Evolution, 56(6), 2002, pp.

1276–1285

THE ORIGINS OF SEXUAL DIMORPHISM IN BODY SIZE IN UNGULATES

F. J. PÉREZ-BARBERı́A,1,2,3 I. J. GORDON,1,4 AND M. PAGEL5,6
1 The
Macaulay Institute, Craigiebuckler, Aberdeen, AB15 8QH, United Kingdom
2 Instituto
de Recursos Naturales y Ordenación del Territorio, 13 Independencia, Oviedo, 33071, Spain
3 E-mail: [email protected]
4 E-mail: [email protected]
5 School of Animal and Microbial Science, University of Reading, Whiteknights, Reading RG6 6AJ, United Kingdom
6 E-mail: [email protected]

Abstract. Jarman (1974) proposed a series of relationships between habitat use, food dispersion, and social behavior
and hypothesized a series of evolutionary steps leading to sexual dimorphism in body size through sexual selection
in African antelope species. The hypothesis states that sexual size dimorphism evolved in a three-step process. Initially,
ancestral monomorphic and monogamous ungulate species occupying closed habitats radiated into open grassland
habitats. Polygynous mating systems then rapidly evolved in response to the aggregation of males and females, perhaps
in relation to the clumped distribution of food resources in open habitats. Subsequently, size dimorphism evolved in
those species occupying open habitats, but not in species that remained in closed habitats or retained monogamy. This
hypothesis has played an important role in explaining the origins of sexual dimorphism in mammals. However, the
temporal sequence of the events that Jarman proposed has never been demonstrated. Here we use a phylogeny of
extant ungulate species, along with maximum-likelihood statistical techniques, to provide a test of Jarman’s hypothesis.

Key words. Body size, habitat use, phylogeny, polygyny, sex.

Received July 2, 2001. Accepted March 22, 2002.

In most mammalian groups males are larger than females Food distribution and anti-predator behavior determine the
(Mace 1979; Reiss 1989; Abouheif and Fairbairn 1997; probability of encounters between individuals in space and
Weckerly 1998; Loison et al. 1999; Pérez-Barberı́a and Gor- time, thus influencing their degree of aggregation (Wrangham
don 2000). Explanations for the degree of sexual dimorphism and Rubenstein 1986). Situations in which species feed on
in body mass (SD) fall into three main groups (Fairbairn low-density food distributed in patches have been linked to
1997). First, sexual selection hypotheses suggest that SD is strong intraspecific and intersexual competition for food. This
a consequence of either intrasexual competition for mates, may favor increased feeding divergence and thus increased
mainly males fighting amongst themselves for mating op- dimorphism in African antelopes (Jarman 1974) and also in
portunities with females, or epigamic selection in which fe- primates (Clutton-Brock 1977; Clutton-Brock et al. 1977).
males choose among males to get mating partners (Trivers But this competition is not only for food. Aggregation of
1972). Second, large body mass may facilitate the evolution animals around clumps of food may determine the strategies
of SD by making polygyny more likely and thus increasing males use in order to maximize their chances of obtaining
the intensity of sexual selection between the sexes (Lack mates. These strategies lead to sexual selection for or against
1968; Selander 1972; Clutton-Brock et al. 1977; Webster sexual dimorphism and affect aspects of a species’ ecology;
1992; Abouheif and Fairbairn 1997; Loison et al. 1999). A for example, adult sex ratio and differential spatial distri-
third group of explanations suggests that SD arises from nat-
bution of the sexes (Geist 1974; Jarman 1974; Pérez-Barberı́a
ural selection by favoring different optimal size for males
and Gordon 1999a). Jarman’s (1974) hypothesis for the evo-
and females because they occupy different ecological niches
lution of sexual dimorphism states that the first antelopes
(Selander 1972; Shine 1989).
were closed-habitat dwelling, monogamous, and monomor-
Darwin (1871) was the first to link sexual dimorphism to
phic species (Janis 1982). These antelopes then evolved into
polygyny. Sexual dimorphism among birds (Lack 1968; Wi-
ley 1974), primates (Clutton-Brock et al. 1977; Ralls 1977; open-habitat dwellers, and once in this habitat, animal ag-
Clutton-Brock and Harvey 1978; Clutton-Brock 1985; Lin- gregation and increased group size favored the evolution of
denfors and Tullberg 1998) and ungulates (Geist 1974; Jar- polygyny. Finally, within the polygynous mating system, sex-
man 1974; Ralls 1977; Bubenik 1985; Loison et al. 1999) is ual selection favored the evolution of sexual dimorphism in
associated with polygyny. However, less is known about the body size, with males being larger than females. Jarman’s
causes of the evolution of sexual size dimorphism within any hypothesis has been widely discussed (Clutton-Brock et al.
given taxa (Jarman 1974; Clutton-Brock and Harvey 1977a; 1977; Ralls 1977; Clutton-Brock and Harvey 1983; Gosling
Clutton-Brock et al. 1977). Fairbairn (1997) pointed out that 1986; Komers 1996a,b; Weckerly 1998; Pérez-Barberı́a and
most attempts to explain the evolution of SD were based on Gordon 1998; Loison et al. 1999) but only the paleoecological
correlations among variables, which make it difficult to sep- approach of Janis (1982) provided evidence in its support.
arate cause from effect. Fairbairn went on to suggest that a Here we provide a direct test of Jarman’s proposed scenario
more promising approach would be to use directional com- for the evolution of sexual size dimorphism in ungulates,
parative analyses that estimate the order of appearance of using a phylogenetic approach that can detect the order and
traits within a clade. This is the approach that we have applied direction of evolutionary change in the key traits of habitat
to our research. use, mating system, and sexual dimorphism in body size.
1276
q 2002 The Society for the Study of Evolution. All rights reserved.
 15585646, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.0014-3820.2002.tb01438.x, Wiley Online Library on [01/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SEXUAL DIMORPHISM IN UNGULATES 1277

MATERIALS AND METHODS ing from the ancestral state, to the derived state of polygyny
and sexual size dimorphism. Because the comparative meth-
Collected Taxa
od used requires binomial categories for all variables (Pagel
The grandorder Ungulata (Kingdon 1997) comprises 238 1994), we coded the three variables into discrete binomial
species (Novak 1991) within five orders (Kingdon 1997). We traits based upon the criteria below.
collated information for three of these orders, Proboscidea Habitat type. Closed-habitat dwelling species are those
(n 5 2 species), Perissodactyla (n 5 5 species) and Artio- that spend most of their year in dense habitats (i.e. forests,
dactyla (n 5 110 species). The number of taxa included was woodlands, bushlands, thickets); open-habitat dwelling spe-
limited by the biological and phylogenetic information avail- cies were those that predominantly use habitats that lack
able. closed-habitat vegetation types, although species in this cat-
egory could be closed-habitat ecotone dwelling species (Fig.
Phylogenetic Information 1). Some taxa have subspecies that differ from each other in
their use of habitat type (e.g. Loxodonta spp., Syncerus spp.).
The phylogeny used in this analysis is a composite based
We chose the most representative habitat for all subspecies
mainly on morphological and paleontological information de-
within species (Pérez-Barberı́a et al. 2001). Information about
rived from a variety of sources (Gentry 1978, 1992; Kingdon
the habitat type occupied came from Novak (1991), Kingdon
1982; Corbet and Hill 1986; Janis and Scott 1987, 1988;
(1997), Loison et al. (1999) and Pérez-Barberı́a et al. (2001).
Gentry and Hooker 1988; Novak 1991; Garland and Janis
Mating system. We considered those species that, during
1993). Two studies based on molecular techniques were also
the mating period or year round, constitute groups of one
used to resolve some of the branching patterns (Essop et al.
male and one or two adult females to be monogamous (see
1997; Gatesy et al. 1997). None of these studies used infor-
Fig. 1). The rest of the species were classified as polygynous.
mation on mating system, habitat, or body size to construct
The polygyny category thus includes a wide range of levels
the phylogenetic tree. Because the phylogeny of this group
of polygyny (Wiley 1974; Ralls 1977; Jarman 1983; Gosling
is still under discussion and there are some discrepancies
1986). Also the polygynous state includes species that show
between the phylogenetic trees obtained when using mor-
no breeding group bonds but in which the male defends short-
phological and molecular techniques (Essop et al. 1997), the
term high-quality food patches as a strategy to increase access
results presented in this paper may be conditional on the
to mating opportunities (sensu Gosling 1986). We consider
phylogeny adopted. The tree is shown in Figure 1.
that even in these cases, contests between males take place
Estimates of divergence times among taxa are not available
during territorial defense (Novak 1991; Kingdon 1997), and
for many nodes in the Ungulata (Pérez-Barberı́a and Gordon
sexual dimorphism could evolve on the basis of the hypoth-
1999b, 2000, 2001). In addition, analysis on branch lengths
esis proposed. The sources used for the mating system clas-
based on the fossil record could not be included, because the
sification come from Jarman (1983), Novak (1991), Kingdon
information is only available for a few species (Garland and
(1997), Loison et al. (1999) and N. Owen-Smith (pers.
Janis 1993). Thus, we calculated branch lengths using three
comm.). There is not enough information in the literature for
different arbitrary methods available in the literature. In Gra-
all of the species used in this study to generate a continuous
fen’s distance method (Grafen 1989), tip species are contem-
variable (e.g., number of females/number of males in mating
poraneous and the depth of each node is set to equal one less
groups) in order to carry out a sensitivity analysis on our
than the number of tip species that descend from it. Pagel’s
binomial classification.
branch-length method (Pagel 1992) was estimated such that
Body size sexual dimorphism. Sexual dimorphism in body
all internode branch segments are set to equal one but the
mass or body size can be measured in a variety of ways
distance from the root to the tips was the same for all species.
(Lovich and Gibbons 1992; Ranta et al. 1994). One widely
The third arbitrary method (Garland et al. 1993) set all branch
accepted procedure is to use the residuals from the regression
lengths equal to 1.0. Sensitivity analysis branch carried out
line of male body mass on female body mass (Ranta et al.
by running all the analyses using the three arbitrary methods
1994), which can also include comparative methods in order
demonstrated that the results did not depend upon which
to account for phylogeny (Abouheif and Fairbairn 1997;
branch lengths were used. Consequently, we present the re-
Pérez-Barberı́a and Gordon 2000). Residuals have the prop-
sults based on branch lengths equal to 1.0.
erty of creating a measure of dimorphism that is uncorrelated
with body mass. However, given that the degree of dimor-
Ecological and Biological Information
phism increases with size (Loison et al. 1999) this method
Three variables were analyzed in this study, namely, the has limited value in ranking individual species in relation to
main habitat type occupied by the species, the main type of their absolute degree of sexual dimorphism (Fairbairn 1997).
mating system adopted by the species, and sexual dimor- To preserve this element of dimorphism, which may arise for
phism in body size within a species. Fossil records and pa- adaptive reasons related to group size or other environmental
leoecological evidence reveal that ancestral ungulates oc- variables (Clutton-Brock et al. 1980; Loison et al. 1999), we
cupied closed habitats and were monogamous and mono- measured dimorphism as the simple ratio of male to female
morphic (Janis 1982). We took this as the starting point (an- body mass. We then classified species as dimorphic or not
cestral state) in our analyses of the sequence of events that based on the size of this ratio for that species. In ungulates,
led to the evolution of sexual dimorphism in body mass. We a reduction of male body mass does not increase male mating
investigated transitions among the states of these three var- success, as appears to occur in some pinniped species (Phoca
iables to estimate the most probable sequence of events lead- and Monachus genera; Stirling 1975; Ralls 1977). Thus, we
15585646, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.0014-3820.2002.tb01438.x, Wiley Online Library on [01/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
F. J. PÉREZ-BARBERÍA ET AL.
1278
 15585646, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.0014-3820.2002.tb01438.x, Wiley Online Library on [01/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SEXUAL DIMORPHISM IN UNGULATES 1279

also classified as monomorphic species those in which the represents the likelihood associated with the model of inde-
male was smaller than the female (n 5 23; Tables 1 and 2). pendent evolution, and L(C) denotes the model of correlated
Because the classification of mono- versus dimorphism is evolution. The model of correlated evolution is justified as a
arbitrary, we examined five different classifications of mono-/ representation of the data if its likelihood is significantly larger
dimorphic based upon different cutoff points for the ratio of than the model of independent change. Significance is assessed
male to female log10-transformed body mass (namely, 1.00, by means of the likelihood ratio statistic LR 5 22 loge[L(I)/
1.12, 1.18, 1.26, 1.32). These are shown in Table 2 along L(C)]. These two models differ by four parameters (see Pagel
with the mean degree of dimorphism associated with each 1994), and the LR statistic is asymptotically distributed as a
and the numbers of species classified as being dimorphic or chi-squared variate with four degrees of freedom (Pagel 1997).
monomorphic. Cutoff points beyond 1.32 gave substantially Tests comparing the size of two transition rate parameters
skewed proportions of dimorphic and monomorphic species. in the model of correlated evolution were used to investigate
We also present the mono-/dimorphism classification cal- hypotheses of the order and direction of evolution (Pagel
culated from the residuals from the regression of male body 1994). For example, we tested whether transitions from the
mass against female body mass, controlling for phylogeny ancestral state of closed habitats and monogamy proceeded
with a generalized least-squares model (Pagel 1997, 1999a), first by a change between habitats rather than a change be-
fitting a major axis regression throughout the origin (Sokal tween mating systems. To do this, we compared the rates of
and Rohlf 1995). The mean value for sexual dimorphism evolution from the joint state of closed-habitat/monogamy to
within the monomorphic category produced by the residuals open-habitat/monogamy with the transition from the same
approach was very similar to that of the monomorphic cat- ancestral state (closed-habitat/monogamy) to the derived
egory produced by the 1.12 cutoff point. However, the mean state of closed-habitat/polygyny. The comparison of these
value for sexual dimorphism within the dimorphic category two parameters identifies which evolutionary route away
produced by the residuals approach was the lowest of the from the ancestral state is more probable. The test of this
five cutoff points used (Table 2). kind of hypothesis compares the log-likelihood of the model
Species-specific body masses for each sex were taken from of correlated evolution, in which all parameters are free to
vary, with one in which the two parameters of interest are
Pérez-Barberı́a and Gordon (2000) and those for Proboscidea,
constrained to be equal. These tests follow a chi-squared
Perissodactyla, Hippopotamidae, and Suidae were derived
distribution with one degree of freedom (Pagel 1997).
from a large list of references (body masses are available
Computer simulations show that the likelihood ratio sta-
from the authors on request).
tistic is conservative when tested across a variety of phylog-
enies and branch length scenarios. The actual null hypothesis
The Comparative Method distribution of the test for correlated evolution often follows
We used Pagel’s (1994) discrete variables method to test a chi-squared distribution with fewer than four degrees of
all hypotheses concerning correlated evolution in two binary freedom (Pagel 1997). We used the chi-squared with four
traits. The method compares two models that are fitted by degrees of freedom criterion for all of our correlated evo-
maximum likelihood to the data and the phylogenetic tree. lution tests, and we used the chi-squared one degree of free-
In the model of independent evolution, two binary characters dom criterion for tests of the order and direction of evolution.
are allowed to evolve on the tree independently. Each char- In addition to testing for associations among the traits, and
acter can adopt two states, 0 or 1; calculating ‘‘forward’’ and for the order of changes with traits, we reconstructed probable
‘‘backward’’ transitions for each requires four parameters. In ancestral states at various nodes of the phylogenetic tree.
the model of dependent or correlated evolution, the method These reconstructions display the probable temporal se-
fits a model in which the two characters evolve in a correlated quence of events. Ancestral states were reconstructed by max-
fashion. This is achieved by considering the four possible imum likelihood following the procedures outlined in Pagel
states that two binary characters can adopt (0,0; 0,1; 1,0; 1,1) (1999b) and as implemented in the computer program DIS-
and allowing one of the variables to change state in any CRETE (available from Mark Pagel, School of Animal and
Microbial Science, University of Reading, Whiteknights,
branch of the tree, yielding eight possible transitions. Tran-
Reading, U.K.).
sitions of both characters are estimated simultaneously as
combinations of single-state transitions. This model therefore RESULTS
requires eight parameters (see also Pagel 1994; Pagel 1997).
The fit of a model to the data and tree is described by the Tracing the Origins of Sexual Dimorphism
log-likelihood. The method tests the hypothesis of correlated All three pairs of traits (habitat-mating system, habitat-
evolution by comparing the two log-likelihoods, where L(I) dimorphism, and mating system-dimorphism) evolve in a cor-

←
FIG. 1. Interspecific topology of the taxa studied and reconstruction of the first transitions across the nodes of the tree (branch lengths
are arbitrary, see Materials and Methods). Habitat: 0, closed-habitat dwelling species; 1, open-habitat dwelling species. Mating system:
0, monogamous; 1 polygynous. Dimorphism: 0, monomorphic species; 1, dimorphic species. See Materials and Methods for definitions.
Open circle symbol indicates the ancestral states closed habitat, monogamy, and monomorphism fixed at the root; triangle symbol indicates
that the transition from closed habitats to open habitats occurs before the transition from monomorphism to dimorphism; closed circle
indicates that the transition from closed habitats to open habitats occurs before the transition from monogamy to polygyny; square symbol
indicates that the transition from monogamy to polygyny occurs before the transition from monomorphism to dimorphism, and diamond
symbol indicates where the dimorphic state arises.
 15585646, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.0014-3820.2002.tb01438.x, Wiley Online Library on [01/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1280 F. J. PÉREZ-BARBERÍA ET AL.

TABLE 1. Distribution of the number of species studied in relation to their habitat use, mating system, and sexual dimorphism in body size.
Percentages in brackets.

Open habitat dweller Polygynous Dimorphic

Open habitat dweller — 77 (89.5) 59 (67) 86 (73)

Polygynous 77 (89.0) — 67 (77) 87 (74)
Dimorphic 59 (83.1) 67 (94.0) — 71 (61)
Number of species 117

related fashion and the results for tests of correlation between Habitat and Size Dimorphism
mating system and dimorphism (Table 3) did not depend on
In Jarman’s (1974) view it was the evolution of polygyny
the particular cutoff points used to define dimorphism. Cor-
in open habitats that created conditions that favored males
related evolution between habitat and dimorphism was sen-
becoming larger than females. If Jarman was correct, then,
sitive to the criterion used to define dimorphism (Table 3).
the evolution of size dimorphism occurred after the occu-
We found that extant ungulate species that occupy open hab-
pation of open habitats. That is, we expect a two-step process
itats tend to have a polygynous mating system and are di-
in which ancestrally monomorphic ungulates first move into
morphic in body mass.
open habitats, and then size dimorphism evolves because the
We can use the fact that there are correlations among the
species that occupy the open habitats are polygynous. Here
traits to test for the most probable sequence of events leading
we test for evidence that size dimorphism arises only after
to sexual size dimorphism. Figures 2a–c show the most prob-
ungulates have occupied open habitats. In the next section
able pathways leading from the closed-habitat dwelling, mo-
nogamous, and monomorphic ancestral ungulate condition to we ask whether it is polygyny that is responsible for the
the open-habitat dwelling, polygynous, and size-dimorphic change in dimorphism.
modern ungulates. We find evidence in our data that dimorphism does indeed
evolve after the occupation of open habitats, although the
precise results are sensitive to how we classify size dimor-
Habitat and Mating System
phism. If our criteria for dimorphism are either unrestrictive
Jarman’s (1974) hypothesis predicts that ancestral monog- (that is, male/female . 1.0), such that species in which the
amous ungulates occupying closed habitats first moved into sexes are only slightly different in size are classified as di-
open habitats and then later evolved polygyny. The data sup- morphic, or our criteria are too strict (male/female . 1.26),
port this two-step process (Fig. 2a). such that only species in which males are approximately 25%
Step 1. Occupation of open habitats preceded the evo- larger or more than females are classified as dimorphic, we
lution of polygyny (Fig. 2a). The rate of evolution from the find that, contrary to Jarman’s view, dimorphism evolved
ancestral state of closed-habitat/monogamy to open-habitat/ while the ungulates still occupied closed habitats. If, how-
monogamy is significantly greater than the rate of evolution ever, we classify as dimorphic species those in which males
to closed-habitat/polygyny (x2 5 9.02, df 5 1, P 5 0.0027). are approximately 10% or more larger than females we find
In fact, the latter transition (closed-habitat/monogamy to that dimorphism evolves following the occupation of open
closed-habitat/polygyny) is not statistically significant. There habitats.
is no evidence that polygyny evolved in closed habitats. The strength of the pathways from Figures 2a and 2c makes
Step 2. Once open habitats appeared they favored the it likely that the evolution of dimorphism occurred after un-
evolution of polygyny (Fig. 2a) and dimorphism (Fig. 2b). gulates occupied open habitats. Polygyny clearly evolves af-
The rate of evolution from open-habitat/monogamy to open- ter open habitats and dimorphism evolves in response to po-
habitat/polygyny is greater (x2 5 11.28, df 5 1, P 5 0.0008) lygyny. For these reasons we used the male/female . 1.12
than that from closed-habitat/monogamy to closed-habitat/ classification in the rest of our analyses. This yields the fol-
polygyny. The additional significant closed-habitat/polygyny lowing set of transitions for habitat and size dimorphism.
to open-habitat/polygyny transition (Fig. 2a) represents two Step 1. Occupation of open habitats preceded the evo-
instances occurring close to the tips of the phylogeny in lution of size dimorphism (Fig. 2b). The rate of evolution
contrast to at least 15 separate open-habitat/monogamy to from the ancestral state of closed-habitat/monomorphic to
open-habitat/polygyny transitions. open-habitat/monomorphic is significantly greater than the

TABLE 2. Mean degree of dimorphism presented as the ratio of male to female log10-transformed body mass for five arbitrary cutoff points
in increasing order of sexual dimorphism. Also, the classification of mono/dimorphic species implied by the residuals from the regression of
male body mass onto female body mass after controlling for phylogeny with a generalized least-squares model and fitting a major axis regression
throughout the origin. The number of species classified as monomorphic or dimorphic in brackets.

Cutoff points 1.00 1.12 1.18 1.26 1.32 Residuals approach

Monomorphic 0.92 (26) 0.98 (46) 1.00 (53) 1.03 (59) 1.06 (67) 0.99 (54)
Dimorphic 1.39 (91) 1.48 (71) 1.52 (64) 1.55 (58) 1.59 (50) 1.10 (63)
 15585646, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.0014-3820.2002.tb01438.x, Wiley Online Library on [01/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SEXUAL DIMORPHISM IN UNGULATES 1281

TABLE 3. Test of correlated evolution and sensitivity analysis for the criterion used to define dimorphism. Dimorphism criterion shows the
five cutoff points used; L(I), likelihood for independent evolution; L(C), likelihood for correlated evolution. Likelihood ratio significance based
on x2-distribution with four degrees of freedom (see Materials and Methods).

Dimorphism criterion
States of evolution Male:Female L(I) L(C) Likelihood ratio P-values

Habitat-Mating system — 299.73 286.02 13.71 ,0.0001

Polygyny-Dimorphism .1.00 292.08 272.20 19.88 ,0.0001
.1.12 2107.04 292.44 14.60 ,0.0001
.1.18 2107.39 295.92 11.47 0.0001
.1.26 2109.50 298.14 11.36 0.0001
.1.32 2111.19 2101.76 9.43 0.0008
Habitat-Dimorphism .1.00 298.66 291.00 7.66 0.0041
.1.12 2113.56 2110.20 3.36 0.1514
.1.18 2113.97 2111.42 2.55 0.2772
.1.26 2116.08 2112.08 4.00 0.0916
.1.32 2117.77 2115.97 1.80 0.4628

nonsignificant rate of evolution to closed-habitat/dimorphic once polygyny arose, this rapidly triggered the evolution of
(x2 5 4.26, df 5 1, P 5 0.0390). Thus, ungulates were most dimorphism across a number of branches.
probably monomorphic when they first occupied open hab-
itats. DISCUSSION
Step 2. Dimorphism was more likely to evolve in open
habitats (Fig. 2b, Step 2, x2 5 7.68, df 5 1, P 5 0.0056) To test Jarman’s (1974) hypothesis for the evolution of
than in closed habitats. As shown below, this is because the sexual dimorphism, we asked which state evolved first, po-
polygynous mating system that evolved in open habitats lygyny or sexual size dimorphism? Critically, the theory as-
serts that polygyny favored the evolution of size dimorphism
strongly selected for size dimorphism.
and not that males of dimorphic species were more likely to
develop polygynous mating systems (Clutton-Brock 1982;
Mating System and Size Dimorphism Reiss 1989; Webster 1992; Clutton-Brock 1994; Abouheif
From Figure 2a we have seen that polygyny evolved after and Fairbairn 1997; Loison et al. 1999).
the occupation of open habitats. Here we test whether po- Our results are consistent with Jarman’s hypothesis and
lygyny is the most likely cause of the evolution of size di- revealed the following scenario for the evolution of sexual
morphism in open habitats. Our analysis revealed a strong size dimorphism. Ancestral monogamous and monomorphic
correlation between polygynous mating system and sexual ungulate species from closed habitats first radiated into open
size dimorphism (x2 5 29.2, df 5 4, P , 0.001). Our tests habitats where polygyny was favored. These were the first
(below) of the likely order of transitions yielded the same two intermediate stages in the development of size dimor-
results for all classifications of dimorphism except for the phism as the early polygynous species were still monomor-
male/female . 1.0 classification, which gave nonsignificant phic. Sexual size dimorphism later evolved in the polygynous
results for any transition from the ancestral state (Fig. 2c). but not in the monogamous species. This suggests that larger
As before, then, we used the 1.12 or greater criterion. body mass was favored when males fought over mating ac-
Step 1. Once polygyny had evolved following the move- cess to females. We cannot strictly rule out the alternative
ment into open habitats, it favored the evolution of size di- hypothesis that larger male body mass arose from male-male
morphism between the sexes (Fig. 2c). Rates of evolution to competition for food resources within a social group (Al-
dimorphism were higher in polygynous species (x2 5 8.82, exander et al. 1979). Although antagonistic behavior between
df 5 1, P 5 0.0030) than in monogamous ones. This rules males for access to patches of food has been observed in
out the alternative that dimorphism evolved once ungulates ungulates (Clutton-Brock et al. 1982), it seems unlikely that
occupied open habitats and then dimorphic and monogamous this competition for a resource such as vegetation, which is
species became polygynous. difficult to defend and generally of low quality, causes the
large degree of sexual dimorphism that is observed in many
Reconstructing the First Transitions extant ungulates species.
Although the sensitivity analysis applied to the criterion
To illustrate the pathway of events in these three traits we used to define dimorphism presented inconsistencies between
reconstructed ancestral states on the phylogeny (Fig. 1). The some transitions for certain cutoff points of dimorphism, the
transition from closed-habitat to open-habitat dwelling is ex- results presented were consistent when the dimorphism cutoff
pected to occur almost simultaneously in two of the main point was chosen to be a medium to large degree of dimor-
branching nodes of the tree, and did so prior to the evolution phism. This corresponds to the fact that the cutoff point used
of polygyny or sexual size dimorphism. The transition from in this study was very close to the third quartile of the dis-
monogamy to polygyny occurred before dimorphism had tribution of body mass sexual dimorphism for 142 species
been established and this happened almost simultaneously in of ungulates studied in Pérez-Barberı́a and Gordon (2000).
three major branches of the tree. Figure 1 also shows that The shift in the habitat use of ungulates early in their
 15585646, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.0014-3820.2002.tb01438.x, Wiley Online Library on [01/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1282 F. J. PÉREZ-BARBERÍA ET AL.

FIG. 2. (a) Flow diagram showing the pathway of changes in two discrete characters from the ancestral closed-habitat/monogamous
state toward the open-habitat/polygynous state via two possible intermediate states. Solid arrows represent significant transition rates (a
5 0.05); dashed arrows indicate transitions that are not significantly different from zero. (b) Flow diagram showing the likely pathway
 15585646, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.0014-3820.2002.tb01438.x, Wiley Online Library on [01/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SEXUAL DIMORPHISM IN UNGULATES 1283

evolution appears to be a consequence of the widespread tested across a range of species to investigate questions not
development of grass species and grasslands in the early Mio- amenable to experimental study.
cene (Leopold 1969; Janis 1982), and of the progressive re-
duction of the area occupied by tropical and subtropical for- ACKNOWLEDGMENTS
ests that took place from the end of Oligocene (Wolfe 1978). We thank C. M. Janis, G. Iason and J. Losos for their
These environmental changes would have revolutionized the helpful comments on the manuscript, N. Owen-Smith for
feeding behavior and pattern of aggregation of ungulates clarifying the mating system of some species, C. Nores for
(Janis 1982; McNaughton 1984). his support during this project, and two anonymous referees.
Our analysis supports the hypothesis that, in ungulates, This paper was supported by funding from the Training and
polygyny was more likely to evolve in open habitats and Mobility of Researchers Scheme of the European Commu-
happened after open habitats had been occupied. Jarman’s nities, The Scottish Executive Environment and Rural Affairs
(1974) hypothesis states that open habitats offered ungulates Department, the Leverhulme Trust, the Biotechnology and
food that was patchily distributed. This encouraged female Biological Sciences Research Council and the Natural En-
aggregation and a polygynous mating system then evolved. vironment Research Council.
The difficulty with objectively testing this hypothesis is de-
termining the spatial structure of the vegetation of the earliest LITERATURE CITED
open habitats. From the fossil record it is not possible to
deduce how widespread or clumped such vegetation resourc- Abouheif, E., and D. J. Fairbairn. 1997. A comparative analysis of
allometry for sexual size dimorphism: assessing Rensch’s rule.
es were (Bradshaw and Mitchell 1999). However, an increase Am. Nat. 149:540–562.
in visibility and reduction of vigilance time as well as anti- Alexander, R. D., J. L. Hoogland, R. D. Howard, K. M. Noonan,
predator strategies have been put forward as factors that favor P. W. Sherman. 1979. Sexual dimorphism and breeding systems
animal aggregation in open habitats (Geist 1974; Pulliam and in pinnipeds, ungulates, primates and humans. Pp. 402–435 in
N. A. Chagnon and W. Irons, eds. Evolutionary biology and
Caraco 1984; Bubenik 1985; Lagory 1986; Wrangham and human social behaviour: an anthropological perspective. Dux-
Rubenstein 1986; Quenette 1990; Pérez-Barberı́a et al. 2001), bury Press, North Scituate, MA.
which could have also happened in the earliest open habitats. Bell, R. H. V. 1969. The use of the herbaceous layer by grazing
There are two main factors favoring the evolution of animal ungulates in the Serengeti National Park, Tanzania. Ph.D. thesis,
aggregation in the ancestral open-habitat dwelling ungulates. University of Manchester, Manchester U.K.
Bergström, R., and K. Danell. 1987. Effects of simulated winter
First, a basal intercalary meristem, which may have been browsing by moose on morphology and biomass of two birch
present in the first grasses (Stebbins 1972), enabled grass species. J. Ecol. 75:533–544.
leaves to recover after grazing because the meristem was not Bradshaw, R., and F. J. G. Mitchell. 1999. The palaeoecological
affected by leaf removal (Hyder 1972; Youngner 1972). This approach to reconstructing former grazing-vegetation interac-
tions. For. Ecol. Manage. 120:3–12.
is not the case in browse plants where the browsing of buds Bubenik, A. B. 1985. Reproductive strategies in Cervids. R. Soc.
would suppress growth for that growing season. Second, an NZ Bull. 22:367–373.
item of browse (e.g., a bud, a leaf, a fruit) tends to be almost Clutton-Brock, T. H. 1982. Sexual selection in the Cervidae. Pp.
completely removed following a bite (Palo et al. 1992), 110–122 in C. M. Wemmer, ed. Biology and management of the
whereas, in most cases, items of grass are not totally depleted Cervidae. Smithsonian Institution Press, Washington, DC.
———. 1985. Size, sexual dimorphism, and polygyny in primates.
after the first bite, due to bite depth constraints (Illius et al. Pp. 51–60 in W. L. Jungers, ed. Size and scaling in primate
1995). This leads to a different spatial distribution of grazable biology. Plenum Press, New York.
and browsable resources following foraging (Hyder 1972; ———. 1994. The cost of sex. Pp. 347–362 in E. Balaban and R.
Youngner 1972; Jarman 1974; Bergström and Danell 1987), V. Short, eds. The differences between the sexes. Cambridge
Univ. Press, Cambridge, U.K.
which is likely to reduce interindividual competition when ———. 1997. Primate ecology: studies of feeding and ranging be-
animals graze in groups in open habitats relative to browsing haviour in lemurs, monkeys, and apes. Academic Press, London.
in closed habitats. This enables large patches of grass to be Clutton-Brock, T. H. and P. H. Harvey. 1977a. Primate ecology
grazed by a succession of different individuals or species and social organization. J. Zool. 183:1–39.
(Gwynne and Bell 1968; Bell 1969; McNaughton 1984), ———. 1978. Mammals, resources and reproductive strategies. Na-
ture 273:191–195.
which would favor aggregation (Wrangham and Rubenstein ———. 1983. The functional significance of variation in body size
1986; Fryxell 1991). Also the regrowth of grass can be en- among mammals. Spec. Publ. Am. Soc. Mammal. 7:633–663.
hanced following grazing (Hyder 1972), which would favor Clutton-Brock, T. H., P. H. Harvey, and B. Rudder. 1977. Sexual
aggregation of individuals in the long-term. dimorphism, socionomic sex ratio and body weight in primates.
Nature 269:797–800.
Our use of phylogenetic information and statistical meth- Clutton-Brock, T. H., S. D. Albon, P. H. Harvey, 1980. Antlers,
ods, which can detect the temporal sequence of evolutionary body size and breeding group size in the Cervidae. Nature 285:
events, shows how explicit hypotheses of evolution can be 565–566.

←
from a closed-habitat/monomorphic state toward an open-habitat/dimorphic state via two possible intermediate states. (c) Flow diagram
showing the likely pathway from a monogamous/monomorphic state towards a polygynous/dimorphic state via two possible intermediate
states. Panels b and c also show the P-values of the sensitivity analysis applied to the dimorphism classification using five different
cutoff points for the ratio of male to female body mass (namely, 1.00, 1.12, 1.18, 1.26, 1.32; numbered from one to five in the figure
in increasing order of sexual dimorphism). Underlined P-values correspond to 1.12 criterion of dimorphism, which was the final criterion
used in this study (see Materials and Methods and Results for further details).
 15585646, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.0014-3820.2002.tb01438.x, Wiley Online Library on [01/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1284 F. J. PÉREZ-BARBERÍA ET AL.

Clutton-Brock, T. H., F. E. Guinness, S. D. Albon. 1982. Red deer: Lack, D. 1968. Ecological adaptations for breeding in birds. Me-
the behavior and ecology of two sexes. Univ. of Chicago Press, thuen, London.
Chicago, IL. Lagory, K. E. 1986. Habitat, group size, and the behaviour of white-
Corbet, G. B., and J. E. Hill. 1986. A world list of mammalian tailed deer. Behaviour 98:168–179.
species. British Museum (Natural History), London. Leopold, E. B. 1969. Late Cenozoic palynology. Pp. 377–438 in
Darwin, C. 1871. The descent of man and selection in relation to R. S. Tschudy and R. A. Scott, eds. Aspects of palynology. Wiley
sex. Murray, London. Interscience, New York.
Essop, M. F., E. H. Harley, and I. Baumgarten. 1997. A molecular Lindenfors, P., and B. S. Tullberg. 1998. Phylogenetic analyses of
phylogeny of some bovidae based on restriction-site mapping primate size evolution: the consequences of sexual selection.
of mitochondrial DNA. J. Mammal. 78:377–386. Biol. J. Linn. Soc. 64:413–447.
Fairbairn, D. J. 1997. Allometry for sexual size dimorphism: pattern Loison, A., J.-M. Gaillard, C. Pelabon, and N. G. Yoccoz. 1999.
and process in the coevolution of body size in male and female. What factors shape sexual size dimorphism in ungulates? Evol.
Annu. Rev. Ecol. Syst. 28:659–687. Ecol. Res. 1:611–633.
Fryxell, J. M. 1991. Forage quality and aggregation by large her- Lovich, J. E. and J. W. Gibbons. 1992. A review of techniques for
bivores. Am. Nat. 38:478–498. quantifying sexual size dimorphism. Growth Dev. Aging 56:
Garland, T., and C. M. Janis. 1993. Does metatarsal femur ratio 269–281.
predict maximal running speed in cursorial mammals? J. Zool. Mace, G. M. 1979. The evolutionary ecology of small mammals.
229:133–151. Ph.D. thesis. University of Sussex, Brighton, U.K.
Garland, T., A. W. Dickerman, C. M. Janis, and J. A. Jones, 1993. McNaughton, S. J. 1984. Grazing lawns: animals in herds, plant
Phylogenetic analysis of covariance by computer simulation. form, and coevolution. Am. Nat. 124:863–886.
Syst. Biol. 42:265–292. Novak, R. M. 1991. Walker’s mammals of the world. Johns Hopkins
Gatesy, J., G. Amato, E. Vrba, G. Schaller, and R. DeSalle. 1997. Univ Press, Baltimore, MD.
A cladistic analysis of mitochondrial ribosomal DNA from the Pagel, M. 1992. A method for the analysis of comparative data. J.
bovidae. Mol. Phylogenet. Evol. 7:303–319. Theor. Biol. 56:431–442.
Geist, V. 1974. On the relationship of social evolution and ecology ———. 1994. Detecting correlated evolution on phylogenies: a
in ungulates. Am. Zool. 14:205–220. general method for the comparative analysis of discrete char-
Gentry, A. W. 1978. Bovidae. Pp. 540–572 in V. Maglio and H. acters. Proc. R. Soc. Lond. B 255:37–45.
Cooke, eds. Evolution of African mammals. Harvard Univ. ———. 1997. Inferring evolutionary processes from phylogenies.
Press, Cambridge, MA. Zool. Scr. 26:331–348.
———. 1992. The subfamilies and tribes of the family bovidae. ———. 1999a. Inferring the historical patterns of biological evo-
Mammal Rev. 22:1–32. lution. Nature 401:877–884.
Gentry, A. W., and J. J. Hooker. 1988. The phylogeny of the Ar- ———. 1999b. The maximum likelihood approach to reconstruct-
tiodactyla. Pp. 235–272 in M. J. Benton, ed. The phylogeny and ing ancestral character states of discrete characters on phylog-
classification of the Tetrapods, Vol. 2. Mammals. Systematics
enies. Syst. Biol. 48:612–622.
Association Special Volume 35B. Clarendon Press, Oxford, U.K.
Palo, R. T., R. Bergstrom, and K. Danell. 1992. Digestibility, dis-
Gosling, L. M. 1986. The evolution of mating strategies in male
antelopes. Pp. 244–281 in D. I. Rubenstein and R. W. Wrangham, tribution of phenols, and fiber at different twig diameters of birch
eds. Ecological aspects of social evolution. Birds and Mammals. in winter: implication for browsers. Oikos: 65:450–454.
Princeton Univ. Press, Princeton, NJ. Pérez-Barberı́a, F. J., and I. J. Gordon. 1998. The influence of sexual
Grafen, A. 1989. The phylogenetic regression. Philos. Trans. R. dimorphism in body size and mouth morphology on diet selec-
Soc. Lond. B 326:119–157. tion and sexual segregation in cervids. Acta Vet. Hung. 46:
Gwynne, M. D., and R. H. V. Bell. 1968. Selection of vegetation 357–367.
components by grazing ungulates in the Serengeti National Park. ———. 1999a. Body size dimorphism and sexual segregation in
Nature 220:390–393. polygynous ungulates: an experimental test with Soay sheep.
Hyder, D. N. 1972. Defoliation in relation to vegetative growth. Oecologia 120:258–267.
Pp. 304–317 in V. B. Youngner and C. M. McKell, eds. The ———. 1999b. The functional relationship between feeding type
biology and utilization of grasses. Academic Press, New York. and jaw and cranial morphology in ungulates. Oecologia 118:
Illius, A. W., I. J. Gordon, J. D. Milne, and W. Wright. 1995. Costs 157–165.
and benefits of foraging on grasses varying in canopy structure ———. 2000. Differences in body mass and oral morphology be-
and resistance to defoliation. Funct. Ecol. 9:894–903. tween the sexes in the Artiodactyla: Evolutionary relationships
Janis, C. M. 1982. Evolution of horns in ungulates: ecology and with sexual segregation. Evol. Ecol. Res. 2:667–684.
paleoecology. Biol. Rev. Camb. Philos. Soc 57:261–317. ———. 2001. Relationships between oral morphology and feeding
Janis, C. M., and K. M. Scott. 1987. The interrelationships of higher style in the Ungulata: a phylogenetically controlled evaluation.
ruminant families with special emphasis on the members of the Proc. R. Soc. Lond. B 268:1021–1030.
Cervoidea. Am. Mus. Novit. 2893:1–85. Pérez-Barberı́a, F. J., I. J. Gordon, and C. Nores. 2001. Evolutionary
———. 1988. The phylogeny of the Ruminantia (Artiodactyla, transitions among feeding styles and habitats in ungulates. Evol.
Mammalia). Pp. 273–282 in M. J. Benton, ed. The phylogeny Ecol. Res. 3:221–230.
and classification of the Tetrapods, Vol. 2. Mammals. System- Pulliam, H. R., and T. Caraco. 1984. Living in groups: is there an
atics Association Special Volume 35B. Clarendon Press, Oxford, optimal group size? Pp. 122–147 in J. R. Krebs and N. B. Davies,
U.K. eds. Behavioural ecology: an evolutionary approach. Blackwell
Jarman, P. J. 1974. The social organisation of antelope in relation Scientific Publications, Oxford, U.K.
to their ecology. Behaviour 48:215–266. Quenette, P. Y. 1990. Functions of vigilance behaviour in mammals:
———. 1983. Mating system and sexual dimorphism in large, ter- a review. Acta Oecol. 11:801–818.
restrial, mammalian herbivores. Biol. Rev. 58:485–520. Ralls, K. 1977. Sexual dimorphism in mammals: avian models and
Kingdon, J. 1982. East African mammals. An atlas of evolution in unanswered questions. Am. Nat. 111:917–938.
Africa. Vol. IIIC. Bovids. Academic Press, London. Ranta, E., A. Laurila, and J. Elmberg. 1994. Reinventing the
———. 1997. The Kingdon field guide to African mammals. Ac- wheel—analysis of sexual dimorphism in body-size. Oikos 70:
ademic Press, San Diego, CA. 313–321.
Komers, P. E. 1996a. Obligate monogamy without paternal care in Reiss, M. J. 1989. The allometry of growth and reproduction. Cam-
Kirk’s dik-dik. Anim. Behav. 51:131–140. bridge Univ. Press, Cambridge, U.K.
———. 1996b. Conflicting territory use in males and females of a Selander, R. K. 1972. Sexual selection and dimorphism in birds.
monogamous ungulate, the Kirk’s dik-dik. Ethology 102: Pp. 180–230 in B. Campbell, ed. Sexual selection and the descent
568–579. of man. Aldine-Altherton Publishing, Chicago, IL.
 15585646, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.0014-3820.2002.tb01438.x, Wiley Online Library on [01/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SEXUAL DIMORPHISM IN UNGULATES 1285

Shine, R. 1989. Ecological causes for the evolution of sexual di- Weckerly, F. W. 1998. Sexual-size dimorphism: influence of mass
morphism: a review of the evidence. Q. Rev. Biol. 64:419–461. and mating systems in the most dimorphic mammals. J. Mammal.
Stebbins, G. L. 1972. The evolution of the grass family. Pp. 1–17 79:33–52.
in V. B. Youngner and C. M. McKell, eds. The biology and Wiley, R. H. 1974. Evolution of social organization and life-history
utilization of grasses. Academic Press, New York. patterns among grouse. Q. Rev. Biol. 49:201–227.
Stirling, I. 1975. Factors affecting the evolution of social behavior Wolfe, J. A. 1978. A paleobotanical interpretation of Tertiary cli-
in the Pinnipedia. Explor. Mer. 169:205–212. mates in the Northern Hemisphere. Am. Sci. 66:694–703.
Sokal, R. R., and F. J. Rohlf, 1995. Biometry: the principles and Wrangham, R. W., and D. I. Rubenstein. 1986. Social evolution in
practice of statistics in biological research. W.H. Freeman, New birds and mammals. Pp. 452–470 in D. I. Rubenstein and R. W.
York. Wrangham, eds. Ecological aspects of social evolution. Birds
Trivers, R. L. 1972. Parental investment and sexual selection. Pp. and mammals. Princeton Univ. Press, Princeton, NJ.
136–179 in B. Campbell, ed. Sexual selection and the descent Youngner, V. B. 1972. Physiology of defoliation and regrowth. Pp.
of man. Aldine-Altherton Publishing, Chicago, IL. 292–303 in V. B. Youngner and C. M. McKell, eds. The biology
Webster, M. S. 1992. Sexual dimorphism, mating system and body and utilization of grasses. Academic Press, New York.
size in new-world blackbirds (Icterinae). Evolution 46:
1621–1641. Corresponding Editor: J. Huelsenbeck