spread of chimpanzees and
orang-utans. SKX 5017 lies more positively, at the margin of the orang-utans, within
the spread of the humans and African apes. On the second principal component,
containing 4.37% of the total variation, the eigenvectors are of both positive
and negative sign indicating a large component of size-independent shape content
(Table 1). On this component both fossils lie negatively to all the apes within the
negative range of the humans and quite close to the orang-utans, occupying
approximately the same space (0.2) on this axis.
Canonical variates analysis
In the analysis of each fossil together with the extant species, the majority of the
discrimination lies within the first two variates, together accounting for at least 92%
of the total discrimination. Subsequent variates contain considerably less variation
and are almost identical for the analysis of both fossils. The third variate contains
between 5.67% and 6.14% of the total discrimination and the fourth variate just
over 1%. The first two variates thus account for most of the discrimination.
On the first canonical variate, the fossil lies well within the spread of the African apes
and humans. The fossil lies with the chimpanzees on the one hand and humans and
gorillas on the other; more specifically between the chimpanzee males and gorilla
females (Fig. 2). The two main dimensions contributing to this discrimination are the
height of the head and breadth of the mid-shaft.
On the second canonical variate there is a clear discrimination between the apes,
humans and the isolated fossil. The fossil lies almost 3 SDU positively from the
female human centroid which lies about halfway between the apes and fossil (Fig. 2).
The main dimensions responsible for this discrimination are the functional length,
height of the head, mid-shaft height and breadth dimensions. The fossil is thus of
distinct form but has the greatest affinity with the human females. This is also borne
out by the Mahalanobis’s distance (Table 2).
On the third canonical variate, the fossil lies broadly between the chimpanzees on
the one hand and the gorillas, humans and orang-utans on the other (Fig. 3). More
specifically, the fossil lies closest to chimpanzee males, gorilla females and human
males. The main dimensions responsible for this discrimination are the height and
breadth of the base, height and breadth of the head and mid-shaft breadth.
On the first canonical variate, the extant species plot in an almost identical manner
to that found above. Along the first variate, the fossil lies a little more negatively
than SKX 5017 (Fig. 4). The orang-utans are clearly separated from the African
apes, humans and the fossil.
On the second canonical variate the fossil lies in essentially the identical position as
that of SKX 5017. The fossil is thus distinct but has the greatest affinity with the human
females (Fig. 4). This is also borne out by the Mahalanobis’s distances contained in
Table 2. However, the Mahalonobis’s distance from SK 1813 to the human females
(as well as the other groups) is more than twice as great as that from SKX 5017.
On the third canonical variate, the fossil lies negatively to all the apes and humans,
approximately 2 SDU from the chimpanzees which lie broadly between the gorillas
and humans on the one hand and the fossil on the other (Fig. 5). SK 1813 occupies a
distinctly unique position on this axis whereas SKX 5017 falls within the overall
spreads of the extant species (Fig. 3). Dimensions contributing to the discrimination
on all three variates are the same as for SKX 5017.
The extant species
On visual comparison of the metatarsal bones of the different Hominoidea, they
appear surprisingly similar apart from the obvious difference of size. They have in
common the broad function of locomotion that involves weight-bearing to a greater
or lesser extent. However, on closer inspection, large-scale differences are obvious.
Firstly, there is variation between the groups, and secondly, variation within each
group. The former reflects largely a functional affinity, being in the broadest sense
one of bipedalism, terrestrial quadrupedalism or arboreal climbing and suspension.
The latter is predominantly as a result of sexual dimorphism that differs between the
species reflecting mating systems and social behaviors (Larsen, 2003). With this in
mind, the exact nature of variation, particularly between the groups may be
As most of the discrimination lies on the first canonical variate, it is possible, to a
large extent, to give a biologically coherent explanation for most of the loaded
coefficients on this axis. However, this is not necessarily true for the subsequent
variates containing successively smaller proportions of discrimination. It is therefore
important to consider the plots of the first against the later variates. Even a brief
scrutiny of plots of the first against later variates reveals that important biological
discrimination is held jointly between variates. Thus the loaded coefficients,
particularly in those subsequent to the first variate, are difficult to interpret
biologically and therefore suggested explanations for these are to some extent
Variate one contains largely size-related discrimination. However, the orang-utans
separate (lying negatively) to such an extent on this axis to suggest some non-sizerelated
variation, discriminating between the orang-utans on the one hand, and the
African apes and humans on the other. On variate two, this discrimination shifts, in
that the humans lie more positively to the apes. Considering the variates together,
two lines of discrimination emerge (Figs. 2 and 4). One line on which the apes lie and
another, the humans, together suggesting a discrimination of locomotor function.
Within the apes, there is a suggestion of increased terrestiality on an oblique line
from the orang-utans lying most negatively, to the gorillas lying most positively. The
humans lie most positively on both variates suggesting a unique discrimination based
on habitual bipedality (Figs. 2 and 4). On variate one, highly weighted coefficients
are associated with the height of the head and breadth of the mid-shaft dimensions.
This is probably attributable to the flatter superior aspect of the ape metatarsal head
and comparatively greater robusticity of the human metatarsal shaft.
On variate two, suggesting largely a form-related discrimination, the functional
length, height of the head, and height and breadth of the shaft contribute most to the
discrimination. These dimensions when taken together contribute largely to the
overall shape of the bone. The particularly heavily weighted maximum length is
obviously very different between the groups, particularly in the orang-utans. The
relative shortness of the bone in this species is the most obvious feature
differentiating the orang-utan foot from that of the African apes and humans. This
is also the functional component that is best adapted to an arboreal lifestyle,
allowing for the comparatively greatest mobility and prehensile capability.
Examining the positions of the different groups with variate three plotted against
variate one, the African apes and humans tend to lie comparatively ‘‘close’’ together
on both axes on a line that clearly separates them from orang-utans. Thus, there is a
clear discrimination based upon geography; this discrimination may be considered to
represent genetic discrimination between Africa and Asia, between the subfamily of
Homininae (gorillas, chimpanzees and humans) and Ponginae (orang-utans) (Figs. 3
Inclusion of the fossil specimens
The first metatarsal in the Hominoidea represents an essential functional
component of the forefoot and plays a major role in the transmission of body
weight during locomotion, be it terrestrial quadrupedal, arboreal or bipedal. Morton
(1924, 1926, 1927, 1928, 1935) and Hicks (1954) demonstrated the importance of the
human first metatarsal segment in the maintenance of the medial longitudinal arch
and facilitating the plantigrade foot posture from mid-stance to toe-off. The
Swartkrans hominin metatarsals offer a unique opportunity to learn more about
early hominin locomotor function in the forefoot as these elements in the fossil
record are extremely rare. However, as these specimens are isolated, without any
context to the remainder of the foot, one should enter a caveat in interpreting these
findings. For example, the relationship of the first metatarsal to the other four in
terms of relative robusticity which in turn reflects different locomotor requirements
(Archibald et al., 1972) cannot be determined. The hindtarsus of the OH8 and
Stw573 fossil assemblages, where a number of bones are present, display mixed
functional affinities that are both ape (primitive) and human (derived) (Kidd and
Oxnard, 2004). Harcourt-Smith and Aiello (2004) have suggested that there may
have been greater diversity in human bipedalism in the earlier phases of our
evolutionary past than previously suspected.
On the plots of variate one against variate two, both the SKX 5017 and SK 1813
first metatarsals do not clearly lie on any of the previously identified two lines of
discrimination, being those of variable ape locomotion on the one hand and the
habitually bipedal humans on the other. As the plots of the apes and humans on
their own suggest discrimination in terms of locomotion, the positions of the fossils
suggest a unique morphology and associated function. The isolated fossils do
however lie closest to the human centroids, perhaps creating a third line of
discrimination, being that of bipedalism, being obligate in humans and facultative to
some extent in the extinct hominins (Figs. 2 and 4). This is in agreement with the
descriptive morphology of these specimens by Susman and Brain (1988) and Susman
and de Ruiter (2004), suggesting that these hominins were bipedal, but not to the
extent or exact manner of modern humans. Examination of the first metatarsal
group means along canonical variates one and two and the Mahalonobis’s distances
confirms the unique nature of the fossil morphology.
On the plots of variate one against variate two, the fossils clearly lie on the line of
describing African apes and humans, discriminating them from the orang-utans
(Figs. 3 and 5). This is concordant with the previously suggested genetic
discrimination between the Homininae, from Africa and the Ponginae from Asia.
However, on variate three, SK 1813 indicates a somewhat greater discrimination
from humans than does SKX 5017. In fact on this axis, the SK 1813 metatarsal lies
furthest from the gorilla males (reflected by the Mahalonobis’s distance in Table 2)
and human females, suggesting that the overall larger Mahalonobis distance values
are as a result of discrimination on this variate. It is also evident that there is a
greater affinity with the chimpanzee morphology on this axis. This brings to mind
some thoughts on the exact nature of the discrimination between the two obviously
similar fossils. The Swartkrans fossils both have similar derived and primitive
features. Susman and de Ruiter (2004) identified the most obvious derived features as
being the distal articular surface extending onto the dorsum of the head, a relatively
well developed dorsoplantar basal diameter associated with a more plantigrade foot
posture from mid-stance to toe-off and a relatively human-like robusticity. Primitive
features on the fossils include a reduced mediolateral dimension of the superior distal
articular surface associated with close-packing of the joint in plantarflexion rather
than dorsiflexion and increased axial torsion of the metatarsal head reflecting an
abducted ape-like position of the hallux during flexion of the metatarsophalangeal
joint. The discrimination found jointly on the axes of variates one and two does not
discriminate between the two fossils and suggests very similar function, that of
bipedalism, though different from that of humans. The shaft of SKX 5017 appears to
be marginally more robust in the sagittal plane than SK 1813. The head of the
former fossil is more ‘‘bulbous’’ and is relatively broader in the mediolateral
dimension in relation to the shaft. This also gives the impression that the shaft of SK
1813 is relatively broader in relation to the distal and proximal portions of the bone.
These subtle differences are of no obvious functional significance and are common
variations within modern humans (Zipfel et al., 2003; Zipfel, 2004). The inclusion of
the fossils, together with the known information regarding functional and structural
correlates of the first metatarsal in the extant species, lends plausibility to the
suggested interpretation of these multivariate analyses. We therefore strongly
support the viewpoint of Susman and de Ruiter (2004) that both primitive and
derived characters should be considered in the study of form and function and not
only functionally relevant characters. Though the dimensions utilized in this
multivariate morphometric study were not designed to capture this, being chosen for
their representation of the broader morphology, it does follow logically that
information of this sort will be captured as a result of intercorrelation between
variables. This does affirm the sometimes subtle relationship between non-metrical
traits and the broader morphology of the bone.
Our grateful thanks go to Professor Francis Thackeray and Stephany Potze of the
Transvaal Museum, Northern Flagship Institution, for allowing us access to the
fossil material. We also wish to acknowledge the comments of two reviewers,
Professor Colin Groves and an anonymous reviewer.
Reproduced with compliments of Dr.Robert Kidd.
Apologies unable to reproduce tables.