Syst. Biol. 52(2):206–228, 2003
Molecular and Morphological Phylogenies of Ruminantia and the Alternative Position
of the Moschidae
ALEXANDRE HASSANIN1,2,3AND EMMANUEL J. P. DOUZERY4
1Laboratoire de Zoologie Mammif `
eres et Oiseaux, USM 601, Origine, Structure, et Evolution de la Biodiversit´
ematique et Evolution,
eum National d’Histoire Naturelle, 55 rue Buffon, 75231 Paris Cedex 05, France; E-mail: email@example.com
2Service de Syst´
eum National d’Histoire Naturelle, 43 rue Cuvier, 75005 Paris, France
3Service de Biosyst´
e Pierre et Marie Curie, 9 quai St-Bernard, 75252 Paris Cedex 05, France
4Laboratoire de Pal´
eobiologie et Phylog´
enie, Institut des Sciences de l’Evolution UMR 5554 CNRS, Universit´
e Montpellier II, CC064,
Place E. Bataillon, 34 095 Montpellier Cedex 05, France; E-mail: firstname.lastname@example.org
Abstract.—The ruminants constitute the largest group of ungulates, with >190 species, and its distribution is widespread
throughout all continents except Australia and Antarctica. Six families are traditionally recognized within the suborder
Ruminantia: Antilocapridae (pronghorns), Bovidae (cattle, sheep, and antelopes), Cervidae (deer), Girafﬁdae (giraffes and
okapis), Moschidae (musk deer), and Tragulidae (chevrotains). The interrelationships of the families have been an area of
controversy among morphology, palaeontology, and molecular studies, and almost all possible evolutionary scenarios have
been proposed in the literature. We analyzed a large DNA data set (5,322 nucleotides) for 23 species including both mito-
chondrial (cytochrome b, 12S ribosomal RNA (rRNA), and 16S rRNA) and nuclear (κ-casein, cytochrome P-450, lactoferrin,
and α-lactalbumin) markers. Our results show that the family Tragulidae occupies a basal position with respect to all other
ruminant families, conﬁrming the traditional view that separates Tragulina and Pecora. Within the pecorans, Antilocapridae
and Girafﬁdae emerge ﬁrst, and the families Bovidae, Moschidae, and Cervidae are allied, with the unexpected placement
of Moschus close to bovids rather than to cervids. We used these molecular results to assess the homoplastic evolution of
morphological characters within the Ruminantia. A Bayesian relaxed molecular clock approach based on the continuous
autocorrelation of evolutionary rates along branches was applied to estimate the divergence ages between the major clades
of ruminants. The evolutionary radiation of Pecora occurred at the Early/Late Oligocene transition, and Pecoran families
diversiﬁed and dispersed rapidly during the Early and Middle Miocene. We propose a biogeographic scenario to explain the
extraordinary expansion of this group during the Cenozoic era. [Bayesian relaxed clock; Bovidae; molecules; morphology;
Moschidae; phylogeny; Ruminantia.]
One of the most spectacular aspects of mammalian
evolution during the Neogene phase of the Cenozoic
era was the rapid diversiﬁcation and expansion of
the ruminant families (Cetartiodactyla, Ruminantia).
These highly specialized herbivorous even-toed ungu-
lates were and are particularly adapted for gathering
and processing plant food. They have a speciﬁc denti-
tion characterized by the presence of an incisiform lower
canine and a horny pad that replaces the upper incisors.
In addition, living ruminants possess a compartmental-
ized stomach that serves as a chamber for fermentation
of cellulose by symbiotic microorganisms.
Ruminantia is the only cetartiodactyl suborder for
which living and fossil members are clearly united by
an osteological apomorphy, the fusion of the cuboid and
navicular bones in the tarsus (Lavocat, 1955; Romer,
1966). Today, Ruminantia constitutes one of the ma-
jor groups of large mammals, with 192 species cur-
rently described (Dung et al., 1993; Grubb, 1993) and a
natural distribution that covers all continents with the
exceptions of Australia and Antarctica. Six extant fami-
lies are traditionally recognized on the basis of charac-
ters of the cranial appendages, limbs, and dentition (e.g.,
Janis and Scott, 1987; Grubb, 1993; Scott and Janis, 1993).
The results of all recent morphopaleontological stud-
ies agree in the designation of the family Tragulidae as
the most basal branch within Ruminantia (Webb and
Taylor, 1980; Bouvrain and Geraads, 1985; Janis and Scott,
1988; Vislobokova, 1990; Scott and Janis, 1993). There-
fore, the basic subdivision of the Ruminantia into the
infraorders Tragulina and Pecora proposed by Flower in
1883 is now widely accepted. Tragulina is represented
by a single extant family, the Tragulidae, which incor-
porates three chevrotain or mouse deer genera of the
tropical lowland forests of the Old World: Hyemoschus in
Central Africa, Moschiola in India, and Tragulus in south-
east Asia (Grubb, 1993). Pecora comprises the ﬁve living
families Antilocapridae (pronghorns; a single species oc-
curring in open countries of North America), Bovidae
(cattle, sheep, and antelopes), Cervidae (deer), Girafﬁdae
(giraffes and okapis from Africa), and Moschidae (musk
deer; a single genus living in highland forests of Asia)
(Janis and Scott, 1987). In modern times, Bovidae
and Cervidae represent the greatest degree of taxo-
nomic and geographical diversity among the Rumi-
nantia, with 48 bovid genera from Africa, most of
Eurasia, and North America and 16 cervid genera from
mainly America and Eurasia. All Pecora exhibit cranial
appendages—permanent unbranched horns for Bovi-
dae, deciduous antlers for Cervidae, skin-covered ossi-
cones for Girafﬁdae, deciduous and branched horns for
antilocaprids—except Hydropotes (a case of secondary
loss of antlers; Randi et al., 1998) and Moschus. These
frontal appendages are probably not homologues, al-
though they have tentatively been used to unite some
pecoran families: Bovidae with Antilocapridae (O’Gara
and Matson, 1975), Bovidae with Girafﬁdae (Hamilton,
1978), Bovidae plus Antilocapridae with Girafﬁdae
(Gentry and Hooker, 1988), and all Pecora except living
Moschidae and extinct “Gelocidae” into horned Eupec-
ora (Webb and Taylor, 1980).
Except for the major dichotomy that separates Pecora
and Tragulina, there is no consensus among morphol-
ogists for interrelationships of the living families, and
2003 HASS ANIN AND DO UZERY—PHYLO GENY O F RUMINANTIA 207
almost all possible evolutionary scenarios have been
proposed in the literature. Gatesy et al. (1992:442) noted
that “there are only 15 possible rooted cladograms
relating the families Bovidae, Cervidae, Antilocapridae,
and Girafﬁdae; 11 have been proposed in the literature”
(Fig. 1). Janis and Scott (1988) pointed out that Rumi-
nantia is a problematic group in phylogeny because it
experienced several evolutionary radiations during the
Tertiary changes from forested to more open habitats,
and different families evolved the same characters in
parallel. The morphological traits that characterize the
extant families become unambiguously detectable in the
fossil record of the Miocene epoch, with the emergence
of Antilocapridae in the New World and Tragulidae,
Bovidae, Cervidae, Girafﬁdae, and Moschidae in the
Old World. However, Eocene and Oligocene deposits
of North America and Eurasia already contained nu-
merous fossil ruminants that are difﬁcult to assign to
either Tragulina or Pecora. Most of these remains have
been generally included into the Hypertraguloidea,
a superfamily that is either the sister group of the
living Tragulidae or is the most primitive group of the
Ruminantia (Vislobokova, 1998).
Several molecular investigations have been conducted
on the suborder Ruminantia. The phylogenetic stud-
ies were initially performed using mitochondrial nu-
cleotide sequences of cytochrome b(Irwin et al., 1991;
Honeycutt et al., 1995; Montgelard et al., 1997; Randi
et al., 1998; Hassanin and Douzery, 1999b; Matthee and
Robinson, 1999; Su et al., 1999), ribosomal RNAs (rRNAs)
(Kraus and Miyamoto, 1991; Miyamoto et al., 1993;
Montgelard et al., 1997), or cytochrome coxidase II (Hon-
eycutt et al., 1995). More recent studies have involved
nuclear sequences of the κ-casein (Cronin et al., 1996),
β-casein (Gatesy et al., 1996), and γ-ﬁbrinogen genes
(Gatesy, 1997) and various genetically independent loci
(Hassanin and Douzery, 1999a; Gatesy and Arctander,
2000; Matthee and Davis, 2001; Matthee et al., 2001). All
these reports indicate that the pecoran families form a
monophyletic assemblage distinct from the Tragulidae,
conﬁrming evidence from such diverse disciplines as
anatomy and palaeontology (e.g., Simpson, 1945; Gen-
try and Hooker, 1988; Scott and Janis, 1993), ethology
(Dubost, 1965), and karyology (Gallagher et al., 1996). In
contrast, relationships among the pecoran families have
remained an especially unresolved issue. The studies on
the mitochondrial cytochrome b(Irwin et al., 1991; Randi
et al., 1998), rRNAs (Kraus and Miyamoto, 1991), and
cytochrome coxidase II (Honeycutt et al., 1995) yielded
an unresolved multifurcation of antilocaprids, bovids,
cervids, and girafﬁds. Moreover, these analyses did not
provide support for the monophyly of the families Bovi-
dae or Cervidae. Kraus and Miyamoto (1991) attributed
this poor resolution to the rapid radiation of the pecoran
lineages over a short period of time in the Late Oligocene
to Early Miocene. A nuclear marker, κ-casein, was ﬁrst
introduced by Chikuni et al. (1995) to analyze Pecora
phylogeny and was used again with greater taxon sam-
pling by Cronin et al. (1996). In the latter study, Antilo-
capridae emerged ﬁrst relative to a trifurcation involving
Bovidae, Cervidae, and Girafﬁdae, but the support
for the different nodes was low. The same result was
observed after combination of the mitochondrial cy-
tochrome band 12S rRNA data (Montgelard et al., 1997).
The use of nuclear κ-casein and β-casein exons with mi-
tochondrial cytochrome bsequences reinforced a sister-
group relationship of antilocaprids relative to bovids,
cervids, and girafﬁds and clustered Cervus with Giraffa
(Gatesy et al., 1996). The association between Cervi-
dae and Girafﬁdae was also recovered with nuclear γ-
ﬁbrinogen sequences in the context of the Cetartiodactyla
phylogeny, but only three pecoran families were repre-
sented in that study (Gatesy, 1997). The comparison of
nuclear amino acid sequences indicated that Giraffa and
Antilocapra are sister groups (Miyamoto and Goodman,
1986), and the use of ribonuclease sequences conﬁrmed
that these two taxa are closer to Bovidae than to Cervidae
(Beintema et al., 1988).
In all of these studies, the taxonomic sample was
not appropriately determined for inferring relationships
among Pecora families because (1) the biodiversity of
the Bovidae and Cervidae was inadequately represented,
with the exception of the studies conducted by Cronin
et al. (1996) and Randi et al. (1998), (2) the isolated
branch leading to Giraffa was not broken by the in-
clusion of the second living girafﬁd genus Okapia, and
(3) no molecular study included the Moschidae and its
single extant genus Moschus, with the exception of the
survey of moschid species by Su et al. (1999). Conclud-
ing their study of the rapid cladogenesis among peco-
rans using 12S and 16S rRNA mitochondrial markers,
Kraus and Miyamoto (1991:128) noted that phylogenetic
hypotheses “must be further tested with comparative in-
formation, which, from a molecular perspective, requires
comparable sequence data from the genus Moschus.”
However, the musk deer has never been sampled despite
the reiterated claim that “molecular studies promise to
be of great use in clarifying the relationships of living
[pecoran] taxa; however, the usefulness of these studies
at present is limited by the lack of data on Moschus” (Scott
and Janis, 1993:300).
Two recent studies by Gatesy and Arctander (2000)
and Matthee et al. (2001) partially circumvent these three
problems by adding Okapia and a large taxon sampling
for Bovidae, and for outgroups for ruminants. Moreover,
these authors combined several mitochondrial and nu-
clear markers and showed that (1) Tragulina clearly sep-
arates from Pecora, (2) Antilocapridae is a major and
earliest diverging lineage among pecorans, and (3) Bovi-
dae and Cervidae cluster together as the sister group of
We sampled at least two distant genera from each
pecoran family and subfamily except for the mono-
generic Moschidae and Antilocapridae. For Cervidae,
we included two members of each of the three ma-
jor lineages: Plesiometacarpalia and Old World and
New World Telemetacarpalia (Randi et al., 1998).
For Bovidae, we incorporated one representative of
each of the major tribes previously identiﬁed as
monophyletic (Hassanin and Douzery, 1999a, 1999b;
208 SYSTEMATIC BIOLOGY VOL. 52
FIGURE 1. Morphological hypotheses for the interfamily relationships within Pecora.
2003 HASS ANIN AND DO UZERY—PHYLO GENY O F RUMINANTIA 209
Gatesy and Arctander, 2000; Matthee et al., 2001), i.e.,
Bovini, Boselaphini, Tragelaphini, Antilopini, Aepyc-
erotini, Alcelaphini, Caprini s.l., Cephalophini, Hippo-
tragini, and Reduncini. We attempted to break the long
Pecora ancestral segment by adding a tragulid species.
This ruminant ingroup, Tragulina +Pecora, was rooted
by one cetacean (Balaenopteridae) and one ancodontan
(Hippopotamidae) because these two taxa are consid-
ered the closest living relatives of Ruminantia (Irwin and
Arnason, 1994; Gatesy et al., 1996; Gatesy, 1997, 1998;
Montgelard et al., 1997; Shimamura et al., 1997).
We sequenced molecular markers that could have
experienced different mutational and selective con-
straints during cetartiodactyl evolution, i.e., mitochon-
drial and nuclear markers belonging to two different
genomes with different rates and patterns of evolu-
tion and protein-encoding, ribosomal, and noncoding
markers with different selective pressures. Because they
have been previously used to decipher the cetartio-
dactyl phylogeny, the following protein-encoding mark-
ers were used: the mitochondrial cytochrome b(Cyb)
gene (e.g., Irwin et al., 1991; Irwin and Arnason, 1994;
Stanley et al., 1994; Tanaka et al., 1996; Montgelard
et al., 1997; Hassanin et al., 1998b; Randi et al., 1998;
Hassanin and Douzery, 1999b; Matthee and Robinson,
1999) and the nuclear κ-casein exon 4 (κCas) (Chikuni
et al., 1995; Cronin et al., 1996; Gatesy et al., 1996).
The ribosomal markers were the mitochondrial 12S and
16S rRNAs (e.g., Miyamoto et al., 1989; Kraus and
Miyamoto, 1991; Allard et al., 1992; Gatesy et al., 1992,
1997; Montgelard et al., 1997), and the non-coding mark-
ers were the nuclear promotor segment of the lactoferrin
gene (Lf ), the 30untranslated region of the aromatase cy-
tochrome P450 gene (Cyp) (Pitra et al., 1997; Hassanin and
Douzery, 1999a), and intron 2 of the α-lactalbumin gene
The aims of this study were (1) to evaluate the con-
tribution of seven mitochondrial and nuclear markers
(Cyb, 12S rRNA, 16 rRNA, αLAlb,Cyp,κCas, and Lf )
and their combination (5,322 characters) to decipher the
phylogeny of all ruminant Pecora families, including
Moschidae; (2) to compare the phylogenetic pictures in-
ferred from molecular and morphological characters;
(3) to determine the time frame for Pecora evolution
under a Bayesian relaxed molecular clock assuming an
Early Miocene (16.4–23.8 million years ago [MYA]) a pri-
ori divergence of both bovid and cervid families; and (4)
to build a biogeographic scenario explaining the current
distribution of ruminants.
MATERIAL AND METHODS
Twenty-three taxa were analyzed in this study
(Table 1). Two taxa, Balaenoptera (family Balaenopteri-
dae) and Hippopotamus (family Hippopotamidae), were
used as outgroups to root the ruminant tree. These out-
group taxa were chosen because molecular investiga-
tions have shown that hippos and cetaceans are closely
related to the ruminants (Gatesy, 1997; Shimamura et al.,
1997). All extant families of the suborder Ruminantia
were represented in this study: the Tragulidae with Trag-
ulus, the Antilocapridae with Antilocapra, the Girafﬁdae
with Giraffa and Okapia, the Moschidae with Moschus, the
Cervidae with 6 species, and the Bovidae with 10 species.
For the Cervidae, we incorporated at least one represen-
tative of each of the major clades of Randi et al. (1998):
Cervus (Cervinae), Muntiacus (Muntiacinae), Capreolus
and Hydropotes (Alcini +Hydropotinae +Capreolini
clade), and Odocoileus and Rangifer (Odocoileini +
Rangiferini clade). For the Bovidae, we integrated
one representative of each of the major lineages of
Hassanin and Douzery (1999a, 1999b): Aepyceros
(Aepycerotini), Bos (Bovini), Boselaphus (Boselaphini),
Cephalophus (Cephalophini), Damaliscus (Alcelaphini),
Gazella (Antilopini), Hippotragus (Hippotragini), Ovis
(Caprini s.l.), Redunca (Reduncini), and Tragelaphus
Total DNA was extracted from blood, hair, skin, mus-
cles, or other soft tissues following the protocols de-
scribed by Sambrook et al. (1989) or Winnepenninckx
et al. (1993) and from bone fragments of museum speci-
mens using the procedure of Hassanin et al. (1998b).
The entire Cyb gene was ampliﬁed by the polymerase
chain reaction (PCR) with the primers described by
Hassanin et al. (1998b) and Hassanin and Douzery
(1999b). For Moschus, a 3.0-kilobase (kb) fragment
spanning the Cyb to the ﬁrst half of the 12S rRNA
was ampliﬁed and cloned as described by Douzery
and Randi (1997). The complete Cyb was sequenced
in four clones using a set of primers described by
Irwin et al. (1991). The ampliﬁcation of the complete 12S
rRNA gene was performed as described by Hassanin
and Douzery (1999a). For Hippopotamus, Okapia, and
Moschus, a 2.8-kb fragment spanning the complete
12S to 16S rDNA was ampliﬁed with primers R1 (50-
AAAGCAAGGCACTGAAAATGCCTAGA-30) and S1
in the tRNAPhe and tRNALeu, respectively, using 30 cycles
of denaturation (94◦C, 45 sec), annealing (50◦C, 45 sec)
and elongation (72◦C, 210 sec), and Taq polymerase from
A.T.G.C. The efﬁciency of PCRs was strongly increased
using DMSO at a 5% ﬁnal concentration. PCR prod-
ucts were cloned in the pGEM-T vector (Promega) us-
ing the Escherichia coli JM109 competent cells for trans-
formations. Recombinant plasmids were puriﬁed from
three, ﬁve, and one positive clone for hippo, okapi,
and musk deer, respectively, and DNA sequencing was
conducted on both strands using [α35S]dATP and the
T7Sequencing mixes kit (Pharmacia Biotech). For Giraffa,
Capreolus, Rangifer, Hippotragus, and Redunca, a 2.0-kb
fragment spanning the end of the 12S to the com-
plete 16S rDNA was ampliﬁed with primers 270(50-
TATACCGCCATCTTCAGCAAAC-30) and S1 and di-
rectly sequenced using [α33P]ddNTP and the Thermo
Sequenase radiolabeled terminator cycle Sequencing kit
TABLE 1. Taxonomy according to Grubb (1993) for the species used in this study. The accession numbers and corresponding references (numbers in parenthesesa) for all sequences
used in this study are provided for the mitochondrial cytochrome b(Cyb), 12S rRNA (12S), and 16S rRNA (16S) and the nuclear α-lactalbumin intron 2 (αLAlb), aromatase cytochrome
P450 (Cyp), lactoferrin promotor (Lf ), and κ-casein exon 4 (κCas).
Species Common name Cyb 12S 16S αLAlb Cyp Lf κCas
Giraffa camelopardalis giraffe AY121992 (1) AY121986 (1) AY122046 (1) AY122015 (1) AY122007 (1) AY122041 (1) U37516 (2)
Okapia johnstoni okapi AY121993 (1) AY121987 (1) AY122044 (1) AY122016 (1) AY122008 (1) AY122042 (1) AY121996 (1)
Moschus moschiferus musk deer AY121995 (1) AY121988 (1) AY122045 (1) AY122033 (1) AY122009 (1) AY122043 (1) AY121997 (1)
Antilocapra americana pronghorn AF091629 (3) AF091706 (3) M55540 (4) AY122014 (1) AF091666 (3) AF091694 (3) U37515 (2)
Cervus elaphus red deer AJ000021 (5) AF091707 (3) M35875 (4)bAY122017 (1) AF091667 (3) AF091636 (3) U37505 (2)
Muntiacus reevesi Chinese muntjak AJ000023 (5) M35877 (6) M35877 (6) AY122018 (1) AY122010 (1) AY122037 (1) U37509 (2)
Odocoileus hemionus black-tailed deer AF091630 (3) AF091708 (3) M35874 (4)cAY122022 (1) AF091665 (3) AF091637 (3) U37360 (2)
Rangifer tarandus reindeer AJ000029 (5) AY121989 (1) AY122047 (1) AY122019 (1) AY122005 (1) AY122038 (1) U37503 (2)
Hydropotes inermis Chinese water deer AJ000028 (5) M35876 (6) M35876 (6) AY122020 (1) AY122006 (1) AY122039 (1) Douzery and
Capreolus capreolus European roe deer AJ000024 (5) AY121990 (1) AY122048 (1) AY122021 (1) no accession AY122040 (1) U37363 (2)
Bos taurus domestic cow V00654 (8) V00654 (8) V00654 (8) AY122023 (1) Z32741 (9) L19985 (10) X14908 (11)
Boselaphus tragocamelus nilgai AJ222679 (12) M86494 (13) M86494 (13) AY122024 (1) AF091672 (3) AF091642 (3) AF030331 (14)
Tragelaphus imberbis lesser kudu AF036279 (12) AF091697 (3) M86493 (13) AY122025 (1) AF091677 (3) AF091649 (3) AF030330 (14)
Ovis aries domestic sheep AF034730 (15) AF091699 (3) AF010406 (16) AY122026 (1) AF091685 (3) AF091651 (3) X51822 (17)
Hippotragus niger sable antelope AF036285 (12) AF091709 (3) AY122049 (1) AY122027 (1) AF091684 (3) AF091652 (3) AY122001 (1)
Damaliscus pygargus blesbok AF036287 (12) M86499 (13) M86499 (13) AY122028 (1) AF091683 (3) AF091653 (3) AY122002 (1)
Gazella granti Grant’s gazelle AF034723 (15) AF091700 (3) M86501 (6)dAY122029 (1) negative PCR AF091654 (3) AY122003 (1)
Cephalophus dorsalis bay duiker AF091634 (3) AF091701 (3) M86498 (6)eAY122030 (1) AF091682 (3) AF091655 (3) AY122000 (1)
Redunca fulvorufula mountain reedbuck AF036284 (12) AF091704 (3) AY122050 (1) AY122031 (1) AF091681 (3) AF091658 (3) AY121999 (1)
Aepyceros melampus impala AF036289 (12) M86496 (13) M86496 (13) AY122032 (1) AF091680 (3) AF091659 (3) AY121998 (1)
Tragulus javanicus lesser Malay chevrotain AY121994 (1) AY121991 (1) M55536 (18)fAY122013 (1) negative PCR AY122034 (1) D14381 (19)
Balaenoptera physalus ﬁn whale X61145 (20) X61145 (20) X61145 (20) AY122011 (1) AY122004 (1) AY122035 (1) U53888 (21)
Hippopotamus amphibius hippo Y08813 (18) Y08810 (18) AJ010813, AY122012 (1) negative PCR AY122036 (1) U53889 (21)
a1=this paper; 2 =Cronin et al. (1996); 3 =Hassanin and Douzery (1999a); 4 =Kraus and Miyamoto (1991); 5 =Randi et al. (1998); 6 =Miyamoto et al. (1990); 7 =Pitra et al. (1997); 8 =Anderson et al.
(1982); 9 =Vanselow and F¨urbass (1995); 10 =Seyfert et al. (1994); 11 =Alexander et al. (1988); 12 =Hassanin and Douzery (1999b); 13 =Allard et al. (1992); 14 =Ward et al. (1997); 15 =Hassanin et al.
(1998b); 16 =Hiendleder et al. (1998); 17 =Furet et al. (1990); 18 =Montgelard et al. (1997); 19 =Chikuni et al. (1995); 20 =Arnason et al. (1991); 21 =Gatesy et al. (1996).
bInterspeciﬁc chimera with Cervus unicolor.
cInterspeciﬁc chimera with Odocoileus virginianus.
dInterspeciﬁc chimera with Gazella thomsoni.
eInterspeciﬁc chimera with Cephalophus maxwelli.
fInterspeciﬁc chimera with Tragulus napu.
2003 HASS ANIN AND DO UZERY—PHYLO GENY O F RUMINANTIA 211
Partial sequences of the nuclear gene encoding aro-
matase cytochrome P450, i.e, positions 2,992–3,185 of
the Bos taurus sequence (accession Z32741), were ac-
quired using the oligonucleotides determined by Pitra
et al. (1997). The promotor segment of the lactoferrin-
encoding gene, i.e., positions 322-647 of the Bos tau-
rus sequence (accession L19985), was generated us-
ing the primers given by Hassanin and Douzery
(1999a). Exon 4 of the κ-casein gene (positions 84–
485 of exon 4 of Bos taurus sequence, accession
X14908) was obtained using the primers designed by
Ettore Randi (unpubl.). Intron 2 of the α-lactalbumin
gene (αLAlb) was ampliﬁed using the primer 50-
ATCTGTAACATCTCCTGTGA-30positioned in exon 2
and the primer 50-TCAGTAAGRTCATCATCCAG-30lo-
cated in exon 3. Both strands of all amplicons were
directly sequenced using the Thermo Sequenase cycle
sequencing kit (Amersham). The sequences have been
deposited in the EMBL/GenBank/DDBJ databases un-
der the accession numbers speciﬁed in Table 1.
Sequences were aligned using the MUST package
(Philippe, 1993). Indels were coded according to Barriel
(1994), with introduction of I and D character states and
question marks representing the methodological conse-
quences of gap coding. The alignment of 12S rRNA and
16S rRNA sequences was reﬁned using the secondary
structure model for mammals (Springer and Douzery,
1996; Schnare et al., 1996). All regions with ambigui-
ties for DNA alignment were excluded from the anal-
yses. Alignments with indels have been deposited in
the EMBL nucleotide database under the following ac-
cession numbers: ALIGN 000484 (Cyp), ALIGN 000487
(12S), ALIGN 000488 (κCas), ALIGN 000489 (αLAlb),
ALIGN 000490 (Lf ), and ALIGN 000491 (16S). To ben-
eﬁt from the maximum number of molecular characters,
the different data sets were combined.
The phylogenetic analyses were primarily per-
formed by rooting with Balaenoptera and Hippopotamus
(sample I =23 taxa). To test the stability of the peco-
ran tree topology, a subsequent analysis was conducted
excluding the two most diverging outgroup taxa. It in-
cluded only genera of the suborder Ruminantia, with
Tragulus used as outgroup (sample II =21 taxa).
Maximum parsimony analyses.—The maximum parsi-
mony (MP) analysis (PAUP 3.1.1.; Swofford, 1993) was
conducted with either equal weighting or differential
weighting of the character-state transformations using
the product of CI ×S (CI =consistency index, S =slope
of saturation) (Hassanin et al., 1998a, 1998b). For each
substitution type (i.e., A to G, C to T, A to C, A to T, C to
G, and G to T), the amount of homoplasy was measured
through the CI and the saturation was assessed graph-
ically by plotting the pairwise number of observed dif-
ferences against the corresponding pairwise number of
inferred substitutions. The saturation analysis was per-
formed using the matrices of patristic distances and ad-
justed character distances calculated by PAUP 3.1.1. The
slope of the linear regression (S) was then used to evalu-
ate the level of saturation. The pairwise genetic and pa-
tristic distances are not independent because they reﬂect
a shared evolutionary history of the taxa compared, but
the linear regression provides a good measure of the sat-
uration intensity. When no saturation is observed, the
slope of the linear regression is equal to 1. When the
level of saturation increases, the slope decreases toward
zero. The CI and S values were calculated for the com-
plete sequences for the three noncoding nuclear markers
(Lf,αLAlb, and Cyp), separately for the three codon po-
sitions for the coding markers (Cyb and κCas) to take
into account the selective constraints, and separately for
stems and loops of the secondary structure for 12S rRNA
(Springer and Douzery, 1996) and 16S rRNA (Schnare
et al., 1996) genes to take into account the functional
constraints. Searches for the shortest tree(s) were per-
formed using default options but with 100 replicates
of the random stepwise addition of taxa. The reliabil-
ity of the nodes was assessed by bootstrap percentages
(BP; Felsenstein, 1985) and by branch support (b or br)
(Bremer, 1988). The bootstrap values were computed af-
ter 1,000 replicates of the closest stepwise addition of
taxa. The Bremer analysis was conducted using topologi-
cal constraints with 100 replicates of the random stepwise
addition of taxa. For the differential weighting analyses,
the branch support values were rescaled (br) with re-
spect to the equally weighted tree length (Gustafsson and
Bremer, 1995). Searches under topological constraints
were used to measure how long a tree must be before
a given group of taxa becomes monophyletic. The num-
ber of additional steps was rescaled with respect to the
equally weighted tree to allow comparisons with the b
and brvalues. DeBry (2001) cautioned that decay indices
must be interpreted in light of branch lengths and that
low values can be meaningless in terms of support.
Maximum likelihood and Bayesian analyses.—A standard
maximum likelihood (ML) approach was ﬁrst conducted
as an alternative to the MP approach because ML is
known to be less sensitive to potential long-branch
attraction artifacts, to take into account the underlying
molecular evolutionary process, and to statistically
compare competing hypotheses (Swofford et al., 2001;
Whelan et al., 2001). ML analyses with PAUP∗4.0b8
(Swofford, 1998) were conducted under the general time
reversible model (GTR; Yang, 1994) with among-site
substitution rate heterogeneity described by a gamma
distribution with eight categories (08; Yang, 1996a)
and a fraction of sites (INV) constrained to be invari-
able. ML parameters were optimized after heuristic
search on a neighbor-joining (NJ) starting tree using
tree bisection-reconnection (TBR) branch swapping.
Reliability of nucleotide-derived trees was estimated
by BP values (Felsenstein, 1985) computed with PAUP∗
using the optimal ML parameters, with NJ starting trees
and a number of TBR branch swapping rearrangements
limited to 1,000 per replicate.
To account for the combination of markers with con-
trasted molecular properties, i.e., nuclear versus mi-
tochondrial and protein coding or ribosomal versus
212 SYSTEMATIC BIOLOGY VOL. 52
noncoding, a partitioned ML analysis was conducted.
Twelve partitions were distinguished in the original data
set according to the structural and functional proper-
ties of the markers: codon positions 1 (380 nucleotides
[nt]), 2 (380 nt), and 3 (380 nt) for Cyb, loops (457 nt)
and stems (446 nt) of the 12S rRNA, loops (857 nt) and
stems (643 nt) of the 16S rRNA, codon positions 1 (125 nt),
2 (125 nt), and 3 (126 nt) for κCas,Lf (a single partition be-
cause of its promotor nature; 354 nt), and αLAlb (a single
partition because of its intron nature; 502 nt), yielding
a total of 4,775 sites. Cyp was not used because three
taxa (Hippopotamus,Tragulus, and Gazella) were not rep-
resented for this marker. For each partition, one inde-
pendent GTR +08model was deﬁned to reﬂect both
within- and between-marker differences in the process
of evolution. The likelihoods of alternative phylogenetic
hypotheses were then computed using PAML version
3.1 (Yang, 1997). We used the method of Yang (1996b)
to combine multiple sequence data: Each of our 12 se-
quence partitions was analyzed with its own estimates
of the GTR +08model and with proportional estimates
of branch lengths from one partition to another. To eval-
uate alternative hypotheses for the location of the root
of the pecoran subtree, the NucML program (MOLPHY
2.3; Adachi and Hasegawa, 1996) was used to write
all the 105 bifurcating trees connecting the ﬁve Pecora
families, starting from the following enforced topology:
rafﬁdae, Moschidae, Bovidae, Cervidae))). To limit the
number of possible trees, the subtopology within each
family was constrained according to the highest likeli-
hood trees reconstructed from the concatenated matrix
of characters. The pecoran families were respectively
represented by Antilocapra,Giraffa +Okapia,Moschus,
gus)))(Gazella +Redunca)))(Boselaphus(Bos +Tragela-
phus))), and (((Capreolus +Hydropotes)(Odocoileus +
Rangifer))(Cervus +Muntiacus)). This choice seems jus-
tiﬁed for well supported nodes (e.g., Bovinae) but more
controversial for weaker nodes (e.g., Bos +Tragelaphus).
The exploration of the tree space was therefore approx-
imated, and it is difﬁcult to predict how much the con-
straints on poorly supported nodes affected the rank-
ing of the different phylogenetic alternatives according
to their log-likelihood. However, this tree space limita-
tion under PAML was the only way to compute 105 log-
likelihoods under the complex combination of indepen-
dent GTR +08models for 12 different gene partitions.
The likelihoods of the 105 trees were then evaluated by
partitioned ML, with a simultaneous estimation of 12 ×
(5 [GTR] +1[0
]) +11 (proportionality rates between
partitions)) +43 (branch lengths) =126 independent pa-
rameters for each topology. The nonparametric test of
Kishino and Hasegawa (1989) was conducted with the
conservative Shimodaira and Hasegawa (1999) correc-
tion for multiple tree comparisons (KH-SH test) to eval-
uate the signiﬁcance of differences in log-likelihood be-
tween the tree with the highest likelihood and the 104 al-
ternative topologies. The ratios between the difference in
log-likelihoods of the best (B) and evaluated (E) trees (δ=
lnL[B] −lnL[E]) and the standard error (σ) of this differ-
ence, and the conﬁdence PSH values have been computed
by PAML 3.1 under the 12 GTR +08models previously
Phylogenetic analyses were also performed using the
Bayesian inference (Huelsenbeck et al., 2001). This new
approach evaluates the posterior probability of a tree
given the character matrix, i.e., the probability that the
tree is correct. The posterior probability is obtained af-
ter combining the prior probabilities of a tree and of the
data with the likelihood of the data given that tree. The
Bayesian approach combines the advantages of deﬁning
an explicit probability model of character evolution and
of obtaining a rapid approximation of posterior probabil-
ities of trees through the use of the Markov chain Monte
Carlo (MCMC) approach. The likelihood model chosen is
always GTR +08, with partition-speciﬁc rates. All analy-
ses were conducted using MrBayes 2.1 (Huelsenbeck and
Ronquist, 2001), with ﬁve independent Markov chains
(one cold chain and four incrementally heated chains)
run for 200,000 metropolis-coupled MCMC generations,
with tree sampling every 20 generations and burn-in
after 5,000 trees.
Ages of divergence between the pecoran clades have
been estimated by the Bayesian relaxed molecular clock
approach developed by Thorne et al. (1998) and Kishino
et al. (2001) in the software DIVTIME 5b. This approach
combines the advantages of relaxing the molecular
clock with a continuous autocorrelation of substitution
rates over evolutionary time and of allowing the si-
multaneous use of several calibration references. As
time constraints, we considered the ﬁrst splits within
Bovidae (Miyamoto et al., 1993) and within Cervi-
dae (Ginsburg, 1988) to have occurred in the Early
Miocene, i.e., between 16.4 and 23.8 MYA. All abso-
lute ages of the geological periods and chronostrati-
graphic references were taken from the 1999 Geolog-
ical Time Scale of the Geological Society of America
The dating procedure involved two steps. First, the
program ESTBRANCHES estimated branch lengths and
the variance-covariance matrix from the concatenated
nucleotide data set (4,775 sites; i.e., the same data ma-
trix as used for the partitioned ML analysis). The F84
nucleotide substitution model was the only one imple-
mented for DNA under ESTBRANCHES and included
the following parameters (as estimated from PAML 3.1):
A=32.9%, C =24.2%, G =17.7%, and T =25.2% for
the base composition; 3.45 for the rate parameter; and
α=0.22 for the shape of the 08distribution that corre-
sponds to eight discrete categories with rates of 0.00004,
0.0026, 0.01696, 0.07533, 0.23829, 0.63007, 1.57486, and
Second, after pruning the outgroup taxa (Balaenoptera
and Hippopotamus), the program DIVTIME estimated
the prior and posterior ages of divergence between ru-
minants, their SDs, and the 95% credibility intervals
(CredI95%). The Markov chain was sampled 10,000 times
2003 HASS ANIN AND DO UZERY—PHYLO GENY O F RUMINANTIA 213
with 100 cycles between each sample and burn-in after
100,000 cycles. The following prior distributions were
adopted: 40 MYA (SD =20 MYA) for the expected time
between tip and root if there had been no constraints
on node times, 0.008 (SD =0.004) substitutions per site
per million years for the rate at root node, 0.025 (SD =
0.025) for the parameter that controls the amount of
rate autocorrelation per million years, and 65.0 MYA
for the highest possible number of time units between
tip and root. Because of the uncertainty about the loca-
tion of the root of the pecoran subtree, molecular datings
were conducted on two topologies, one with Antilocapri-
dae as sister of the remaining pecorans and one with
Antilocapridae +Girafﬁdae as sister group of Moschi-
dae +Bovidae +Cervidae. To check for convergence of
MCMC analyses, two independent runs were performed
for the same data and same prior distributions but with
different starting points, one at a randomly selected state
and the other at a potentially better state.
Forty-eight morphological characters were analyzed
to infer phylogenetic relationships among the ﬁve pec-
oran families Antilocapridae, Bovidae, Cervidae, Giraf-
ﬁdae, and Moschidae. The family Tragulidae was used
as an outgroup because all recent morphological stud-
ies have shown its placement as sister taxon of Pecora
(e.g., Webb and Taylor, 1980; Bouvrain and Geraads,
1985; Gentry and Hooker, 1988; Janis and Scott, 1988;
Vislobokova, 1990; Scott and Janis, 1993). Most of the
characters used in analyses came from the studies per-
formed by Janis and Scott (1987), Gentry and Hooker
(1988), and Scott and Janis (1993). All the character states
were coded using the following assumptions (Table 2):
All the ruminant families were considered mono-
phyletic, and all the character states examined were con-
TABLE 2. Morphological characteramatrix (a =0or1;b=1or2;?=unknown; - =not applicable).
aSoft anatomy: 1 =abdominal musk gland; 2 =ruminating stomach with four chambers (including a well-developed omasum); 3 =cardiac oriﬁce of rumen
more ventrally situated; 4 =gall bladder; 5 =ileocecal gland; 6 =placenta diffuse (0) or cotyledonary with few (1) or many (2) cotyledons; 7 =extensible tongue;
8=females with four (0) or two (1) mammae. Postcranial: 9 =ﬁbular facet on calcaneum with proximal concavity large, distal concavity small; 10 =astragalus
with proximal and distal portions parallel; 11 =anterior longitudinal metatarsal gully including a bridged distal end; 12 =posterior metatarsal tuberosity; 13 =
cubonavicular facet very ﬂat and broad; 14 =internal cuneiform independent; 15 =metatarsals II and V lost; 16 =metacarpals III and IV fused from proximal end
for most or entire length; 17 =metacarpals II and V incomplete (1) or complete loss of lateral digits (2); 18 =complete distal metapodial keels; 19 =limbs much
elongated; 20 =number of sacral vertebrae: 5(0), 4(1); 21 =number of ribs: ≤13 (0), 14 (1), ≥15 (2). Cranial: 22 =double lacrimal oriﬁce; 23 =tympanohyal vagina
fully enclosed; 24 =auditory bulla hollowed internally; 25 =anteorbital (ethmoidal) vacuity; 26 =lacrimal fossa (a depression in the lacrimal bone that contains
the preorbital gland); 27 =foramen ovale large; 28 =postglenoid process absent; 29 =cranial appendages; 30 =cranial appendages present in postorbital position
(1) or supraorbital position (2); 31 =branched bone core; 32 =deciduous bone core; 33 =bone core with keratinous sheath; 34 =deciduous horn sheath; 35 =forked
horn sheath; 36 =cranial appendages only in males; 37 =cranial appendages are formed from an outgrowth of the frontal bone (1) or from a dermal ossiﬁcation
center (2). Dental: 38 =long and curved upper canines; 39 =upper molar entostyle; 40 =upper molar metastyle; 41 =upper M3 metaconule large; 42 =upper molar
lingual cingulum; 43 =anterior cingulum on lower molar; 44 =metaconid on lower premolars simpliﬁed; 45 =metastylid on the lower molars; 46 =lower canine
bilobed; 47 =postentocristid complete on the lower molars; 48 =ectostylids on the lower molars.
sidered the ancestral states within each of the six rumi-
nant families. For Bovidae and Cervidae, ancestral states
were estimated by using consensus topologies derived
from previous molecular investigations (Randi et al.,
1998; Hassanin and Douzery, 1999a, 1999b; Matthee and
Robinson, 1999; Matthee and Davis, 2001). A question
mark indicates that the primitive character state was
not determined, and a dash indicates that the charac-
ter was not applicable. All question marks and dashes
were treated as missing character states for the parsi-
MP analyses were based on CI ×S weighting of
the character-state changes because this weighted ap-
proach allows the deﬁnition of different stepmatrices
to reﬂect the contrasted evolutionary properties within
(e.g., different codon positions) and between (e.g., mi-
tochondrial versus nuclear) markers. When MP analy-
ses based on the CI ×S weights are conducted with
Balaenoptera and Hippopotamus included as outgroups,
one most-parsimonious tree of 1,675,124 steps was re-
covered (Fig. 2). All members of the suborder Ruminan-
tia cluster together (BP =100; br=+199), with Tragulus
clearly separated from the other ruminant genera (BP =
100; br=+100). Within Pecora, the families Bovidae,
Cervidae, and Girafﬁdae are monophyletic with high
bootstrap and branch support values (respectively, BP =
100/100/100; br=+28/+91/+74), and Antilocapra ap-
pears to branch ﬁrst with respect to all other lineages
(BP =81; br=+9). The families Bovidae, Cervidae, and
Moschidae cluster together (BP =85; br=+12), with
Moschus unexpectedly closer to bovids (BP =95; br=
+16) than to cervids (residual BP =1; rescaled additional
steps =+16). Two markers individually support the
FIG URE 2. Maximum parsimony tree reconstructed from the combination of the seven markers. Weighted parsimony analysis was based on the product of homoplasy ×saturation
indices (length =1,675,124 steps). Bootstrap percentages/rescaled branch support values found with the combination of the seven markers are indicated in bold. Underlined values
are bootstrap percentages (>50%) obtained independently on the seven different markers: Cyb/12S/16S/αLAlb/Cyp/Lf/κCas (Cyb =cytochrome b; 12S =12S rDNA, 16S =16S rDNA;
Cyp =cytochrome oxidase P450; Lf =lactoferrin, αLAlb =α-lactalbumin; κCas =κ-casein). Asterisk indicates that the node was not supported by the bootstrap analysis (<50%).
Thick branches indicate higher taxa corresponding to the order, suborders, and families.
2003 HASS ANIN AND DO UZERY—PHYLO GENY O F RUMINANTIA 215
position of Moschus close to the Bovidae: αLAlb (BP =
92) and Lf (BP =96). Within the family Cervidae, there
are two major clades: the Plesiometacarpalia, which as-
sociates Cervus with Muntiacus (BP =100; br=+25), and
the Telemetacarpalia (BP =96; br=+13), which includes
two clades: Rangifer +Odocoileus (BP =100; br=+34)
and Hydropotes +Capreolus (BP =100; br=+54). Within
Bovidae, two major clades previously named Bovinae
and Antilopinae sensu lato were recovered with high
support (respectively, BP =100/100; br=+18/+26).
The tribes Caprini (Ovis), Alcelaphini (Damaliscus), and
Hippotragini (Hippotragus) are strongly associated (BP =
100; br=+22) with Hippotragini and Alcelaphini as
sister taxa (BP =99; br=+21). When MP analyses
were performed excluding the more distant genera Bal-
aenoptera and Hippopotamus and keeping only Tragulus
as an outgroup, three most-parsimonious trees of
1,322,933 steps were recovered (data not shown). The
topology of these trees is identical to that of the tree in
the previous analysis, except for lack of resolution in the
branching pattern of Aepyceros,Cephalophus, Gazella, and
Redunca. All interfamily nodes were again recovered but
with higher support: the basal position of Antilocapra
(BP =97), the clade Cervidae +Moschus +Bovidae (BP =
95), and the association of Moschus with the Bovidae
When MP analyses were performed with equal
weighting of the character-state transformations, two
most-parsimonious trees of 5,687 steps were recovered.
The interfamily relationships are identical to those from
the weighted MP analyses except for the unresolved
position of Moschus, which appears associated with ei-
ther Bovidae (BP =45) or Cervidae (BP =43) (data
not shown). When MP analyses were performed ex-
cluding the more distant genera Balaenoptera and Hip-
popotamus and using only Tragulus as the outgroup, two
most-parsimonious trees of 4,832 steps were recovered
(data not shown). The topology of these trees is identi-
cal to that of the tree in the previous analysis. Moschus
is allied with either Bovidae (BP =40) or Cervidae
ML and Bayesian Phylogenetic Analyses
Different nucleotide substitution models implemented
under both PAUP∗and PAML were compared. The log-
likelihood (lnL) of the best tree was −35848.43 under
HKY85, −35543.16 under TN93, and −35360.37 under
GTR. Incorporation of the gamma distribution of substi-
tution rate heterogeneity among sites yielded −31 980.92
under GTR +08and −31978.43 after adding a fraction of
invariable sites. All these log-likelihood increases were
signiﬁcant (P<0.05) under likelihood ratio tests com-
paring simpler to more complex models. Therefore, the
single GTR +08+INV model best explained our combi-
nation of 7 markers and was used in subsequent analyses.
The highest likelihood and maximum posterior prob-
ability phylograms (Fig. 3) are mostly in agreement
with the most well-supported nodes of the MP anal-
yses and recovered with high support (BP >98;
Bayesian posterior probability [PPB]=1.00) the follow-
ing clades and their subclades: Pecora; Girafﬁdae; Cervi-
dae, Plesiometacarpalia, Telemetacarpalia, Old World
Telemetacarpalia, and New World Telemetacarpalia;
Bovidae, Bovinae, Antilopinae, Caprini +Alcelaphini +
Hippotragini, and Alcelaphini +Hippotragini. The
Moschidae appears robustly associated with Bovidae
(BP =94; PPB=1.00), and this group then branches
with Cervidae (BP =71;PP
B=0.99). One topology
disagreement between MP and ML or Bayesian anal-
yses concerns the location of Antilocapridae. MP sug-
gests that Antilocapridae is the sister group of the re-
maining pecoran families, but ML suggests an alternative
hypothesis with Antilocapridae weakly clustering with
Girafﬁdae (BP =52 ; PPB=0.57). Residual BP =46 and
PPB=0.43 deﬁne a basal position of Antilocapra among
When the closer outgroup Tragulus was used instead
of the more distant Balaenoptera and Hippopotamus, all
the clades previously identiﬁed were recovered, includ-
ing the sister-group relationship between Moschidae and
Bovidae (BP =92;PP
B=1.00) and their association with
Cervidae (BP =71; PPB=1.00). However, Antilocapra
is in the basal position with respect to all other Pecora
(BP =77; PPB=1.00), as seen in MP analyses.
Differences in replacement patterns and rates between
mitochondrial and nuclear sequences can lead to biased
phylogenetic results (Whelan et al., 2001) and could hin-
der us from using the powerful approach to analyze the
seven markers as a concatenated whole. The simulta-
neous use of protein-encoding (Cyb,κCas), ribosomal
(12S and 16S rRNAs), and noncoding (αLAlb,Cyp,Lf )
sequences may even exacerbate this problem. Thus, we
conducted a partitioned ML evaluation of the phylo-
genetic relationships among Antilocapridae, Girafﬁdae,
Cervidae, Moschidae, and Bovidae. To reach this goal,
the KH-SH test was used to evaluate the null hypothe-
sis stating that all 105 trees compared, including the ML
tree, are equally good explanations of the data. Because
the use of a fraction of invariable sites is not implemented
in PAML and to homogenize the models used across the
different partitions, a GTR +08model was attributed to
each of the 12 character partitions.
We computed the log-likelihood of the 105 possibil-
ities of connecting the ﬁve pecoran families into a bi-
furcating subtree, rooted by Balaenoptera,Hippopotamus,
and Tragulus. These 15 ×7=105 rootings arise because
there are 15 possible unrooted ﬁve-taxon trees, each with
seven potential locations for the root. The best rooting,
i.e., the one yielding the highest likelihood, is along the
branch leading to Antilocapra +Girafﬁdae (Fig. 4). The
104 other topologies were ranked by increasing level of
rejection by the KH-SH test (PSH values ranging from
0.93 to 0.004). Among them, 74 are rejected (PSH <0.05).
The 30 other topologies correspond to only ﬁve peco-
ran subtrees among 15 possible. In these ﬁve subtrees,
Moschidae were never associated with either Antilo-
capridae or Girafﬁdae, providing further evidence for
the tight afﬁnities of Moschus with bovids and cervids
(Fig. 4). Moreover, we observed that the ﬁrst rootings
216 SYSTEMATIC BIOLOGY VOL. 52
FIGURE 3. Maximum likelihood tree reconstructed from the combination of the seven markers. This highest likelihood tree (lnL =−31978.43)
has been reconstructed under the following model of sequence evolution: GTR with rate parameters of 2.47 (A to C), 7.71 (A to G), 1.74 (A to
T), 0.75 (C to G), 20.35 (C to T), and 1.00 (G to T), 14% invariable sites, and a gamma rate heterogeneity of α=0.33. The maximum posterior
probability tree of the Bayesian approach displays an identical topology. ML bootstrap percentages/posterior probabilities are given for each
node. Thick branches indicate higher taxa corresponding to the order, suborders, and families.
breaking the Moschidae–Bovidae clade were numbers
12 and 13 (PSH =0.34). Owing to the conservative be-
havior of the KH-SH test (Shimodaira, 2002), the latter
PSH value should also be viewed as a reasonably good in-
dication of the phylogenetic afﬁnity between Moschidae
and Bovidae. This hypothesis is supported by the fact
that all rootings separating the musk deer from bovids
were rejected by the Kishino-Hasegawa (1989) test
Molecular Estimates of Divergence Ages
The Bayesian relaxed molecular clock method of
Thorne et al. (1998) and Kishino et al. (2001) was used
to estimate the divergence ages within the pecoran sub-
tree. Two calibration constraints also were simultane-
ously used: 16.4–23.8 MYA for the ﬁrst split within bovids
and within cervids. Comparison of the prior and poste-
rior divergence ages show that these distributions are
different for most nodes, with narrower credibility in-
tervals for posterior ages (Table 3), indicating that much
information regarding node times is attributable to the
concatenated markers. If we had observed that prior and
posterior distributions were about the same, then most
molecular dating information would seem to be coming
from the priors rather than the data.
The a priori calibration constraints were set to the
Early Miocene (16.4–23.8 MYA) for both Cervidae and
Bovidae, i.e., for the split between Cervus +Muntiacus
and the other cervids and the split between Boselaphus +
Bos +Tragelaphus and the other bovids. Whereas the di-
vergence of the two families was assumed to be con-
temporary, posterior estimates of the divergence ages
for cervids and bovids appear stuck to the lower bound
(CredI95% =16.4–19.0 MYA; Table 3) and upper bound
(CredI95% =20.4–23.8 MYA), respectively, of the Early
Miocene, suggesting that the evolutionary radiation of
bovids occurred before that of cervids.
The difﬁculty in ﬁnding the root of the pecoran subtree
is illustrated by the short time intervals between sub-
sequent divergences. Pecorans diversiﬁed between 26.0
and 35.6 MYA, whereas antilocaprids and girafﬁds ap-
pear between 23.9 and 32.5 MYA and bovids, moschids,
and cervids diverged 25.1–30.3 MYA during the Late
Oligocene (cf CredI95%; Table 3). The split between
Moschidae and Bovidae occurred between 23.6 and
2003 HASS ANIN AND DO UZERY—PHYLO GENY O F RUMINANTIA 217
FIGURE 4. Conﬁdence levels of the conservative Shimodaira–Hasegawa test for the location of the root of the pecoran subtree under a
12-partition ML analysis. The 105 possibilities of rooted trees for relationships among the ﬁve pecoran families were explored and ranked
according to their increasing level of rejection (i.e., circled numbers on the trees and numbers on histogram bars correspond to decreasing
probability values [PSH]). For the computation of log-likelihoods, each of the 12 character partitions had its own GTR +08model. Relative to
the highest likelihood tree, 74 rooting possibilities were statistically rejected. The 30 remaining root positions that were not rejected are located
on the ﬁve illustrated unrooted subtrees. Among them, Moschidae is never directly associated with either Antilocapridae or Girafﬁdae. The
vertical dotted line corresponds to the PSH =0.05 rejection level. For the present data, a probability value of 0.30 for the Shimodaira–Hasegawa
test corresponds to a Pvalue of 0.05 for the Kishino–Hasegawa test. ANT, BOV, CER, GIR, and MOS are Antilocapridae, Bovidae, Cervidae,
Girafﬁdae, and Moschidae, respectively.
28.5 MYA during the Late Oligocene. All subsequent
cladogenesis within girafﬁds, cervids, and bovids then
occurred during the Early and Middle Miocene (Table 3).
The SDs for splitting ages ranged from 1.4 to 4.0 MYA
for the two most basal dichotomies among ruminants
and from 0.7 to 1.7 MYA for cladogenic events within
The use of two different reference topologies because
of the ambiguous location of the pecoran root seems
not to have affected the posterior distributions of di-
vergence ages by >0.1 MYA for divergence within the
Bovidae +Moschidae +Cervidae clade and by >0.9–
2.7 MYA for Girafﬁdae and the base of the ruminant
subtree (Table 3).
Analysis of Morphological Characters
Of 48 total characters, only 12 were parsimony in-
formative for the relationships among pecoran families
(highlighted sites, Table 2): position of the cardiac ori-
ﬁce of rumen (character 3), bridge on the metatarsal
gully (11), presence of posterior metatarsal tuberosity
(12), fusion of the internal cuneiform (14), loss of lateral
metacarpals (17), number of ribs (21), size of the foramen
ovale (27), presence of cranial appendages (29), bone core
covered by keratinous sheath (33), cranial appendages
only in males (36), long and curved upper canines (38),
and presence of metastylid on the lower molars (45).
Of 105 trees evaluated, 6 most-parsimonious trees
of 55 steps were retained (CI excluding uninformative
characters =0.684, retention index =0.539). The differ-
ences between the six MP trees involve the position of
Moschus, which appears close to either Bovidae or Cervi-
dae or basal with respect to Bovidae and Cervidae, and
the position of Antilocapra, which appears either basal
within Pecora or as sister taxon of the clade composed
of Bovidae, Moschidae, and Cervidae. The strict con-
senus (Fig. 1) is congruent with the molecular analyses:
Cervidae, Moschidae, and Bovidae are enclosed together
(b =+1) and this result is supported by a single exclusive
synapomorphy, a large foramen ovale (27).
218 SYSTEMATIC BIOLOGY VOL. 52
TABLE 3. Molecular dating of the main splitting events within Pecora. The Bayesian relaxed molecular clock approach of continuous auto-
correlation of evolutionary rates along branches developed by Thorne et al. (1998) and Kishino et al. (2001) was used. Two calibration points
allowed the estimate of absolute divergence ages: the ﬁrst splits within Bovidae and within Cervidae were constrained to occur in the Early
Miocene, i.e. 16.4–23.8 million years ago (MYA). Values correspond to the prior and posterior divergence ages ±1 SD, calculated according to
two slightly different topologies: Antilocapra as sister group to Girafﬁdae or in the basal position among pecorans. The 95% credibility intervals
(CredI95%) for prior and posterior divergence ages are given in parentheses. The corresponding geologic time scale is also given. The following
posterior distributions were obtained (values for the Antilocapra +Girafﬁdae topology): 44.3 (SD =3.5, CredI95% =37.6–51.7) MYA for the
expected time between tip and root, 0.0078 (SD =0.0009, CredI95% =0.0061–0.0097) substitutions per site per million years for the rate at root
node, and 0.0039 (SD =0.0021, CredI95% =0.0012–0.0094) for the parameter that controls the amount of rate autocorrelation per million years.
Antilocapra +Girafﬁdae Antilocapra basal
Divergences between taxa Prior Posterior Prior Posterior Epoch
Tragulina/Pecora 33.0 ±7.5 44.3 ±3.6 35.8 ±8.0 46.3 ±4.0 Middle Eocene
(22.4–51.9) (37.6–51.7) (23.8–55.6) (39.1–54.8)
Radiation of Pecora 29.4 ±6.2 28.9 ±1.4 32.4 ±7.0 31.6 ±2.0 Early Oligocene
(20.7–45.1) (26.0–31.7) (22.2–49.2) (27.9–35.6)
Antilocapridae/Girafﬁdae 19.5 ±8.2 27.0 ±1.6 Late Oligocene
Girafﬁdae/(Cervidae 29.0 ±5.7 29.4 ±1.5 Early Oligocene
+Moschidae +Bovidae) (20.7–43.2) (26.4–32.5)
Giraffa/Okapia 9.6 ±7.4 19.0 ±1.6 14.5 ±8.9 19.9 ±1.7 Early Miocene
(0.3–26.6) (16.0–22.3) (0.8–32.8) (16.7–23.2)
Cervidae/(Moschidae +Bovidae) 25.7 ±4.5 27.7 ±1.3 25.5 ±4.3 27.8 ±1.3 Late Oligocene
(19.2–36.9) (25.1–30.1) (19.2–36.3) (25.1–30.3)
Cervidae (constraints: 16.4–23.8 MYA) 20.1 ±2.1 17.2 ±0.7 20.1 ±2.1 17.2 ±0.7 Early Miocene
(16.6–23.6) (16.4–19.0) (16.6–23.6) (16.4–19.0)
Cervus/Muntiacus 10.1 ±6.0 13.8 ±0.9 10.0 ±5.9 13.7 ±0.9 Middle Miocene
(0.5–21.1) (12.1–15.7) (0.5–20.8) (12.1–15.7)
Hydropotes/Capreolus 6.7 ±4.8 10.3 ±0.9 6.6 ±4.8 10.3 ±0.9 Late Miocene
(0.2–17.3) (8.6–12.3) (0.2–17.4) (8.6–12.2)
Rangifer/Odocoileus 6.6 ±4.8 11.5 ±0.9 6.7 ±4.9 11.5 ±0.9 Middle/Late
(0.2–17.3) (9.8–13.5) (0.2–17.4) (8.6–12.2) Miocene
(Hydropotes +Capreolus)/(Rangifer 13.4 ±5.0 16.2 ±0.8 13.4 ±5.0 16.2 ±0.8 Early/Middle
+Odocoileus) (3.1–21.6) (14.9–18.1) (3.1–21.5) (14.9–18.0) Miocene
Moschidae/Bovidae 22.8 ±3.5 26.1 ±1.2 22.6 ±3.4 26.2 ±1.2 Late Oligocene
(17.7–31.3) (23.6–28.3) (17.6–30.7) (23.6–28.5)
Bovidae (constraints: 16.4–23.8 MYA) 19.8 ±2.1 22.6 ±0.9 19.8 ±2.1 22.6 ±0.9 Early Miocene
(16.6–23.5) (20.4–23.8) (16.6–23.5) (20.4–23.8)
Bovinae 13.2 ±4.9 19.3 ±1.1 13.2 ±4.8 19.3 ±1.1 Early Miocene
(3.1–21.4) (17.0–21.2) (3.3–21.4) (17.0–21.3)
Antilopinae 16.5 ±3.3 19.7 ±1.0 16.5 ±3.3 19.6 ±1.0 Early Miocene
(9.0–22.2) (17.4–21.5) (9.1–22.3) (17.4–21.4)
Hippotragus/Damaliscus 3.3 ±2.8 13.1 ±1.1 3.3 ±2.8 13.1 ±1.1 Middle Miocene
(0.1–10.3) (11.0–15.3) (0.1–10.5) (11.0–15.2)
Ovis/(Hippotragus +Damaliscus) 6.5 ±3.6 15.1 ±1.1 6.6 ±3.6 15.1 ±1.1 Middle Miocene
(1.0–14.3) (13.0–17.2) (1.0–14.4) (12.9–17.2)
Monophyly of the Pecora
The ﬁve living pecoran families, Antilocapridae, Bovi-
dae, Cervidae, Girafﬁdae, and Moschidae, are classically
uniﬁed as higher ruminants, or pecorans, and are dis-
tinguished from tragulids by numerous morphological
characters (Janis and Scott, 1987). This distinction be-
tween Tragulina and Pecora is here conﬁrmed by our
DNA data, with Tragulus (Tragulidae) diverging ﬁrst
with respect to all other ruminant families. The mono-
phyly of Pecora, here represented by an exhaustive sam-
pling of the ﬁve living families, is highly supported in
terms of bootstrap and Bayesian support (100, what-
ever the method applied), and 23 autapomorphies were
found for this clade, including three diagnostic indels
(deletion of T at position 401 of the 12S rRNA; inser-
tion of A at position 222 of the Lf, and a deletion of
nine nucleotides [GTAGGGCYA] at position 98 of the Lf ).
This huge number of exclusive synapomorphies proba-
bly reﬂects the long divergence time between the Trag-
ulina/Pecora split (44.3–46.3 MYA) and the pecoran ra-
diation (28.9–31.6 MYA), which spans a minimum of 12.7
million years as here estimated by the Bayesian relaxed
Monophyly of the Pecoran Families and Subfamilies
All our molecular results strongly support the mono-
phyly of all the families for which several representa-
tives are available: Bovidae, Cervidae, and Girafﬁdae
(BP >98, PPB=1.00). Many molecular autapomorphies
characterize these families: 5 for Bovidae, 16 for Giraf-
ﬁdae, and 17 for Cervidae. The most striking molecu-
lar signatures are in Girafﬁdae, with two long deletions
(6 nucleotides [ATAAGC] at position 401 of αLAlb and
2003 HASS ANIN AND DO UZERY—PHYLO GENY O F RUMINANTIA 219
8 nucleotides [GCCCCAGG] at position 113 of Lf ), and
Cervidae, with two insertions (A at position 760 of
16S rRNA and T at position 7 of Cyp) and one large
deletion of 16 nucleotides (CATAAAAGGCAACAGG at
position 381 of αLAlb). Two distinct clades, previously
proposed on the basis of the structure of the metacarpals,
were recovered within the family Cervidae: the Ple-
siometacarpalia (Cervus +Muntiacus;BP=99–100,
PPB=1.00), which are morphologically characterized
by the retention of both distal and proximal parts of the
metacarpals (of digits II and V), and the Telemetacarpalia
(Rangifer +Odocoileus +Hydropotes +Capreolus;BP=96–
98, PPB=1.00), in which the metacarpals are greatly re-
duced with only the distal parts remaining. This distinc-
tion is also supported by signatures such as a deletion
of 12 nucleotides for the Telemetacarpalia (TAATACC-
CTGTA at position 259 of αLAlb). Within Bovidae, two
major clades previously named Bovinae and Antilop-
inae sensu lato (Hassanin and Douzery, 1999a, 1999b)
were recovered with high support (BP =97–100, PPB=
1.00) but without diagnostic signatures. Although some
workers have placed Antilocapra americana within Bovi-
dae (e.g., O’Gara and Matson, 1975), our results conﬁrm
its family status in the monotypic family Antilocapridae
(e.g., Simpson, 1945) because it appears clearly outside
of Bovidae. Similarly, the genus Moschus, represented by
the living musk deer species, has traditionally been re-
garded as constituting the subfamily Moschinae within
Cervidae (e.g., Simpson, 1945), but a number of authors
have suggested that Moschus, should be placed in its
own family within the superfamily Cervoidea (e.g., Gray,
1821; Brooke, 1878; Flerov, 1952; Webb and Taylor, 1980;
Groves and Grubb, 1987). Our molecular tree supports
the family status of Moschidae because Moschus was not
grouped within Cervidae.
Interfamily Relationships within the Pecora
Morphology.—The Pecora are generally recognized as a
monophyletic group, but their interfamily relationships
are poorly understood, and recent morphological studies
have produced numerous conﬂicting hypotheses (Fig. 1).
Many of the characters used for pecoran phylogeny are
known to be homoplastic (Scott and Janis, 1993), and de-
pending on the authors, different character states can be
deﬁned for each of the various families. Given that these
problems are especially critical in MP analyses, the def-
inition and/or the value of several characters tradition-
ally used for pecoran taxonomy should be examined.
Cranial appendages have been widely used in the re-
construction of pecoran phylogeny. Webb and Taylor
(1980) proposed the term Eupecora to designate all the
Pecora characterized by the presence of cranial ap-
pendages, i.e., all families except the Moschidae. How-
ever, the absence of cranial appendages in Moschus can
be alternatively interpreted as a secondary loss, ex-
actly as demonstrated for Hydropotes in Cervidae (Randi
et al., 1998). Other authors have allied bovids with an-
tilocaprids because both share horns composed of a
bone core covered with a keratinous sheath (O’Gara
and Matson, 1975). Nevertheless, deciduous keratinous
sheaths are also encountered in some specimens of
Okapia (Frechkop, 1955), which is why we coded this
character as unknown in the ancestor of Girafﬁdae
(character 33, Table 2). Bovids have also been grouped
with girafﬁds because they are supposed to share a sim-
ilar developmental pathway with a separate epiphyseal
os cornu (Frechkop, 1955). Nonetheless, the supposed
homology between bovid horns and girafﬁd ossicones
appears to be based on the erroneous assumption that a
separate ossiﬁed os cornu is routinely formed in bovids.
In bovids, as in cervids, the cranial appendages are
formed from an outgrowth of the frontal bone, whereas
in girafﬁds these appendages are formed from a der-
mal ossiﬁcation center (Janis and Scott, 1987). Unfortu-
nately, the developmental condition in antilocaprids has
not been determined.
Simpliﬁed and high-crown or hypsodont cheek teeth
has traditionally been used to group Antilocapridae
with Bovidae whereas has been used to group Giraf-
ﬁdae with Cervidae the possession of low-crown or
brachyodont cheek teeth with many accessory styles and
ribs (e.g., Romer, 1966). Although the height of cheek
teeth has been widely used in ruminant phylogeny,
Janis and Scott (1987) considered this character to be
without taxonomic value because it has evolved in par-
allel many times within herbivorous mammals (e.g., Pro-
boscidea, Rhinocerotidae, Equidae, Suidae, and Camel-
idae). Moreover, this character is problematic because
of the ambiguity in its deﬁnition. For example, Moschus
and Hydropotes were considered hypsodont by Gentry
and Hooker (1988) but brachyodont by Scott and Janis
(1993). Furthermore, hypsodonty is variable within Bovi-
dae, some species being more brachyodont than oth-
ers. Considering these problems, hypsodonty was not
included in our morphological character data matrix.
The possession of two lacrimal oriﬁces situated on the
orbital rim has been used to associate Antilocapridae
with Cervidae (Leinders, 1979; Leinders and Heintz,
1980). The number of lacrimal oriﬁces is a very ambigu-
ous character since it is highly variable in bovids—with
one, two or even three oriﬁces—and since it may
even be variable within a species. For example, one
or two oriﬁces are found for Antilocapra americana and
Moschus moschiferus (Scott and Janis, 1987; personal
The association of Antilocapridae with Cervidae and
Moschidae was also proposed on the basis of the
presence of a closed metatarsal gully (Leinders, 1979;
Leinders and Heintz, 1980; Janis and Scott, 1987; Scott
and Janis, 1987). The mode of fusion of the metatarsal
gully does not include a bridged distal end in bovids and
girafﬁds, whereas a closed metatarsal gully is present in
cervids and antilocaprids (Leinders and Heintz, 1980).
This character is not constant in Moschus; an open
gully was observed in one specimen by Janis and Scott
(1987). This character is also variable in Tragulidae; the
metatarsal gully of Tragulus is open and that of Hye-
moschus is closed.
The complete loss of side toes has been used to unite
Antilocapridae with Bovidae, but it does not constitute
220 SYSTEMATIC BIOLOGY VOL. 52
a real synapomorphy. Actually, the complete loss of side
toes is encountered in antilocaprids and some bovid
species but also in girafﬁds. In bovids, the second and
ﬁfth digits are generally not completely lost because of
the retention of proximal remnants (dewclaws). Because
the loss of lateral digits can be functionally associated
with increasing cursoriality in open habitats, this charac-
ter is expected to be homoplastic for ruminant phylogeny
(Scott and Janis, 1993) and has apparently evolved in par-
allel many times within cetartiodactyls (e.g., Camelidae
The retention of a gall bladder has been used to unite
Antilocapridae with Bovidae (e.g., O’Gara and Matson,
1975; Gentry and Hooker, 1988), but this organ is also
present in Tragulidae, Moschidae, and some Girafﬁdae
(Frechkop, 1955). Its absence is probably a derived char-
acter of the Cervidae that has also been acquired inde-
pendently by some girafﬁds (Groves and Grubb, 1987).
We built a new morphological matrix (Table 2) using
the two following assumptions for the coding of char-
acter states: (1) all the six ruminant families were con-
sidered monophyletic and (2) all the character states ex-
amined here were considered the primitive states within
each of the families. Our analysis based on 48 morpho-
logical characters revealed that only two of these charac-
ters were phylogenetically informative for establishing
interfamily relationships within Ruminantia. The results
suggest that Antilocapridae and Girafﬁdae emerged ﬁrst
with respect to a clade composed of the families Bovidae,
Moschidae, and Cervidae (Fig. 1). Although this topol-
ogy is in perfect agreement with our molecular trees, the
nodes are poorly supported and only one unambigu-
ous synapomorphy supports the grouping of bovids,
moschids, and cervids: the presence of a large foramen
ovale (character 27, Table 2). Although this character
was implicitly coded as discontinuous by Scott and Janis
(1993), and in our data matrix, this character is contin-
uous and is therefore of less interest because its coding
can be variable depending on the observer.
Molecular phylogeny: Moschus +Bovidae.—The living
musk deer (Moschus) was regarded as a member of
the family Cervidae (e.g., Simpson, 1945; Viret, 1961;
Eisenberg, 1987; Nowak, 1991) because it possesses a
cervidlike closed metatarsal gully. Because of the absence
of antlers and the possession of sabrelike upper canines,
the Moschinae were often considered a basal offshoot
within the Cervidae, as it was assumed for Hydropotes
(Chinese water deer), which also lacks antlers and pos-
sesses sabrelike upper canines. Contrary to Hydropotes,
Moschus exhibits several features of the soft anatomy, that
are supposed to be primitive with respect to all living
cervids: a gall bladder, an ileocecal gland, an intestine
with three and a half (as opposed to two and a half)
colic coils, and a placenta with many (rather than few)
cotyledons. For these reasons, Moschus is more gener-
ally included in its own family, the Moschidae. Because
it lacks the cervid feature of a double lacrimal oriﬁce,
Leinders (1979) and Leinders and Heintz (1980) consid-
ered Moschus to be less closely related than Antilocapra
to cervids. Webb and Taylor (1980) considered all living
Pecora except the Moschidae (the Eupecora) to be united
by the presence cranial appendages.
The present morphological analysis indicates that
Moschus is allied with the families Bovidae and Cervi-
dae. This result conﬁrms the view of several authors
who have noted that the musk deer exhibits a mixture
of bovid and cervid characters (Leinders and Heintz,
1980). However, the molecular placement of Moschus as
the sister group of bovids rather than cervids is unex-
pected, because morphologists have never proposed this
evolutionary possibility. Phylogenetic analyses using a
large sample of pecoran species for Cyb sequences have
already suggested that Moschus could be more closely
related to Bovidae than Cervidae (Hassanin, 1999), but
these relationships were not robustly supported. Here,
the association of Moschus with Bovidae is supported
by (1) Bayesian, ML, and MP analyses, (2) the separate
analyses of two nuclear markers (MP, Fig. 2; Lf:BP=
92 and αLAlb:BP=96), and (3) the analysis combining
seven markers (BP =95, Fig. 2; BPML =94, PPB=1.00,
Fig. 3). The log-likelihoods of the best tree under models
with a single partition versus 12 partitions are −31980.15
and −30322.09, respectively. Therefore, the 12-partition
model better describes the evolution of the seven mark-
ers (P<0.001 under a likelihood ratio test), and the phy-
logenetic results observed (i.e., the sister group relation-
ship between Moschidae and Bovidae) does not reﬂect
the use of an oversimpliﬁed ML model (Whelan et al.,
Moreover, the position of Moschus close to Bovidae re-
mains strongly supported when phylogenetic relation-
ships are reconstructed using a less divergent outgroup,
such as Tragulus. Two diagnostic molecular signatures
characterized the grouping of Moschus with Bovidae
(C →T, Lf, position 128; and C →G, αLAlb, position 80),
whereas no signature was detected for Moschus consid-
ered either sister to Cervidae or basal to the clade joining
bovids and cervids. From a paleontological point of view,
the case of Hispanomeryx might reﬂect the close relation-
ships between Moschus and Bovidae; this Miocene genus
of Spain, which does not possess cranial appendages, has
been alternatively included in Moschidae (Morales et al.,
1981) and Bovidae (Moy`a-Sol`a, 1986).
Hoplitomeryx, from the late Miocene of Italy, seems
crucial for a better understanding of the interrelation-
ships between Cervidae, Moschus, and Bovidae. This en-
demic genus of the island fauna of Monte Gargano was
placed in a new family, Hoplitomerycidae, on the basis
of two unusual properties of its skull (Leinders, 1983):
(1) it possesses ﬁve cranial appendages, i.e., one in me-
dial position on the nasal bone and four in supraorbital
position on the frontal bone, and (2) it bears daggerlike
upper canines, whereas the presence of large upper ca-
nines is always correlated with the absence of cranial
appendages in all extant pecorans. Leinders (1983:41)
described the cranial appendages of Hoplitomeryx: “they
show typical features of horncores as present in the Bovi-
dae and they almost certainly once were covered by a
keratine sheath.” Janis and Scott (1987:57) reported that
“the deep grooving of the horn cores of Hoplitomeryx is
2003 HASS ANIN AND DO UZERY—PHYLO GENY O F RUMINANTIA 221
reminiscent of the condition of bovid horn cores with a
permanent keratin sheath, rather than the spongy tex-
ture of the antilocaprid horn core.” Nevertheless, Lein-
ders concluded that horns developed independently in
Bovidae and placed Hoplitomerycidae as the sister fam-
ily of Cervidae because they share a double lacrimal ori-
ﬁce and a closed metatarsal gully. However, these two
characters are highly variable in taxa other than cervids.
For instance, specimens of Moschus have either one or
two lacrimal oriﬁces, and the gully on the anterior side
of metatarsal III/IV is either open or closed. These two
characters are therefore very doubtful for linking Hoplit-
omeryx with Cervidae. In contrast, there is no evidence
of parallel evolution of bovidlike horns. Hence, we sug-
gest that Hoplitomerycidae is closer to Bovidae than to
Cervidae. Hoplitomeryx differs from Bovidae in having
ﬁve rather than two (most bovids) or four (Tetracerus
quadricornis) horns and by the presence of large sabrelike
upper canines. Males of Moschus and Hydropotes, which
are the sole extant pecorans that lack cranial appendages,
use their upper canines as weapons in intraspeciﬁc com-
bat (Grzimek, 1968) and probably as means of defense
against predators. On the basis of the absence of antlers,
Hydropotes was considered as the sister group of all living
antlered cervids (e.g., Groves and Grubb, 1987; Janis and
Scott, 1987), but recent molecular analyses have clearly
shown that Hydropotes lost the antlers secondarily (Randi
et al., 1998; this study). The development of large up-
per canines may be therefore interpreted as a secondary
adaptation for intraspeciﬁc ﬁghting. Similarly, the acqui-
sition of sabrelike upper canines in Moschus may have
occurred after the loss of horns from a bovidlike ances-
tor, i.e., one without large upper canines. Small upper
canines have been found in Miocene bovids, e.g., Pro-
tragocerus labidotus (Gentry, 1970). In addition, vestiges
of upper canine alveoli can be seen in some extant genera
of Bovidae (e.g., Tetracerus, Cephalophus, Oreotragus), and
the occasional occurrence of rudimentary upper canines
has been recorded in various species (e.g., Neotragus pyg-
maeus, Ourebia ourebi, Gazella granti, Rupicapra rupicapra)
(Dekeyser and Derivot, 1956; Gentry, 1970). However,
Hoplitomeryx is unique among extant and fossil rumi-
nants because it possesses a large set of weapons, with
ﬁve horns and large upper canines. Leinders (1983) sug-
gested that this remarkable adaptation could have de-
veloped as a consequence of insularity to protect against
large birds of prey, which were the exclusive predator of
Molecular phylogeny: basal position of Antilocapridae and
Girafﬁdae.—Our DNA results suggest two possibilities
about the phylogenetic position of antilocaprids and gi-
rafﬁds. First, MP analyses suggest a basal position for
Antilocapra, in agreement with previous molecular in-
vestigations that suggested the ﬁrst divergence of An-
tilocapridae within Pecora (BP =75–81) (Cronin et al.,
1996; Montgelard et al., 1997; Gatesy and Arctander, 2000;
Matthee et al., 2001). Two autapomorphies that appear
to be transversions support the placement of Girafﬁdae
with Cervidae, Moschus, and Bovidae (T →A[Gfor
Giraffa], 16S, position 541; and T →G, κCas, position
314). Second, ML analyses suggest the grouping of An-
tilocapra with the Girafﬁdae. Two autapomorphies, again
transversions, deﬁne this association (A →T, 16S, posi-
tion 586; and A →C[TforGiraffa], Cyb, position 897).
The ML conclusion of the phylogenetic analysis of the
six markers, described as a combination of 12 partitions,
is that two competing phylogenetic hypotheses for the
radiation of Pecora families are observed. The ﬁrst one,
supported by the MP and some ML analyses is that An-
tilocapridae is the ﬁrst offshoot among Pecora, followed
by Girafﬁdae, and then the three remaining families.
The second one is that Antilocapridae and Girafﬁdae are
sister groups and represent the sister clade of Bovidae,
Cervidae, and Moschidae. The conservative SH test does
not discriminate (PSH =0.93) between these alternative
rootings of the pecoran subtree (Fig. 4).
Despite the fact that the relative position of Antilo-
capridae and Girafﬁdae is not resolved, it seems clear
that both families diverged before the divergence of the
families Bovidae, Cervidae, and Moschidae. This result is
supported by all of our analyses, and three nuclear DNA
signatures are diagnostic for the clade uniting bovids,
moschids, and cervids (C →T, Cyp, position 199; G →
C, αLAlb, position 281; and C →T, Lf, position 276). This
result is also corroborated by our morphological anal-
ysis (Fig. 1). The single morphological synapomorphy
uniting Cervidae, Moschidae, and Bovidae is the pres-
ence of a large foramen ovale. Other poorly explored
characters tend to give more credit to this clade. With
regard to the stomach anatomy, the entrance of the
oesophagus to the rumen is more ventrally positioned in
bovids and cervids (Hofmann, 1973) than in the giraffe
(Giraffa). Unfortunately, the position of the oesophagus in
Antilocapra, Okapia, and Moschus is not known. The major
volatile compounds in the interdigital glands of Antilo-
capra are clearly different from those discovered in vari-
ous species of Bovidae and Cervidae (Wood, 2001). These
biochemical differences may also reﬂect the distinctness
of bovids and cervids with respect to antilocaprids. Inter-
digital gland compounds have not been determined for
Moschidae and Girafﬁdae.
Paleontology and Biogeography
What are the diagnostic characters of Ruminantia in the
fossil record?—Numerous recent molecular analyses have
shown that the order Cetartiodactyla is composed of
ﬁve major groups: Tylopoda (camels and llamas), Suina
(pigs and peccaries), Hippopotamidae, Cetacea, and Ru-
minantia. There is a strong consensus among molec-
ular systematists to unite Cetacea with Hippopotami-
dae, to associate this clade with Ruminantia, and to
consider Tylopoda as the ﬁrst offshoot among Cetar-
tiodactyla (e.g., Graur and Higgins, 1994; Irwin and
Arnason, 1994; Gatesy et al., 1996, 1999; Gatesy, 1997;
Montgelard et al., 1997; Shimamura et al., 1997, 1999;
Nikaido et al., 1999; Madsen et al., 2001; Murphy et al.,
2001a, 2001b). These molecular results therefore in-
validated the grouping of Tylopoda and Ruminantia
into Selenodontia, as assumed by many paleontologists
222 SYSTEMATIC BIOLOGY VOL. 52
(e.g., Lavocat, 1955; Romer, 1966; Webb and Taylor, 1980;
Scott and Janis, 1987; Gentry and Hooker, 1988). In this
context, all the skeletal characters used to unite these two
suborders must be considered to have evolved in par-
allel, probably as the consequence of similar ecological
Ruminants and tylopods are terrestrial herbivores and
feed primarily on ﬁbrous vegetation. They indepen-
dently acquired a complex stomach with multiple cham-
bers that permits rumination, i.e., they are able to digest
cellulose because of a symbiotic relationship with mi-
croorganisms. The dentition of ruminants also reﬂects
the evolutionary trends toward an herbivorous diet. The
cheek teeth are selenodont (and not bunodont as in Suina
and Hippopotamus), and there is a spacious diastema be-
tween labial teeth (incisors and canine) and cheek teeth
(molars and premolars). Although the latter characters
evolved in parallel in tylopods, ruminants are easily dis-
tinguishable by two dental synapomorphies: the upper
incisors are absent and replaced by a horny pad, and
the lower canine is incisiform and contiguous with the
three incisors. Moreover, the upper canine is absent or its
presence is sexually dimorphic, i.e., large and curved in
males and smaller in females (in Tragulidae, Moschidae,
Muntiacus, and Hydropotes).
Evolution of the limb skeleton of Ruminantia parallels
changes in stomach anatomy and dentition. Ruminants
and tylopods live in more open habitats than do other
cetartiodactyls, and as a result they have developed con-
vergent adaptations to better cursoriality: Lateral digits
II and V are either absent or reduced, and in the hind feet
metatarsals III and IV are always fused into a canon bone.
However,ruminants possess two derived limb character-
istics: in the tarsus the navicular and cuboid are fused,
and in the carpus the magnum and trapezoid are fused.
The most recent common ancestor of all extant rumi-
nants is expected to display the following diagnostic fea-
tures: (1) cubonavicular, (2) fusion of the magnum and
trapezoid, (3) absence of upper incisors, and (4) inclusion
of the lower canine in the incisor row. All other characters
must be regarded with caution to signal the ﬁrst occur-
rence of Ruminantia in the fossil record.
Middle Eocene origin of ruminants.—In the Middle/Late
Eocene, a peak of herbivore diversity occurred for mam-
mals in Europe, Asia, and North America, reﬂecting
the diversiﬁcation of selenodont cetartiodactyls, perisso-
dactyls, rodents, and lagomorphs. This epoch was char-
acterized by increasing seasonal dryness that favored the
development of more open, less forested habitat than oc-
curred earlier in the Eocene (Janis, 2000b).
In agreement with our mean molecular estimation of
44.3–46.3 MYA for the Tragulina/Pecora split (Fig. 5),
the fossil record indicates that ruminant traits ap-
peared suddenly in the Middle–Late Eocene of the
Northern Hemisphere: Amphimerycidae in Europe (Am-
phimeryx, France), Archaeomerycidae in Central Asia
(Archaeomeryx, Mongolia), and Leptomerycidae (Lep-
tomeryx) and Hypertragulidae (Hypertragulus and Sim-
imeryx) in North America (Webb and Taylor, 1980;
Vislobokova, 1997; Webb, 1998). All members of these
families have a cuboid fused with the navicular (Viret,
1961; Geraads et al., 1987; Gentry and Hooker, 1988; Scott
and Janis, 1993) and incisiform lower canines (Gentry
and Hooker, 1988). Leptomerycidae are however more
derived than other families; contrary to Amphimeryci-
dae, Archeomerycidae, and Hypertragulidae, they do
not possess upper incisors, and contrary to Hypertrag-
ulidae, they present a fusion of the trapezoid and mag-
num. Leptomerycidae alone possess all four diagnostic
features expected in the common ancestor of all extant
ruminants. The close relationships of Leptomerycidae
with Ruminantia has been supported by cladistic anal-
yses; they have been hypothesized as sister taxa to ei-
ther Tragulina (Geraads et al., 1987) or Pecora (Janis and
Scott, 1987, Gentry and Hooker, 1988; Scott and Janis,
At the beginning of the Middle Eocene, Europe and
Asia were separated from Africa, and North America
was seperated from South America. The rifting of the
North Atlantic cut off the dispersal routes between
Europe and North America, and in parallel the sea
level rose and seas invaded much of Siberia (Raven and
Axelrod, 1974; McKenna, 1983; MacFadden, 1992). This
inundation created a signiﬁcant oceanic barrier to fau-
nal interchange, and as a consequence the mammalian
faunal similarity between Palaearctica and Nearctica de-
creased dramatically. This divergence is illustrated by
early perissodactyls, for which there was a split result-
ing in Equidae in North America and Palaeotheriidae in
Eurasia (MacFadden, 1992). A similar vicariant separa-
tion may be inferred also for fossils related to ruminants
at this time, with Hypertragulidae and Leptomeryci-
dae in North America, Amphimerycidae in Europe, and
Archaeomerycidae in Asia.
First occurrence of Pecora in the early oligocene of
Eurasia.—The major division of Ruminantia between
Pecora and Tragulina is widely accepted among mor-
phologists and molecular systematists, with Pecora in-
cluding Antilocapridae, Bovidae, Cervidae, Girafﬁdae,
and Moschidae and Tragulina including only Tragulidae.
However, there is no consensus among paleontologists
for establishing the phylogenetic afﬁnities of the fossils
from Eocene and Oligocene deposits. Therefore, it seems
particularly important to deﬁne the morphological dif-
ferences between extant Pecora and Tragulina.
All modern families of Pecora are associated with more
open habitats than are Tragulidae, which are exclusively
found in overgrown tropical forests in southeast Asia
and Africa. Consequently, limb features of Pecora re-
veal better adaptation to cursoriality: (1) third and fourth
metacarpals are completely fused into a canon bone,
whereas in tragulids they are either unfused (African
tragulids) or partially fused (Asian tragulids); (2) the as-
tragalus is compact and has parallel sides; and (3) there
are complete distal metapodial keels (Frechkop, 1955;
Scott and Janis, 1993). In addition, tragulids are small,
with a shoulder height that ranges from 20 to 35 cm,
whereas pecorans are generally much larger: 81–104 cm
for antilocaprids, 150–370 cm for girafﬁds (Okapia: 150–
170 cm; Giraffa: 250–370 cm), 50–61 cm for moschids,
2003 HASS ANIN AND DO UZERY—PHYLO GENY O F RUMINANTIA 223
FIGURE 5. Consensus phylogenetic tree from the MP, ML, and Bayesian analyses and a time scale for ruminant evolution as inferred from a
Bayesian relaxed molecular clock approach. Only the nodes strongly supported by the three methods (BP >70, PPB >0.99) are indicated in this
224 SYSTEMATIC BIOLOGY VOL. 52
24–220 cm for bovids, and 25–190 cm for cervids (Nowak,
1991). Small pecorans are represented by bovid and
cervid species secondarily adapted to forests or areas
of dense vegetation, e.g., dwarf antelope in Africa (24–
41 cm), muntjacs in Asia (40–80 cm), or pudus in South
America (25–43 cm). All extant pecorans except Moschus
and Hydropotes have cranial appendages. However, the
presence of cranial appendages in the common ancestor
of modern pecorans is not certain; several authors have
suggested that horns, antlers, and ossicones are not ho-
mologous (e.g., Janis and Scott, 1987; Gentry and Hooker,
1988; Scott and Janis, 1993).
The Eocene/Oligocene boundary at around 34 MYA
is correlated with the expansion and diversiﬁcation of
grasslands, when the warm and damp environment of
the Eocene changed to colder and dryer Oligocene con-
ditions (Prothero and Heaton, 1996; Meng and McKenna,
1998). During the Late Eocene and Oligocene, the num-
ber of small herbivores declined severely, whereas a sig-
niﬁcant increase was recorded in medium-size animals,
almost entirely represented by tylopods, equids, and ru-
minants in North America (Janis, 2000b). Our molecular
estimation indicates that the evolutionary radiation of
Pecora occurred at the Early Oligocene (Fig. 5) between
28.9 and 31.6 MYA, i.e., just after the Eocene/Oligocene
transition referred as the Grande Coupure (Stehlin, 1909).
This estimation is in accordance with the fossil record,
since the ﬁrst occurrence of medium-size mammals un-
ambiguously related to Pecora is in Eurasia with Eumeryx
in the Early Oligocene of Mongolia about 32 MYA and
Dremotherium in France at approximately 29–30 MYA
(Blondel, 1997; Vislobokova, 1997). Dremotherium ap-
pears more derived than Eumeryx because it possesses
all characteristics of modern pecorans, i.e., a canon bone
in the four limbs, complete distal metapodial keels (not
in Eumeryx), and a compact and parallel-sided astragalus
(Scott and Janis, 1993).
After the Middle Eocene, Beringia became the dom-
inant Holarctic intercontinental connection between
North America and Eurasia. The Bering land bridges
periodically emerged during times of low sea level,
permitting sporadic intercontinental exchanges between
North America and Asia. On the basis of the available
fossil record, extant ruminants originated in the Mid-
dle/Late Eocene of North America, with the ﬁrst appear-
ance of Leptomerycidae. Because pecorans emerged in
Europe during the Oligocene, it must be assumed that
their common ancestor crossed the Bering Strait from
North America to Asia before this period, perhaps at the
Eocene/Oligocene boundary, which was probably char-
acterized by a major sea level regression due to severe
cooling (Prothero and Heaton, 1996).
Evolutionary radiation and dispersion of the pecoran
families during the Miocene and Plio-Pleistocene.—The
greatest radiation of Pecora occurred during the
Miocene, when much of the Earth’s forest habitats were
replaced by grasslands because of the widespread cool-
ing and drying of the climate. The Miocene was also
an important epoch for dispersal events between Africa,
Eurasia, and North America. The characters deﬁning all
extant families of Pecora appeared in the Early Miocene
(Gentry, 1994; Gentry et al., 1999).
The Antilocapridae are endemic to North America,
and the sole extant species is the pronghorn (Antilocapra
americana). This family was ﬁrstly recorded in the Early
Miocene with diverse genera included into the subfam-
ily Merycodontinae, but species closely related to the
pronghorn and placed into the subfamily Antilocapri-
nae ﬂourished in the Middle/Late Miocene (Janis and
Manning, 1998). The origin of this family may lie in Eura-
sia, because there are no pecorans in North America prior
to the Early Miocene. Paleoclimatological studies indi-
cate that the Early Miocene was a relatively cold time and
presumably a time of low sea level, which opened up the
previously inundated land bridges of Beringia (Opdyke,
1990). At about 24 MYA, horses dispersed from North
America to Eurasia (MacFadden, 1992). The earliest en-
try of Antilocapridae into North America from Eurasian
ancestors may have been at the same time.
The Girafﬁdae are today conﬁned to Africa, where they
are represented by only two genera, Giraffa and Okapia.
The ossicones and bilobed lower canines are two autapo-
morphic features of extant girafﬁds. Bilobed lower ca-
nines were found in the Early Miocene of Africa in gen-
era such Climacoceras and Canthumeryx (Hamilton, 1978;
Geraads, 1986). However, some possible ossicone frag-
ments from Lorancameryx were found in more ancient
deposits in the Early Miocene of Spain (Morales et al.,
1993). These ﬁndings suggest that girafﬁds originated
in Europe in the Early Miocene and dispersed rapidly
in Africa. This hypothesis is supported by the fact that
Africa was isolated from Eurasia until the Early Miocene
and by the highly probable Eurasian origin of pecorans.
Girafﬁds may have entered Africa during a period of
low sea level, at about 21 MYA (Van der Made, 1999),
after which they split off in the African–Arabian–Indian
faunal realm and enjoyed a certain early success. Our
molecular results suggest that the split between Giraffa
and Okapia occurred at that time (Fig. 5), between 19.0
and 19.9 MYA. Some girafﬁds later dispersed to the rest
of Eurasia, but they declined with the diversiﬁcation of
bovids later in the Miocene (Gentry, 1994).
The Cervidae are today found in America, Eurasia, and
North Africa. The possession of antlers in males is a di-
agnostic feature for this family. Antlers appeared in the
Early Miocene of Eurasia with Procervulus, Ligeromeryx,
Dicrocerus, Lagomeryx, and Stephanocemas (Eisenberg,
1987; Gentry, 1994; Gentry et al., 1999). Molecular dat-
ings then indicate that most of the major splits within
cervids took place during the Early and Middle Miocene
(Fig. 5). Later, Cervidae colonized North America in the
Early Pliocene, and at the Pliocene completion of the Isth-
mus of Panama they entered South America, where they
experienced a dramatic adaptative radiation (Eisenberg,
1987). Later they made marginal penetrations into north-
ern parts of Africa, and they entered India at the end of
the Pliocene, at about 2.5 MYA (Barry and Flynn, 1989).
The Moschidae are now conﬁned in Asia, but nu-
merous fossils from Oligocene and Miocene deposits of
Eurasia, America, and Africa have been included in this
2003 HASS ANIN AND DO UZERY—PHYLO GENY O F RUMINANTIA 225
family on the basis of three major characters: small to
medium body size, absence of cranial appendages, and
presence of long upper canines. However, these char-
acters are not diagnostic for Moschidae because they
are also encountered in the living cervid Hydropotes and
in most ancient fossil pecorans. Accordingly, it is difﬁ-
cult to establish the ﬁrst occurrence of moschids in the
fossil record. The use of a Bayesian relaxed molecular
clock with the combination of nuclear and mitochondrial
markers suggests that moschids originated before the
Oligocene/Miocene transition at 26.1–26.2 MYA (Fig. 5).
The Bovidae have a very large distribution; they are
found on all continents except Australia, Antarctica, and
South America. In the fossil record, they are easily iden-
tiﬁed by the presence of horn cores. These typical ap-
pendages appeared with Eotragus in the Early Miocene
of western Europe and Pakistan at about 18 MYA and
thereafter in Africa (Vrba, 1985; Barry and Flynn, 1989;
Gentry, 1994). According to our molecular estimations,
the split between Antilopinae and Bovinae took place
during the beginning of the Early Miocene, around 22.6
MYA, and may be explained by a vicariant barrier of sea
level change and/or development of an arid belt between
Africa and Eurasia (Hassanin and Douzery, 1999b). These
two subfamilies simultaneously ﬂourished during the
Early Miocene, around 19.3–19.7 MYA, with the appear-
ance at this time of all extant tribes of the family (Fig. 5).
Among antilopines, the splits separating Caprini, Al-
celaphini, and Hippotragini occurred during the Middle
Miocene, around 15.1 MYA. During the Plio-Pleistocene,
three caprine genera (Oreamnos, Ovibos, and Ovis) dis-
persed to North America by crossing the Bering Strait,
but none of them entered in South America.
CONCLUS IO NS
Molecular analyses of seven mitochondrial and nu-
clear markers have shown that families Bovidae, Cervi-
dae and Moschidae are closely related, with the musk
deer as the sister group of bovids rather than cervids.
However, the relative phylogenetic afﬁnities of Antilo-
capridae and Girafﬁdae remain unsettled. Morpholog-
ical characters currently used in the literature exhibit
too much homoplasy to be powerful for inferring phy-
logenetic relationships among extant ruminant fami-
lies. The evolutionary radiation of Pecora occurred at
the Early/Late Oligocene transition as estimated from
a Bayesian relaxed molecular clock approach, and pec-
oran families diversiﬁed and dispersed rapidly during
the Early and Middle Miocene.
A.H. thanks Christiane Denys, Jean Deutsch, and Simon Tillier
for laboratory facilities, Jacques Cuisin, Jean-Marc Pons, Francis
Renoult, Daniel Robineau, and Michel Tranier for providing access
to the MNHN collections, and Denis Geraads, Gr´egoire M´etais, and
Salvador Moya-Sola for help with the bibliography. E.J.P.D. thanks
Fran¸cois Catzeﬂis for introducing him in the Pecora world. We both
thank Philippe Chardonnet and Bertrand des Clers for tissues of Aepyc-
eros melampus, Gazella granti, Redunca fulvorufula, and Tragelaphus im-
berbis, Eric Pasquet for Cephalophus dorsalis skin, Jean-Claude Thibault
for Ovis aries blood samples, Francoise Hergueta-Claro for Hippotra-
gus niger and Damaliscus pygargus cells, Jaap Beintema, Am´elie Bonnet,
Jean-Yves Dubois, Francoise Hergueta-Claro, and David Irwin for
Giraffa camelopardalis and Okapia johnstoni tissue and DNA, Anne Collet
for Balaenoptera physalus tissues, Dan Gallagher,Arlene Kumamoto, and
Jim Womack for Tragulus javanicus DNA, Conrad Matthee and Bettine
Jansen van Vuuren for Hippotragus niger DNA, Knut Roed for Rangifer
tarandus DNA, Diana R. Reynolds for Hippopotamus amphibius tissues,
Harry Scherthan for Muntiacus reevesi cells, and Antoine Semp´er´e
for Moschus moschiferus tissues. We also acknowledge Chris Simon,
Jeffrey Thorne, Hirohisa Kishino, and one anonymous reviewer for
their helpful comments and suggestions on the manuscript. This work
has been supported by funds from the TMR Network Mammalian Phy-
logeny (contract FMRX-CT98-022) of the European Community, the
Genopole Montpellier Languedoc-Roussillon, and the Action Bioinfor-
matique inter-EPST of the CNRS to E.J.P.D. and represents contribution
no. 2002-047 of the Institut des Sciences de l’Evolution de Montpellier
ADACHI, J., AND M. HAS EGAWA. 1996. Model of amino acid substitution
in proteins encoded by mitochondrial DNA. J. Mol. Evol. 42:459–468.
ALEXANDER, L. J., A. F. STEWART ,A.G.MACKINLAY,T.V.
APELINS KAYA,T.M.TKACH,AND S. I. GORODETSKY. 1988. Isola-
tion and characterization of the bovine kappa-casein gene. Eur. J.
ALLARD, M. W., M. M. MIYAMO TO,L.JAR ECKI ,F.KRAUS ,AND M. R.
TENNANT. 1992. DNA systematics and evolution of the artiodactyl
family Bovidae. Proc. Natl. Acad. Sci. USA 89:3972–3976.
ANDERS ON, S., M. H. L. DEBRUIJN,A.R.COULS ON,I.C.EPERS ON,
F. SANG ER,AND I. G. YOUNG. 1982. Complete sequence of bovine
mitochondrial DNA. Conserved features of the mammalian mito-
chondrial genome. J. Mol. Biol. 156:683–717.
ARNASON, U., A. GULLB ERG,AND B. WIDEG REN. 1991. The complete
nucleotide sequence of the mitochondrial DNA of the ﬁn whale,
Balaenoptera physalus. J. Mol. Evol. 33:556–568.
BARRIEL, V. 1994. Phylog´enies mol´eculaires et insertions-d´el´etions de
nucl´eotides. C.R. Acad. Sci. Paris Life Sci. 317:693–701.
BARRY,J.C.,AND L. FLYNN. 1989. Key biostratigraphic events in the
Siwalik sequence. Pages 557–571 in European Neogene mammal
chronology. (E. H. Lindsay, V. Fahlbusch, and P. Mein, eds.). Plenum,
BEINTEM A, J. J., C. SCH ¨
ULLER,M.IRIE,AND A. CARS ANA. 1988. Molec-
ular evolution of the ribonuclease superfamily. Prog. Biophys. Mol.
BLONDEL, C. 1997. Les ruminants de Pech Desse et de Pech du Fraysse
(Quercy; MP28); ´evolution des ruminants de l’Oligoc`ene d’Europe.
Geobios (Lyon) 30:575–591.
BOUVR AIN, G., AND D. GERAADS . 1985. Un squelette complet de Bachith-
erium (Artiodactyla, Mammalia) de l’Oligoc`ene de C´ereste (Alpes
de Haute-Provence). Remarques sur la syst´ematique des Ruminants
primitifs. C.R. Acad. Sci. Paris II 300:75–78.
BREMER,K. 1988. The limits of amino acid sequence data in angiosperm
phylogenetic reconstruction. Evolution 42:795–803.
BROOKE, V. 1878. On the classiﬁcation of the Cervidae, with a synopsis
of the existing species. Proc. Zool. Soc. Lond. 1878:883–928.
CHIKUNI, K., Y. MORI,T.TABATA,M.SAITO,M.MOMMA,AND
M. KOS UGIYAM A. 1995. Molecular phylogeny based on the κ-casein
and cytochrome bsequences in the mammalian suborder Ruminan-
tia. J. Mol. Evol. 41:859–866.
CRONIN, M. A., R. STUART,B.J.PIERSON,AND J. C. PATTON. 1996.
κ-Casein gene phylogeny of higher ruminants (Pecora, Artiodactyla).
Mol. Phylogenet. Evol. 6:295–311.
DEBRY, R. W. 2001. Improving interpretation of the decay index for
DNA sequence data. Syst. Biol. 50:742–752.
DEKEYS ER,P.L.,AND J. DER IVOT. 1956. Sur la pr ´esence de canines
sup´erieures chez les Bovid´es. Bull. Inst. Fr. Afr. Noire 18:1272–1281.
DOUZERY, E., AND E. RAND I. 1997. The mitochondrial control region
of Cervidae: Evolutionary patterns and phylogenetic content. Mol.
Biol. Evol. 14:1154–1166.
DUBOS T, G. 1965. Quelques traits remarquables du comportement de
Hyaemoschus aquaticus. Biol. Gabonica 1:282–287.
226 SYSTEMATIC BIOLOGY VOL. 52
DUNG, V. V., P. M. GIAO,N.N.CHINH,D.TUOC ,P.ARCTANDER,AND
J. MACKINNON. 1993. A new species of living bovid from Vietnam.
EISENBERG, J. F. 1981. The mammalian radiations. An analysis of
trends in evolution, adaptation, and behavior. Univ. Chicago Press,
EISENBERG, J. F. 1987. The evolutionary history of the Cervidae with
special reference to the South American radiation. Pages 60–64 in Bi-
ology and management of the Cervidae (C. M. Wemmer, ed.). Smith-
sonian Institution Press, Washington, D.C.
FELSENS TEI N, J. 1985. Conﬁdence limits on phylogenies: An approach
using the bootstrap. Evolution 39:783–791.
FLEROV , K. K. 1952. Fauna of USSR mammals, Volume I, Number 2.
Musk deer and deer. Institute of Zoology publication 55. Academy
of Sciences of the USSR, Moscow.
FLOWER, W. H. 1883. On the arrangement of the orders and families of
existing Mammalia. Proc. Zool. Soc. Lond. 1883:178–186.
FRECHKOP, S. 1955. Super-ordre des Ongul´es. Pages 484–1167 in
Trait´e de zoologie. Anatomie, syst´ematique, biologie. Tome XVII,
premier fascicule. Mammif`eres. Les ordres: anatomie, ´ethologie,
syst´ematique. P. P. Grass ´e, Masson, and Cie, Paris.
FURET, J. P., J. C. MERCIER,S.SOULIER,P.GAYE,D.HUE-DELAHAIE,
AND J. L. VILOTTE. 1990. Nucleotide sequence of ovine kappa-casein
cDNA. Nucleic Acids Res. 18:5286.
GALLAGHER, D. S., M. L. HOUCK,A.M.RYAN,J.E.WOMACK ,AND A. T.
KUMAMO TO. 1996. A karyotypic analysis of the lesser Malay chevro-
tain, Tragulus javanicus (Artiodactyla: Tragulidae). Chromosome Res.
GATES Y, J. 1997. More DNA support for a Cetacea/Hippopotamidae
clade: The blood-clotting protein gene γ-ﬁbrinogen. Mol. Biol. Evol.
GATES Y, J. 1998. Molecular evidence for the phylogenetic afﬁnities of
Cetacea. Pages 53–111 in The emergence of whales: Evolutionary
patterns in the origin of Cetacea (J. G. M. Thewissen, ed.). Plenum,
GATES Y, J., G. AMATO,E.VRBA,G.SCHALLER,AND R. DES ALLE. 1997. A
cladistic analysis of mitochondrial ribosomal DNA from the Bovidae.
Mol. Phylogenet. Evol. 7:303–319.
GATES Y, J., AND P. A RCTANDER. 2000. Hidden morphological sup-
port for the phylogenetic placement of Pseudoryx nghetinhensis with
bovine bovids: A combined analysis of gross anatomical evidence
and DNA sequences from ﬁve genes. Syst. Biol. 49:515–538.
GATES Y, J., C. HAYASHI ,M.A.CRONIN,AND P. A RCTANDER. 1996. Ev-
idence from milk casein genes that cetaceans are close relatives of
hippopotamid artiodactyls. Mol. Biol. Evol. 13:954–963.
GATES Y, J., M. MILINKOVITCH,V.WADD ELL,AND M. STANHOPE. 1999.
Stability of cladistic relationships between Cetacea and higher-level
artiodactyl taxa. Syst. Biol. 48:6–20.
GATES Y, J., D. YELON,R.DESALLE,AND E. S. VRBA. 1992. Phylogeny
of the Bovidae (Artiodactyla, Mammalia), based on mitochondrial
DNA sequences. Mol. Biol. Evol. 9:433–446.
GENTRY, A. W. 1970. The Bovidae (Mammalia) of the Fort Ternan fos-
sil fauna. Pages 243–323 in Fossil vertebrates of Africa, Volume 2.
(L. S. B. Leakey and R. J. G. Savage, eds.). Academic Press, London.
GENTRY, A. W. 1994. The Miocene differentiation of Old World Pecora
(Mammalia). Hist. Biol. 7:115–158.
GENTRY, A. W. 2000. The ruminant radiation. Pages 11–25 in Antelopes,
deer, and relatives. Fossil record, behavioral ecology, systematics,
and conservation (E. S. Vrba and G. B. Schaller, eds.). Yale Univ.
Press, New Haven, Connecticut.
GENTRY,A.W.,AND J. J. HOOKER. 1988. The phylogeny of Artiodactyla.
Pages 235–271 in The phylogeny and classiﬁcation of the tetrapods,
Volume 2. Mammals (M. J. Benton, ed.). Clarendon Press, Oxford,
GENTRY, A. W., G. E. R¨
OSSNER,AND E. P. J. HEIZM ANN. 1999. Suborder
Ruminantia. Pages 225–253 in The Miocene land mammals of
Europe. (G. E. R¨ossner and K. Heissig, eds.). Verlag Dr. Friedrich
GERAADS , D. 1986. Remarques sur la syst ´ematique et la phylog ´enie des
Girafﬁdae (Artiodactyla, Mammalia). Geobios 19:465–477.
GERAADS , D., G. BOUV RAIN,AND J. SUDRE. 1987. Relations phyl ´etiques
de Bachiterium Filhl, ruminant de l’Oligoc`ene d’Europe occidentale.
GINS BURG , L. 1988. La faune des mammif`eres des sables mioc`enes du
synclinal d’Esvres (Val-de-Loire). C.R. Acad. Sci. Paris S ´er. II 307:319–
GRAUR, D., AND D. G. HIGGINS. 1994. Molecular evidence for the in-
clusion of cetaceans within the order Artiodactyla. Mol. Biol. Evol.
GRAY, J. E. 1821. On the natural arrangement of vertebrate mammals.
Lond. Med. Repos. 15:296–310.
GROVES,C.P.,AND P. G RUBB. 1987. Relationships of living deer. Pages
21–59 in Biology and management of the Cervidae. (C. M. Wemmer,
ed.). Smithsonian Institution Press, Washington, D.C.
GRUBB, P. 1993. Order Artiodactyla. Pages 377–414 in Mammal species
of the world. A taxonomic and geographic reference, 2nd edition
(D. E. Wilson and D. A. M. Reeder, eds.). Smithsonian Institution
Press, Washington, D.C.
GRZIMEK, B. 1968. Grzimek’s animal life encyclopedia. Van Nostrand
Reinhold Co., New York.
GUST AFSS ON,M.H.G.,AND K. BREMER. 1995. Morphology and
phylogenetic interrelationships of the Asteraceae, Calyceraceae,
Campanulaceae, Goodeniaceae, and related families (Asterales).
Am. J. Bot. 82:250–265.
HAMILT ON, W. R. 1978. Fossil giraffes from the Miocene of Africa and
a revision of the phylogeny of the Giraffoidea. Philos. Trans. R. Soc.
Lond. B 283:165–229.
HASSANIN, A. 1999. Phylog´enie des Bovidae (Mammalia, Artio-
dactyla). Apports de l’ADN ancien, ´evolution mol´eculaire et
strat´egies de pond´eration. Th`ese, Mus´eum National d’Histoire
HASSANIN, A., AND E. DOUZER Y. 1999a. Evolutionary afﬁnities of
the enigmatic saola (Pseudoryx nghetinhensis) in the context of the
molecular phylogeny of Bovidae. Proc. R. Soc. Lond. B 266:893–900.
HASSANIN, A., AND E. DOUZER Y. 1999b. The tribal radiation of the fam-
ily Bovidae (Artiodactyla) and the evolution of the mitochondrial
cytochrome bgene. Mol. Phylogenet. Evol. 13:227–243.
HASSANIN, A., G. LECOINTRE,AND S. TILLI ER. 1998a. The ‘evolution-
ary signal’ of homoplasy in protein-coding gene sequences and its
phylogenetic consequences for weighting in phylogeny. C.R. Acad.
Sci. S´er. III 321:611–620.
HASSANIN, A., E. PAS QUET,AND J.-D. VIGNE. 1998b. Molecular sys-
tematics of the subfamily Caprinae (Artiodactyla, Bovidae) as deter-
mined from cytochrome bsequences. J. Mammal. Evol. 5:217–236.
HIENDLED ER, S., H. LEWALS KI,R.WASS MUTH,AND A. JANK E. 1998.
The complete mitochondrial DNA sequence of the domestic sheep
(Ovis aries) and comparison with the other major ovine haplotype.
J. Mol. Evol. 47:441–448.
HOFMANN, R. R. 1973. The ruminant stomach: Stomach structure
and feeding habits of East African game ruminants. East African
monographs in biology, Volume 2 (T. R. Odhiambo, ed.). East
African Literature Bureau, Nairobi.
HONEYCUTT , R. L., M. A. NEDB AL,R.M.ADKINS,AND L. L. JANECEK.
1995. Mammalian mitochondrial DNA evolution: A comparison of
the cytochrome band cytochrome coxidase II genes. J. Mol. Evol.
HUELSENB ECK,J.P.,AND F. RONQUIST. 2001. MrBayes: Bayesian
inference of phylogenetic trees. Bioinformatics 17:754–755.
HUELSENB ECK,J.P.,F.RONQUIST,R.NIELSEN,AND J. P. BO LLBACK. 2001.
Bayesian inference of phylogeny and its impact on evolutionary
biology. Science 294:2310–2314.
IRWIN,D.M.,AND U. ARNASON. 1994. Cytochrome bgene of marine
mammals: Phylogeny and evolution. J. Mammal. Evol. 2:37–55.
IRWIN, D. M., T. D. KOCHER,AND A. C. WILSON. 1991. Evolution of the
cytochrome bgene of mammals. J. Mol. Evol. 32:128–144.
JANIS , C. M. 2000a. The endemic ruminants of the Neogene of North
America. Pages 26–37 in Antelopes, deer, and relatives.Fossil record,
behavioral ecology, systematics, and conservation (E. S. Vrba and
G. B. Schaller, eds.). Yale Univ. Press, New Haven, Connecticut.
JANIS , C. M. 2000b. Patterns in the evolution of herbivory in large
terrestrial mammals: The Paleogene of North America. Pages 168–
222 in Evolution of herbivory in terrestrial vertebrates. Perspectives
from the fossil record (H. D. Sues, ed.). Cambridge Univ. Press,
JANIS ,C.M.,AND E. MANNING. 1998. Antilocapridae. Pages 491–507
in Evolution of Tertiary mammals of North America, Volume 1.
2003 HASS ANIN AND DO UZERY—PHYLO GENY O F RUMINANTIA 227
Terrestrial carnivores, ungulates, and ungulatelike mammals (C. M.
Janis, K. M. Scott, and L. L. Jacobs, eds.). Cambridge Univ. Press,
JANIS ,C.M.,AND K. M. SCOTT. 1987. The interrelationships of higher
ruminant families with special emphasis on the members of the
Cervoidea. Am. Mus. Novit. 2893:1–85.
JANIS ,C.M.,AND K. M. SCOTT. 1988. The phylogeny of the Ruminantia
(Artiodactyla, Mammalia). Pages 273–282 in The phylogeny and
classiﬁcation of the tetrapods, Volume 2. Mammals (M. J. Benton,
ed.). Clarendon Press, Oxford, U.K.
KISHINO, H., AND M. HASEGAWA. 1989. Evaluation of the maximum
likelihood estimate of the evolutionary tree topologies from DNA
sequence data, and the branching order in Hominoidea. J. Mol.
KISHINO, H., J. L. THORNE,AND W. J. BRUNO. 2001. Performance of a
divergence time estimation method under a probabilistic model of
rate evolution. Mol. Biol. Evol. 18:352–361.
KRAUS ,F.,AND M. MIYAMO TO. 1991. Rapid cladogenesis among the
pecoran ruminants: Evidence from mitochondrial DNA sequences.
Syst. Zool. 40:117–130.
LAVOCAT , R. 1955. Ruminants fossiles. Pages 668–693 in Trait´ede
zoologie. Anatomie, syst´ematique, biologie, Tome XVII, Premier fas-
cicule. Mammif`eres. Les ordres: Anatomie, ´ethologie, syst´ematique.
P.P. Grass´e, Masson, and Cie, Paris.
LEINDER S, J. J. M. 1979. On the osteology and function of the digits
of some ruminants and their bearing on taxonomy. Z. S ¨augetierk.
LEINDER S, J. J. M. 1983. Hoplitomerycidae fam. nov. (Ruminantia,
Mammalia) from Neogene ﬁssure ﬁllings in Gargano (Italy). Scr.
LEINDER S,J.J.M.,AND E. HEINTZ. 1980. The conﬁguration of the
lacrimal oriﬁces in pecorans and tragulids (Artiodactyla, Mam-
malia) and its signiﬁcance for the distinction between Bovidae and
Cervidae. Beaufortia 30:155–162.
MACFADDEN, B. J. 1992. Fossil horses. Systematics, paleobiology, and
evolution of the family Equidae. Cambridge Univ. Press, New York.
MADS EN, O., M. SCALLY,C.J.DOUADY,D.J.KAO,R.W.DEBRY,
R. ADKINS,H.AMRINE,M.J.STANHOPE,W.W.DE JONG,AND M. S.
SPRING ER. 2001. Parallel adaptative radiations in two major clades
of placental mammals. Nature 409:610–614.
MATTHEE, C. A., J. D. BURZLAFF,J.F.TAYLO R,AND S. K. DAVIS . 2001.
Mining the mammalian genome for artiodactyl systematics. Syst.
MATTHEE, C. A., AND S. K. DAVIS . 2001. Molecular insights into the
evolution of the family Bovidae: A nuclear DNA perspective. Mol.
Biol. Evol. 18:1220–1230.
MATTHEE,C.A.,AND T. J. ROBINSON. 1999. Cytochrome bphylogeny
of the family Bovidae: Resolution within Alcelaphini, Antilopini,
Neotragini, and Tragelaphini. Mol. Phylogenet. Evol. 12:31–46.
MATTHEW, W. D. 1934. A phylogenetic chart of the Artiodactyla.
J. Mammal. 15:207–209.
MCKENNA, M. C. 1983. Holarctic landmass rearrangement, cosmic
events, and cenozoic terrestrial organisms. Ann. Mo. Bot. Gard.
MENG, J., AND M. C. MCKENNA. 1998. Faunal turnovers of Palaeogene
mammals from the Mongolian plateau. Nature 394:364–367.
MIYAMO TO,M.M.,AND M. GOODMAN. 1986. Biomolecular systemat-
ics of eutherian mammals: Phylogenetic patterns and classiﬁcation.
Syst. Zool. 35:230–240.
MIYAMO TO, M. M., F. KRAUS ,P.J.LAI PIS ,S.M.TANHAUSER,AND S. D.
WEBB. 1993. Mitochondrial DNA phylogenies within Artiodactyla.
Pages 268–281 in Mammal phylogeny: Placentals (F. S. Szalay, M. J.
Novacek, and M. C. McKenna, eds.). Springer-Verlag, New York.
MIYAMO TO, M. M., F. KRAUS ,AND O. A. RYDER. 1990. Phylogeny and
evolution of antlered deer determined from mitochondrial DNA
sequences. Proc. Natl. Acad. Sci. USA 87:6127–6131.
MIYAMO TO, M. M., S. M. TANHAUS ER,AND P. J. LAIPIS. 1989. System-
atic relationships in the artiodactyl tribe Bovini (family Bovidae),
as determined from mitochondrial DNA sequences. Syst. Zool.
MONTG ELARD , C., F. CATZEFLIS,AND E. DO UZERY. 1997. Phyloge-
netic relationships of artiodactyls and cetaceans as deduced from
the comparaison of cytochrome band 12S rRNA mitochondrial
sequences. Mol. Biol. Evol. 14:550–559.
MORALES , J., S. MOY ´
A,AND D. SORIA. 1981. Presencia de la
familia Moschidae (Artiodactyla, Mammalia) en el Vallesiense de
Espana: Hispanomeryx duriensis novo gen. nova sp. Estud. Geol.
MORALES , J., M. PICKFORD,AND D. SORIA. 1993. Pachyostosis in a
lower Miocene giraffoid from Spain, Lorancameryx pachyostoticus nov.
gen. nov. sp. and its bearing on the evolution of bony appendages
in artiodactyls. Geobios 26:207–230.
A, S. 1986. El g´enero Hispanomeryx Morales et al. (1981):
Posicion ﬁlogen´etica y sistematica. Su contribucion al conocimiento
de la evolucion de los Pecora (Artiodactyla, Mammalia). Paleontol.
I Evol. 20:267–287.
MURPHY, W. J., E. EIZIRIK,W.E.J
AND S. O’BRIEN. 2001a. Molecular phylogenetics and the origins of
placental mammals. Nature 409:614–618.
MURPHY, W. J., E. EIZIRIK,S.J.O’BRIEN,O.MADS EN,M.SCALLY,C.
OUADY,E.TEELI NG,O.A.RYD ER,M.J.STANHOPE,W.W.DE JONG,
AND M. S. SPRING ER. 2001b. Resolution of the early placental mam-
mal radiation using Bayesian phylogenetics. Science 294:2348–2351.
NIKAIDO, M., A. P. ROONEY,AND N. OKADA. 1999. Phylogenetic
relationships among cetartiodactyls based on insertions of short and
long interpersed elements: Hippopotamuses are the closest extant
relatives of whales. Proc. Natl. Acad. Sci. USA 96:10261–10266.
NOWAK, R. M. 1991. Walker’s mammals of the word, 5th Edition,
Volume II. Johns Hopkins Univ. Press, Baltimore, Maryland.
O’GARA,B.W.,AND G. MAT SON. 1975. Growth and casting of horns by
pronhorns and exfoliation of horns by bovids. J. Mammal. 56:829–
OPDYKE, N. D. 1990. Magnetic stratigraphy of Cenozoic terrestrial
sediments and mammalian dispersal. J. Geol. 98:621–637.
PHILIPPE, H. 1993. MUST: A computer package of management utilities
for sequences and trees. Nucleic Acids Res. 21:5264–5272.
PITRA, C., R. F ¨
URBAS S ,AND H.-M. SEYFERT. 1997. Molecular phylogeny
of the tribe Bovini (Mammalia: Artiodactyla): Alternative placement
of the Anoa. J. Evol. Biol. 10:589–600.
PROTHERO,D.R.,AND T. H. HEATON. 1996. Faunal stability during
the early Oligocene climatic crash. Palaeogeogr. Palaeoclimatol.
RANDI, E., N. MUCCI ,M.PIERPAOLI,AND E. DOUZERY. 1998. New
phylogenetic perspectives on the Cervidae (Artiodactyla) are
provided by the mitochondrial cytochrome bgene. Proc. R. Soc.
Lond. B 265:793–801.
RAVEN,P.H.,AND D. I. AXELRO D. 1974. Angiosperm biogeography
and past continental movements. Ann. Mo. Bot. Gard. 61:539–
ROMER, A. S. 1966. Vertebrate paleontology, 3rd edition. Univ. Chicago
SAMBROOK, J., E. F. FRITSCH,AND T. MANI ATIS. 1989. Molecular
cloning. Cold Spring Harbor Laboratory Press, Cold Spring Harbor,
SCHNARE, M. N., S. H. DAMBER GER ,M.W.GRAY,AND R. R. GUT ELL.
1996. Comprehensive comparison of structural characteristics in
eukaryotic cytoplasmic large subunit (23S-like) ribosomal RNA. J.
Mol. Biol. 256:701–719.
SCOTT,K.M.,AND C. M. JANIS . 1987. Phylogenetic relationships of the
Cervidae, and the case for a superfamily “Cervoidea.” Pages 3–20
in Biology and management of the Cervidae (C. M. Wemmer, ed.).
Smithsonian Institution Press, Washington, D.C.
SCOTT,K.M.,AND C. M. JANI S. 1993. Relationships of the Ruminantia
(Artiodactyla) and an analysis of the characters used in ruminant
taxonomy. Pages 282–302 in Mammal phylogeny: Placentals (F. S.
Szalay, M. J. Novacek, and M. C. McKenna, eds.). Springer-Verlag,
SEYFERT, H. M. A., A. TUCKOR ICZ,H.INTERTHAL,D.KOCZAN,AND
G. HOBOM. 1994. Structure of the bovine lactoferrin-encoding gene
and its promotor. Gene 143:265–269.
SHIMAMURA, M., H. ABE,M.NIKAIDO,K.OHSHI MA,AND N. OKADA.
1999. Genealogy of families of SINEs in cetaceans and artiodactyls:
The presence of a huge superfamily of tRNA(Glu)-derived families
of SINEs. Mol. Biol. Evol. 16:1046–1060.
228 SYSTEMATIC BIOLOGY VOL. 52
SHIMAMURA, M., H. YASUE,K.OHS HIM A,H.ABE,H.KATO ,T.KISHIRO,
M. GOTO,I.MUNECHIKA,AND N. OKADA. 1997. Molecular evidence
from retroposons that whales form a clade within even-toed
ungulates. Nature 388:666–670.
SHIMO DAIR A, H. 2002. An approximately unbiased test of phylogenetic
tree selection. Syst. Biol. 51:492–508.
SHIMO DAIR A, H., AND M. HASEG AWA. 1999. Multiple comparisons of
log-likelihoods with applications to phylogenetic inference. Mol.
Biol. Evol. 16:1114–1116.
SIMPSON, G. G. 1945. The principles of classiﬁcation and a classiﬁcation
of mammals. Bull. Am. Mus. Nat. Hist. 85:1–350.
SPRING ER,M.S.,AND E. DO UZERY. 1996. Secondary structure and
patterns of evolution among mammalian mitochondrial 12S rRNA
molecules. J. Mol. Evol. 43: 357–373.
STANLEY, H. F., M. KADWELL,AND J. C. WHEELER. 1994. Molecular
evolution of the family Camelidae: A mitochondrial DNA study.
Proc. R. Soc. Lond. B 256:1–6.
STEHLIN, H.-G. 1909. Remarques sur les faunules de mammif`eres des
couches ´eoc`enes et oligoc`enes du Bassin de Paris. Bull. Soc. G´eol.
SU, B., Y.-X. WANG ,H.LAN,W.WANG,AND Y. Z HANG . 1999. Phylo-
genetic study of complete cytochrome bgenes in musk deer (genus
Moschus) using museum samples. Mol. Phylogenet. Evol. 12:241–249.
SWOFFORD, D. L. 1993. PAUP: Phylogenetic analysis using parsimony,
version 3.1.1. Computer program distributed by the Illinois Natural
History Survey, Champaign.
SWOFFORD, D. L. 1998. PAUP∗: Phylogenetic analysis using par-
simony (∗and other methods), version 4. Sinauer, Sunderland,
SWOFFORD, D. L., P. J. WAD DELL,J.P.HUELSENBECK,P.G.FOSTER,
P. O. LEWIS ,AND J. S. ROG ERS . 2001. Bias in phylogenetic estimation
and its relevance to the choice between parsimony and likelihood
methods. Syst. Biol. 50:525–539.
TANAKA, K., C. D. SOLIS,J.S.MAS ANGK AY, K.-I. MAEDA,Y.
AWAMOT O,AND T. NAMIKAWA. 1996. Phylogenetic relationship
among all living species of the genus Bubalus based on DNA
sequences of the cytochrome bgene. Biochem. Genet. 34:443–452.
THORNE, J. L., H. KIS HINO,AND I. S. PAINTER. 1998. Estimating the
rate of evolution of the rate of molecular evolution. Mol. Biol. Evol.
VAN DER MAD E, J. 1999. Intercontinental relationship Europe–Africa
and the Indian continent. Pages 457–472 in The Miocene land
mammals of Europe (G. E. R¨ossner and K. Heissig, eds.). Verlag Dr.
Friedrich Pfeil, M ¨unchen.
VANSELO W, J., AND R. F ¨
URBAS S . 1995. Novel aromatase transcripts from
bovine placenta contain repeated sequence motifs. Gene 154:281–286.
VIRET, J. 1961. Artiodactyla. Pages 887–1021 in Trait´e de pal´eontologie,
Tome VI en deux volumes. L’origine des mammif`eres et les aspects
fondamentaux de leur ´evolution, Volume 1. Mammif`eres, origine
reptilienne, ´evolution (J. Piveteau, ed.). Masson et Cie, Paris.
VISLOBOKOVA, I. 1990. The basic features of historical development
and classiﬁcation of the Ruminantia. Paleontol. J. 4:3–14.
VISLOBOKOVA, I. 1997. Eocene—Early Miocene ruminants in Asia.
M´em. Trav. E.P.H.E. Inst. Montpellier 21:215–223.
VISLOBOKOVA, I. 1998. A new representative of the Hypertraguloidea
(Tragulina, Ruminantia) from Khoer-Dzan locality in Mongolia,
with remarks on the relationships of the Hypertragulidae. Am. Mus.
VRBA, E. S. 1985. African Bovidae: Evolutionary events since the
Miocene. S. Afr. J. Sci. 81:263–266.
WARD, T. J., R. L. HONEYCUT T,AND J. N. DERR . 1997. Nucleotide
sequence evolution at the κ-casein locus: Evidence for positive
selection within the family Bovidae. Genetics 147:1863–1872.
WEBB, S. D. 1998. Hornless ruminants. Pages 463–476 in Evolution
of Tertiary mammals of North America. Volume 1. Terrestrial
carnivores, ungulates, and ungulatelike mammals (C. M. Janis, K.
M. Scott, and L. L. Jacobs, eds.). Cambridge Univ. Press, New York.
WEBB,S.D.,AND B. E. TAYLOR. 1980. The phylogeny of hornless
ruminants and a description of the cranium of Archeomeryx. Bull.
Am. Mus. Nat. Hist. 167:120–157.
WHELAN, S., P. LIO,AND N. GOLDMAN. 2001. Molecular phylogenetics:
State-of-the-art methods for looking into the past. Trends Genet.
WINNEPENNINCK X, B., T. BACKELJAU,AND R. DEWACHTER . 1993.
Extraction of high molecular weight DNA from molluscs. Trends
WOOD, W. F. 2001. Antibacterial compounds in the interdigital glands
of pronghorn, Antilocapra americana. Biochem. Syst. Ecol. 29:417–419.
YANG, Z. 1994. Estimating the pattern of nucleotide substitution. J.
Mol. Evol. 39:105–111.
YANG, Z. 1996a. Among-site rate variation and its impact on phyloge-
netic analyses. Trends Ecol. Evol. 11:367–372.
YANG, Z. 1996b. Maximum-likelihood models for combined analyses
of multiple sequence data. J. Mol. Evol. 42:587–596.
YANG, Z. 1997. PAML: A program package for phylogenetic analysis
by maximum likelihood. CABIOS 13:555–556.
First submitted 22 July 2002; reviews returned 6 October 2002;
ﬁnal acceptance 23 November 2002
Associate Editor: Jeff Thorne