Indications of habitat association of Australopithecus robustus in the Bloubank Valley, South Africa

Abstract

Establishing the habitat preferences of early hominin taxa is a necessary, though difficult, requirement for understanding the interaction between environmental change and hominin evolution. The environments typically associated with Australopithecus robustus have been reconstructed as predominantly open grasslands situated within a habitat mosaic that included a more wooded component with a nearby perennial water source. Most studies have concluded that the open grassland component represents the habitat preference of the hominins. In this study we investigate indicators of habitat association of A. robustus that are preserved in the animal paleocommunities represented in a series of fossil cave infills in the Bloubank Valley of South Africa, including Swartkrans, Sterkfontein, Kromdraai, and Coopers. Testing for conditions of isotaphonomy reveals a potential bias relating to depositional matrix and perhaps accumulating agent, though such a bias has not unduly influenced the taxonomic composition the assemblages. Correspondence analysis of census data from modern African nature reserves demonstrates that carnivore predation patterns are indicative of animal communities, which in turn are representative of habitats. As a result, modern census data are used to document patterns of habitat preference of large herbivores, thus allowing assignment of fossil taxa to a series of broadly defined habitat categories. Correspondence analysis of fossil assemblages reveals that the abundance profile of A. robustus is most similar to that of woodland-adapted taxa. In addition, fluctuations in the relative abundance of taxa assigned to the broad habitat categories reveal a significant negative correlation between A. robustus and open grassland-adapted taxa, indicating that the more grassland-adapted taxa there are in a given assemblage, the fewer hominins there tend to be. Thus, it appears that the open grasslands that comprise the majority of the paleoenvironments associated with A. robustus do not necessarily indicate the habitat preference of the hominins. Rather, it would appear that in addition to being dietary generalists, A. robustus were also likely to have been habitat generalists.

Full text

Page 1

Indications of habitat association of Australopithecus robustus
in the Bloubank Valley, South Africa
Darryl J. de Ruitera,*, Matt Sponheimerb, Julia A. Lee-Thorpc
aDepartment of Anthropology, Texas A&M University, College Station, TX 77843-4352, USA
bDepartment of Anthropology, University of Colorado at Boulder, Boulder CO 80309, USA
cDivision of Archaeological, Geographical and Environmental Sciences, University of Bradford, Bradford BD7 1DP, UK
a r t i c l e i n f o
Article history:
Received 7 November 2007
Accepted 6 June 2008
Keywords:
Paleoecology
Animal paleocommunity
Correspondence analysis
Swartkrans
Sterkfontein
Kromdraai
Coopers
Faunal analysis
a b s t r a c t
Establishing the habitat preferences of early hominin taxa is a necessary, though difficult, requirement
for understanding the interaction between environmental change and hominin evolution. The
environments typically associated with Australopithecus robustus have been reconstructed as pre-
dominantly open grasslands situated within a habitat mosaic that included a more wooded component
with a nearby perennial water source. Most studies have concluded that the open grassland component
represents the habitat preference of the hominins. In this study we investigate indicators of habitat
association of A. robustus that are preserved in the animal paleocommunities represented in a series of
fossil cave infills in the Bloubank Valley of South Africa, including Swartkrans, Sterkfontein, Kromdraai,
and Coopers. Testing for conditions of isotaphonomy reveals a potential bias relating to depositional
matrix and perhaps accumulating agent, though such a bias has not unduly influenced the taxonomic
composition the assemblages. Correspondence analysis of census data from modern African nature
reserves demonstrates that carnivore predation patterns are indicative of animal communities, which in
turn are representative of habitats. As a result, modern census data are used to document patterns of
habitat preference of large herbivores, thus allowing assignment of fossil taxa to a series of broadly
defined habitat categories. Correspondence analysis of fossil assemblages reveals that the abundance
profile of A. robustus is most similar to that of woodland-adapted taxa. In addition, fluctuations in the
relative abundance of taxa assigned to the broad habitat categories reveal a significant negative corre-
lation between A. robustus and open grassland-adapted taxa, indicating that the more grassland-adapted
taxa there are in a given assemblage, the fewer hominins there tend to be. Thus, it appears that the open
grasslands that comprise the majority of the paleoenvironments associated with A. robustus do not
necessarily indicate the habitat preference of the hominins. Rather, it would appear that in addition to
being dietary generalists, A. robustus were also likely to have been habitat generalists.
Published by Elsevier Ltd.
Introduction
The dolomitic cave infills of the former Transvaal in South Africa
have long been known as significant hominin fossil repositories.
Apart from Taung in the North West Province and Makapansgat in
the Northern Province, all of the early hominin-bearing caves are
located in or near the Bloubank Valley, Krugersdorp District,
Gauteng Province (approx. 26?000S, 27?450E). Vegetation in the
Bloubank Valley is a type of false grassveld known as the central
variation of the Bankenveld (Acocks,1988: 113). A false grassveld is
a relatively open grassland with summer rains averaging approxi-
mately 750 mm and frosty winters that result in particularly sour,
wiry grasses that become relatively unpalatable in winter. Trees are
mainly restricted to river courses and around the openings of so-
lution cavities and sinkholes. Although the fossil cave infills of the
Bloubank Valley area are currently poorly temporally constrained,
several deposits have revealed large and well-documented faunal
assemblages associated with the hominin taxon Australopithecus
robustus. To date, A. robustus fossils have been recovered from six
discrete localities in the Bloubank Valley area, though only four of
these localities (Sterkfontein, Kromdraai, Swartkrans, Coopers),
comprising eight distinct faunal assemblages, have produced
sufficiently large and/or well-documented samples to be included
in this analysis (Table 1).
In his initial announcement of A. robustus, Broom (1938)
concluded that these hominins inhabited an environment much
like that of the present Bloubank Valley. He went on to suggest that
A. robustus lived ‘‘.among the rocks and on the plains’’ (Broom,
* Corresponding author.
E-mail addresses: deruiter@tamu.edu (D.J. de Ruiter), msponheimer@yahoo.com
(Matt Sponheimer), J.A.Lee-thorp@Bradford.ac.uk (J.A. Lee-Thorp).
Contents lists available at ScienceDirect
Journal of Human Evolution
journal homepage: www.elsevier.com/locate/jhevol
0047-2484/$ – see front matter Published by Elsevier Ltd.
doi:10.1016/j.jhevol.2008.06.003
Journal of Human Evolution 55 (2008) 1015–1030

Page 2

1943: 79), though he later allowed the possibility that the envi-
ronment might have been somewhat wetter and more vegetated in
the past (Broom and Robinson, 1952). Examining the mammalian
faunas associated with the Transvaal hominins, Cooke (1952,1963)
agreed that they indicated an environment analogous to that of the
area today, supporting Broom’s interpretation of the robust aus-
tralopiths as open plains dwellers. Robinson (1963) speculated that
the expansion of open grassland habitats through the Plio-Pleis-
tocene was a significant evolutionary factor propelling many of the
adaptive developments seen in the robust australopiths, in partic-
ular in relation to alterations in dentition and cognitive capacities.
More recent studies have utilized significantly augmented
faunal assemblages fromthe Bloubank Valleyarea to reconstruct an
environment for A. robustus that was predominantly an open to
lightly wooded grassland (Vrba, 1975, 1976, 1980, 1985a,b; Brain,
1981a; Brain et al., 1988; Shipman and Harris, 1988; McKee, 1991;
Denys, 1992; Avery, 1995, 2001; Watson, 2004), perhaps with
a nearby edaphic grassland (Reed, 1997; Reed and Rector, 2006),
though one study has suggested a mesic, closed woodland for
Member 1 of Swartkrans (Benefit and McCrossin, 1990). Although
relatively open grasslands are primarily indicated, several of these
studies have concluded that these grasslands were part of a larger
habitat mosaic that included a woodland component with a nearby
perennial water source (Brain et al., 1988; Avery, 1995; Reed, 1997;
Watson, 2004). Given the probable linkage between environmental
and evolutionary change in the hominin lineage (Robinson, 1963;
Foley, 1987), disentangling which portions of the environmental
mosaic can be associated with A. robustus is an important albeit
difficult endeavor.
Paleoecological analyses of A. robustus localities generally
operate under the reasonable assumption that the relatively open
grassland environments that are typically reconstructed represent
the habitat preference of the hominins. However, the close
geographical and perhaps temporal proximity of the South African
cave infills has caused some to question whether this type of
environment really does represent the habitat preference of the
hominins (Shipman and Harris, 1988; White, 1988; Wood and
Strait, 2004). Nevertheless, the association between A. robustus and
open grassland habitats remains a persistent component of our
current understanding of the paleoecology of this species.
The aim of the present study is to investigate whether any
indicators of habitat association of A. robustus are preserved in the
faunal assemblages of the Bloubank Valley area. Owing to the
potentially significant influence of biasing factors, such as accu-
mulating agent and depositional environment, strict taphonomic
control is of the utmost importance. Therefore, particular attention
is paid to testing for isotaphonomic conditions between the
assemblages. In this study we document fluctuations in the abun-
dance of A. robustus relative to a series of ecologically sensitive taxa
whose habitat preferences are used to model the ecological
composition of A. robustus’ surrounding animal paleocommunity.
Habitat preferences for these fossil taxa are established via
comparison with animal communities from a series of modern
African nature reserves. Reliance on death assemblages to model
once-living animal communities can be problematical, though
studies have demonstrated close correspondence between the two
(Behrensmeyer et al., 1979; Reed, 1997). In this paper we further
investigate the association between modern carnivore assemblages
and animal community composition to test whether animals tend
to die where they live, and thus whether carnivore-derived
assemblages can be used to model animal communities and, in
turn, environments.
Materials and methods
Faunal assemblage data were recorded for the A. robustus-
bearing deposits of Swartkrans Members 1–3 (SKLB, SKHR, SKM2,
SKM3), Kromdraai B (KB), Coopers D (COD), and Sterkfontein
Member 5-Oldowan Infill (ST5OL). Although no hominins have
been recovered from Kromdraai A (KA), for comparative purposes it
is included in this analysis as it has produced a large and well-
documented faunal assemblage. Kromdraai A and B represent
distinct depositional units, probably derived from significantly
different time periods. Based on the fauna from Kromdraai A, a date
of approximately1.5 millionyears of age (Ma) is evident (White and
Harris, 1977; Delson, 1984). Using a single magnetic reversal, and
assuming a faunal age between 1.5–2.0 Ma, Thackeray et al. (2002)
suggest that Kromdraai B is at least 1.9 Ma. The presence of a rela-
tively complete Hexaprotodon protamphibius cranium (Vrba, 1981),
a taxon which disappears in East Africa by approximately 1.9 Ma,
supports such a magnetostratigraphic age. The site of Swartkrans
has produced the largest concentration of specimens attributable to
A. robustus. The geology of the site has been well-documented, and
comprises four separate hominin-bearing faunal assemblages
extracted from three discrete members (Brain, 2004). The earliest
of the Swartkrans deposits is Member 1, which has been divided
into two separate subdeposits. The Lower Bank of Member 1
represents the oldest of the Swartkrans assemblages, biostrati-
graphically dated to approximately 1.7 Ma (de Ruiter, 2003; Brain,
2004). Its companion deposit, the Hanging Remnant, has been
biostratigraphically dated to about 1.6 Ma (White and Harris,1977;
Delson,1984; Vrba,1985a; de Ruiter, 2003; Brain, 2004), a date that
accords well with an ESR estimate of 1.63 Ma (Curnoe et al., 2001).
Although ages as young as 1.0 and 0.7 Ma have been proposed for
Members 2 and 3 of Swartkrans, respectively (Vrba, 1995), these
deposits are more consistent with a date of approximately 1.5 Ma in
terms of biostratigraphy (White and Harris, 1977; Delson, 1984; de
Ruiter, 2003; Brain, 2004). The Oldowan Infill of Sterkfontein
Member 5 has been dated to approximately 1.7–2.0 Ma (Kuman
and Clarke, 2000). The abundant suids and bovids derived from
Coopers D indicate an age estimate of 1.6–1.9 Ma (Berger et al.,
2003), consistent with a recent U-Pb date of approximately 1.62 Ma
(Steininger et al., 2008).
We have arranged the fossil deposits into what we consider to
be the most probable chronological sequence: KB-ST5OL-COD-
SKLB-SKHR-SKM2-SKM3-KA (see Table 1 for abbreviations used in
the text). All of these assemblages were examined by us with the
exception of ST5OL. This latter deposit was analyzed by Pickering
(1999) using data collection techniques consistent with those
employed for the current study. Data collection involved a manual
overlap approach as recommended by Bunn (1982, 1986) to
document both the minimum number of elements (MNE) and the
comprehensive minimum number of individuals (cMNI: Pickering,
1999) in each assemblage. Details of the procedure are presented
in de Ruiter (2004). In short, the technique involves a specimen-
by-specimen comparison of fossils to obtain the most accurate
Table 1
Faunal assemblages examined in this study with probable age estimatesa
SiteMember/deposit Abbreviation
used in text
Age
estimate
Sterkfontein
Kromdraai
Member 5, Oldowan Infill
Kromdraai A
Kromdraai B
Member 1, Lower Bank
Member 1,
Hanging Remnant
Member 2
Member 3
Coopers D
ST5OL
KA
KB
SKLB
SKHR
1.7–2.0
1.5
1.9
1.7
1.6
Swartkrans
SKM2
SKM3
COD
1.5
1.5
1.6–1.9 Coopers
aFaunal assemblage data for ST5OL from Pickering (1999). See text for derivation
of age estimates.
D.J. de Ruiter et al. / Journal of Human Evolution 55 (2008) 1015–1030 1016

Page 3

approximation of the number of skeletal elements and individuals
as possible. This is accomplished by laying out all specimens of
a particular skeletal element and/or taxonomic group on a large
table (or floor), and comparing them individually to determine
whether they are likely to have come from a single element or
animal. In cases where large numbers of specimens are involved
(e.g., bovid dentitions, bovid postcrania) samples were subdivided
into nonoverlapping dental wear stage categories or body size
groupingsbefore proceeding
comparisons.
Although early collection procedures at Kromdraai were highly
selective (Broom,1951), later excavations (Brain,1981b; Vrba,1981;
Berger et al.,1994) adopted a complete fossil recovery strategy. The
same biased collection procedure is true of early work in
the Hanging Remnant of Member 1 and Member 2 of Swartkrans in
the late 1940s (Broom, 1951), though complete recovery practices
were exercised in subsequent excavations under the direction of
C.K. Brain (Brain, 1981b, 2004). In particular, Brain’s in situ exca-
vations of uncalcified and decalcified sediments in the Lower Bank
of Member 1, Member 2, and Member 3 of Swartkrans were so
precise that they allowed a workable GIS to be constructed (Nigro
et al., 2003). Excavations at Coopers D (Berger et al., 2003) and in
the Oldowan Infill of Member 5 at Sterkfontein (Kuman and Clarke,
2000) have employed total recovery excavation procedures since
these respective operations were inaugurated. Such consistent
fossil collection procedures minimize the potential influence of
different sampling strategies on assemblage composition.
with specimen-by-specimen
Testing for taphonomic bias
The aim of this study is to investigate the habitat association of
A. robustus in relation to its surrounding animal paleocommunity.
This amounts to examining a biological signal that we assume is
reflected in estimates of taxonomic abundance. However, biological
responses to differing environmental conditions, as mirrored in
taxonomic abundance data, can be masked by taphonomic factors
(Badgley, 1986; Bobe et al., 2002). Such factors must be controlled
for in any comparative analysis of fossil assemblages if meaningful
interpretations are to be drawn. The potential biases introduced as
a result of bone accumulating agent and depositional matrix have
been well-documented in the South African cave infills (Brain,
1981b). Our approach is to first determine whether there is evi-
dence of taphonomic bias(es), and then assess the potential impact
of any recognized taphonomic bias(es) on assemblage composition.
A variety of bone-collecting agents have been implicated in the
accumulation of the South African cave infills (Brain, 1981b;
Pickering, 1999; de Ruiter and Berger, 2000; de Ruiter, 2004;
Newman, 2004; Pickering et al., 2004, 2007). Carnivore prey
acquisition tends to be highly selective (Pienaar, 1969; Wilson,
1981; de Ruiter and Berger, 2001), resulting in potentially biased
bone accumulations. However, it is likely that the South African
assemblages are the result of the combined operation of multiple
agents intermittently utilizing the caves over long timespans
(Brain, 1980). Bone surface modifications provide direct evidence
for the involvement of bone-accumulating agents, including hom-
inins, carnivores, and rodents (Brain, 1981b; Pickering, 2002;
Newman, 2004). The presence of culturally modified materials,
such as stone and bone tools, can likewise serve as an indication of
a hominin accumulating agent. Coprolites can be used to implicate
specific donors, typically hyenas (Pickering, 2002). Additionally, the
ratio of carnivores toungulates has been cited as a reliable indicator
of carnivore involvement in an accumulation, specifically that of
brown hyenas (Brain, 1981b; Cruz-Uribe, 1991; Pickering, 2002).
The relative destruction of bones by carnivores is taxonomically
mediated, with accumulators such as hyenas doing moredamage to
carcasses thancollectors such as leopards(Brain, 1981b;
Blumenschine and Marean, 1993). Relative levels of fragmentation
are also affected by differences in depositional matrix, which will in
turn impact the taxonomic identifiability of fossil materials. In the
South African cave infills fossils are derived from three principal
depositional matrices. Hard breccia deposits are heavily calcified
sediments cemented together into a solid mass, requiring labor
intensive manual or chemical preparation (KA, KB, SKHR).
Uncalcified sediments are those which were never cemented by
calcium carbonate (SKLB). Decalcified sediments are breccia
deposits where the cementing calcium carbonate has been leached
out by the activities of tree roots, leaving loose soil and fossils
behind (SKM2, SKM3, COD, ST5OL). Marean (1991) recommended
examining the completeness of ungulate compact bones (carpals,
tarsals, lateral malleolus of the fibula) to determine the relative
severity of postdepositional fragmentation in faunal assemblages,
creating what he termed the completeness index. The procedure
involves assigning a completeness value (percentage complete) to
ungulate compact bones lacking evidence of bone surface modifi-
cation, summing these completeness values, and dividing by the
NISP of compact bones. Marean (1991) suggested that complete-
ness values be computed per bone and per body size class.
However, in several of the South African fossil deposits carpals and
tarsals are not common, and division into skeletal element and
body size groupings oftenproduced particularly small sample sizes.
In order to facilitate comparison of assemblages using a maximum
of available fossil material, the completeness index was computed
amalgamating all ungulate body sizes and compact bones.
Given the potential for taphonomic biases arising via accumu-
lating agent and depositional matrix, it is of particular importance
that we investigate the impact of any taphonomic overprint that
might be evident. Skeletal part representation has long been
considered to be a useful indicator of potential taphonomic
overprinting in faunal assemblages (Voorhies,1969; Behrensmeyer,
1991; Bobe and Eck, 2001; Bobe et al., 2002), as changes in the
proportional representation of skeletal elements between deposits
would likely signal the existence of a taphonomic bias. We there-
fore compare the relative abundance of a selection of skeletal
elements that span a range of transportability, destructibility, and
carnivore attraction in order to test for isotaphonomic conditions
across assemblages. The particular skeletal elements examined
include cranial, dental, and postcranial remains, incorporating both
fore- and hind limb elements. They represent a variety of different
shapes and structural densities, and thus include a range of
potential taphonomic influences. In order to evaluate the statistical
significance of differences in the relative abundance of skeletal
elements, 95% confidence intervals were constructed based on the
formula:
p?1.96 * SQRT[(p*q)/(n–1)],
where p is the proportion of a given skeletal element, q is equal
to 1?p, and n represents the total sample size (Buzas, 1990).
A final test for isotaphonomic conditions utilizes chord distance
(CRD), a measure of faunal dissimilarity, to compare the tapho-
nomic and taxonomic composition of the assemblages (Ludwig and
Reynolds, 1988; Bobe et al., 2002). Chord distance measures
emphasize relative proportions
abundances (Ludwig and Reynolds, 1988), making them particu-
larly useful for comparing assemblages comprised of varying
sample sizes. Chord distance values are computed between
assemblage j and assemblagek by the formula:
CRDjk¼SQRT[2(1–ccosjk)]
with ccosjk¼SS(Xij*Xik)/SQRT[SSX2ijSSX2ik]
where Xij represents the abundance of the ithtaxon or skeletal
element in the jthassemblage, Xik represents the abundance of the
ithtaxon or skeletal element in the kthassemblage, and S is the total
number of taxa or skeletal elements common to the two assem-
blages. Chord distance values range from zero for assemblages with
ofcategories over absolute
D.J. de Ruiter et al. / Journal of Human Evolution 55 (2008) 1015–10301017

Page 4

identical composition, to the square root of 2 (z1.414) for
assemblages with nothing in common. These values will allow us to
explore whether there is any link between taphonomic conditions
and taxonomic abundance across assemblages, or if the two factors
vary independently.
Taxonomic abundance and faunal change
After controlling for taphonomic factors, taxonomic abundance
data are used to test for biological responses of animal paleo-
communities to changes in environmental conditions over time in
the Bloubank Valley. Several studies have documented the utility of
taxonomic abundance data for signaling environmental or climatic
changes, as responses of animal communities to external alter-
ations are more likely to be reflected in fluctuations in relative
abundance than in speciation or extinction events, in particular at
local scales (Bobe and Eck, 2001; Bobe et al., 2002; Alemseged,
2003). While fluctuations in taxonomic abundance in faunal
assemblages can elucidate patterns of change in paleoenviron-
ments, this is not to say that taxonomic abundance in a given fossil
assemblage reproduces the actual composition of the original
paleocommunity. Nonetheless, differences in proportional repre-
sentation of mammalian taxa can be used to investigate changes in
paleocommunity composition over time (Klein, 1980; Bobe and
Behrensmeyer, 2004).
Sample size varies significantly across the assemblages exam-
ined in this study, potentially confounding analyses based on
taxonomic abundance (Magurran, 1988; Bobe and Eck, 2001).
Rarefaction analysis is a technique for estimating the number of
species expected in a given assemblage if all assemblages were of
equal size (Magurran, 1988), thereby allowing us to detect the
presence of sample size biasing. We test for the relative influence of
sample size on animal paleocommunity composition by doc-
umenting species richness and species evenness in each of the
assemblages. Species richness is a measure of the numberof species
in an assemblage relative to sample size. For this study we use the
Fisher’s log series (a) as our measure of species richness; Fisher’s
log series (a) allows a goodness-of-fit test (c2) to determine if there
is a significant difference between observed and expected species
distributions. Species evenness is a measure of the relative
dominance of the most abundant species in an assemblage, since
assemblages characterized by one or a few very common animals
are differently distributed than assemblages where many species
exist in similar abundances. We use the Berger-Parker index as our
estimate of species evenness to estimate the impact of variations in
abundance within assemblages; Berger-Parker values are typically
presented as reciprocal values (1/d), such that greater values
indicate less dominance of the most common species in an
assemblage.
For this study we focus on the predominantly herbivorous taxa
from the Bloubank Valley sites, including representatives of the
Cercopithecidae, Equidae, Suidae, and Bovidae, in relation to the
Hominidae (Table 2). Most of these herbivorous taxa are dependent
on the relative distribution of the vegetation that forms the basis of
their diet (Jarman, 1974; Skinner and Smithers, 1990; Estes, 1991).
The habitat dependence of these various taxa means that they tend
to be particularly responsive to fluctuations in vegetational distri-
bution, which in turn are influenced by such climatic factors as
temperature and moisture levels. As such, they provide a useful
proxy for prevailing environmental conditions, in particular
relating to changes in these conditions over time. Because of their
higher trophic position, carnivores tend to have wide habitat
tolerances (Skinner and Smithers,1990; Estes,1991). Owing to this,
they are unlikely to aid in resolving habitat structure in the fossil
assemblages; therefore, they are excluded from this analysis.
Because of the likelihood of differing taphonomic histories, smaller
mammals, such as the Hyracoidea, Rodentia, and Lagomorpha, are
not included. Very rare animals (i.e., those with fewer than eight
individuals in the eight combined assemblages) are excluded owing
to their rarity: Elephantidae, Giraffidae, Hippopotamidae, Orycter-
opodidae, and Manidae. A total of 24,211 specimens were identified
to skeletal part and taxonomic family, representing a minimum of
1,266 individuals animals included in the subset of materials
analyzed in this study. These combined individuals represent
approximately 74% of the 1,719 macromammals recorded in the
respective assemblages (Table 2), thus encompassing the majority
of available faunal information.
Animal census data from a series of 33 African nature reserves
are utilized to document the habitat preferences of modern
herbivores (Table 3). Species are grouped into genera for the
primates, equids, and suids, and into tribes for the bovids. Census
data are taken from original published reports wherever possible,
and represent as accurate a compendium of animal abundance
information as is possible for the reserves included. We conducted
a correspondence analysis to examine the association between taxa
and habitats in the modern nature reserves to document the re-
lationship between taxonomic abundance and habitat preference.
Correspondence analysis is a visual ordination technique designed
to graphically display relationships between variables. Utilizing
data arranged in bivariate contingency tables, correspondence
analysis visually displays clusters of points representing similar,
closely related variables, while dissimilar variables appear farther
apart fromeach other (Greenacreand Vrba,1984; Greenacre, 2007).
For instance, when applied to animal communities or faunal
assemblages, taxa are grouped with the localities in which they are
well-represented, while at the same time each locality is grouped
with the taxa which are prominent in it. The resulting clusters of
similar variables are interpreted by examining their spread across
each axis in search of the underlying features that unite them.
Employinga taxonomic uniformitarian
relatives of modern taxa are assumed to have similar habitat
preferences as their modern counterparts as determined via
correspondence analysis. Isotopic (Sponheimer, 1999; Sponheimer
et al., 1999, 2003; Luyt, 2001; Harris and Cerling, 2002), dental
microwear (El-Zaatari et al., 2005), and ecological functional mor-
phological (Reed,1997; Sponheimer et al.,1999) evidence is used to
test this assumption. For instance, specimens of Metridiochoerus
argument, fossil
Table 2
Comprehensive minimum numbers of individuals of mammalian families recovered
from the breccia cave infills examined in this study. Taxonomic families in bold are
included in this analysis (see text for details)
Taxonomic
family
Fossil deposit
KB ST5OL CODSKLBSKHR SKM2SKM3 KATotal
Bovidae
Cercopithecidae
Procaviidae
Hominidae
Equidae
Felidae
Canidae
Hyaenidae
Leporidae
Suidae
Viverridae
Hystricidae
Mustelidae
Giraffidae
Pedetidae
Hippopotamidae
Elephantidae
Orycteropodidae
Manidae
Total
14
39
1
6
1
4
4
4
2
1
2
0
1
0
0
1
0
0
0
80
33
8
5
2
3
2
2
1
0
2
4
0
0
0
0
0
0
0
0
62
88
20
6
2
2
13
7
7
12
10
7
4
1
2
2
0
0
0
0
183
70
18
29
9
4
8
5
4
10
2
1
2
2
0
2
1
2
1
1
171
182
70
31
58
8
18
7
17
0
7
1
3
0
1
0
1
1
0
0
405
150
27
19
8
17
8
14
12
7
8
6
2
2
2
1
1
0
1
0
285
139
30
24
6
9
9
15
9
9
2
11
3
2
1
1
1
1
1
1
274
149
28
16
0
31
8
12
6
3
3
2
1
0
0
0
0
0
0
0
259
825
240
131
91
75
70
66
60
43
35
34
15
8
6
6
5
4
3
2
1719
D.J. de Ruiter et al. / Journal of Human Evolution 55 (2008) 1015–10301018

Page 5

Table 3
Census data for modern African game parks and modern carnivore kill data
Country Game Park
Papio Chlorocebus Equus Phacochoerus Potamochoerus Alcelaphini Antilopini Aepycerotini Tragelaphini Reduncini Bovini Hippotragini Neotragini Cephalophini Source
Benin Pendjari4,0005000 5,0000 4,22400 100 13,2815,8152,3250 4,633 Milligan et al., 1982;
Sayer and
Green, 1984
Kabaija, 2005
Kabaija, 2005
Kabaija, 2005
Kabaija, 2005
Green, 1979
Milligan et al., 1982
Milligan et al., 1982
Esser and Van
Lavieren, 1979
Van Lavieren and
Bosch, 1977; Van
Lavieren and
Esser, 1979
Milligan et al., 1982
Botswana
Botswana
Botswana
Botswana
Burkina Faso
Burkina Faso
Burkina Faso
Cameroon
Chobe
Makgadikgadi
Kgalagadi
Moremi
Arli
Deux Bale
Po
Waza
331
0
0
2,205
1,890
0
0
0
0
0
0
0
100
0
0
0
2,121
15,640
0
1,674
0
0
0
0
170
0
0
1,542
2,960
74
187
200
0
0
0
0
0
0
0
0
854
3,155
8,102
4,343
1,916
453
543
605
0
4,668
4,814
0
0
0
0
10
868
296
0
18,615
0
0
0
0
320
592
15,487
1,111
800
198
108
0
539
0
0
12,332
8,500
227
290
13,277
3,773
0
0
40,160 232
650
40
248
0
1,185
0
0
1,135
477
6,382
43
0
0
0
0
0
0
1,678
0
2,240
651
482
0
1,920
1,200
777
223
CameroonBouba Ndjida1,500 2500 2,19606,98800 1,100 7,0462,000 4,3560 5,400
Central African
Republic
Democratic
Republic Congo
Ethiopia
Gabon
Saint-Floris000 5003,022000 3,224 1,81350400
Virunga000 60335 1,19900 53 5,7977,402001 Bourlie `re, 1963
Omo
SW Gabon
0
0
0
70
983
0
8
0
0
5020
2,093
0
646
0
0
0
950
1,820
0
420
404
3,570
0
0
0
0
3
3,710
Baba et al., 1982
Prins and
Reitsma, 1989
Geerling and
Bokdam, 1973
Kutilek, 1974
Darling, 1960
Bourlie `re, 1963;
Foster and Kearney,
1966; Foster and
Coe, 1968
Joubert and
Mostert, 1975
Joubert and
Mostert, 1975;
Greenacre and
Vrba, 1984
Milligan et al., 1982
Ayeni, 1980;
Milligan et al., 1982
Afolayan and Ajayi,
1980; Milligan
et al., 1982
Mentis, 1970;
Taylor, 1998
Pienaar et al., 1966;
Pienaar, 1969;
Schaller, 1972;
Greenacre and
Vrba, 1984
Greenacre and
Vrba, 1984
Hirst, 1969
Lamprey, 1962
Ivory CoastComoe 3,000 2,00000 15008,00000 1,000 7,510450 1,0000 13,000
Kenya
Kenya
Kenya
Lake Nakuru
Masai Mara
Nairobi
50
0
165
25
0
22
0
12,000
1,929
20
1,000
230
0
0
0
0
20,000
3,977
250
12,500
690
260
5,000
655
22
650
91
1,135
750
143
27
4,000
0
0
0
0
12
200
4
3
50
2
Namibia Tsumeb1,500 501061,98403,738 293 2716,313000 400280
NamibiaEtosha0014,0001,50004,600 12,00002,50000 296 500250
Niger
Nigeria
W
Kainji
0
0
0
0
0
0
2,130
1,200
0
0
1,440
2,500
0
0
0
0
240
950
6,120
4,800
4,140
275
2,850
2,200
0
0
0
1,525
Nigeria Yankari171 140 1130 74006 175376703
South AfricaiMfolozi4,202 1401,4265,52104,30704,894 16,4473,9273,1950 937715
South Africa Kruger10,000 5,000 14,4005,000 50013,7500 153,0008,3955,33510,614 1,5875,5001,300
South AfricaMkuzi 500 500001,39709,394 533 690064
South Africa
Tanzania
Timbavati
Tarangire
500
0
0
0
980
2,500
287
400
0
20
3,044
1,600
0
300
8,569
3,100
821
530
302
270
0
1,400
0
10
0
280
0
0
(continued on next page)

Page 6

Table 3 (continued)
CountryGame Park
Papio Chlorocebus Equus Phacochoerus Potamochoerus Alcelaphini Antilopini Aepycerotini Tragelaphini Reduncini Bovini Hippotragini Neotragini Cephalophini Source
Tanzania Lake Manyara 5000 255950 6750 1505037 2,097000 Mwayalosi, 1977;
Prins and Douglas-
Hamilton, 1990
Kruuk, 1972;
Schaller, 1972;
Estes and
Small, 1981
Kruuk, 1972;
Schaller, 1972;
Greenacre and
Vrba, 1984; Sinclair
and Arcese, 1995
Sheppe and
Osborne, 1971
Dasmann and
Mossman, 1962;
Greenacre and
Vrba, 1984
Tanzania Ngorongoro400 2004,50000 16,6355,2350 214 120661000
TanzaniaSerengeti 8,700 5,000 280,000 17,0000 455,000190,00065,000 9,5005,500 50,000 5,00000
Zambia Kafue Flats00 1,2000 503,00000 213 37,620250 25000
ZimbabweHwange1,0000 1,9004000 2,6300 8,000 5,450 1,250 13,000 2,5003,000 2,000
Modern bone-accumulating agent data
Nossobporcupine den
Makgadikgadi brown hyena den
0
0
0
0
0
12
0
0
0
0
14
7
40
5
0
0
0
0
0
0
0
0
0
0
5
2
2
1
Brain, 1981b
Lacruz and
Maude, 2005
Skinner et al., 1986
Pienaar, 1969
Pienaar, 1969
Pienaar, 1969
Le Roux and
Skinner, 1989
Zuberbuhler and
Jenny, 2002
Kruuk, 1972
Schaller, 1972
Kruuk, 1972
Kruger
Kruger
Kruger
Kruger
Londolozi
(Kruger)
Taı ¨ Forest
spotted hyena den 0
spotted hyena kills 0
brown hyena kills
leopard kills
leopard kills
0
0
0
0
7
27
1
8
8
0
4
1
1
12
5
1
0
0
11
0
18
21
7
9
0
0
0
0
0
0
111
110
49
789
77
24
24
80
40
9
1
25
46
39
1
28
2
2
3
0
0
0
4
0
0
1
1
0
22
2
0
0
1
14
17
7
11
2
leopard kills0900200000000 82
Ngorongoro
Serengeti
Serengeti
spotted hyena kills 0
leopard kills
hyaena kills
0
0
0
54
1
68
0
0
4
0
0
0
206
17
169
21
114
157
0
0
1
0
2
2
0
20
1
1
0
3
0
0
0
0
0
0
0
0
0
1
0

Page 7

exhibit isotope values indicating significant C4resources in its diet
(Harris and Cerling, 2002), similar to modern Phacochoerus. As
a result, Metridiochoerus is assigned to a grassland category
(ecological assignments detailed below). In cases where modern
census data are unavailable, we again assume a taxonomic unifor-
mitarian argument. For instance, gelada baboons are unknown in
any of the modern nature reserves included in this study. However,
the dietary preference of the extinct taxon Theropithecus oswaldi
indicates a predominantly grassland-based diet, similar to modern
Theropithecus (Lee-Thorp et al., 1989). In this case, Theropithecus,
like its living descendants, is assigned to the grassland category.
Modern carnivore predation patterns are examined in order to
test the association between animal communities and death
assemblages. Data on bone accumulations of modern carnivoresare
limited, and most published reports are derived from areas
exhibiting considerable human disturbance. We therefore rely on
two carnivore lairs located in areas evincing minimal human
disturbance to investigate whether death assemblages mirror the
habitats from which they are recovered (Table 3). We also examine
the composition of a modern porcupine den, as these rodents are
known to be proficient bone accumulators (Brain, 1981b). In
addition, modern leopard and hyena kill data from the Serengeti,
Ngorongoro, Kruger, and Taı ¨ Forest national parks are examined to
test if carnivore predation patterns are reflective of the animal
communities from which they are drawn. Although these modern
carnivore predation patterns do not represent discrete faunal
assemblages, the resultant skeletal remains can nonetheless pro-
vide us with valuable ecological information (e.g., Behrensmeyer
et al., 1979).
The habitat preferences of the modern herbivores are used to
assign the select fossil taxa (minus the hominins) from the
Bloubank Valley sites to a series of broadly defined habitat cate-
gories in order to investigate the ecological composition of the
faunal assemblages. Fluctuations in the relative abundance of A.
robustus are investigated to document the correlation between
numbers of hominins and numbers of animals assigned to habitat
categories. The intent is not to search for any temporal patterning
across assemblages, but rather to investigate whether there is
a consistent relationship between A. robustus and any particular
habitat category.
Results
Taphonomic conditions
In order to search for evidence relating to particular accumu-
lating agents, details of a series of taphonomic indicators are
presented inTable 4. The total NISP presented inTable 4 relates only
to those specimens that are identifiable to skeletal part and
taxonomic family. Hominin produced damage is rare; the large
number of hominin modified materials in SKM3 includes 270 bone
fragments bearing evidence of burning (Brain and Sillen,1988). The
stone tools found in all of the deposits are indicative of hominin
activity, though it is not possible to determine whether these
materials were deposited within the cave itself or in the catchment
area immediately surrounding the cave (Butzer, 1984; Pickering,
1999). Carnivore damage is evident in all deposits, though such
indications are infrequent (typically less than 5% of the respective
assemblages). Rodent gnawed bones, although rare, also reveal
some level of contribution from these bone collectors. Coprolites
are present in several of the assemblages, indicating that carnivores
(probably hyenas) were active in the immediate vicinity of the
caves. The carnivore to ungulate ratio also indicates that carnivores
were involved in the accumulations, pointing to hyaenas as
accumulators of at least some portion of the material.
When completeness index values are computed as a measure of
fragmentation (Table 5), there is no appreciable difference between
decalcified and uncalcified sediments in terms of bone destruction.
As a result, for this study they are considered together as a unit. A
t-test (t ¼3.25, p¼0.02, df¼5) reveals a significant difference in
the levels of fragmentation between hard breccia and uncalcified/
decalcified breccia. It appears that hard breccia-derived fossils tend
to be less fragmented than uncalcified/decalcified breccia-derived
fossils. These differing levels of fragmentation are likely to
influence the relative identifiability of fossil remains.
Skeletal element abundance data are presented in Table 6, and
Fig. 1 illustrates the relative abundance of these skeletal elements
across the faunal assemblages. Because of the likelihood of differing
taphonomic histories for very small animals, only data from body
size class II, III, and IV individuals (based on Brain, 1981b) are
included. Although not strongly indicated, the pattern thatemerges
from the skeletal part distributions confirms some level of bias
relating to depositional matrix. In broad terms, hard breccia-
derived assemblages, in particular SKHR, tend to have too many
craniodental remains and too few postcranial remains relative to
the uncalcified/decalcified assemblages. Since taxonomic identifi-
cation depends on fossil preservation and extraction, in particular
of the more diagnostic craniodental elements, this difference
represents a potentially important taphonomic bias. Isolated teeth
show a relatively even distribution across the assemblages (c2¼
5.99, p¼0.54), while all other elements display relatively uneven
distributions. Since isolated teeth account for the bulk of the faunal
material in each assemblage, thus forming the basis of most
taxonomic identifications, their relatively even distribution might
mitigate the potential taphonomic bias relating to depositional
matrix. Nonetheless, it is apparent that depositional matrix
Table 4
Taphonomic indicators diagnostic of bone accumulating agentsa
Fossil deposit
KB ST5OLCODSKLB SKHR SKM2 SKM3 KA
Stone tools
Hominin-modified bone
Carnivore-modified bone
Rodent-gnawed bone
Coprolites
Carnivore:carnivoreþungulate ratio
Total NISP (identifiable specimens)
4
0
14 (0.28)
1 (0.02)
4 (0.08)
0.49
4985
483
1 (0.03)
174 (4.66)
6 (0.16)
0
0.19
3731
50
0
121 (1.59)
13 (0.17)
2 (0.03)
0.26
7574
62
13 (0.22)
131 (2.17)
22 (0.36)
59 (0.98)
0.21
6040
1
0
45 (0.47)
6 (0.06)
0
0.18
9583
132
31 (0.37)
72 (0.86)
24 (0.29)
8 (0.10)
0.20
8416
73
375 (5.96)
197 (3.13)
41 (0.65)
0
0.23
6293
45
0
36 (1.95)
5 (0.27)
6 (0.32)
0.13
1847
aThe stone tool category excludes debitage and naturally occurring stone. Hominin modified bone includes cut- and hammerstone percussion-marked bones and bones
with probable traces of burning, regardless of the potential author(s) of these traces. Numbers in parentheses indicate percentage of total NISP of taxonomically identifiable
fossils recovered from each deposit. Carnivore-modified bone includes bones with tooth markings and with evidence of gastric etching. Carnivore coprolites are considered to
be highly diagnostic of hyena activity (Pickering, 2002). A carnivore:carnivoreþungulate ratio of 0.20 or greater is generally considered to be indicative of carnivore, probably
hyena, activity (Cruz-Uribe, 1991; Pickering, 2002), though lower values do not necessarily exclude hyaenas as accumulating agents (Lacruz and Maude, 2005). Data on
hominin modified bones for ST5OL from Pickering (1999) and for SKLB, SKM2 and SKM3 from Pickering et al. (2007).
D.J. de Ruiter et al. / Journal of Human Evolution 55 (2008) 1015–10301021

Page 8

presents a potentially significant bias; therefore, it is necessary to
test whether there is any linkage between taphonomic conditions
and taxonomic identification.
In order totest for the impact of taphonomic bias(es) introduced
as a result of accumulating agent and/or depositional matrix, chord
distances were computed for pairs of assemblages using taxonomic
abundance (Table 7) and skeletal element abundance data (Table 6)
from the Bloubank Valley assemblages; chord distance values are
presented in the bottom rows of these respective tables. In
addition, because of the current uncertainty over age estimates in
the South African cave infills, we have produced a matrix of chord
distances (Table 8) that can be consulted should there be a signifi-
cant change in the age assessment of any particular site. However,
we would note that current efforts at dating the robust australopith
sites using radiogenic isotopes do not contradict the arrangement
of the deposits as listed in this study (e.g., Steininger et al., 2008).
Because isolated teeth represent such a preponderance of skeletal
elements, the possibility exists that they are masking more subtle
taphonomic signals in the skeletal part data. We therefore compute
chord distance values between pairs of assemblages both with and
without isolated teeth included as a category. Removing the
isolated teeth from consideration results in greater chord distance
differences between assemblages, particularly for SKHR and KA
(Fig. 2). Both of these assemblages were recovered exclusively from
a hard breccia matrix, likely influencing their respective tapho-
nomic compositions. However, when we correlate the taphonomic
chord distances with isolated teeth against the chord distances
without isolated teeth, we see that there is a strong, significant
correlation between them (Spearman’s rs¼0.96, p¼0.00). In other
words, although the removal of isolated teeth results in greater
apparent taphonomic dissimilarity, the relative ranking of the
assemblages remains effectively unchanged, indicating a broadly
commensurate level of change across all the assemblages.
Taxonomic chord distances demonstrate that faunal turnover
between assemblages was considerable, with a peak reached
between KB and ST5OL, two assemblages with little in common in
terms of relative faunal representation (Fig. 2). In spite of this
marked taxonomic difference, there is little difference in tapho-
nomic conditions between KB and ST5OL, in particular when
isolated teeth are included in the chord distance computation. At
the same time, the greatest taphonomic chord distance is seen be-
tween SKLB and SKHR, though the taxonomic chord distance be-
tween these assemblages is relatively low; both are derived from
Member 1 of Swartkrans, representing uncalcified and hard breccia
deposits, respectively. It would thus appear that the difference in
taphonomic chord distance values can be attributed to depositional
matrix. There is no correlation between taxonomic chord distances
and taphonomic chord distances either with (Spearman’s rs¼0.29,
p¼0.54) or without (Spearman’s rs¼0.11, p¼0.82) isolated teeth
included,confirmingthattaphonomicandtaxonomicchord distance
values are not linked. In other words, taxonomic and taphonomic
chord distance values vary independently. These data demonstrate
that although a taphonomic bias likely exists relating to depositional
matrix and perhaps bone accumulating agent, such biasing has not
consistently influenced the taxonomic composition of the assem-
blages in any particular direction. In the absence of a consistent
taphonomic bias, we conclude that taxonomic abundance data from
the assemblages represent reasonable reflections of original animal
paleocommunity composition and, therefore, fluctuations in taxo-
nomic abundance across the assemblages can be interpreted as an-
imal community responses to changing environmental conditions.
Taxonomic and ecological composition
Sample size can have an impact on estimates of taxonomic
abundance, and in the case of the assemblages examined in this
study there is a significant correlation between cMNI and the
number of species counted in an assemblage (Spearman’s rs¼0.78,
p¼0.02; data from Table 9). Rarefaction analysis generates a series
of curves for the respective assemblages if they are artificially
reduced (rarefied) in size to that of the smallest assemblage. In this
case, all of the assemblages are rarefied to the size of ST5OL at 48
individuals, resulting in plots comprised of predicted numbers of
species relative to the numbers of individuals counted in each
assemblage (Fig. 3). The largest assemblage, SKHR, plots close tothe
majority with an estimated 14 species when it is reduced to 48
individuals. The two smallest assemblages, ST5OL and KB, appear to
have fewer recognized species than the majority of the assem-
blages, while SKM2, the second largest assemblage, includes more
species than might be expected. Although these latter sites do
Table 6
Abundance of a selection of skeletal elements (MNE) for body size II, III, and IV individuals in each of the faunal assemblages (body size categories based on Brain, 1981b)a
Fossil deposit
KB ST5OL CODSKLBSKHRSKM2 SKM3KA Total
Maxilla
Mandible
Isolated teeth
Humerus
Radius
Metacarpal
Femur
Tibia
Metatarsal
Astragalus
Chord distance (teeth included)
Chord distance (teeth excluded)
70 17
50
10
28
172
261
657
37
26
27
36
25
27
33
14
31
23
92
65
137
524
30
16
13
33
12
17
15
308
633
3415
315
217
228
179
173
297
245
23
176
12
17
16
10
10
21
10
11
265
25
21
23
12
17
20
24
425
39
34
47
34
31
54
44
389
41
23
23
15
20
42
45
279
44
22
24
12
15
17
24
700
87
58
55
27
43
99
50
0.114
0.470
0.105
0.332
0.087
0.284
0.381
0.967
0.344
0.836
0.090
0.340
0.217
0.766
aChord distance values presented in the last two rows are computed for assemblages with isolated teeth included in the anaysis and isolated teeth excluded (see text for
details). Chord distance values are calculated between pairs of sites and are listed for the site at the head of the column and the site in the column to the left.
Table 5
Depositional matrix and associated completeness index values (based on Marean,
1991) for the assemblages included in this studya
DepositBreccia type Completeness index
KA
KB
SKHR
SKLB
SKM2
SKM3
ST5OL
COD
hard
hard
hard
uncalcified
decalcified
decalcified
decalcified
decalcified
0.81
0.78
0.82
0.72
0.76
0.65
nd
0.71
aCompleteness index computed by assigning percentage completeness values to
ungulate compact bones, and dividing the summed completeness value by the total
NISP of compact bones in an assemblage.
D.J. de Ruiter et al. / Journal of Human Evolution 55 (2008) 1015–10301022

Page 9

separate out from the majority of the assemblages, the difference is
not large; therefore, it is unclear how great an impact sample size
might have on the taxonomic composition of the assemblage.
Notwithstanding, because sample size appears to be linked to the
number of species identified, it is necessary to examine the in-
fluence of sample size on the ecological composition of the
assemblages.
Species diversity indices allow us to document the ecological
compositionofassemblages(Ludwigand Reynolds,1988;
Magurran,1988), to test whether these assemblages are reasonable
reflections of coherent animal communities. To investigate
species richness we use Fisher’s log series (a); values are presented
in Table 9. There is no significant correlation between cMNI and
Fisher’s log series (a; Spearman’s rs¼0.19, p¼0.65), suggesting
that larger sample sizes do not necessarily result in significantly
richer (i.e., more speciose) faunal assemblages. Turning to the
goodness of fit test (c2) for the Fisher’s log series (a), none of the
assemblages shows an observed distribution that significantly
a
b
cd
ef
gh
ij
Maxilla
0
0.1
0.2
KB ST5OL COD SKLB SKHR SKM2 SKM3KA
Mandible
0
0.1
0.2
KB ST5OL COD SKLB SKHR SKM2 SKM3 KA
Isolated teeth
0
0.2
0.4
0.6
0.8
KBST5OL CODSKLB SKHR SKM2 SKM3 KA
Humerus
0
0.1
0.2
KBST5OL COD SKLB SKHR SKM2 SKM3 KA
Radius
0
0.1
0.2
KBST5OL COD SKLB SKHR SKM2 SKM3KA
Metacarpal
0
0.1
0.2
KBST5OL COD SKLB SKHR SKM2 SKM3KA
Femur
0
0.1
0.2
KB ST5OL COD SKLB SKHR SKM2 SKM3KA
Tibia
0
0.1
0.2
KBST5OL COD SKLB SKHR SKM2 SKM3KA
Metatarsal
0
0.1
0.2
KB ST5OL CODSKLB SKHR SKM2 SKM3 KA
Astragalus
0
0.1
0.2
KBST5OL COD SKLB SKHR SKM2 SKM3KA
Fig. 1. Relative abundances of a selection of skeletal elements for body size II, III, and IV individuals in each of the assemblages. Values are calculated from MNE data in Table 6.
Binomial error bars indicate 95% confidence intervals. Shaded boxes denote hard breccia assemblages; unshaded boxes denote uncalcified/decalcified assemblages.
D.J. de Ruiter et al. / Journal of Human Evolution 55 (2008) 1015–10301023

Page 10

varies from expected. Likewise, there is no significant relation
between cMNI and the Berger-Parker index (Spearman’s rs¼0.52,
p¼0.18). This latter point suggests that increases in sample size do
not necessarily produce faunal assemblages that are more evenly
distributed in terms of species dominance. These diversity data
combine to demonstrate that although there is a relationship
between sample size and the number of species in an assemblage,
there is no indication that increasing sample size unduly influences
the ecological composition of the assemblages.
Results of a correspondence analysis of the taxonomic abun-
dance of large herbivores from a series of modern African nature
reserves are presented in Fig. 4 (data from Table 3). Three distinct
clusters representing three habitat types are evident. The first is
a clustering of taxa and parks from closed or wet habitats, including
Potamochoerus, Cephalophini, Reduncini, and Hippotragini. This
cluster groups together animals that require very dense vegeta-
tional coverage [Potamochoerus, Cephalophini (i.e., tree coverage of
greater than 40% of available land surface)] with those requiring
somewhat less coverage (Reduncini, Hippotragini) in the form of
thick stands of tall grasses and sedges at water’s edge. These taxa
are all linked by their need for a permanent, ample water supply.
For the purpose of this study they are all grouped together into
a single habitat category (closed/wet), as they are consistently
associated in modern nature reserves (see also Alemseged, 2003).
The second clustering represents parks predominated by wood-
lands that are characterized by tree coverage of 20–40% of available
Table 7
Comprehensive minimum numbers of individuals (cMNI) of the select mammalian taxa from the Bloubank Valley cave infills with reconstructed habitat associationsa
Fossil deposit
KB ST5OLCODSKLB SKHR SKM2SKM3 KA Associated habitat
Australopithecus robustus
Papio hamadryas robinsoni
Papio angusticeps
Papio (Dinopithecus) ingens
Gorgopithecus major
Theropithecus oswaldi
Small papionin
Cercopithecoides williamsi
Equus burchelli
Equus capensis
Eurygnathohippus lybicum
Phacochoerus sp.
Metridiochoerus andrewsi
Megalotragus sp.
Connochaetes cf. taurinus
Medium-sized alcelaphine
Damaliscus sp.
Antidorcas marsupialis
Antidorcas recki
Antidorcas bondi
Gazella sp.
Oreotragus oreotragus
Raphicerus campestris
Ourebia ourebi
Syncerus sp.
Simatherium kohllarseni
Pelorovis sp.
Taurotragus oryx
Tragelaphus strepsiceros
Tragelaphus scriptus
Hippotragus sp.
Kobus cf. leche
Redunca arundinum
Redunca fulvorufula
Pelea sp.
Total
Chord distance
62
0
0
0
0
1
7
0
3
0
0
0
2
0
1
6
29 58
30
860
1
–
woodland
woodland
woodland
woodland
grassland
woodland
woodland
grassland
grassland
grassland
grassland
grassland
grassland
grassland
grassland
grassland
grassland
grassland
grassland
grassland
woodland
woodland
closed/wet
woodland
woodland
woodland
woodland
woodland
woodland
closed/wet
closed/wet
closed/wet
grassland
grassland
18
14
0
2
0
0
5
1
0
0
1
0
0
5
0
0
0
3
2
1
0
0
0
1
0
0
1
0
0
0
0
0
1
0
61
12
0
0
0
8
0
0
1
1
0
0
10
3
15
18
7
18
12
0
0
0
2
0
0
1
0
1
3
2
2
0
0
2
2
12
0
1
0
4
0
1
0
3
1
1
1
3
23
11
7
13
0
3
5
1
1
0
2
0
0
0
0
0
0
0
0
0
1
103
20 23
00
1
0
2
0
4
9
7
1
7
1
4
0
0
0
7
0
0
1
7
1
1
1
4
15
0
10
0
2
0
7
23
1
1
2
4
13
28
56
0
18
9
0
0
1
0
3
0
0
3
6
1
2
0
1
0
4
211
17
0
16
7
0
0
6
2
0
7
7
48
37
20
0
12
33
7
1
1
0
2
0
0
0
7
0
3
0
1
0
3
325
19
24
29
19
3
0
5
3
7
3
2
0
1
1
6
4
9
1
0
0
10
210
33
19
17
28
18
3
0
0
0
1
2
0
0
0
0
2
0
0
0
0
0
0
0
48
1.385
5
0
14
1
4
0
3
0
0
2
2
0
4
1
0
0
2
122 186
1.0660.6160.6290.8320.4980.983
aThe last row for each column gives the chord distances computed from taxonomic abundance data. Chord distance values are calculated for pairs of sites and are listed for
the site at the head of the column and the site in the column to the left. See text for derivation of associated habitats.
Table 8
Matrix of chord distances computed between pairs of assemblages for taphonomic (upper right) and taxonomic (lower left) dataa
DepositKB ST5OLCODSKLBSKHRSKM2SKM3KA
KB–
0.114
0.0640.1040.331 0.1120.0650.198 Taphonomic
chord distances
ST5OL
COD
SKLB
SKHR
SKM2
SKM3
KA
1.385
1.129
1.029
1.016
1.101
1.058
1.253
–
1.066
1.106
1.110
0.800
1.039
0.649
0.105
–
0.616
0.866
0.639
0.521
0.993
0.067
0.087
–
0.629
0.610
0.303
1.052
0.414
0.350
0.381
–
0.832
0.808
1.026
0.105
0.103
0.092
0.344
–
0.498
0.755
0.118
0.071
0.086
0.344
0.090
–
0.983
0.271
0.227
0.245
0.171
0.214
0.217
–
Taxonomic chord
distances
aNote: values in bold are those presented in Tables 6 and 7; taphonomic chord distances based on skeletal part data including isolated teeth.
D.J. de Ruiter et al. / Journal of Human Evolution 55 (2008) 1015–10301024

Page 11

land surface. This group includes such taxa as Papio, Chlorocebus,
Phacochoerus, and the bovid tribes Tragelaphini, Bovini, Neotragini
and Aepycerotini. The particularly high numbers of Aepyceros in the
closely geographically-spaced Kruger, Timbavati, and Mkuzi parks
pull them away from the remaining woodland parks, though all
three nonetheless represent woodland habitats. The final clustering
represents open grassland parks and taxa, typified by relatively
sparse tree coverage of less than 20% of available land surface, and
ranging from bushy grasslands to open savannas. This grouping
includes Equus, the Alcelaphini, and the Antilopini. The close
clustering evident in this latter group indicates that the abundance
profiles of these taxa are notably similar across modern nature
reserves.
Data on modern carnivore kill patterns, modern carnivore dens,
and the porcupine den from Table 3 are inserted into the corre-
spondence analysis presented in Fig. 4 as supplementary points.
This insertion of supplementary points is a standard procedure in
correspondence analysis whereby additional row values can be
subsequently incorporated to demonstrate where they plot in the
computation, but without influencing the outcome of the original
analysis.When these supplementary points are added it is apparent
that thecarnivore predationpatterns are stronglyindicative of their
surrounding habitats, as are the bone accumulations from the two
modern lairs. The porcupine den is also strongly representative of
its surrounding habitat. In all cases, the environment that would be
reconstructed from these modern data corresponds closely with
the actual environmental setting. As a result, we conclude that
carnivore kill data are representative of the taxonomic composition
of the of the surrounding animal communities, and that these data
can be used to reconstruct habitats at this scale of analysis.
Taxonomic abundance data for the select large mammals from
the Bloubank Valley fossil assemblages are presented in Table 7. All
fossil specimens employed in this analysis were identifiable to at
least the level of the genus with two exceptions. First, the category
‘small papionin’ is comprised of individuals that have been
variously identified as Parapapio, Cercocebus, and perhaps even
Lophocebus, as it is difficult to reliably identify these small primates
(Frost and Delson, 2002). Second, the category ‘medium-sized
alcelaphine’ consists of fossils that might be referred to a variety of
taxa, including Parmularius, Beatragus, Rabaticeras, and larger-
bodied species of Damaliscus. Since many of these medium-sized
alcelaphine species are diagnosed based on horn cores, and since
horn cores are poorly represented in the South African cave infills,
more precise taxonomic identification is presently not possible. As
a result, they are counted as a single taxonomic category, likely
resulting in an underestimate of the actual numbers of individuals
if more than one medium-sized alcelaphine species was originally
deposited in a given assemblage. The three habitat groupings rec-
ognized in Fig. 4 are applied to the A. robustus-bearing faunal as-
semblages to test the ecological composition of the animal
paleocommunities, with the inferred habitat preferences provided
in Table 7. The mountain reedbuck (Redunca fulvorufula) is re-
assigned to the grassland category, as it does not share the extreme
water dependence of the remaining Reduncini. The oribi (Ourebia
ourebi) prefers a more closed/wet habitat than other Neotragini.
Apart from these two taxa, the remaining bovids show strong
correspondence in ecological requirements at the tribal level.
The numbers of individuals from the fossil assemblages
assigned to each of the three habitat categories were summed (data
from Table 7) and a correspondence analysis performed (Fig. 5). In
this analysis, the habitat categories were analyzed together, with A.
robustus values inserted as supplementary points so that they
would not influence the outcome of the habitat separation. There
does not appear to be any temporal trend in the ordering of the
fossil deposits along either axis. KB plots as a distinct outlier along
axis 1, a positioning which is strongly influenced by the large
number of primates in this assemblage. The category ‘closed/wet’
plots as an outlier along axis 2, though it remains closely aligned
with the ‘grassland’ category along axis 1; notwithstanding, it is
apparent that none of the fossil assemblages group with the
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
KB-
ST5OL
ST5OL-
COD
COD-
SKLB
SKLB-
SKHR
SKHR-
SKM2
SKM2-
SKM3
SKM3-
KA
Fossil deposits
CRD values
CRD Taxonomy
CRD Taphonomy with teeth
CRD Taphonomy without teeth
Fig. 2. Plots of chord distances between pairs of assemblages for taphonomic (Table 6)
and taxonomic (Table 7) data. Correlation of taphonomic chord distances with and
without isolated teeth (Spearman’s rs¼ 0.96, p¼ 0.00) indicates strong similarity
betweenthe twosets of data. Correlationof taphonomic and taxonomic chorddistances
reveals no significant relationship (Spearman’s rs¼0.29, p¼0.54). The lack of correla-
tion between taphonomic and taxonomic chord distances indicate that fluctuations in
taxonomic abundance vary independently of changes in taphonomic conditions.
Table 9
Measures of species richness and species diversitya
Fossil deposit
KBST5OL CODSKLBSKHRSKM2SKM3 KA
cMNI
# species
Fishers log series (a)
c2value
p-value
Berger-Parker index
61
14
5.69
4.17
0.38
3.39
48
12
122
20
103
20
325
22
210
29
186
24
211
23
5.14
4.23
0.38
2.67
6.80
2.72
0.61
6.78
7.40
3.82
0.43
4.48
5.33
4.35
0.50
5.60
9.12
2.79
0.59
7.24
7.34
5.71
0.34
5.64
6.57
3.82
0.56
3.77
aThe Fisher’s log series (a) computation allows for a goodness of fit test; c2and
probability values are presented with the a values.
2610 14 1822 263034384246
Number of individuals
0
2
4
6
8
10
12
14
16
18
20
Predicted number of species
SKM2 (18)
SKM3 (15)
KA (14)
KB (12)
ST5OL (12)
SKLB (15)
COD (15)
SKHR (14)
Fig. 3. Rarefaction curves computed for the assemblages examined in this study.
Rarefaction analysis predicts the number of species that might be present if sample
sizes of all assemblages are artificially standardized to that of the smallest assemblage
(ST5OL). KB and ST5OL have relatively smaller assemblages, and SKM2 has a relatively
larger assemblage, compared to the remaining deposits. Values in parentheses after
site names indicate the predicted number of species per assemblage.
D.J. de Ruiter et al. / Journal of Human Evolution 55 (2008) 1015–10301025

Page 12

‘closed/wet’ category. Most of the assemblages group nearest to the
‘grassland’ category, and farther away from the ‘woodland’
category. This relative positioning is especially apparent along axis
1, which accounts for approximately 80% of the inertia (variance) in
the data. This proximity of fossil assemblages to the ‘grassland’
category is consistent with the majority of reconstructions of the
paleoenvironment typically associated with A. robustus. However,
when A. robustus values are inserted as supplementary points, it is
evident that the hominins plot closer to the ‘woodland’ category,
a grouping that is inconsistent with a close association between the
hominins and a grassland habitat. Instead, these data demonstrate
that ‘woodland’ taxa share the most comparable abundance profile
relative to the hominins. In other words, the relative representation
of A. robustus is most similar to the relative representation of
‘woodland’ taxa across the assemblages.
The relative abundance of taxa assigned to the three habitat
categories was computed (data from Table 7) and plotted (Fig. 6).
The proportions of taxa representing the three habitat categories
tend to be relatively consistent across the assemblages, with three
principal exceptions. First, the fauna represented at KB is
dramatically different from that seen in the other assemblages,
with an abundance of woodland-adapted taxa and a relative
paucity of grassland-adapted taxa. Second, there is a slight
underrepresentation of grassland taxa in SKHR, relating to the
abundance of hominins from this assemblage. Third, although they
are not common, SKM2 has significantly more taxa indicative of
a closed/wet habitat than the remaining assemblages. Apart from
these departures, there are no significant differences in termsof the
relative representation of fauna adapted to the various habitat
categories over time. Excluding KB, grassland-adapted taxa clearly
predominate, generally representing greater than 60% of animals
in a given assemblage; woodland taxa are moderately well-
represented, typically accounting for slightly more than 20% of
animals. These data are in accordance with paleoenvironmental
reconstructions indicating predominantly grasslands for the fossil
cave infills.
Correlating the proportions of A. robustus with proportions of
taxa assigned to the different habitat categories, we see a strong,
statistically significant, negative association between the hominins
and the ‘grassland’ category (rs¼?0.86, p¼0.007; Table 10). At the
same time there are only weak, insignificant correlations with the
‘woodland’ and ‘closed/wet’ categories. The significant, negative
correlation between A. robustus and grassland-adapted taxa
indicates that the more grassland animals there are in a given
assemblage, the fewer hominin individuals there tend to be.
Although these correlations do not clearly indicate the habitat
preference of the hominins, they do demonstrate an inverse
relationship between the hominins and grassland-adapted fauna.
We interpret this to mean that although theylived in environments
predominantly characterized by open grasslands, they were not
closely tied to such environments, thus the predominant environ-
mental signal does not necessarily indicate a habitat preference for
the robust australopiths.
0
Axis 1 (28.76% of Inertia)
0
Axis 2 (20.80% of Inertia)
Modern game parks
Modern carnivore kills
Modern taxa
Tai Forest leopard kills
Potamochoerus
Reduncini
Cephalophini
Hippotragini
Po
BoubaNdjida
Phacochoerus
Bovini
Papio
Chobe
Tragelaphini
Cercopithecus
Tarangire
Kgalagadi
Tsumeb
Manyara
Hwange
iMfolozi
Moremi
Kruger leopard kills
Londolozi leopard kills
Neotragini
Aepycerotini
Mkuzi
Kruger
Timbavati
Kruger spotted hyena den
Kruger spotted hyena kills
Nossob
porcupine den
Serengeti leopard kills
Alcelaphini
Makgadikgadi
Serengeti
hyena kills
Omo
Mara
Nairobi Serengeti
Antilopini
Equus
Makgadikgadi
brown hyena den
Etosha
Ngorongoro
Waza
KafueFlats
SWGabon
Yankari
Saint-Floris
LakeNakuru
Arli
Virunga
DeuxBale
Comoe
Penjari
Kainji
W
Kruger brown hyena kills
Ngorongoro spotted
hyena kills
Fig. 4. Correspondence analysis of modern nature reserve census counts and associated carnivore kill data (data from Table 3). Modern carnivore kill data were inserted as
supplementary points so as not to influence the outcome of the analysis. The geographically closely-spaced Kruger, Timbavati, and Mkuzi parks all have very high numbers of impala
(Aepyceros melampus), pulling them away from the remaining woodland habitat parks; nonetheless, all three are comprised of woodland habitats.
0
Axis 1 (79.85% of Inertia)
0
Axis2 (20.15% of Inertia)
KB
ST5OL
SKLB
COD
SKHR
KA
Grassland
SKM3
SKM2
Woodland
Closed/wet
A. robustus
Fig. 5. Correspondence analysis of habitat categories derived from the faunal assem-
blages examined in this study (data from Table 7). A. robustus values were inserted as
supplementary points so as not to influence the outcome of the analysis. The close
proximity of A. robustus and the ‘woodland’ category along axis 1 demonstrates that
the abundance profile of A. robustus is most similar to that of the ’woodland’ category.
D.J. de Ruiter et al. / Journal of Human Evolution 55 (2008) 1015–10301026

Page 13

Discussion
Taphonomic data implicate a variety of bone accumulating
agents in each of the fossil assemblages, including carnivores,
rodents, and hominins. In addition, the relative influence of abiotic
factors such as slopewash cannot be discounted, as significant
numbers of bones would have been mobilized into the caves from
theirsurrounding catchmentareas (Butzer,1984). However,none of
these lines of evidence are sufficient to implicate a predominant, or
consistent, bone accumulating agent across the assemblages.
Moreover, the time-averaged nature of the fossil cave infills
enhances the likelihood that numerous different agents were
involved over time. Consequently, it is apparent that a variety of
accumulating agents were active in the vicinity of the caves during
the time they were open to the surface. As Brain (1980: 107) has
pointed out, ‘‘any cave which has been open for thousands of years
is likely to have had bones brought to it in a variety of different
ways’’. We assume that the combined impact of numerous agents
over long spans of time minimized the idiosyncratic influence of
any individual accumulating agent in the fossil assemblages. Since
there is no consistent taphonomic pattern relating to accumulating
agent, we further assume that any taphonomic biases introduced as
a result of bone accumulating agents had an approximately equiv-
alent impact across assemblages, as no one assemblage appears to
have been significantly more biased relative to the others.
Testing for conditions of isotaphonomy does, however, reveal
a bias relating to depositional matrix. Fossils from hard breccia
deposits appear less fragmented than fossils from uncalcified/
decalcified deposits, and are characterized by an overabundance
of craniodental remains. This results in a potentially significant
bias relating to identifiability of specimens, as it is the cranio-
dental remains that form the bulk of specific taxonomic assign-
ments. This difference is influenced by the relative difficulty
encountered when manually preparing fossils out of hard breccia,
in particular fragmented postcranial specimens that are often not
removable (pers. obs.). However, since taxonomic identification is
based principally upon craniodental remains and since the most
common skeletal elements (isolated teeth) are relatively evenly
distributed across assemblages, we conclude that this taphonomic
bias has not irretrievably masked the underlying biological signal
relating to animal paleocommunity composition. Indeed, in terms
of chord distances, there is no relationship between taphonomic
conditionsandtaxonomiccomposition.
taxonomic representation varies independently of taphonomic
conditions. We interpret this to mean that changes in taxonomic
abundance over time do indeed signal animal paleocommunity
responses to alterations in environmental conditions, allowing us
to investigate fluctuations in animal paleocommunity composi-
tion in the Bloubank Valley.
The faunal assemblage from KB consistently stands out as
unique relative to the other Bloubank Valley cave infills. Brain
(1981b) suggested that KB had been collected by large carnivores,
while Vrba (1981) concluded that the cave represented a death trap
for ungulates and primates that were then opportunistically scav-
enged by visiting carnivores. The notable prevalence of primates in
this assemblage might indicate some form of accumulation bias,
perhaps a situation that rendered primates particularly susceptible
to incorporation in the assemblage (e.g., a specialized predator of
primates). The small numbers of carnivore modified bones and
coprolites do not aid in resolving this issue, and a high level of
comminution of bones has potentially obscured indications of bone
surface modification (Brain, 1981b). While the high number of
primates in and of itself is insufficient to implicate a specialized
primate predator, the high carnivore:ungulate ratio does imply
significant carnivore activity. However, the dissimilar taxonomic
composition of KB is not coupled with a notable taphonomic dif-
ference, thus the unique nature of this assemblage cannot be solely
a result of taphonomic bias. We therefore hypothesize that KB
samples a paleoenvironment that was unlike that seen in the other
Thisindicatesthat
a
b
c
Grassland
0.0
0.2
0.4
0.6
0.8
1.0
KB ST5OLCOD SKLBSKHR SKM2SKM3 KA
Woodland
0.0
0.2
0.4
0.6
0.8
1.0
KB ST5OLCOD SKLBSKHRSKM2 SKM3 KA
Closed/wet
0.0
0.2
0.4
0.6
0.8
1.0
KBST5OL CODSKLBSKHR SKM2 SKM3 KA
Fig. 6. Relative abundance of taxa assigned to the three habitat categories utilized in
this study. Values are calculated from cMNI data in Table 7. Binomial error bars indicate
95% confidence intervals.
Table 10
Relative abundance values (proportions) of select mammalian taxa assigned to habitat categoriesa
Fossil deposit
KB ST5OL COD SKLBSKHR SKM2 SKM3KA Spearman’s rs p-level
A. robustus
Closed/wet
Woodland
Grassland
0.10
0.00
0.67
0.23
0.04
0.00
0.25
0.71
0.02
0.02
0.17
0.80
0.09
0.00
0.17
0.74
0.18
0.01
0.20
0.61
0.04
0.06
0.23
0.67
0.03
0.03
0.19
0.75
0.00
0.01
0.20
0.79
–
?0.59
0.48
?0.86
–
0.13
0.23
0.007*
aSpearman’s rscorrelation coefficients are computed for each category compared to A. robustus. Probability value with an asterisk (*) is statistically significant.
D.J. de Ruiter et al. / Journal of Human Evolution 55 (2008) 1015–10301027

Page 14

deposits; more research is required to confirm this suggestion,
preferably with an augmented faunal assemblage.
Results of the correspondence analysis of modern nature
reserves demonstrate groupings of taxa and habitats that are
consistent with those of previous correspondence analyses
(Greenacre and Vrba, 1984; Alemseged, 2003), even though we
utilize a different set of nature reserves and census counts, and
include taxa not previously incorporated (Cercopithecidae, Equi-
dae, Suidae). In their bone transect study in the Amboseli National
Park, Behrensmeyer et al. (1979) determined that bone distribu-
tions of certain taxa did not match their live census data. However,
at the grosser scale of our analysis, we do see close correspondence
between modern carnivore kill data and animal community com-
position. This includes both data from bone accumulations as well
as animal kill data from a series of different carnivorous agents.
These results indicate that animals do tend to die where they live,
thus it would appear that carnivore-produced bone accumulations
are broadly representative of animal communities, which in turn
are good indicators of environment.
Taxonomic abundance data demonstrate that the paleoenvir-
onments of all but KB can be reconstructed as predominantly open
grasslands. The preponderance of grassland-living taxa in the
majority of the Bloubank Valley assemblages is in agreement with
paleoecological analyses that reconstruct a predominantly open
and relativelyarid environment with nearby edaphic grasslands for
A. robustus (Vrba, 1975, 1976, 1980, 1985a,b; Brain, 1981a; Brain
et al., 1988; Shipman and Harris, 1988; McKee, 1991; Denys, 1992;
Avery, 1995, 2001; Reed, 1997; Watson, 2004; Reed and Rector,
2006). The results of this study differ, however, in that they indicate
that these open grasslands do not reflect the habitat preference of
the hominins. Although A. robustus is consistently associated with
open grassland environments, they exhibit a strong, statistically
significant, negative relationship with the taxa that occupy this
habitat. In other words, the more open grassland-adapted
taxa there are in an assemblage, the fewer hominins there are
in that assemblage. Such a conclusion contrasts with the notion of
A. robustus as an open grassland specialist.
If the unique nature of the fauna from KB is not exclusively the
result of taphonomic bias, the predominantly wooded environment
that is indicated by this assemblage might, in fact, represent
a habitat favored by the hominins. However, because correlations
between the hominins and the remaining habitat categories are
insignificant, statistical support for an actual habitat preference
remains elusive. One line of evidence that does support awoodland
habitat preference for A. robustus is the correspondence analysis
that groups the hominins more closely with this particular habitat.
The proximity of A. robustus to the ‘woodland’ category along axis 1
in Fig. 4 indicates that this is the category with the most compa-
rable abundance profile relative to the hominins. In other words,
the relative representation of A. robustus is most similar to the
relative representation of ‘woodland’ taxa across the assemblages.
Although not conclusive, this close association between A. robustus
and the ‘woodland’ category is suggestive that the conditions that
were sufficient for woodland-adapted animals were also favored by
the hominins.
Several studies of the isotope chemistry of A. robustus dental
enamel have demonstrated a preponderance of C3 resources,
indicative of a principally forest- or woodland-based diet (Lee-
Thorp et al., 1994; Sponheimer et al., 2005, 2006a,b). This isotopic
evidence is supported by studies of enamel microwear patterns
that demonstrate consumption of hard food items, such as seeds
and nuts, that are typically associated with forest-based food
sources (Grine, 1986; Grine and Kay, 1988). At the same time,
isotopic analysis has demonstrated that a significant proportion of
A. robustus’ diet was comprised of C4 grass-based resources,
accounting for an average 35% of the diet, perhaps in the form of
fallback foods (Sponheimer et al., 2005, 2006b). Although isotope
data for sedges, termites, and numerous African mammals exist
(Sponheimer et al., 2003, 2005), there is currently little data
regarding the isotopic composition of other potential fallback
foods, such as underground storage organs, in Africa. Nonetheless,
the hominins appear to have preferred a forest-based diet, though
they were also capable of consuming sometimes considerable
amounts of resources extracted from the surrounding grasslands
that comprised the major portion of the habitat mosaic.
The patterns of habitat utilization documented in this study
present us with several potential ecological implications. It is
possible that the assemblages are time-averaged, and that the
hominins have been artificially lumped in death alongside taxa
that they might never have encountered in life. This would imply
that the hominins were itinerant occupants of the area, present
during the rarer occasions when conditions were particularly
favorable (expanded woodlands), and absent when conditions
were unfavorable (expanded grasslands). However, the environ-
mental mosaics reconstructed for several of the deposits indicate
a variety of habitats, including woodlands potentially capable of
sustaining hominin populations (Brain et al., 1988; Avery, 1995;
Reed, 1997; Watson, 2004). The likelihood therefore exists that
the hominins were habitat generalists capable of living in a vari-
ety of environments, but perhaps preferring woodlands over the
less-favored grasslands when conditions were sufficient. As large-
bodied, mobile, intelligent apes, the hominins would have been
able to respond to environmental oscillations by altering their
behavioral patterns in numerous ways. Among the apes, hominins
are unique in their capacity to modify their diet to consume
significant quantities of C4-based resources (Sponheimer et al.,
2005). In fact, A. robustus is marked by the ability to dramatically
alter its dietary behavior on both seasonal and interannual scales
(Sponheimer et al., 2006b). The capacity to subsist on less-favored
dietary items likely allowed the hominins to survive periods of
resource stress by resorting to fallback foods that might be un-
available to other occupants of the area, as well as by altering
their population densities.
Summary and conclusions
The aim of this study was to investigate whether any indicators
of the habitat association of A. robustus were preserved in the
faunal assemblages of the Bloubank Valley of South Africa.
Notwithstanding evidence of limited taphonomic biasing relating
to depositional matrix and perhaps accumulating agents, it appears
that these potential biases have not unduly influenced the
ecological composition of the faunal assemblages. Correspondence
analysis of census data from a series of modern nature reserves
displayed the habitat preferences of a select group of large mammal
taxa, in turn allowing assignment of fossil taxa from the Bloubank
Valley assemblages to a series of broadly defined habitat categories.
Subsequent correspondence analysis of the faunal assemblages
reveals thatA. robustus has an abundance profile most similar tothe
‘woodland’ habitat category, meaning that the relative represen-
tation of the hominins corresponds most closely to that of wood-
land-adapted taxa. Additionally, the strong, negative correlation
that is evident between A. robustus and grassland-adapted taxa
contrasts with reconstructions of these hominins as open grassland
habitat specialists. Rather, our admittedly limited dataset from
a small number of closely spaced fossil localities nonetheless
suggests that A. robustus was a habitat generalist. These data,
coupled with recent evidence demonstrating a highly generalized
diet, indicate that the commonly held perception that the specialist
adaptations of A. robustus doomed it to extinction in the face of
fluctuating environmental conditions during the Plio-Pleistocene
requires rethinking.
D.J. de Ruiter et al. / Journal of Human Evolution 55 (2008) 1015–10301028

Page 15

Acknowledgements
We thank Bob Brain, Francis Thackeray, Stephany Potze, and
Teresa Kearney of the Transvaal Museum for access to fossil and
modern comparative materials in their care. Mike Raath, Bruce
Rubidge, Lee Berger, and Rodrigo Lacruz gave us access to the fossil
materials housed at the University of the Witwatersrand. We thank
Kathy Kuman for sharing the details of the excavations in the
Oldowan Infill of Sterkfontein, and the nature of its depositional
matrix. Sheela Athreya and David Carlson provided invaluable
statistical advice, though any errors remain our responsibility.
Earlier drafts of this paper were greatly improved through the
helpful comments and insights of Travis Pickering and two
anonymous reviewers. Steve Frost and Jason Heaton provided
informative discussions regarding the fossil baboons of South
Africa. The rarefaction analysis was performed using software
developed by Steven M. Holland in the University of Georgia Stra-
tigraphy Lab (available online at: http://www.uga.edu/~strata/
software/Software.html). This research was funded by the Wenner-
Gren Foundation (USA), the National Research Foundation (RSA),
the Paleoanthropology Scientific Trust (RSA), as well as the Faculty
Research Enhancement Program and the International Research
Travel Assistance Grant program of Texas A&M University.
References
Acocks, J.P.H., 1988. Veld types of South Africa, third ed. Mem. Bot. Surv. S. Afr. 57,
1–146.
Afolayan, T.A., Ajayi, S.S., 1980. The influence of seasonality on the distribution of
large mammals in the Yankari Game Reserve, Nigeria. Afr. J. Ecol. 18, 87–96.
Alemseged, Z., 2003. An integrated approach to taphonomy and faunal change
in the Shungura Formation (Ethiopia) and its implications for hominid evolu-
tion. J. Hum. Evol. 44, 451–478.
Avery, D.M., 1995. Southern savannas and Pleistocene hominid adaptations: the
micromammalian perspective. In: Vrba, E.S., Denton, G.H., Partridge, T.C.,
Burckle, L.H. (Eds.), Paleoclimate and Evolution, with Emphasis on Human
Origins. Yale University Press, New Haven, pp. 459–478.
Avery, D.M., 2001. The Plio-Pleistocene vegetation and climate of Sterkfontein and
Swartkrans, South Africa, based on micromammals. J. Hum. Evol. 41, 113–132.
Ayeni, J.S.O.,1980. Management problems of the Kainji National Park, Nigeria. Afr. J.
Ecol. 18, 97–111.
Baba, M., Doi, T., Ikeda, H., Iwamoto, T., Ono, Y., 1982. A census of large mammals in
Omo National Park, Ethiopia. Afr. J. Ecol. 20, 207–210.
Badgley, C., 1986. Counting individuals in mammalian fossil assemblages from
fluvial environments. Palaios 1, 328–338.
Behrensmeyer, A.K., 1991. Terrestrial vertebrate accumulations. In: Allison, P.A.,
Briggs, D.E.G. (Eds.), Taphonomy: releasing the data locked in the fossil record.
Plenum Press, New York, pp. 291–335.
Behrensmeyer, A.K., Western, D., Dechant-Boaz, D., 1979. New perspectives in ver-
tebrate paleoecology from a recent bone assemblage. Paleobiology 5, 12–21.
Benefit, B.R., McCrossin, M.L., 1990. Diet, species diversity and distribution of
African fossil baboons. Kroeber Anthropol. Soc. Pap. 71–72, 77–93.
Berger, L.R., de Ruiter, D.J., Steininger, C., Hancox, J., 2003. Preliminary results of
excavations at the newly investigated locality Coopers D, Gauteng, South Africa.
S. Afr. J. Sci. 99, 276–278.
Berger, L.R., Menter, C.G., Thackeray, J.F., 1994. The renewal of excavation activities
at Kromdraai, South Africa. S. Afr. J. Sci. 90, 209–210.
Blumenschine, R.J., Marean, C.W., 1993. A carnivore’s view of archaeological bone
assemblages. Center for Archaeological Investigations. In: Hudson, J. (Ed.), From
bones to behavior. SIUC, Carbondale, IL, pp. 273–300.
Bobe, R., Behrensmeyer, A.K., 2004. The expansion of grassland ecosystems in Africa
in relation to mammalian evolution and the origin of the genus Homo. Paleo-
geogr., Paleoclimatol., Paleoecol. 207, 399–420.
Bobe, R., Behrensmeyer, A.K., Chapman, R.E., 2002. Faunal change, environmental
variability, and late Pliocene hominin evolution. J. Hum. Evol. 42, 475–497.
Bobe, R., Eck, G.G., 2001. Responses of African bovids to Pliocene climatic change.
Paleobiology Memoirs 2. Paleobiology 27 (Suppl. to No. 2), 1–47.
Bourlie `re, F., 1963. Observations on the ecology of some large African mammals. In:
Howell, F.C., Bourlie `re, F. (Eds.), African Ecology and Human Evolution. Methuen
& Co., London, pp. 43–54.
Brain, C.K., 1980. Some criteria for the recognition of bone-collecting agencies in
African caves. In: Behrensmeyer, A.K., Hill, A.P. (Eds.), Fossils in the Making.
University of Chicago Press, Chicago, pp. 107–130.
Brain, C.K., 1981a. Hominid evolution and climatic change. S. Afr. J. Sci. 77, 104–105.
Brain, C.K.,1981b. The Hunters or the Hunted? University of Chicago Press, Chicago.
Brain, C.K., 2004. Structure and stratigraphy of the Swartkrans Cave in light of the
new excavations. In: Brain, C.K. (Ed.), Swartkrans: A Cave’s Chronicle of Early
Man, second ed. Transvaal Museum Monograph No. 8, Pretoria, pp. 7–22.
Brain, C.K., Sillen, A.,1988. Evidence from the Swartkrans cave for the earliest use of
fire. Nature 336, 464–466.
Brain, C.K., Churcher, C.S., Clark, J.D., Grine, F.E., Shipman, P., Susman, R.L., Turner, A.,
Watson, V., 1988. New evidence of early hominids, their culture and environ-
ment from the Swartkrans Cave, South Africa. S. Afr. J. Sci. 84, 828–835.
Broom, R., 1938. The Pleistocene anthropoids apes of South Africa. Nature 142,
377–379.
Broom, R., 1943. South Africa’s part in the solution of the problem of the origin of
man. S. Afr. J. Sci. 40, 68–80.
Broom, R., 1951. Finding the Missing Link. Watts, London.
Broom, R., Robinson, J.T., 1952. Swartkrans Ape-Man Paranthropus crassidens.
Transvaal Museum Memoir No. 6, Pretoria.
Bunn, H.T., 1982. Meat-eating and human evolution: studies on the diet and sub-
sistence patterns of Plio-Pleistocene hominids in East Africa. Ph.D. Dissertation,
University of California, Berkeley.
Bunn, H.T., 1986. Patterns of skeletal representation and hominid subsistence
activities at Olduvai Gorge, Tanzania, and Koobi Fora, Kenya. J. Hum. Evol. 15,
673–690.
Butzer, K.W., 1984. Archeogeology and Quaternary environment in the interior of
southern Africa. In: Klein, R.G. (Ed.), Southern African Prehistory and
Palaeoenvironments. A.A. Balkema, Rotterdam, pp. 1–64.
Buzas, M.A., 1990. Another look at confidence limits for species proportions.
J. Paleontol. 64, 842–843.
Cooke, H.B.S., 1952. Mammals, ape-men, and stone age men in southern Africa.
S. Afr. Archaeol. Bull. 26, 59–69.
Cooke, H.B.S., 1963. Pleistocene mammal faunas of Africa, with particular reference
to southern Africa. In: Howell, F.C., Bourlie `re, F. (Eds.), African Ecology and
Human Evolution. Wenner-Gren Foundation, New York, pp. 65–116.
Cruz-Uribe, K., 1991. Distinguishing hyena from hominid bone accumulations. J.
Field Archaeol. 18, 467–486.
Curnoe, D., Grun, G., Taylor, L., Thackeray, J.F., 2001. Direct ESR dating of a Pliocene
hominin from Swartkrans. J. Hum. Evol. 40, 379–391.
Darling, F.F.,1960. An ecological reconnaissance of the Mara Plains in Kenya Colony.
Wildl. Monogr. 5, 1–41.
Dasmann, R.F., Mossman, A.S., 1962. Abundance and population structure of
wild ungulates in some areas of southern Rhodesia. J. Wildl. Manage. 26,
262–268.
Delson, E., 1984. Cercopithecid biochronology of the African Plio-Pleistocene:
correlation among eastern and southern hominid-bearing localities. Cour.
Forsch.-Inst. Senckenberg 69, 199–218.
Denys, C., 1992. Les analyses multivarie ´es: une aide a ` l’interpre ´tation des pale ´o-
environments. L’example des rongeurs Plio-Ple ´istoce `ne d’Afrique australe.
Geobios 14, 209–217.
de Ruiter, D.J., 2003. Revised faunal lists for Members 1–3 of Swartkrans, South
Africa. Annls. Transv. Mus. 40, 29–41.
de Ruiter, D.J., 2004. Relative abundance and skeletal part representation of
macromammals from Swartkrans. In: Brain, C.K. (Ed.), Swartkrans: A Cave’s
Chronicle of Early Man, second ed. Transvaal Museum Monograph No. 8,
Pretoria, pp. 265–278.
de Ruiter, D.J., Berger, L.R., 2000. Leopards as taphonomic agents in dolomitic
cavesdimplications for bone accumulations in the hominid-bearing deposits of
South Africa. J. Archaeol. Sci. 27, 665–684.
de Ruiter, D.J., Berger, L.R., 2001. A note on leopard (Panthera pardus Linnaeus) cave
caching related to anti-theft behavior in the John Nash Nature Reserve, South
Africa. Afr. J. Ecol. 39, 396–398.
El-Zaatari, S., Grine, F.E., Teaford, M.F., Smith, H.F., 2005. Molar microwear and
dietary reconstructions of fossil Cercopithecoidea from the Plio-Pleistocene
deposits of South Africa. J. Hum. Evol. 49, 180–205.
Esser, J.D., Van Lavieren, L.P.,1979. Importance, repartition et tendance evolutive des
populations des grand mammiferes herbivous et de l’autruche dans le Parc
National de Waza, Cameroun. La Terre et la Vie 33, 3–26.
Estes, R.D., 1991. The Behavior Guide to African Mammals. University of California
Press, Berkeley.
Estes, R.D., Small, R., 1981. The large herbivore populations of Ngorongoro Crater.
Afr. J. Ecol. 19, 175–185.
Foley, R., 1987. Another Unique Species: Patterns in Human Evolutionary Ecology.
Longman Scientific and Technical, Essex.
Foster, J.B., Coe, M.J., 1968. The biomass of game animals in Nairobi National Park,
1960–66. J. Zool., Lond. 155, 413–425.
Foster, J.B., Kearney, D., 1966. Nairobi National Park game census. E. Afr. Wildl. J. 5,
112–120.
Frost, S.R., Delson, E., 2002. Fossil Cercopithecidae from the Hadar Formation and
surrounding areas of the Afar Depression, Ethiopia. J. Hum. Evol. 43, 687–748.
Geerling, C., Bokdam, J., 1973. Fauna of the Comoe National Park, Ivory Coast. Biol.
Conserv. 5, 251–257.
Green, A.A., 1979. Density estimates of the larger mammals of Arli National Park,
Upper Volta. Mammalia 43, 59–70.
Greenacre, M.J., 2007. Correspondence Analysis in Practice. Chapman and Hall,
London.
Greenacre, M.J., Vrba, E.S., 1984. Graphical display and interpretation of antelope
census data in African wildlife areas, using correspondence analysis. Ecology 65,
984–997.
Grine, F.E., 1986. Dental evidence for dietary differences in Australopithecus and
Paranthropus. J. Hum. Evol. 15, 783–822.
Grine, F.E., Kay, R.F., 1988. Early hominid diets from quantitative image analysis of
dental microwear. Nature 333, 765–768.
D.J. de Ruiter et al. / Journal of Human Evolution 55 (2008) 1015–10301029