FirstCite?
e-publishing
Natural selection on molar
size in a wild population
of howler monkeys
(Alouatta palliata)
David DeGusta1*, Melanie A. Everett2
and Katharine Milton3
1Laboratory for Human Evolutionary Studies, Museum of Vertebrate
Zoology, and Department of Integrative Biology, University of
California, Berkeley, CA 94720-3160, USA
2Department of Anthropology, Indiana University, Bloomington,
IN 47405, USA
3Division of Insect Biology, Department of Environmental Science,
Policy, and Management, University of California, Berkeley,
CA 94720-3112, USA
*Author for correspondence (degusta@uclink.berkeley.edu).
Recd 17.12.02; Accpd 15.01.03; Online
Dental traits have long been assumed to be under
selection in mammals, based on the macroevolu-
tionary correlation between dental morphology and
feeding behaviour. However, natural selection act-
ing on dental morphology has rarely, if ever, been
documented in wild populations. We investigated
the possibility of microevolutionary selection on
dental traits by measuring molar breadth in a sam-
ple of Alouatta palliata (mantled howler monkey)
crania from Barro Colorado Island (BCI), Panama.
The age at death of the monkeys is an indicator of
their fitness, since they were all found dead of natu-
ral causes. Howlers with small molars have signifi-
cantly decreased fitness as they die, on average, at
an earlier age (well before sexual maturity) than
those with larger molars. This documents the exist-
ence of phenotypic viability selection on molar tooth
size in the BCI howlers, regardless of causality or
heritability. The selection is further shown to occur
during the weaning phase of A. palliata life history,
establishing a link between this period of increased
mortality and selection on a specific morphological
feature. These results provide initial empirical sup-
port for the long-held assumption that primate
molar size is under natural selection.
Keywords: dentition; fitness; microevolution; selection;
viability; weaning
1. INTRODUCTION
The evolutionary relationships of fossil mammals are esti-
mated based largely on dental morphology, due to the fre-
quent preservation of teeth and their apparently strong
phylogenetic signal. Molar size, in particular, has been a
key character for palaeomammalian systematics, especially
in primates (Hlusko et al. 2002). As such, it is a long and
widely held assumption that at least some aspects of dental
morphology, such as molar size, are generally subject to
natural selection in mammals, including primates (Osborn
1907). Currently, the evidence for such selection is limited
to the macroevolutionary correlation of dental mor-
Proc. R. Soc. Lond. B (Suppl.) 02bl0005.S1 ? 2003 The Royal Society
DOI 10.1098/rsbl.2003.0001
phology and feeding behaviour across taxa (e.g. Kay 1975;
Swindler 2002). To our knowledge, selection on a dental
trait has not been demonstrated in a natural primate popu-
lation (Endler 1986; Kingsolver et al. 2001), as previous
attempts using palaeontological samples were confounded
by temporal and taphonomic issues (Marcus 1969; Perzig-
ian 1975). One of the authors (K.M.) initiated and over-
saw the collection of a skeletal sample that, by virtue of
being a ?found dead? collection, allows for microevolution-
ary tests of natural selection on dental traits. We carried
out such a test and report the main result.
2. MATERIAL AND METHODS
(a) Howler monkey crania
Our study is based on a sample (n = 215) of Alouatta palliata
(mantled howler monkey) crania from monkeys found dead on Barro
Colorado Island (BCI), Panama between 1986 and 1996
(DeGusta & Milton 1998; Jones et al. 2000). BCI is one of the most
studied tracts of tropical forest in the world, and abundant infor-
mation is therefore available on numerous aspects of the environment
and fauna (e.g. Milton 1980; Leigh & Wright 1990). Howlers on
BCI are free ranging and are not provisioned, hunted or otherwise
manipulated by humans, nor are they subject to vertebrate predation
(DeGusta & Milton 1998). While the skeletal sample is biased against
recovery of immature individuals, there is absolutely no expectation
or evidence of a dental-size recovery bias within age categories
(DeGusta & Milton 1998).
(b) Theoretical issues in identifying selection
Since all monkeys were found dead of natural causes, their age at
death is an indicator of their fitness. Individuals that die before reach-
ing sexual maturity have lower fitness (zero) than monkeys that live
to adulthood (potential for non-zero fitness). In this sample, then, a
correlation between age-at-death and dental morphology is potential
evidence of mortality-based (viability) selection on the dentition
(Endler 1992).
However, such a correlation can be caused by factors other than
selection acting directly on the studied trait. First, environmental
variation can create correlations between fitness values and traits that,
while selectively neutral, are also influenced by such variation
(Rauscher 1992). Quantitative methods exist for detecting environ-
mental variation effects on selection (Rauscher 1992), but require
data rarely available from skeletal remains (e.g. breeding values).
Theoretically, however, it is unlikely that highly mobile howler mon-
keys, effectively confined to BCI, should experience significant intra-
population environmental variation (though this is admittedly
speculative). Furthermore, since dental morphology is determined
early in life and is not thereafter remodelled (except by mechanical
wear), such variation would have to occur in the early stages of life
(i.e. before the animal reached ca. 1 year of age). All crania were
recovered from within 6 km2, further indicating that significant
environmental variation is unlikely in this sample. Finally, even if
such variation is responsible for a correlation between a trait and fit-
ness, there is still phenotypic selection despite the fact that it will not
lead to evolutionary change (cf. Kruuk et al. 2002). This highlights
the important distinction between phenotypic selection and the evol-
utionary response to selection (Arnold & Wade 1984).
The inference of selection from a statistical correlation between a
trait and survival can also be confounded by linkages between that
trait and others. Specifically, the trait of interest may not be under
direct selection, but instead be genetically or functionally linked to a
trait that is under such selection (Lande & Arnold 1983). Quantitat-
ive methods exist for determining the relative contributions of indi-
vidual traits to overall fitness, but these require that the trait(s)
directly under selection are included as one of the measured traits
(Lande & Arnold 1983; Arnold & Wade 1984). For primates, this
would clearly require the simultaneous inclusion of numerous vari-
ables (behavioural, physiological, genetic and morphological), whose
measurement would itself substantially impact upon the population.
In addition, most such variables are not available from skeletal
material. Furthermore, when considering phenotypic selection
(rather than the evolutionary response to that selection), there is no
distinction between ?direct? and ?indirect? selection?if a trait is corre-
lated with fitness, then it is under selection (Endler 1992).
As such, it is clear that a significant correlation between a dental
trait and age-at-death in the howler sample would be evidence of
phenotypic selection on that trait (Endler 1986). Lacking a range of
other data, as discussed above, it is not currently possible to predict
the evolutionary response (if any) to such selection. To identify
02bl0005.S2 D. DeGusta and others Natural selection on molar size in howler monkeys
Table 1. Width of maxillary permanent first molar by age stage in the BCI howler sample.
age stage age in monthsa no. of individuals mean (mm) median (mm) s.d. range (mm)
2 6?12 16 7.09 7.64 1.06 5.31?8.22
3 12?24 21 7.56 7.44 0.56 5.94?8.41
4 24?36 16 7.61 7.63 0.37 7.01?8.42
5 
 36 162 7.75 7.74 0.39 6.95?9.10
a Chronological correlations are approximate (DeGusta & Milton 1998).
whether such a correlation exists in our sample requires that we meas-
ure an appropriate dental trait and estimate the age of each individual
in the sample.
(c) The dental trait, ageing and sexing
The dental trait we measured in the BCI A. palliata sample was
the buccolingual width of the permanent maxillary first molar
(UM1). Tooth crowns do not grow once fully formed, and all indi-
viduals in our sample have fully formed UM1s. Thus, any size differ-
ences between age stages are not due to growth (drastically reducing
the problem of environmental variance, as discussed above). The
buccolingual width was selected for measurement because it is at least
partially heritable (Jernvall & Jung 2000; Hlusko et al. 2002) and is
not altered by occlusal wear in BCI A. palliata ( Jones et al. 2000).
The measurement error was negligible (1.7%). The UM1 was chosen
for study because it is one of the first permanent teeth to erupt, thus
permitting its longitudinal examination. The other early erupting
teeth in A. palliata are incisors, which are frequently missing in the
BCI sample and whose dimensions are altered by occlusal wear. A
number of specimens in the BCI sample also lack the mandible, so
study of the lower first molar (LM1) is limited by sample size. Even
so, the limited results from the LM1 are congruent with the results
reported here for the UM1.
Specimens were assigned to one of five dental eruption age stages
(1?5, youngest to oldest): 1?only deciduous dentition erupted; 2?
permanent I1, I2 and M1 erupted; 3?permanent M2 erupted; 4?
permanent P2, P3 and P4 erupted; and 5?permanent C and M3
erupted. A tooth was considered erupted if its crown penetrated the
alveolar plane. Dental eruption stages are a robust relative aging
method, and the approximate correspondence between the dental age
stages and age in months (DeGusta & Milton 1998) is given in table
1. Sexual maturity clearly occurs at stage 5 in BCI A. palliata
(DeGusta & Milton 1998). Data on the timing of tooth crown forma-
tion (amelogenesis) in Alouatta indicate that the UM1 crown begins
to form before birth (Tarrant & Swindler 1973) and is probably com-
pleted some weeks following birth.
Adult specimens were sexed based on canine dimorphism, but the
sex of immature skulls is, at present, unknown. Since male BCI how-
lers have significantly larger mean UM1 breadth than females (Jones
et al. 2000), differences in sex ratios across immature age-stage sub-
samples can introduce a spurious trend in molar size. However, since
the degree of sexual dimorphism in this characteristic is known for
this population (mean difference of 0.25 mm; Jones et al. (2000)),
we can determine whether a difference in male-to-female ratios across
age-stage sub-samples is responsible for any significant correlations
detected.
3. RESULTS
The youngest howlers (stage 2, about 6?12 months old)
have significantly smaller UM1s than the other age stages
using the t-test with Bonferroni?s correction (figure 1;
table 1). There are no other statistically significant differ-
ences in UM1 size between age stages (table 2). The
absolute mean differences in UM1 breadth between age
stage 2 and the older age stages are much greater than
0.25 mm (table 1) which, as discussed above, means that
differing sex ratios between the age stages cannot explain
this difference.
The smaller mean size in age stage 2 is due to the pres-
ence of specimens with very small molars in that age
group, rather than the absence of larger individuals (figure
2). Of the six smallest UM1 breadths in the sample, the
Proc. R. Soc. Lond. B (Suppl.)
2 3 4 5
age stage
m
ea
n 
m
ol
ar
 s
iz
e 
(m
m
)
6.8
7.0
7.2
7.4
7.6
7.8
Figure 1. A graph of mean molar size (buccolingual breadth
of maxillary first molar) by age stage (2?5, youngest to
oldest) for a sample of 215 Alouatta palliata (howler
monkey) crania from Barro Colorado Island, Panama.
five smallest are from age stage 2 individuals and the sixth
is from an age stage 3 individual (figure 2). There is a gap
of 1 mm in the distribution of UM1 breadths between
these six individuals (less than 5.9 mm) and the remaining
209 individuals in the sample (greater than 6.94 mm).
The six smallest crania were recovered in four different
years, spanning the decade of collection, ruling out a
cohort effect. Since molars do not grow once fully formed
(Jernvall & Jung 2000), and since all individuals in our
sample have fully formed and erupted UM1s, the differ-
ence in mean UM1 size between age stage 2 howlers and
older howlers cannot be attributed to growth.
4. DISCUSSION
Upper first molar breadth is correlated with age-at-
death in BCI howlers. Specifically, BCI howlers with the
smallest molars have greatly increased mortality at about
6?12 months of age. This corresponds with the weaning
period in A. palliata (particularly for males; Clarke
(1990)), which is known to be a time of frequent infant
mortality in A. palliata (especially for males; Clarke
(1990)). Interestingly, however, molar size is not at all cor-
related with longevity beyond that age. It might have been
expected that there would be a consistent trend of
?larger molars = longer survival?. Instead, we have shown
that once the very smallest individuals are weeded out at
age stage 2 (suggesting stabilizing selection), molar size no
longer correlates with age-at-death (figures 1 and 2). This
documents the presence of an early selective bottleneck in
BCI howlers, associated with weaning, that selects against
individuals with small molars, thus linking a life-history
Natural selection on molar size in howler monkeys D. DeGusta and others 02bl0005.S3
Table 2. Comparisons of mean differences in maxillary perma-
nent first molar breadth between age stages using the t-test
with Bonferroni?s correction.
comparison mean differencea p valueb
age stage 2 and 3 0.46 0.005?
age stage 2 and 4 0.51 0.003?
age stage 2 and 5 0.69  0.0001?
age stage 3 and 4 0.05 0.755
age stage 3 and 5 0.20 0.085
age stage 4 and 5 0.14 0.258
a Difference is first listed age stage minus second.
b Asterisks denote significant comparisons (adjusted alpha value of
0.008).
2
3
4
5
4 5 6 7 8 9 10
ag
e 
st
ag
e
molar size (mm)
Figure 2. A plot of individual molar sizes (buccolingual
breadth of maxillary first molar) by age stage (2?5, youngest
to oldest) for a sample of 215 Alouatta palliata (howler
monkey) crania from Barro Colorado Island, Panama.
episode long known to have high mortality with a specific
morphological feature.
The correlation between small molar size and decreased
fitness means that molar size is under phenotypic selec-
tion, specifically viability selection, regardless of causality
or heritability (as discussed above). Even so, since BCI A.
palliata are generally folivorous (Milton 1980), expanded
molar size has plausible adaptive value as one of the fac-
tors influencing rate of food processing and caloric intake.
Therefore, molar size is unlikely to be adaptively neutral
and the documented selection is unlikely to be entirely
due to the correlation of molar size with some other factor
under selection. However, molar size is also unlikely to be
entirely uncorrelated with other factors under selection.
Maternal nutritional status (itself a result of numerous
factors) is likely to play an important role in determining
molar size, since the UM1 crown begins to form shortly
after birth. Overall size may also be a factor, though this
cannot be tested directly in the BCI sample since howler
crania grow throughout age stages 2?5. The notable gap
(1 mm, or ca. 13% of total size) in the distribution
between the smallest molars and the rest of the sample
may indicate the influence of a more discrete variant.
These results document phenotypic viability selection
on molar tooth size in the BCI A. palliata population. To
Proc. R. Soc. Lond. B (Suppl.)
our knowledge, this is the first time that a dental trait has
been shown to be under selection in a wild primate popu-
lation. This selection is further shown to occur during the
weaning phase of A. palliata life history, providing a docu-
mented link between this period of increased mortality
and a specific morphological feature. In total, these results
provide initial support for the long-held assumption that
aspects of mammalian dental morphology are under natu-
ral selection.
Acknowledgements
The authors thank Henry Gilbert, F. Clark Howell and Tim White
(all of UC Berkeley) for helpful comments, as well as the three anony-
mous reviewers for their excellent suggestions. They especially thank
all those who collected dead monkeys on BCI to create this collec-
tion, particularly Bonifacio de Leon and Raul Rios of the Smithsonian
Tropical Research Institute, as well as Elena Ceden?o, Sharon
Kitchner and Beth King. The collection of these skulls was supported
in part by NSF grant no. 85-12634 to K.M. and NSF grant no. 90-
20058 to K.M., Nick Lerche, Linda Lowenstine and John Anderson.
Arnold, S. J. & Wade, M. J. 1984 On the measurement of natural
and sexual selection: theory. Evolution 38, 709?719.
Clarke, M. R. 1990 Behavioral development and socialization of
infants in a free-ranging group of howling monkeys (Alouatta
palliata). Folia Primatol. 54, 1?15.
DeGusta, D. & Milton, K. 1998 Skeletal pathologies in a population
of Alouatta palliata: behavioral, ecological, and evolutionary impli-
cations. Int. J. Primatol. 19, 615?650.
Endler, J. A. 1986 Natural selection in the wild. Princeton University
Press.
Endler, J. A. 1992 Natural selection: current usages. In Keywords in
evolutionary biology (ed. E. Keller & E. Lloyd), pp. 220?224. Cam-
bridge, MA: Harvard University Press.
Hlusko, L. J., Weiss, K. M. & Mahaney, M. C. 2002 Statistical gen-
etic comparison of two techniques for assessing molar crown size
in pedigreed baboons. Am. J. Phys. Anthropol. 117, 182?189.
Jernvall, J. & Jung, H. 2000 Genotype, phenotype, and developmen-
tal biology of molar tooth characters. Ybk Phys. Anthropol. 43,
171?190.
Jones, A. L., DeGusta, D., Turner, S. P., Campbell, C. J. & Milton,
K. 2000 Craniometric variation in a population of mantled howler
monkeys (Alouatta palliata): evidence of size selection in females
and growth in dental mature males. Am. J. Phys. Anthropol. 113,
411?434.
Kay, R. F. 1975 The functional adaptations of primate molar teeth.
Am. J. Phys. Anthropol. 43, 195?216.
Kingsolver, J. G., Hoekstra, H. E., Hoekstra, J. M., Berrigan, D.,
Vignieri, S. N., Hill, C. E., Hoang, A., Gibert, P. & Beerli, P. 2001
The strength of phenotypic selection in natural populations. Am.
Nat. 157, 245?261.
Kruuk, L. E. B., Slate, J., Pemberton, J. M., Brotherstone, S., Guin-
ness, F. & Clutton-Brock, T. 2002 Antler size in red deer: herita-
bility and selection but no evolution. Evolution 56, 1683?1695.
Lande, R. & Arnold, S. J. 1983 The measurement of selection on
correlated characters. Evolution 37, 1210?1226.
Leigh Jr, E. G. & Wright, S. J. 1990 Barro Colorado Island and trop-
ical biology. In Four neotropical forests (ed. A. H. Gentry), pp. 28?
47. New Haven, CT: Yale University Press.
Marcus, L. F. 1969 Measurement of selection usiing distance stat-
istics in the prehistoric orang-utan Pongo pygmaeus palaeosuma-
trensis. Evolution 23, 301?307.
Milton, K. 1980 The foraging strategy of howler monkeys: a study in
primate economics. New York: Columbia University Press.
Osborn, H. F. 1907 Evolution of mammalian molar teeth. New York:
Macmillan.
Perzigian, A. J. 1975 Natural selection on the dentition of an arikara
population. Am. J. Phys. Anthropol. 42, 63?70.
Rauscher, M. D. 1992 The measurement of selection on quantitative
traits: biases due to environmental covariances between traits and
fitness. Evolution 46, 616?626.
Swindler, D. R. 2002 Primate dentition: an introduction to the teeth of
non-human primates. Cambridge University Press.
Tarrant, L. H. & Swindler, D. R. 1973 Prenatal dental development
in the black howler monkey (Alouatta caraya). Am. J. Phys. Anthro-
pol. 38, 255?260.