Content uploaded by V. Pilbrow
Author content
All content in this area was uploaded by V. Pilbrow on May 30, 2014
Content may be subject to copyright.
2. Patterns of molar variation in great apes and their
implications for hominin taxonomy
V. PILBROW
University of Melbourne
Zoology Department
Victoria 3010 Australia
vpilbrow@unimelb.edu.au.
Keywords: Pan, Pongo, Gorilla, models, molar metrics
Abstract
In studying the nature of variation and determining the taxonomic composition of a hominin fossil assem-
blage the phylogenetically closest and thus the most relevant modern comparators are Homo and Pan and
following these, Gorilla and Pongo. Except for Pan, however, modern hominids lack taxonomic diversity,
since by most accounts each one is represented by a single living species. Pan is the sister taxon to modern
humans and it is represented by two living species. As such the species of Pan have greater relevance for
studying interspecific variation in fossil hominin taxonomy. Despite their relatively impoverished species repre-
sentations Pan troglodytes, Gorilla gorilla and Pongo pygmaeus are, nevertheless, represented by subspecies.
This makes them relevant for studying the nature of intraspecific variation, in particular for addressing the
question of subspecies in hominin taxonomy. The aim of this study is to examine the degree and pattern of
molar variation in species and subspecies of P. pygmaeus, G. gorilla, P. troglodytes and P. paniscus. I test the
hypothesis that measurements taken on the occlusal surface of molars are capable of discriminating between
species and subspecies in commingled samples of great apes. The results of this study are used to draw inferences
about our ability to differentiate between species and subspecies of fossil hominins. The study samples include
P t troglodytes (n = 152) P t verus (n = 64) P t schweinfurthii (n = 79) G g gorilla (n = 208)
G g graueri (n = 61) G g beringei n =30 P p pygmaeus (n = 140)and P p abelii n =25.
Measurements taken from digital images were used to calculate squared Mahalanobis distances between
subspecies pairs. Results indicate that molar metrics are successful in differentiating between the genera,
species and subspecies of great apes. There was a hierarchical level of differentiation, with the greatest
separation between genera, followed by that between species within the genus Pan and finally between
subspecies within species. The patterns of molar differentiation showed excellent concordance with the
patterns of molecular differentiation, which suggests that molar metrics have a reasonably strong phyloge-
netic signal. Pan troglodytes troglodytes and P. troglodytes schweinfurthii were separated by the least dental
distance. P. troglodytes verus was separated by a greater distance from these two, but on the whole the
distances among subspecies of P. troglodytes were less than among subspecies of G. gorilla and P. pygmaeus.
The dental distance between G. g. gorilla and G. g. graueri was greater than that observed between
P. troglodytes and P. paniscus. With size adjustment intergroup distances between gorilla subspecies were
reduced, resulting in distances comparable to subspecies of P. troglodytes. A contrast between size-preserved and
9
S.E. Bailey and J.-J. Hublin (Eds.), Dental Perspectives on Human Evolution, 9–32.
© 2007 Springer.
10 Pilbrow
size-adjusted analyses reveal that size, sexual dimorphism and shape are significant factors in the patterning of
molar variation in great apes. The results of this study have several implications for hominin taxonomy, including
identifying subspecies among hominins. These implications are discussed.
Introduction
Molars make up a disproportionately large
part of early hominin fossil collections and
figure prominently in taxonomic assessments.
When determining whether the differences
observed among sets of fossil hominin molars
can be attributed to that of a species, or are
part of the variation to be expected within
a species, paleoanthropologists generally look
to modern analogs. Because the fossil record
does not present us with the requisite numbers
of specimens, or the anatomical, behav-
ioral and ecological details needed to gauge
the nature of variation in fossil hominins,
extant hominids1, namely humans and great
apes, provide the next best alternative for
modeling variation. The justification behind
using closely related extant taxa for models
is fairly sound: through recency of common
ancestry in closely related taxa are likely
to have shared similar patterns and ranges
of variation. Therefore, they likely provide
reasonably accurate estimates of the type of
variation to be expected in the fossils (see
papers in Kimbel and Martin, 1993). Just
as important, we have a fairly good under-
standing of patterns of variation in extant
hominids in external morphology, breeding
patterns, habitat preferences and genetic
structure and we can see how these match
up with variation in fossilizable attributes
such as cranial and dental features. The
extant hominids thus provide a compre-
hensive comparative model for understanding
variation in the molars of fossil hominins. If
we are to ultimately reconstruct the biology
and lifeways of fossil forms in a manner that
is consistent with living forms this modeling
of variation is essential.
A consensus opinion emerging from
molecular systematists (Goodman, 1962;
Goodman et al., 1982; 1998; Caccone and
Powell, 1989; Ruvolo, 1994; 1997), corrob-
orated by morphological data (Begun, 1992;
Gibbs et al., 2000; Guy et al, 2003; Lockwood
et al., 2004), is that chimpanzees are the
closest extant relatives of modern humans.
Chimpanzees are therefore especially relevant
for studying the nature of variation in fossil
hominins. Additionally, chimpanzee patterns
of variation are well documented. Two
species of chimpanzees, Pan paniscus and
Pan troglodytes, are recognized by a plethora
of morphological, behavioral, ecological
and genetic studies (Coolidge, 1933;
Johanson, 1974; Shea, 1981, 1983a, b, c,
1984; Sibley and Ahlquist, 1984; Caccone and
Powell, 1989; Kinzey, 1984; Shea et al., 1993;
Wrangham et al., 1994; Uchida, 1996; Guy
et al., 2003; Taylor and Groves, 2003;
Lockwood et al., 2004; Pilbrow, 2006a, b;
but see Horn, 1979). Diversity within
P. troglodytes is also substantial. The tradi-
tional taxonomy recognizes three subspecies
(Hill, 1967; 1969), but mtDNA studies have
suggested that an additional one should be
recognized (Gonder et al., 1997). They also
suggest that the West African subspecies P.
t. verus should be recognized as a distinct
species, P. verus (Morin et al., 1994). Certain
morphological data sets have registered
the distinctiveness of P. t. verus compared
to the other two subspecies (Braga, 1995;
Uchida, 1996; Pilbrow, 2003, 2006a, b;
Taylor and Groves, 2003). The substantial
diversity exhibited by Pan at the inter- and
intra-specific level, together with a close
phylogenetic relationship to humans makes
them particularly germane to discussions
about hominin taxonomy.
Gorillas and orangutans are more distantly
related to humans. According to the consensus
view (above) gorillas are a sister taxon to
Molar Variation in Great Apes 11
the chimp-human clade and orangutans are
the closest outgroup (for a contrary viewpoint
see Schwartz, 1984). Regardless, both are
members of the family Hominidae making
them phylogenetically appropriate taxa for
studying fossil hominin variation. Gorillas and
orangutans differ from chimpanzees in their
taxonomic diversity. According to the tradi-
tional taxonomy, the diversity within Gorilla
gorilla is no more than can be accommo-
dated within a single species with three recog-
nized subspecies (Coolidge, 1929; Groves,
1970). The same is true for Pongo pygmaeus,
where two subspecies are traditionally recog-
nized (Courtney et al., 1988). There have been
several recent attempts to revise the conven-
tional taxonomy, particularly by molecular
systematists. They advocate that the east
and west African gorillas best represent two
distinct species, and likewise that the Bornean
and Sumatran orangutans should be distin-
guished at the species level (Ruvolo et al.,
1994; Garner and Ryder, 1996; Xu and
Arnason, 1996; Zhi et al., 1996; Salton-
stall et al., 1998; Jensen-Seaman and Kidd,
2001). Several studies also propose that one
or more additional subspecies of G. gorilla
be recognized (Sarmiento and Butynski, 1996;
Sarmiento and Oates, 2000; Stumpf et al.,
2003). The proposal of two species within
Gorilla and Pongo (Groves, 2001) has not
gained wide acceptance, however, because
several molecular and morphological studies
show that relative variation within the tradi-
tional subspecies of Gorilla gorilla and Pongo
pygmaeus is high relative to that between the
subspecies (Courtenay et al., 1988; Gagneux
et al., 1999; Muir et al., 2000; Jensen-Seaman
et al., 2003; Leigh et al., 2003).
Even if we adhere to a conventional
taxonomy (e.g., Jenkins, 1990) while awaiting
final word on alternative taxonomic scenarios,
the diversity within and among great ape
species is impressive. If considered to
be a single species, both G. gorilla and
P. pygmaeus exhibit considerable variation
at the infraspecific level. Together with
the species of Pan and subspecies of
P. troglodytes, they provide comprehensive
data for models of inter- and intraspecific
variation, which can be applied to under-
standing variation in fossil hominins.
Great ape patterns of diversity are especially
relevant to discussions about identifying
subspecies from the hominin fossil record.
Subspecies are an unresolved quandary
in systematics and taxonomy. Defined as
geographically circumscribed, phenotypically
distinct units (Futuyma, 1986; Smith, et al.,
1997), or the point at which we no
longer lump populations (Groves, 1986),
subspecies are more difficult to identify
than species (Tattersall, 1986; Kimbel, 1991;
Shea et al., 1993; Templeton, 1999; Leigh
et al., 2003). As explained by Templeton
(1999), this is because the criterion of repro-
ductive isolation, which gives ontological
strength to the concept of a species, is
lacking for the subspecies. A subspecies, can
encompass everything from a population to a
species. Kimbel (1991), citing Mayr’s (1982)
view that subspecies cannot be treated as
incipient species, reasoned that subspecies
are an arbitrary tool of taxonomy and are
best ignored in paleoanthropology, being
a hindrance to the task of determining
the phylogeny of fossil hominins. Tattersall
(1986, 1991, 1993) demonstrated that only
a few subtle morphological features differ
between closely related species of Lemur.He
proposed that when morphological distinc-
tions are observed in the fossil record they
should be considered as evidence for species,
rather than lower-order taxonomic units. Thus
Neandertals and modern human are likely to
have been distinct species.
Evolutionary biologists who study the
nature of variation in extant primates argue,
to the contrary, that subspecies provide
important information about the structure
of the gene pool, and patterns of genetic
contact and evolutionary divergence among
12 Pilbrow
populations. From a neontologist’s
perspective these are meaningful aspects
of the population biology and history
of a species, vital enough to advocate
conservation status for endangered taxa
(Templeton, 1999; Leigh et al., 2003).
Several researchers have called for a need to
identify subspecies among extinct hominins
so as to gain a better appreciation of their
population dynamics (Shea et al., 1993; Jolly,
1993, 2001; Leigh et al, 2003). There is a
strong contingent of paleoanthropologists
who believe that Neandertals constitute
an extinct subspecies of modern humans
(Wolpoff et al., 2001). Disagreeing with
Tattersall (1991), Jolly (1993) argued that
Tattersall’s criteria for recognizing species
based on morphological distinction cannot be
applied to baboons because baboon popula-
tions achieve morphological distinctiveness
without reproductive isolation, and such
populations may never become extinct in
the same sense as phylogenetic species.
According to Jolly if craniodentally distinct
baboon taxa were identified as species in the
paleontological context, significant attributes
of their genetic structure, or zygostructure, as
he calls it, would be obscured. Lemurs and
baboons clearly have contrasting signatures
of genetic isolation relative to morphological
distinctiveness. Using baboons as models
Jolly (2001), in his turn, has suggested
that the interactions between hominins like
Neandertals and modern humans could have
involved a certain amount of interbreeding.
Although this implies that Neandertals are
a subspecies of modern humans, at least
according to the dominant biological species
concept, Jolly prefers to use the term allotaxa
(Grubb, 1999) to describe them, to avoid
the distinction between the biological and
phylogenetic species concepts, and to place
the emphasis on population history rather
than on naming names.
As phylogenetic kin, great ape patterns
of taxonomic diversity should have greater
relevance than lemur or baboon patterns
for addressing the question of infraspecific
diversity in extinct hominins. The purpose
of this paper is to document the appor-
tionment of molar variation among species
and subspecies of extant great apes. Phenetic
distances separating the traditional subspecies
of P. troglodytes are compared with those
separating P. troglodytes from P. paniscus.
These in turn are compared with the distances
between subspecies within G. gorilla and
P. pygmaeus. The patterns of dental diver-
gence are then compared with the patterns
derived from previous cranial, dental and
genetic studies to evaluate the relevance of
this material for understanding diversity and
as models for understanding variation in
fossil hominin molars. A close match with
the patterns revealed by selectively neutral
molecular data is particularly important
because this helps to determine whether dental
data can reveal patterns of genetic divergence
(Collard and Wood, 2000, 2001; Guy et al.,
2003; Lockwood et al., 2004).
Because great apes range considerably in
size and because G. gorilla and P. pygmaeus
also display marked sexual dimorphism, both
raw and size-adjusted measurements, and
sex-pooled and sex-regregated samples were
used in the analysis. The following specific
questions were addressed:
(1) Are molar metrics able to differentiate
between the four species of great apes?
(2) How successful are molar metrics
in classifying subspecies within each
species?
(3) How do inter- and intra-species dental
distances compare within and across
species?
(4) In what way do the phenetic distances
between taxa change when adjusted for
size and sex?
(5) How do dental patterns of divergence
compare with molecular patterns of
divergence?
Molar Variation in Great Apes 13
Several studies have previously examined the
nature of inter- and intraspecific morpho-
logical variation in the great apes. Shea et al.
(1993) were able to differentiate between
P. paniscus and P. troglodytes and between
subspecies of P. troglodytes using cranio-
metric data. Braga (1995) demonstrated the
differences between these same groups using a
larger data set of discrete cranial traits. Groves
(1967, 1970) and more recently, Stumpf et al.
(2003) used craniometric data to differentiate
between gorilla subspecies. Taylor and Groves
(2003) found that patterns of diversity based
on molecular data in the African apes were
discernible using mandibular measurements.
Guy et al. (2003) and Lockwood et al. (2004)
used a geometric morphometric analysis of the
craniofacial complex and the temporal bone
to differentiate among great ape subspecies.
Dental data have also been used to
address the question of variation in the
great apes (Mahler, 1973; Johanson, 1974;
Swindler, 1976, 2002; Kinzey, 1984; Scott
and Lockwood, 2004). The most significant
study from the perspective of the present one
is Uchida’s (1992, 1996). Uchida measured
molar cusp base areas on photographs of
the occlusal surface to examine patterns of
variation within and among great ape species
and subspecies. She demonstrated that molar
cusp areas successfully discriminate between
subspecies within great ape species. Her work
provided significant insight into the nature of
dental diversity in great apes.
This study borrows from Uchida’s in its
technique for taking measurements from the
occlusal surface of molars. It differs from it
and the others in several important respects.
Molar crest lengths, which have not been used
by previous studies, were used to test whether
other aspects of molar morphology are as
effective as cusp base areas or length/breadth
dimensions in differentiating taxa. In addition,
this study seeks to determine whether molar
metrics are capable of differentiating between
species and subspecies of great apes when
these are commingled. Consequently, nine
taxa, including P. paniscus, and the subspecies
of P. troglodytes, G. gorilla, and P. pygmaeus
were included in a single analysis. Most
importantly, in this study samples were drawn
from populations representing the entire range
of distribution for great apes. This provides an
understanding of how variation is partitioned
at infraspecific levels of the species, rather
than in select populations representing the
species, as has often been done. This provides
a comprehensive hierarchical model from
which to approach fossil hominin variation
(Albrecht et al., 2003; Miller et al., 2004).
Materials and Methods
The analysis is based on the unworn denti-
tions of 804 adult individuals, including 341
chimpanzees, 298 gorillas and 165 orangutans
(Table 1). Only individuals with third molars
erupted or erupting were selected. Samples
were obtained from major museums in the USA
and Europe taking care to select specimens
from the known geographic distribution of
the great apes (for locality information and
museum listings, see Pilbrow, 2003). Locality
data from museum records were verified
against the United States Geographic Names
Database and compared with previous museum
based studies (Groves, 1970; Röhrer-Ertl,
1984;Sheaetal.,1993;Braga,1995).Localities
Table 1. Sample sizes and sex proportions of taxa used in
this study
Taxon N M/F %
P. t. verus 64 47/53
P. t. troglodytes 152 46/54
P. t. schweinfurthii 79 52/48
P. paniscus 46 39/61
G. g. gorilla 208 66/34
G. g. graueri 60 67/33
G. g. beringei 30 52/48
P. p. pygmaeus 140 44/56
P. p. abelii 25 44/56
Total 804 53/47
14 Pilbrow
were then aggregated into the traditionally
recognized species and subspecies (Jenkins,
1990): P. paniscus, P. t. verus, P. t. troglodytes,
P. t. schweinfurthii, G. g. gorilla, G. g. graueri,
G. g. beringei, P. p. pygmaeus, P. p. abelii.
The overall sex ratios for the study
samples are fairly well balanced (Table 1).
However, within individual subgroups the
sexes are not distributed equally. There is
also an imbalance in the sample sizes repre-
senting each subspecies. This is because
museum collections are biased towards certain
locales and sexes. To provide an estimate
of overall variation I carried out separate
analyses on pooled and segregated sexes
to evaluate if and how variation differs
between them.
Molar dimensions were measured on a
digital image of the occlusal surface of the
molar. Measurements consisted of mesiodistal
and buccolingual dimensions and the length
of molar crests. The technique for taking
photographs is described in detail elsewhere
(Pilbrow, 2003; Bailey et al., 2004) and
will not be elaborated here. The mesiodistal
dimension was identified as the longest
dimension across the tooth crown. Two
buccolingual dimensions were taken at the
mesial and distal cusps and identified as the
widest dimensions across the tooth crown at
these points. To measure crest (or cristid)
lengths I first identified cusp boundaries using
the longitudinal, lingual and buccal devel-
opment grooves for the upper molar, and
the longitudinal, lingual, mesiobuccal and
distobuccal developmental grooves for the
lower molar (Figure 1). I then measured
crest lengths from cusp boundary to cusp
tip. If accessory cuspules were encountered
they were not included in the crest length
measurement. A total of 13 dimensions were
taken on the lower molar and 11 on the upper.
Figure 1 illustrates the measurements taken.
NIH Image, a public domain image analysis
program was used to take measurements
(http://rsb.info.nih.gov/nih-image/).
In an intra-observer error study using 23
gorilla molars and premolars, I found that
the average error in measuring the length
of the tooth on the actual specimen versus
measuring it on a digitized image was 1.36%
SD =053%range =012–276%.An
inter-observer study (Bailey et al., 2004)
was also undertaken, comparing cusp base
area measurements, on images obtained using
slightly different photographic techniques,
photo equipment and measurement software.
There were no statistically significant differ-
ences in the measurements taken by two
observers.
The dental measurements were size adjusted
by indexing each measurement against the
geometric mean of all measurements for that
tooth (Mosimann and James, 1979; Darroch
and Mosimann, 1985; James and McCulloch,
1990; Falsetti et al., 1993). Separate analyses
were performed using raw and size-adjusted
measurements. This helped to evaluate how
molar size contributes to subspecies differ-
ences. Because of missing teeth and differ-
ential wear patterns the sample sizes differed
for the molars. To maximize sample sizes the
data set was subdivided according to molar
type and separate analyses were performed
for each molar. The results from six molars
were then averaged to get an overall pattern
of molar differentiation. However, the role of
each molar in contributing to the differences
was also evaluated.
A step-wise discriminant analysis (SPSS
12.0) was used to see how accurately molar
morphometrics differentiate the nine taxa.
The percentage accuracy by which individuals
were classified helped to verify the precon-
ceived separation of the taxa. The loading
of the variables on the discriminant functions
helped determine which variables influenced
the differentiation. Group centroids were
used to calculate Mahalanobis distances or
squared generalized distances D2between
taxonomic pairs. This provided a phenetic
distance between groups. The F statistic was
Molar Variation in Great Apes 15
Figure 1. Measurements taken on digital image of upper molar (A, B) and lower molar (C, D). A, C:
1, Length; 2, breadth across mesial cusps; 3, breadth across distal cusps. B: 1, Preparacrista;
2, Postparacrista; 3, Premetacrista; 4, Postmetacrista; 5, Preprotocrista; 6, Postprotocrista;
7, Prehypocrista; 8, Posthypocrista. D: 1, Preprotocristid; 2, Postprotocristid; 3, Prehypoconidcristid;
4, Posthypoconidcristid; 5, Prehypconulidcristid; 6, Posthypoconulidcristid; 7, Premetaconidcristid;
8, Postmetaconidcristid; 9, Preentoconidcristid; 10, Postentoconidcristid. Both molars from the
collections of the Zoologische Staatssaammlung, Munich.
used to test for the significance of pair-
wise distances. The first two discriminant
functions, which most often accounted for a
large proportion of the variance, were used in
two-dimensional scatter-plots to show within-
group variance, as well as between-group
separation. Finally, the geometric mean was
used as a generalized size factor and Pearson’s
correlations between the scores on discrim-
inant functions and the geometric mean were
used to determine the role of size or allometry
in discriminating groups.
Results
Raw Variables
When raw variables were used in the
analysis the classification accuracy for the
nine taxa, averaged over six molars, was
around 70% (Table 2). Classification accuracy
was low for the UM3 (50%), but higher
for the other molars (63% to 73%, not
shown in table). Although the percentages of
cases correctly assigned did not differ much
between the sexes, classification accuracy
was lower when the sexes were combined.
Mayr (1942) suggested that if at least 75%
of individuals within populations can be
accurately differentiated from other popula-
tions within the species, these intraspecific
groups may be described as subspecies.
Classification accuracy was highest for
P. paniscus, the only taxon not differen-
tiated at the subspecies level, indicating that
this taxon is most distinct from the others.
Accuracy of classification was lowest for
P. t. troglodytes and P. t. schweinfurthii. The
16 Pilbrow
Table 2. Average classification accuracy using raw variables
Ptv Ptt Pts Pp Ggg Gggr Ggb Ppp Ppa
Sex combined
P. t. verus 70 10 16 3 0 0 0 0 1
P. t. troglodytes 16 47 24 10 0 0 0 1 2
P. t. schweinfurthii 17 16 56 900001
P. paniscus 33390 00001
G. g. gorilla 000072 916 2 1
G. g. graueri 00009 71 19 0 0
G. g. beringei 0 0 0 0 19 20 59 10
P. p. pygmaeus 322030166 23
P. p. abelii 04111011676
Average classification accuracy: 67%
Males
P. t. verus 72 11 15 1 0 0 0 0 1
P. t. troglodytes 18 48 24900001
P. t. schweinfurthii 17 17 56 900001
P. paniscus 34390 00000
G. g. gorilla 000078 710 3 2
G. g. graueri 00005 74 20 1 0
G. g. beringei 0 0 0 0 16 22 62 00
P. p. pygmaeus 120020172 22
P. p. abelii 04000002571
Average classification accuracy: 69%
Females
P. t. verus 70 14 12 2 0 0 0 1 1
P. t. troglodytes 17 52 18 11 0 0 0 1 1
P. t. schweinfurthii 12 16 59 1100020
P. paniscus 35191 00000
G. g. gorilla 000077 614 2 1
G. g. graueri 00005 83 12 0 0
G. g. beringei 0 0 0 0 14 21 65 00
P. p. pygmaeus 311020170 22
P. p. abelii 13010002075
Average classification accuracy: 71%
Cross-matrix shows percentage of cases of taxa from column one classified into taxa from row one. Correct classifications.
rate of misclassifications, which can be deter-
mined by noting the percentage of cases from
each taxon in the row, classified into one
or the other taxa from each column, demon-
strates that misclassified subspecies are likely
to be assigned to other subspecies within the
same species. Average classification accuracy
for the four great ape species, calculated
by summing the percentage accuracy for the
subspecies within each species, was close to
95%. This demonstrates that there is greater
overlap among subspecies than species, which
fits with the understanding that the gene pools
of subspecies are not as completely segregated
as that of species, allowing greater genetic
exchange. It should be remembered, of course,
that apart from P. paniscus and P. troglodytes,
the great ape species are considered to be
distinct genera.
Mahalanobis distances (Table 3) show that
the greatest distance is between the smallest
(P. paniscus) and the largest (G. g. graueri)
of the great apes. The distances between
subspecies of P. troglodytes and those of
G. gorilla are only marginally lower. Inter-
mediate distances separate the subspecies of
Molar Variation in Great Apes 17
Table 3. Average mahalanobis distances using raw variables
Ptv Ptt Pts Pp Ggg Gggr Ggb Ppp Ppa
Sex combined
P. t. verus 000 ∗∗∗∗∗∗∗
P. t. troglodytes 182 000 ∗∗∗∗∗∗
P. t. schweinfurthii 159 081 000 ∗∗∗∗∗∗
P. paniscus 556 306 422 000 ∗∗∗∗∗
G. g. gorilla 4251 4494 4206 6304 000 ∗∗∗∗
G. g. graueri 6956 7290 6970 9588 739 000 ∗∗∗
G. g. beringei 5598 5934 5644 8062 392 293 000 ∗∗
P. p. pygmaeus 1324 1311 1360 2382 1912 3816 2673 000 ∗
P. p. abelii 1509 1440 1439 2477 1883 4011 2834 299 000
Males
P. t. verus 000 ∗∗∗∗∗∗
P. t. troglodytes 161 000 ∗∗∗∗∗∗
P. t. schweinfurthii 152 081 000 ∗∗∗∗∗∗
P. paniscus 623 386 452 000 ∗∗∗∗∗
G. g. gorilla 4634 4869 4741 7053 000 ∗∗∗∗
G. g. graueri 7986 8283 8173 11120 858 000 ∗∗
G. g. beringei 6787 7128 6996 9802 536 292 000 ∗∗
P. p. pygmaeus 1676 1714 1833 3111 1710 3805 2886 000
P. p. abelii 1751 1719 1759 2944 1752 4147 3121 329 000
Females
P. t. verus 000 ∗∗∗∗∗∗∗∗
P. t. troglodytes 246 000 ∗∗∗∗∗∗
P. t. schweinfurthii 221 124 000 ∗∗∗∗∗∗
P. paniscus 749 392 571 000 ∗∗∗∗∗
G. g. gorilla 5405 5977 5394 8469 000 ∗∗∗∗
G. g. graueri 8799 9665 8986 12886 961 000 ∗∗∗
G. g. beringei 6680 7403 6832 10302 504 412 000 ∗∗
P. p. pygmaeus 1433 1500 1475 2935 2591 5070 3310 000 ∗
P. p. abelii 1560 1596 1509 2949 2714 5535 3740 355 000
Asterisk above the diagonal indicates that pair wise distances are significant p<005.
P. pygmaeus from other groups. Subspecies
withinaspeciesdisplaythelowestMahalanobis
distances, which are not always statistically
significant (Table 3).
Distances between subspecies within
species differ markedly among the great
apes. The distances separating subspecies
of P. troglodytes are remarkably low.
Pan troglodytes verus is separated by a
greater distance from P. t. troglodytes and
P. t. schweinfurthii than either of these are
from each other. Distances that are three
to four times greater separate Pan paniscus
from the subspecies of P. troglodytes. Of the
P. troglodytes subspecies, P. t. troglodytes
is closest to P. paniscus. The two subspecies of
P. pygmaeus are separated by a greater distance
than are the subspecies of P. troglodytes.
However, the distance is less than that
separating P. paniscus from the subspecies
of P. troglodytes. Relative to the other
great apes, the subspecies of G. gorilla are
separated by large distances. The Mahalanobis
distances between the western gorilla
(G. g. gorilla) and the eastern gorillas (G.
g. graueri and G. g. beringei) are even
greater than that separating P. paniscus from
P. troglodytes.
Using the lower second molar (LM2) as
an example, Figure 2 shows the distribution
18 Pilbrow
Figure 2. Dispersion of great apes along first two discriminant functions in the analysis of raw
measurements on the LM2. Sexes combined.
of the taxa along the first two discriminant
functions of sex-combined samples. The LM2
lies in the middle of the range for classifi-
catory result for the molars, and therefore,
for consistency, the LM2 is used to illustrate
the spread of data in scatter plots throughout
this paper. The four species are enclosed
by 95% confidence ellipsoids, which helps
to illustrate the distribution of the species
and the relative variance within each. The
major separation along function 1, with non-
overlapping distributions, is between gorillas
at one end of the axis and chimpanzees,
including bonobos, at the other end.
P. pygmaeus is separated from the others
along function 1 and 2 although it shows
slight overlap with P. troglodytes.P. paniscus
is tightly clustered with P. troglodytes
showing greater overlap. The relative size
of the confidence ellipsoids indicates that P.
paniscus and P. troglodytes are characterized
by lower overall variance in raw molar metrics
than P. pygmaeus and G. gorilla. Gorilla
molars display comparatively higher variance.
Size strongly influences the segregation
when raw molar metrics are used. The first
discriminant function accounts for 85% to
90% of the variance, depending on molar
position. Pearson’s correlations between the
discriminant scores for this function and the
geometric mean, a proxy for size, are between
0.89 and 0.95, p <001. The second function
accounts for 5–10% of the variance, but has
lower correlations with the geometric mean
(0.01 to 0.26, not always significant).
Figure 3 presents the same data as Figure 2
but males and females of the four species
are separated and encircled by 95% confi-
dence ellipsoids. The close clusters formed
by P. paniscus and P. troglodytes, which
are not dimorphic in molar dimensions,
contrast strongly with the larger cloud of data
points formed by the dimorphic orangutans
and gorillas. The larger ellipses surrounding
gorilla and orangutan males suggests that the
variance in males is greater than in females.
Even among the dimorphic apes though, there
is greater overlap between male and female
data points of the same species than between
Molar Variation in Great Apes 19
Figure 3. Spread of data along functions 1 and 2 of the LM2 using raw variables. 95% confidence
ellipsoids surrounding males and females of P. paniscus, P. troglodytes, P. pygmaeus and G. gorilla.
females of a larger species and males of
a smaller one. At all molar positions the
variables most strongly affecting the segre-
gation are the length and the breadth at mesial
and distal cusps.
Size Adjusted Variables
Since size appears to be a strong factor
affecting separation, the results of the size-
adjusted discriminant analyses are instructive.
Table 4 shows the percentage of cases
correctly classified using size-adjusted dimen-
sions averaged over all molars. Once again,
accuracy is higher with sex-segregated molars
but only slightly. Accuracy is lower for the
UM3 (44%) than the other molars (50%
to 61%, not shown). The correct classifi-
cation is reduced by about 14% after size
adjustment. Of all groups the decrease in
accuracy is greatest for P. paniscus (Table 4).
In this taxon, which had the highest accuracy
using raw variables, classification accuracy is
about 35% lower when scaled variables are
employed. Misclassified individuals are likely
to be assigned to other subspecies within the
same species, with P. paniscus individuals
assigned to subspecies of P. troglodytes. The
percentage accuracy for species groups is
about 80%; 15% lower than in the previous
analysis.
Table 5 shows the size-adjusted
Mahalanobis distances averaged over all
molars for combined and sex-separated
samples. After size adjustment, the distance
between Pan and Gorilla is reduced consid-
erably. However, in this analysis the greatest
overall distance is between P. t. verus and
G. g. graueri. Distances among subspecies
within species also differ with size-scaled
data. All inter-group distances are lower,
except for the distances between subspecies
20 Pilbrow
Table 4. Average classificatory accuracy using shape variables
Ptv Ptt Pts Pp Ggg Gggr Ggb Ppp Ppa
Sex combined
P. t. verus 61 9151000131
P. t. troglodytes 17 34 22 13 1 2 2 7 3
P. t. schweinfurthii 14 13 54 1022222
P. paniscus 13 12 10 56 20141
G. g. gorilla 112161 14 14 3 4
G. g. graueri 111211 63 19 2 1
G. g. beringei 0 2 3 1 14 23 52 23
P. p. pygmaeus 374433353 20
P. p. abelii 04139241563
Average classification accuracy: 55%
Males
P. t. verus 65 813901211
P. t. troglodytes 16 37 25 11 0 2 2 6 1
P. t. schweinfurthii 14 14 54 722133
P. paniscus 12 11 11 54 41151
G. g. gorilla 102257 13 16 3 6
G. g. graueri 000310 60 20 3 4
G. g. beringei 1 0 2 1 13 17 59 34
P. p. pygmaeus 274524255 19
P. p. abelii 02033362360
Average classification accuracy: 56%
Females
P. t. verus 63 10 12 10 0 0 1 3 1
P. t. troglodytes 12 36 21 15 1 2 2 7 4
P. t. schweinfurthii 14 13 50 1223132
P. paniscus 12 17 9 56 00330
G. g. gorilla 102169 12 11 2 2
G. g. graueri 011016 65 15 1 1
G. g. beringei 0 3 3 2 13 20 53 33
P. p. pygmaeus 363434454 19
P. p. abelii 13335141070
Average classification accuracy: 57%
Cross-matrix shows the percentage of cases for taxa in the first column classified into taxa in the first row. Accurate classifications in bold.
of P. troglodytes, which are higher following
size adjustment. The overall distance between
P. t. verus and the other two subspecies is
similar to their distance from P. paniscus.
In lower molars P. t. verus was placed
further away from the central and east
African chimpanzees than P. paniscus.
The distances between subspecies of P.
troglodytes and P. paniscus although reduced
are statistically significant for the analyses
with combined sexes and females, but not
with males.
As was the case with Pan, distances
among subspecies of G. gorilla are also
reduced, so that the distances between east
and west African gorillas are now lower
than the distances separating P. paniscus
from P. troglodytes subspecies. However, they
are still higher than the distances separating
P. troglodytes subspecies. Similarly, the
relative distance among gorilla subspecies
does not change after size adjustment. Eastern
gorilla subspecies are closely clustered, well
separated from the western subspecies.
Molar Variation in Great Apes 21
Table 5. Mahalanobis distances using shape variables
Ptv Ptt Pts Pp Ggg Gggr Ggb Ppp Ppa
Sex combined
P. t. verus 000 ∗∗∗∗∗∗∗∗
P. t. troglodytes 214 000 ∗∗∗∗∗∗∗
P. t. schweinfurthii 225 110 000 ∗∗∗∗∗∗
P. paniscus 324 209 253 000 ∗∗∗∗∗
G. g. gorilla 1326 1076 1022 1265 000 ∗∗∗
G. g. graueri 1360 1060 1063 1199 231 000 ∗∗
G. g. beringei 1244 999 999 1205 194 138 000 ∗∗
P. p. pygmaeus 915 628 828 834 907 1057 836 000 ∗
P. p. abelii 1157 860 972 1122 787 1093 848 265 000
Males
P. t. verus 000 ∗∗ ∗∗∗∗∗
P. t. troglodytes 244 000 ∗ ∗∗∗∗∗
P. t. schweinfurthii 241 122 000 ∗∗∗∗∗
P. paniscus 407 234 313 000 ∗∗∗∗∗
G. g. gorilla 1378 1148 1074 1239 000 ∗∗∗
G. g. graueri 1513 1183 1182 1234 261 ∗∗∗
G. g. beringei 1400 1171 1132 1270 243 197 ∗∗∗
P. p. pygmaeus 1042 743 935 890 910 1095 946 ∗
P. p. abelii 1410 1094 1183 1284 732 1018 815 364 ∗
Females
P. t. verus 000 ∗∗∗∗∗∗
P. t. troglodytes 275 000 ∗∗∗∗∗∗
P. t. schweinfurthii 299 130 000 ∗∗∗∗∗∗
P. paniscus 346 240 268 000 ∗∗∗∗∗
G. g. gorilla 1456 1135 1073 1362 000 ∗∗∗
G. g. graueri 1308 1022 1028 1191 271 000 ∗∗
G. g. beringei 1235 957 978 1198 258 155 000 ∗∗
P. p. pygmaeus 920 602 810 867 1004 1047 831 000
P. p. abelii 1117 790 928 1110 955 1209 994 327 000
Asterisk above diagonal shows statistically significant p<005pair-wise distances.
D2values between P. p. pygmaeus and
P. p. abelii do not change substantially
after size adjustment. Mahalanobis distances
are higher between Bornean and Sumatran
orangutans than the intraspecific distances
within P. troglodytes and G. gorilla.
The first discriminant function after size
adjustment accounts for a lower proportion of
variance than before. The variance explained
by the first function is between 55% to 73%
in sex-pooled samples, depending on molar
position. In contrast, the second function
accounts for a higher proportion of the
variance, increasing from 20% to 28%,
depending on the molar position. With this
change in the distribution of variance, the
separation of taxa is not as distinct on the
first axis as it was before. Figure 4 illustrates
the spread of data on the first two functions
of the LM2 using size-adjusted measures on
sex-pooled samples. Compared to the non-
size-adjusted data (Figure 2) the first axis
now separates P. troglodytes from G. gorilla,
with P. paniscus occupying an intermediate
position. Whereas in the analysis of raw data
P. troglodytes and P. paniscus formed small,
tight clusters, size-adjusted data reveals that
the ellipsoid surrounding P. troglodytes is
similar in size to that of G. gorilla. P. paniscus
is virtually confined within P. troglodytes,
22 Pilbrow
Figure 4. Distribution of great ape subspecies along first two discriminant functions in the analysis
using shape variables on the LM2. Sexes combined.
which is also reflected in the low classi-
fication accuracy. The second function
separates G. g. gorilla from G. g. graueri
and G. g. beringei and the subspecies of
P. pygmaeus from the other great apes,
although there is considerable overlap. Size-
adjusted P. pygmaeus data display greater
variance than the other groups.
Pearson’s correlations of the discriminant
scores on function 1 with the geometric mean
are negative but quite high on all size-adjusted
molars (−063 to −074p<001), except
the LM1 and LM2 (−037 and −049, respec-
tively, p <001). The variance explained
by the first function is also lower in these
two molars (57% and 55%), although the
classification accuracy is relatively high (56%
and 58%, respectively). The second function
has a fairly strong correlation with tooth size
in the LM1 and LM2 (0.43 and 0.38, respec-
tively, p <001), but not in other molars.
This suggests that a size component is still
preserved in great ape taxa even after the
effect of isometric size is reduced; for the LM1
and LM2 this effect is seen in the first two
functions.
In Figure 5 the four great ape species are
identified by sex in a scatter plot of the
first two discriminant functions of size-scaled
variables in the LM2. In contrast with the
scatter using raw data (Figure 3), the sexes
overlap considerably in their distributions.
This is because the size difference between
male and female orangutans and gorillas is
adjusted, contributing to a reduction in overall
variance in these taxa compared to Pan.
Molar Variation in Great Apes 23
Figure 5. Distribution of sexes of great ape species along factor 1 and 2 of the LM2 using shape data.
Ellipsoids not displayed because of extensive overlap between sexes.
After size adjustment, the variables most
responsible for causing separation among
groups are the length of the preprotocrista on
the upper molars and the lengths of postpro-
tocristid and prehypocristid on the LM1 and
LM2. The mesiodistal length of the molars
also has a fairly strong correlation with
function 1 in all analyses.
Discussion
The results of this study open up the following
questions for further discussion: how reliable
are molar metrics for recognizing patterns of
diversity and taxonomy in the great apes?
How do size differences among the great apes
affect molar discrimination and patterns of
variation? How does sexual dimorphism and
shape affect the nature of molar variation
among great apes? Can great ape models
be used to understand the taxonomy of
fossil hominins? What inferences can we
draw from great ape patterns of intraspe-
cific variation for recognizing subspecies
among fossil hominins? Each of these is
addressed below.
Concordance Between Dental and
Molecular Patterns of Divergence
A primary conclusion of this study is
that molar metrics are able to discrim-
inate between Pan, Gorilla and Pongo,
and their constituent species, P. paniscus,
P. troglodytes, G. gorilla and P. pygmaeus.
These are well-established taxa with very
little debate regarding their taxonomic status.
Regarding subspecies, P. t. verus is clearly
distinguishable from P. t. troglodytes and
P. t. schweinfurthii, but the latter two are not
as easily differentiated from each other. All
three G. gorilla subspecies are easily recog-
nizable and G. g. gorilla is well differen-
tiated from G. g. graueri and G. g. beringei.
24 Pilbrow
Finally, P. p. pygmaeus is well differen-
tiated from P. p. abelii. These patterns
of differentiation deviate slightly from the
traditionally understood taxonomy, in the
separation of P. t. verus from the other
subspecies of P. troglodytes, and in the clear
distinction between east and west African
gorillas. However, they are consistent with
the conclusions of several recent molecular
studies (Morin et al., 1994; Ruvolo et al.,
1994; Garner and Ryder, 1996; Saltonstall
et al., 1998; Gagneux et al., 1999; Jensen-
Seaman and Kidd, 2001). Since molecular
data ideally provide accurate information
regarding patterns of divergence (Collard and
Wood, 2000, 2001), a close match between the
infraspecific measures of divergence revealed
by this study and those of molecular studies
suggests that molar metrics are successful
at revealing taxonomically relevant patterns
of divergence in great apes. Similar results
obtained by several morphological studies
(Braga, 1995; Uchida, 1996; Guy et al., 2003;
Stumpf et al., 2003; Taylor and Groves,
2003; Lockwood et al.; 2004) support this
conclusion.
The Role of Size
Molar size plays a significant role in classi-
fying great ape taxa in this study. This is
seen in the high classification accuracy when
raw variables are used and a strong corre-
lation between the most significant discrim-
inant functions and the geometric mean in
these analyses. The importance of molar
size in driving dispersion is most evident
in the loss of the distinctive status of
P. paniscus following size adjustment. The
distance between gorilla subspecies is also
reduced, so that the distance between the
eastern and western gorillas is no longer
greater than the distance between P. paniscus
and P. troglodytes. Furthermore, following
size adjustment gorilla and orangutan male
and female molars do not form discrete
clusters. Previous multivariate studies of the
cranium (Shea et al., 1993) and mandible
(Taylor and Groves, 2003) have reported
a reduction in classification accuracy for
P. paniscus and subspecies of G. gorilla
following size correction. To explain their
finding, Shea et al. (1993) suggested that
overall size difference is an important
criterion by which morphological differ-
entiation is achieved. They argued (Shea
et al., 1993: 278) that “size and allometric
effects are not merely troublesome obfus-
cation to be removed to yield a clearer
view of “true” biological or phyloge-
netic distinctions; indeed size differences
have so many pervasive effects on myriad
biological systems and levels (e.g., Lindstedt
and Calder, 1981) that they should be
central to systematic conclusions in and of
themselves.”
Indeed, size appears to be an important
element in the niche partitioning and adaptive
divergence of the African apes. Sympatric
chimpanzees and gorillas differ in their
fallback food preferences (fruits versus
leaves), habitat preferences (greater versus
lesser arboreality), and social systems (multi-
male versus single-male groups), all of
which are associated with size differences
in primates. Size related differences between
P. pansicus and P. troglodytes are achieved
and maintained through ontogenetic scaling
(Shea 1981; 1983a, b, c; 1984). However, the
differences are not proportional in all body
systems: the face and teeth of P. paniscus
are relatively smaller whereas the hindlimbs
are relatively longer than P. troglodytes
(Jungers and Susman, 1984; Morbeck and
Zihlman, 1989). Also, as pointed out by
Shea et al. (1993), size differences do not
exhaust the differences between African apes;
they differ quite significantly in size-adjusted
morphological systems (Taylor and Groves,
2003; Guy et al., 2003; Lockwood et al.,
2004), was as found in this study.
Molar Variation in Great Apes 25
The Nature and Pattern of Molar
Variation
The contrast between size-preserved and size-
adjusted analyses provides important insights
into the nature and pattern of molar variation
in great apes. For example, size-preserved
molars demonstrate how sexual dimorphism
affects dispersion. It was seen that the highly
dimorphic gorillas and orangutans, especially
the males in these taxa, display higher molar
size variance, which increased the divergence
between the species. When molar size was
adjusted by indexing all variables against the
geometric mean, the size difference between
males and females was reduced, as was the
overall variance in gorillas. This resulted in
similar levels of variance among chimpanzees
and gorillas.
It is noteworthy, however, that despite these
high levels of molar sex-dimorphism, the
males and females of each species clustered
together in size-preserved analyses (Figure 3).
Even in the dimorphic gorillas and orangutans
there was greater overlap between males and
females than between members of distinct
species. This suggests that in fossil hominin
molars where sex attribution and level of
dimorphism is unknown, variation due to
sexual dimorphism is likely to produce a
pattern of two overlapping distributions with
a bimodal peak rather than two mostly non-
overlapping distributions.
Size-adjusted analyses also help to highlight
the finding that subspecies of P. troglodytes
show greater distinction when shape data
are used. The distances among subspecies of
P. troglodytes were greater, comparable to
those among subspecies of G. gorilla and
P. pygmaeus. In particular, the separation
of P. t. verus from P. t. troglodytes and
P. t. schweinfurthii was heightened, so that
in some analyses it was more divergent than
P. paniscus.
A comparison of raw and shape analyses
also revealed that molar variance in
orangutans was affected by shape as well as
size. Although shape data reduced the size
difference between sexes it did not result in a
substantial change in inter-subspecies distance
or overall variance in P. pygmaeus.
A final point worth noting in this regard
is that even after size conversion there
was clear separation between east and
west African gorillas, lowland and mountain
gorillas, west African chimpanzees and central
and east African chimpanzees, and Bornean
and Sumatran orangutans. Strong correlations
with the geometric mean even in the size-
adjusted analyses indicates that allometric
size is important in distinguishing between
these taxa.
Great Apes as Models
for Hominin Variation
A major finding that emerges from this
study is that no single pattern can be used
to characterize great ape molar variation.
Isometric size, sexual dimorphism and
allometric size all contribute to variable
extents. Several previous studies have under-
scored the disparity in great apes in patterns of
variation and the signals revealed by different
body systems (Shea and Coolidge, 1988;
Ruvolo et al., 1994; Gagneux et al., 1999;
Ackermann, 2002; Taylor and Groves, 2003;
Guy et al., 2003; Schaefer et al., 2004; Scott
and Lockwood, 2004; Taylor, 2006).
A related finding, that chimpanzees are
characterized by low variance in dental
metrics, so that the distances between
P. paniscus and P. troglodytes are similar or
lower than the distances between subspecies
of gorillas and orangutans, also fits with
existing studies. Shea and Coolidge, 1988;
Ruvolo et al., 1994; Guy et al., 2003; Taylor
and Groves, 2003; Lockwood et al., 2004.
Ruvolo et al. (1994) used their findings
to advocate for a species level distinction
between east and west African gorillas.
Jolly et al. (1995), however, pointed out
26 Pilbrow
that it is problematic to use the distance
from one taxon to judge the distance from
another since no consistent standards of
distances can be established among verte-
brates. This study reveals that chimpanzees
and gorillas are characterized by funda-
mental differences in patterns of variation,
in particular in sexual dimorphism, which
affects the magnitude of variance within and
among subgroups (see also, Leigh et al.,
2003; Pilbrow and Bailey, 2005). Previous
studies have alluded to the different biogeo-
graphic and evolutionary histories of the
great apes resulting in differential patterns
of ontogenetic development, mating behaviors
and migratory patterns (Shea and Coolidge,
1988; Kingdon, 1989; Goldberg, 1998; Leigh
and Shea, 1995; Gagneux et al., 1999;
Leigh et al., 2003).
Corroboration with other studies helps to
strengthen the conclusions of this study, that
great ape patterns of dental variation cannot
easily be applied to understanding patterns of
molar variation in extinct hominin groups (see
also Ackermann, 2002; Taylor, 2006). My
results suggest that chimpanzees will provide
a different interpretation of hominin patterns
of molar variation compared to gorillas and
orangutans. Specifically, chimpanzee molars
could falsify a single species hypothesis
where gorilla and orangutan molars may not.
Greater Mahalanobis distance between two
subspecies of G. gorilla than between P.
troglodytes and P. paniscus in this study
suggests that even if both chimpanzee species
are combined, so as to model combined-
species ranges of variation, they could falsify a
single-species hypothesis in a situation where
gorillas may uphold it. The right choice of
model could be a difficult question here. A
greater reliance on the chimpanzee interpre-
tation may be justified by the closer phylo-
genetic affinity between chimpanzees and
hominins. However, this must be based on
the assumption that the pattern and magnitude
of variance in chimpanzees and hominins is
similar, and is part of their shared evolutionary
history. Given that closely related great apes
differ so remarkably in their patterns of
molar variation, this may not be a valid
assumption.
On the other hand, chimpanzee molars
may not falsify the converse null hypothesis,
namely that variation in a fossil hominin
assemblage is more than can be accommo-
dated within a single species, when gorilla
and orangutan molars may. If combined
chimpanzee species are still unable to falsify
this hypothesis, the likelihood that more than
one species is present is likely to become
more robust.
The different interpretations for hominin
taxonomy provided by chimpanzees and
gorillas are illustrated by the analysis of
Harvati et al. (2004). In their study pair-wise
distances between Neandertal and modern
human populations were compared with
distances between subspecies in twelve species
of Old World monkeys and apes. The distances
between G. g. gorilla and G. g. beringei were
not significantly smaller than the distances
between Neandertals and modern humans, but
thedistancesbetweenP.troglodytessubspecies
and P. paniscus were, as were the distances
between the subspecies of Old World monkeys.
Harvatietal.(2004)concludedthatNeandertals
and modern humans constitute distinct species,
and suggested that the gorilla subspecies might
be distinct species as also advocated by Groves
(2001). However, based on the present study,
it appears that their results were influenced by
the differing patterns of variation in the African
apes, whereby low variation within and among
chimpanzee species could falsify the single
species hypothesis but higher variation among
gorilla subspecies could not (see also Shea and
Coolidge, (1988). Their reliance on a wider
comparative sample of Old World monkeys
upholds their conclusion, nonetheless, and
suggests that a wider comparative database
from which to model hominin variation is a
likely solution to the use of phylogenetically
Molar Variation in Great Apes 27
affiliated great ape species. Thus baboons may
be used as ecological models along with the
great apes (Jolly, 2001).
Implications for Differentiating Subspecies
Among Fossil Hominins
Finally, the hierarchical levels of differen-
tiation evident among great ape molars in
this study has several implications for the
ability to recognize subspecies among fossil
hominins. Discrimination among great ape
genera is better than among species, which in
turn is better than among subspecies (Tables 2
and 4). Subspecies have the lowest classifi-
cation accuracy, but they are distinguishable
even when shape data are used. The distances
among subspecies within P. troglodytes,
G. gorilla, and P. pygmaeus are not compa-
rable, but that is to be expected because
subspecies, by nature, are transitory units,
documenting the fluid nature of population
dynamics. Nevertheless, they provide infor-
mation about historical processes under-
lying diversification. These results suggest a
potential for similar levels of diversity to
be recognized among extinct hominins. In
light of the foregoing discussion, however,
where subspecies of one great ape group
are equivalent to species of another, no
firm taxonomic standards can be set for
identifying subspecies, or species, of fossil
hominins. It might be preferable then to
use terms such as operational taxonomic
units or allotaxa (Grubb, 1999) to describe
distinct units from the hominin fossil context,
Jolly (2001). Especially in the paleoanthropo-
logical context, however, direct corroborating
evidence for population-level interactions is
lacking. Our interpretations of these interac-
tions, including diversification, reticulation,
genetic exchange and isolation are neces-
sarily based on how we designate popula-
tions, in taxonomic terms based on criteria
we establish from the comparative context. I
would suggest, based on the results of this
study, that paleoanthropologists are likely to
have greater success in identifying and appre-
ciating intraspecific variation if the compar-
ative taxa are sorted into infraspecific units
such as subspecies. This will enable them to
contrast interspecific patterns of variation with
those that are observed within the species. The
present study provides such a dataset.
Conclusions
Great apes vary substantially in the magnitude
and patterns of variation in molar occlusal
morphology. Size, shape, and sexual dimor-
phism are among the factors that affect the
nature of variation. Molar dimensions and
crest lengths provide clear signals of diver-
sification among subspecies within species,
among species and among genera of great
apes. These hierarchical levels can be differ-
entiated. The patterns of diversification match
those of molecular studies. In general, sex-
segregated samples provide clearer discrim-
ination than sex-pooled samples, but the
difference between the two is negligible.
Despite high sexual dimorphism sexes within
species cluster together. This indicates that
in mixed species samples of fossil hominin
molars of unknown sex variation due to
sexual dimorphism is unlikely to be confused
with interspecific variation. The difference
between great apes in the patterns of molar
variation signals caution when applying these
models for studying alpha taxonomy in
fossil hominins. The choice of model will
affect interpretations regarding the number of
species and patterns of variation proposed for
the fossils. Keeping this caveat in mind, the
great apes provide an exhaustive database of
patterns and ranges of molar variation in taxa
closely related to the fossil hominins. When
used in combination with other ecological
models they provide an excellent under-
standing of species-level diversification, but
more importantly of patterns of intraspecific
28 Pilbrow
variation, which provide an understanding of
the processes of interaction among hominins
Acknowledgments
Heartfelt thanks to Shara Bailey and Jean-
Jacques Hublin for inviting me to attend the
dental anthropology conference in Leipzig. I
really enjoyed the conference, especially the
efficiency and hospitality of the Max Planck
Institute. For this I should also thank Silke
Streiber, Diana Carstens, Allison Cleveland,
Michelle Hänel and Jörn Scheller. An extra
thanks goes to Shara for going beyond the call
of duty to organize my consular visit to Berlin
while I was there.
For help with this paper I would like
to thank the museums that made their
collections available for study: American
Museum of Natural History, NY; Anthropol-
ogisches Institüt und Museum der Univer-
sität Zürich-Irchel, Zürich; British Museum
of Natural History, London; Field Museum
of Natural History, Chicago; Museum of
Comparative Zoology, Harvard; Muséum
National d’Histoire Naturelle, Paris; Powell-
Cotton Museum, Kent; Peabody Museum
of Anthropology, Harvard; United States
National Museum, Washington, D.C.; Musée
Royal de l’Afrique Centrale, Tervuren; Zoolo-
gisches Museum, Berlin; Anthropologische
und Zoologische Staassammlung, Münich.
The project was funded by grants from the
LSB Leakey Foundation, National Science
Foundation (SBR-9815546), and the Wenner-
Gren Foundation. The paper benefited greatly
from comments by an anonymous reviewer,
Leslea Hlusko, and Shara Bailey. I thank them
for that.
Note
1. The terms hominin and hominid used here follow
Wood and Richmond (2000). However, when
referring to chimpanzees, gorillas and orangutans
together, the conventional term great apes is used.
References
Ackermann, R.R., 2002. Patterns of covariation in the
hominoid craniofacial skeleton: implications for
paleoanthropological models. Journal of Human
Evolution 42, 167–187.
Albrecht, G.H., Gelvin, B.R., Miller, J.M.A., 2003. The
hierarchy of intraspecific craniometric variation
in gorillas: a population-thinking approach
with implications for fossil species recognition
studies. In: Taylor, A.B., Goldsmith, M.L.
(Eds.), Gorilla Biology. Cambridge University
Press, Cambridge, pp. 62–103.
Bailey, S.E., Pilbrow, V.C., Wood, B.A., 2004. Inter-
observer error involved in independent attempts
to measure cusp base areas of Pan M1s. Journal
of Anatomy 205, 323–331.
Begun, D.R., 1992. Miocene fossil hominids and the
chimp-human clade. Science 257, 1929–1933.
Braga, J.C., 1995. Définition de certains caractères
discrets crâniens chez Pongo, Gorilla, et Pan.
Perspectives taxonomiques et phylogénétiques.
Ph.D. Dissertation, University of Bordeaux.
Caccone, A., Powell, J.R., 1989. DNA divergence
among hominoids. Evolution 43, 925–942.
Collard, M., Wood, B., 2000. How reliable are human
phylogenetic hypothesis? Proceedings of the
National Academy of Sciences of the USA 97,
5003–5006.
Collard, M., Wood, B., 2001. Homoplasy and the early
hominid masticatory system: inferences from
analyses of extant hominoids and papionins.
Journal of Human Evolution 41, 167–94.
Coolidge, H.J., 1929. A revision of the genus Gorilla.
Memoirs of the Museum of Comparative
Zoology Harvard 50, 293–381.
Coolidge, H.J., 1933. Pan paniscus: Pygmy
chimpanzee from south of the Congo River.
American Journal of Physical Anthropology
18, 1–57.
Courtenay, J., Groves, C.P., Andrews, P., 1988. Inter–
or intra-island variation? An assessment of
the differences between Bornean and Sumatran
orang-utans. In: Schwartz, J.H. (Ed.), Orangutan
Biology. Oxford University Press, Oxford,
pp. 19–29.
Darroch, J.N., Mosimann, J.E., 1985. Canonical and
principal components of shape. Biometrika 72,
241–252.
Falsetti, A.B., Jungers, W.L., Cole III, T.M., 1993.
Morphometrics of the callitrichid forelimb: a
case study in size and shape. International
Journal of Primatology 14, 551–572.
Molar Variation in Great Apes 29
Futuyma, D.J., 1986. Evolutionary Biology. Sinauer
Associates, Inc., Sunderland, MA
Gagneux, P., Wills, C., Gerloff, U., Tautz, D.,
Morin, P.A., Boesch, C., Fruth, B., Hohmann, G.,
Ryder, O., Woodruff, D.S., 1999. Mitochon-
drial sequences show diverse evolutionary
histories of African hominoids. Proceedings of
the National Academy of Sciences, USA 96,
5077–5082.
Garner, K.J., Ryder, O.A., 1996. Mitochondrial DNA
diversity in gorillas. Molecular Phylogenetics
and Evolution 6, 39–48.
Gibbs, S., Collard, M., Wood, B., 2000. Soft-tissue
characters in higher primate phylogenetics.
Proceedings of the National Academy of
Sciences, USA 97, 11130–11132.
Goldberg, T.L., 1998. Biogeographic predictors of
genetic diversity in populations of eastern
African chimpanzees (Pan troglodytes schwe-
infurthii). International Journal of Primatology
19, 237–254.
Gonder, M.K., Oates, J.E., Disotell, T.R.,
Forstner, M.R.J., Morales, J.C., Melnick, D.J.,
1997. A new west African chimpanzee
subspecies? Nature 388, 337.
Goodman, M., 1962. Evolution of the immunologic
species specificity of human serum proteins.
Human Biology 34, 104–150.
Goodman, M., Romero-Herrera, A.E., Dene, H.,
Czelusniak, J., Tashian, R.E., 1982.
Macromolecular sequences in systematic and
evolutionary biology. In: Goodman, M. (Ed.),
Macromolecular Sequences in Systematic and
Evolutionary Biology. Plenum Press, New York,
pp. 115–191.
Goodman, M., Porter, C.A., Czelusniak, J., Page, S.L.,
Schneider, H., Shoshani, J., Gunnell, G.,
Groves, C.P., 1998. Toward a phylogenetic
classification of primates based on DNA
evidence complemented by fossil evidence.
Molecular Phylogenetics and Evolution,
585–598.
Groves, C.P., 1967. Ecology and taxonomy of the
Gorilla. Nature 213, 890–893.
Groves, C.P., 1970. Population systematics of the
Gorilla. Journal of Zoology. London. 161,
287–300.
Groves, C.P., 1986. Primate systematics. In:
Swindler, D.R., Erwin, J. (Eds.), Compar-
ative Primate Biology, Volume 1. Systematics,
Evolution, and Anatomy. Alan R. Liss, New
York, pp. 187–218.
Groves, C.P., 2001. Primate Taxonomy. Smithsonian
University Press, Washington, D.C.
Grubb, P., 1999. Evolutionary processes implicit
in distribution patterns of modern African
mammals. In: Bromage, T.J., Schrenk, F. (Eds.),
African Biogeography, Climate Change, and
Human Evolution. Oxford University Press,
New York, pp. 150–164.
Guy, F., Brunet, M., Schmittbuhl, M., Viriot, L.,
2003. New approaches in hominoid taxonomy:
morphometrics. American Journal of Physical
Anthropology 121, 198–218.
Harvati, K., Frost, S.R., McNulty, K.P., 2004.
Neanderthal taxonomy reconsidered: implica-
tions of 3D primate models of intra- and
interspecific differences. Proceedings of the
National Academy of Sciences, USA 101,
1147–1152.
Hill, W.C.O., 1967. The taxonomy of the genus Pan.
In: Starck, D. Schneider, R., Kuhn, H.-J. (Eds.),
Neue Ergebnisse der Primatologie. Fischer,
Stuttgart, pp. 47–54.
Hill, W.C.O., 1969. The nomenclature, taxonomy and
distribution of Chimpanzees. In: Bourne, G.H.
(Ed.). The Chimpanzee. Vol. 1. Karger, Basel,
pp. 22–49.
Horn, A.D., 1979. The taxonomic status of the bonobo
chimpanzee. American Journal of Physical
Anthropology 51, 273–282.
James, F.C., McCullough, C.E., 1990. Multivariate
statistical methods in ecology and systematics:
Panacea or Pandora’s Box. Annual Review of
Ecology and Systematics 211, 129–166.
Jenkins, P., 1990. Catalogue of Primates in the British
Museum (Natural History). Part V. Natural
History Museum Publications, London.
Jensen-Seaman, M.I., Kidd, K.K., 2001. Mitochon-
drial DNA variability and biogeography
of eastern gorillas. Molecular Ecology 10,
2241–2247.
Jensen-Seaman, M.I., Deinard, A.S., Kidd, K.K., 2003.
Mitochondrial and nuclear DNA estimates
of divergence between western and eastern
gorillas. In: Taylor, A.B., Goldsmith, M.L.
(Eds.), Gorilla Biology. Cambridge University
Press, Cambridge, pp. 247–268.
Johanson, D.C., 1974. An odontological study
of the chimpanzee with some implications
for hominoid evolution. Ph.D. Dissertation,
University of Chicago.
Jolly, C.J., 1993. Species, subspecies, and
baboon systematics. In: Kimbel, W.H.,
Martin, L.B. (Eds.) Species, Species Concepts,
30 Pilbrow
and Primate Evolution. Plenum Press,
New York, pp.67–107.
Jolly, C.J., 2001. A proper study for mankind: analogies
from the papionin monkeys and their impli-
cations for human evolution. Yearbook of
Physical Anthropology 44, 177–204.
Jolly, C.J., Oates, J.F., Disotell, T.R., 1995.
Chimpanzee kinship. Science 268, 185–187.
Jungers, W.L. Susman, R.S., 1984. Body size
and skeletal allometry in African apes. In:
Susman, R.L. (Ed.), Evolutionary Morphology
and Behavior of the Pygmy Chimpanzee.
Plenum Press, New York. pp. 131–178.
Kimbel, W.H., 1991. Species, species concepts and
hominid evolution. Journal of Human Evolution
20, 355–372.
Kimbel, W.H., Martin, L.B., 1993. (Eds). Species,
Species Concepts and Primate Evolution.
Plenum Press, New York.
Kingdon, J., 1989. Island Africa. Princeton University
Press, Princeton.
Kinzey, W.G., 1984. The dentition of the pygmy
chimpanzee, Pan paniscus. In: Susman, R.L.
(Ed.), The Pygmy Chimpanzee: Evolutionary
Biology and Behavior. Plenum Press, New
York, pp. 65–88.
Leigh, S.R., Shea, B.T., 1995. Ontogeny and
the evolution of adult body size dimor-
phism in apes. American Journal of
Primatology 36, 37–60.
Leigh, S.R., Relethford, J.H., Park, P.B.,
Konigsberg, L.W., 2003. Morphological differ-
entiation of Gorilla subspecies. In: Taylor, A.B.,
Goldsmith, M.L. (Eds.), Gorilla Biology.
Cambridge University Press, Cambridge,
pp. 104–131.
Lindstedt, S.L., Calder, W.A.III., 1981. Body
size, physiological time, and longevity of
homeothermic animals. Quaterly Review of
Biology 56, 1–16.
Lockwood, C.A., Kimbel, W.H., Lynch, J.M.,
2004. Morphometrics and hominoid phylogeny:
Support for a chimpanzee-human clade and
differentiation among great ape subspecies.
Proceedings of the National Academy of
Sciences, USA101, 4356–4360.
Mahler, P.E., 1973. Metric variation in the pongid
dentition. Ph.D. Dissertation, University of
Michigan.
Mayr, E., 1942. Systematics and the Origin of Species.
Columbia University Press, New York.
Mayr, E., 1982. Of what use are subspecies? The Auk
99, 593–595.
Miller, J.M.A., Albrecht, G.H., Gelvin, B.R., 2004.
Craniometric variation in early Homo compared
to modern gorillas: a population-thinking
approach. In: Anapol, F., German, R.Z.,
Jablonski, N.G. (Eds.), Shaping Primate
Evolution. Cambridge University Press,
Cambridge, pp. 66–96.
Morbeck, M.E., Zihlman, A.L., 1989. Body size
and proportions in chimpanzees, with special
reference to Pan troglodytes schweinfurthii
from Gombe National Park, Tanzania. Primates
30, 369–382.
Morin, P.A., Moore, J.J., Chakraborthy, R., Jin, L.,
Goodall, J., Woodruff, D.S., 1994. Kin
selection, social structure, gene flow, and
the evolution of chimpanzees. Science 265,
1193–1201.
Mosimann, J.E., James, F. C., 1979. New statis-
tical methods for allometry with application to
Florida red-winged blackbirds. Evolution 33,
444–459.
Muir C.C., Galdikas B.M., Beckenbach A.T., 2000.
MtDNA sequence diversity of orangutans from
the islands of Borneo and Sumatra. Journal of
Molecular Evolution 51, 471–80.
Pilbrow, V.C., 2003. Dental variation in African apes
with implications for understanding patterns of
variation in species of fossil apes. Ph.D. Disser-
tation, New York University.
Pilbrow, V.C., 2006a. Lingual incisor waits in the
modern hominoids and their utility for fossil
hominoid taxonomy. American Journal of
Physical Anthropology 129, 323-338.
Pilbrow, V.C., 2006b. Population systematics of
chimpanzees using molar morphometrics.
Journal of Human Evolution 51, 646-662.
Pilbrow, V.C. Bailey, S.E., 2005. Patterns of
intraspecific variation in dental metrics
of Gorilla gorilla, Pan troglodytes and
Homo sapiens: a comparison using Wright’s
(1969) Fst statistic. PaleoAnthropology 3, 13.
http://pa.pennpress.org/
Röhrer-Ertl, O., 1984. Research history, nomen-
clature and taxonomy of the orang-utan.
In: Schwartz, J.H., (Ed.), Orang-utan
Biology. Oxford University Press, Oxford,
pp.7–18.
Ruvolo, M., 1994. Molecular evolutionary processes
and conflicting gene trees: the hominoid case.
American Journal of Physical Anthropology
94, 89–113.
Ruvolo, M., 1997. Molecular phylogeny of
the hominoids: inferences from multiple
Molar Variation in Great Apes 31
independent DNA sequence data sets.
Molecular Biology and Evolution 14, 248–265.
Ruvolo, M., Pan, D., Zehr, S., Goldberg, T.,
Disotell, T.R., 1994. Gene trees and hominoid
phylogeny. Proceedings of the National
Academy of Sciences of the USA 91,
8900–8904.
Saltonstall, K., Amato, G., Powell, J., 1998. Mitochon-
drial DNA variability in Grauer’s gorillas
of Kahuzi-Biega National Park. Journal of
Heredity 89, 129–135.
Sarmiento, E.E., Butynski, T., 1996. Present problems
in gorilla taxonomy. Gorilla Journal 19, 5–7.
Sarmiento, E.E., Oates, J.F., 2000. The Cross River
gorillas: a distinct subspecies, Gorilla gorilla
diehli Matschie 1904. Am. Mus. Novitates
3250, 56 pp.
Schaefer, K., Mitteroecker, P., Gunz, P., Bernhard, M.,
Bookstein, F.L., 2004. Craniofacial sexual
dimorphism patterns and allometry among
extant hominids. Annals of Anatomy 186, 471–
478.
Schwartz, J.H., 1984. The evolutionary relationships of
man and orang-utans. Nature 308, 501–505.
Scott, J.E., Lockwood C.A., 2004. Patterns of tooth
crown size and shape variation in great apes and
humans and species recognition in the hominid
fossil record. American Journal of Physical
Anthropology 125, 303–319.
Shea, B.T., 1981. Relative growth of the limbs and
trunk in the African apes. American Journal of
Physical Anthropology 56, 179–201.
Shea, B.T., 1983a. Size and diet in the evolution of
African ape craniodental form. Folia Primato-
logica 40, 32–68.
Shea, B.T., 1983b. Allometry and heterochrony in the
African apes. American Journal of Physical
Anthropology 62, 275–289.
Shea, B.T., 1983c. Paedomorphy and neoteny in the
pygmy chimpanzee. Science 222, 521–522.
Shea, B.T., 1984. An allometric perspective on
the morphological and evolutionary relation-
ships between pygmy (Pan paniscus) and
common (Pan troglodytes) chimpanzees. In:
Susman, R.L. (Ed.), The Pygmy Chimpanzee:
Evolutionary Biology and Behavior. Plenum
Press, New York, pp. 89–130.
Shea, B.T., Coolidge, H.J. Jr., 1988. Cranio-
metric differences and systematics in the
genus Pan. Journal of Human Evolution 17,
671–685.
Shea, B.T., Leigh, S.R., Groves, C.P., 1993. Multi-
variate craniometric variation in chimpanzees:
implications for species identification in
paleoanthropology. In: Kimbel, W.H.,
Martin, L.B. (Eds.), Species, Species Concepts,
and Primate Evolution. Plenum Press, New
York, pp. 265–296.
Sibley, C., Ahlquist, J., 1984. The phylogeny of the
hominoid primates, as indicated by DNA-DNA
hybridization. Journal of Molecular Evolution
20, 1–15.
Smith, H., Chiszar, M.D., Montanucci, R.R., 1997.
Subspecies and classification. Herpetological
Review 28, 13–16.
Stumpf, R.M., Polk, J.D., Oates, J.F., Jungers, W.L.,
Heesy, C.P., Groves, C.P., Fleagle, J. G.,
2003. Patterns of diversity in gorilla cranial
morphology. In: Taylor, A.B., Goldsmith, M.L.
(Eds.), Gorilla Biology. Cambridge University
Press, Cambridge, pp. 35–61.
Swindler, D.R., 1976. Dentition of Living Primates.
Academic Press, London.
Swindler, D.R., 2002. Primate Dentition: An Intro-
duction to the Teeth of Non-Human Primates.
Cambridge University Press, Cambridge.
Taylor, A.B., 2006. Size and shape dimorphism
in great ape mandibles and implications for
fossil species recognition. American Journal of
Physical Anthropology 129, 82–98.
Taylor, A.B., Groves, C.P., 2003. Patterns of
mandibular variation in Pan and Gorilla and
implications for African ape taxonomy. Journal
of Human Evolution 44, 529–561.
Tattersall, I., 1986. Species recognition in human
paleontology. Journal of Human Evolution 15,
165–175.
Tattersall. I., 1991. What was the human revolution?
Journal of Human Evolution 20, 77–83.
Tattersall, I., 1993. Speciation and morphological
differentiation in the genus Lemur. In:
Kimbel, W.H., Martin, L.B. (Eds.), Species,
Species Concepts, and Primate Evolution.
Plenum Press, New York, pp. 163–176.
Templeton, A.R., 1999. Human Races: A genetic and
evolutionary perspective. American Anthropol-
ogist 100, 632–650.
Uchida, A., 1992. Intra-species variation among the
great apes: Implications for taxonomy of
fossil hominoids. Ph.D. Dissertation, Harvard
University.
Uchida, A., 1996. Craniodental variation among the
great apes. Peabody Museum Bulletin, 4.
Harvard University.
Wolpoff, M.H., Hawks, J., Frayer, D.W., Hunley, K.
2001. Modern human ancestry at the
32 Pilbrow
peripheries: a test of the replacement theory.
Science. 291, 293–297.
Wood, B., Richmond, B.G., 2000. Human evolution:
taxonomy and paleobiology. Journal of
Anatomy 197, 19–60.
Wrangham, R.W., de Waal, F.B.M., McGrew, W.C.,
1994. The challenge of behavioral diversity. In:
Wrangham, R.W., McGrew, W.C., de Waal,
F.B.M., Heltne, P.G. (Eds.), Chimpanzee
Cultures. Harvard University Press, Cambridge,
Mass., pp. 1–18.
Xu, X., Arnason, U., 1996. The mitochondrial
DNA molecule of Sumatran orangutan and
a molecular proposal for two (Bornean and
Sumatran) species of orangutan. Journal of
Molecular Evolution 43, 431–437.
Zhi, L., Karesh, W.B., Janczewski, D.N., Frazier-
Taylor, H., Sajuthi, D., Gombek, F., Andau, M.,
Martenson, J.S., O’Brien, S.J., 1996. Genomic
differentiation among natural populations of
orang-utan (Pongo pygmaeus). Current Biology
6, 1326–1336.
