Phylogeny and evolution of cats (Felidae)

Full text

CHAPTER 2
Phylogeny and evolution of cats (Felidae)
Lars Werdelin, Nobuyuki Yamaguchi, Warren E. Johnson, and
Stephen J. O’Brien
Introduction
Cats, wild as well as domestic, fossil as well as living,
are familiar to people around the world. The family
Felidae has a worldwide distribution and has been
associated with humans in various ways throughout
history (Quammen 2004). Their functional mor-
phology, ecology, and behaviour have been the sub-
ject of intense scrutiny by scientists for over 200
years. The fossil record of cats is extensive and some
of its members are among the most recognizable of
extinct animals. Despite all this, the phylogeny and
evolution of the family Felidae, and even the content
of the family, have remained poorly understood. In
this review, we will first present the current state of
knowledge with regard to the interrelationships
of living Felidae and the timing of the radiation of
modern cats. We will also present the fossil record of
Felidae in broad outline, focusing first on describing
the different groups of species and their characteris-
tics, and then discussing the general patterns of cat
evolution that we can deduce from current data.
Provided with this overview, we will attempt to iden-
tify those areas most in need of further research in
order to achieve the aim of a fuller understanding of
felid evolution, especially that of the living felids and
their ecological and functional relationship to the
extinct sabre-toothed felids.
In this discussion, we will synthesize the available
data, distinguishing as far as possible monophyletic
groups of taxa, suggesting the most likely interrela-
tionships of the fossil lineages, but also pointing out
that there are many problem areas that need to be
resolved. This section should be viewed as a chal-
lenge to investigators to use old data or discover
new data to corroborate or refute the scenarios pro-
posed herein. We end the paper with a small section
demonstrating some evolutionary patterns among
extant Felidae, suggesting that there is much to be
gained from the deeper analysis of the current phy-
logenetic information.
Felid morphology is described and discussed else-
where (Kitchener et al., Chapter 3, this volume) and
will not be reiterated here except as needed. Teeth of
the upper jaw are referred to in upper-case letters
(I, C, P, and M) and teeth of the lower jaw in lower-
case letters (i, c, p, and m), followed bythe appropriate
number in the sequence. Character mapping on
cladograms was carried out with Mesquite, version
1.12 (Maddison and Maddison 2004). Stratigraphic
Artist’s reconstruction of the sabre-toothed cat Megantereon cultridens stalking its prey. (Illustration courtesy
of Mauricio Anto
´n.)
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

ages of taxa as given in the text and figures were
obtained from either primary literature or (for North
America) the Paleobiology Database (www.paleodb.
org) and (for Eurasia) the NOW database (www.hel-
sinki.fi/science/now/database.html).
Phylogeny
Many attempts have been made to investigate the
interrelationships of Felidae. These have followed
two broad approaches. Some, like Matthew (1910),
Kretzoi (1929a, b) and Beaumont (1978) have
incorporated both fossil and extant felids in their
analyses, while others, such as Pocock (1917a), Her-
rington (1986), and Salles (1992) have focused exclu-
sively on the living members of the family. A new era
in felid phylogenetics was ushered in with the intro-
duction of molecular evidence (Collier and O’Brien
1985; O’Brien et al. 1985a; Johnson et al. 1996),
while the first study to use a total evidence approach
was that of Mattern and McLennan (2000).
All of these approaches have had their problems.
In the case of fossil studies, confounding factors have
included the relatively poor fossil record, the prob-
lem of finding useful characters in fragmentary ma-
terial and the convergence between Nimravidae and
Felidae. Though previously included in the Felidae
(Matthew 1910; Piveteau 1961), the former, Nimra-
vidae, is now known to be diphyletic. Its Paleogene
(65.5–23.0 million years ago [Ma]; Gradstein et al.
2004) members form a basal clade within either Feli-
formia or Carnivora as a whole (Neff 1983; Hunt
1987; Morlo et al. 2004), while its Neogene (23.0
Ma—recent) members are placed in a separate family,
Barbourofelidae, with affinities to Felidae (see
below). Morphological studies of extant felids have
been hampered by the very uniform morphology of
the members of the family, making it difficult to find
and polarize characters for phylogenetic analysis.
Molecular studies, on the other hand, have been
particularly hampered by the apparently short time-
span during which the clades of modern felids
evolved. Thus, clades of closely related taxa have
been identified but the interrelationships of these
clades have been difficult to pinpoint.
Recently, two of us (Warren E. Johnson and Ste-
phen J. O’Brien) published a phylogeny of Felidae
based on a data set of 22,789 base pairs of DNA,
including autosomal, Y-linked, X-linked, and mito-
chondrial gene segments (Johnson et al. 2006b). The
results of this study, while not immutable, provide a
firm basis for understanding the interrelationships
and evolution of the extant Felidae. The results con-
firm some prior results, both molecular and morpho-
logical, while providing new insights and surprises.
The study distinguishes eight clades of extant fe-
lids (Fig. 2.1). The first of these to split off from the
stem lineage is the Panthera lineage (genera Neofelis
and Panthera)atc. 10.8 Ma (Fig. 2.1, node A). Most
previous studies of felid phylogeny have placed
Panthera as the crown group, but a few (Turner and
Anto
´n 1997; Mattern and McLennan 2000) also have
the Panthera lineage as basal to other cats. Within this
P. linsang
N. nebulosa
P. leo
P. pardus
P. tigris
P. uncia
P. onca
P. marmorata
P. badia
P. temmincki
L. serval
C. caracal
C. aurata
L. pardalis
L. wiedii
L. colocolo
L. jacobita
L. tigrinus
L. geoffroyi
L. guigna
L. rufus
L. canadensis
L. pardinus
L. lynx
A. jubatus
P. concolor
P. yagouaround
i
F. chaus
F. nigripes
F. silvestris
F. margarita
O. manul
P. rubiginosus
P. planiceps
P. bengalensis
P. viverrinus
N. diardi
51035
Million years before present
A
B
C
D
E
F
G
Figure 2.1 The phylogeny of the extant Felidae. Thick
lines indicate the presence of a fossil record, thin lines
indicate the absence of a fossil record. Node labels as in
the main text. (Based on the work of Johnson et al. 2006b.)
60 Biology and Conservation of Wild Felids
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

lineage, the clouded leopard, Neofelis, with the two
species N. nebulosa and N. diardi (Buckley-Beason
et al. 2006; Kitchener et al. 2006) is placed basally,
as would be expected from its distinctive morphol-
ogy implying a long separate evolutionary lineage
(Christiansen 2006), with the rest of the pantherines
radiating within the last 4 million years.
The next clade to branch off, at c. 9.4 Ma (Fig. 2.1,
node B), is the bay cat lineage (genus Pardofelis). This
clade consists of the poorly known bay cat (P. badia),
Asian golden cat (P. temminckii), and marbled cat
(P. marmorata). The last mentioned species has been
linked to the Panthera lineage (e.g. Herrington 1986)
and this is reflected in its position here, as basal
member of the clade branching off closest to the
Panthera lineage.
The third lineage is the Caracal lineage, with two
genera, Caracal and Leptailurus, incorporating three
African species: caracal (Caracal), African golden cat
(C. aurata), and serval (S. leptail urus serval). This
lineage branches off at c. 8.5 Ma (Fig. 2.1, node C),
with the serval basal to the other two species.
The next lineage is the ocelot lineage (genus Leo-
pardus), including most of the South American small
cats (Seymour 1999). This lineage branches off at
c. 8.0 Ma (Fig. 2.1, node D). The beginning of this
lineage is thus independent of the formation of the
land bridge between South and North America about
3 Ma (Marshall et al. 1982). However, the radiation of
the extant species within this lineage shows dates
that are compatible with a single origin of the extant
radiation from a North American ancestor, as previ-
ously proposed (Werdelin 1989).
The fifth lineage comprises the genus Lynx,
splitting off at c. 7.2 Ma (Fig. 2.1, node E). This
lineage has also often been linked to Panthera (e.g.
Collier and O’Brien 1985; Salles 1992), but the recent
more robust study by Johnson et al. (2006b) indicates
that the relationship is more distant than previously
thought. Within the clade, L. rufus is basal as has
generally been thought, but L. canadensis and L.
lynx are not reconstructed as sister taxa, unlike in
previous analyses (Werdelin 1981).
The next lineage is the Puma lineage, including
the genera Puma and Acinonyx which split off at
c. 6.7 Ma (Fig. 2.1, node F). This lineage has previously
been recognized in both morphological (Herrington
1986; Van Valkenburgh et al. 1990) and molecular
(Johnson and O’Brien 1997) studies. It is worth not-
ing that the puma and jaguarundi probably split be-
fore the Great American Biotic Interchange that
followed the formation of the land bridge between
South and North America (Marshall et al. 1982), and
thus both are of North American origin.
The seventh and eighth lineages are the small cats
of the Old World—the leopard cat and domestic cat
lineages. They split from each other at c. 6.2 Ma (Fig.
2.1, node G). The former includes the genera Otoco-
lobus and Prionailurus and the latter the genus Felis.
The splits within the former are much deeper than
within the latter, suggesting that the genus Felis may
be oversplit. This is also the conclusion of Driscoll
et al. (2007), who distinguish only four species in Felis:
F. chaus, F. nigripes, F. margarita,andF. silvestris.The
last mentioned species now also includes F. or n a t a ,
F. bie t i ,andF. lybica, making it one of the most wide-
spread small cat species.
Most of the nodes in this phylogeny are robustly
supported (Johnson et al. 2006b). A few, however, are
still unstable, showing either low support or incon-
gruence between different analyses and data sets.
These as yet incompletely resolved nodes are: the rela-
tive positions of Panthera leo, P. pardus,andP. onca,
as well as the relative positions of P. tigris and P. uncia
within this clade; the position of L. jacobita;the
position of O. manul; the position of F. nigripes;and
the clade uniting Felis and Prionailurus/Otocolobus to
the exclusion of Puma/Acinonyx.
The most notable fact about this phylogeny of
extant cats lies in the short time intervals between
the splits of the eight lineages. The radiation of
lineages along the entire stem of the felid clade occurs
within the Late Miocene (over a period of c. 6.3 Ma)
and such a short space of time suggests the occur-
rence of some sort of functional or ecological release,
but what that may be isat present unknown. We shall
return to the fossil record of extant cats below.
The fossil record
According to available molecular data, the Felidae
originated some time at or just after the end of the
Eocene (Gaubert and Ve
´ron 2003). This accords well
with the fossil record. The earliest forms placed in
the felid lineage, Proailurus and possibly Stenogale
Phylogeny and evolution of cats (Felidae) 61
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

and Haplogale (Hunt 1998; Peigne
´1999), occur after
the ‘Grande Coupure’ marking the Eocene/Oligo-
cene boundary (c. 33.9 Ma; Gradstein et al. 2004).
In the Mammals Paleogene (MP) level system of
Paleogene terrestrial mammal stratigraphy in Europe,
this boundary is placed between MP 20 and MP 21
(Schmidt-Kittler 1990). In the fissure fillings of the
Quercy region, France, where most of our knowledge
of early European carnivorans originates, feliforms
are not known before MP 21 (Hunt 1998). Owing to
the scarcity of their remains, modern excavations
have yet to establish the first occurrence of the Feli-
dae. What we know, however, suggests that some
older known finds may be from the Early Oligocene,
that is, before 28.4 Ma (Gradstein et al. 2004). Thus,
the earliest felids appeared sometime between c. 35
Ma (age of the sister group) and 28.5 Ma (minimum
age of the earliest fossils).
It is well established on morphological grounds,
basicranial as well as dental, that Proailurus, known
from the Quercy fissure fills, but also from excellent
material from the Early Miocene site of Saint-Ge
´r-
and-le-Puy, France, (MN 2 in the Neogene mammal
zonation of Europe; 22.8–20 Ma) is a felid. Despite
this, the morphological path leading to the felid
condition is not well delineated. Hunt (1998) dis-
cusses changes to the auditory bulla seen in a variety
of early feliforms, including Haplogale and Stenogale,
and leading to the bulla of Proailurus. However, the
placement of Asiatic linsangs (genus Prionodon)as
the sister group to Felidae on molecular grounds by
Gaubert and Ve
´ron (2003), instead of with the Viver-
ridae, in which they have traditionally been placed,
adds complexity to the story. Hunt (2001) placed
Prionodon in a clade with ‘true’ viverrids, for example
Genetta, on the basis of basicranial anatomy (but
without consideration of other features). What this
conflict between separate data sets consisting of
non-overlapping characters means for our under-
standing of the fossil record of the precursors of
Felidae and for the origins of the family has yet to
be established.
Early felids
As noted, the earliest well-established felid is Proai-
lurus (Figs. 2.2, letter A; 2.3, and 2.4). Peigne
´(1999)
provides a discussion of the evolution of this species
and its relationship to other early putative felids.
M
i
l
l
i
o
n
Y
e
a
r
s
B
e
f
o
r
e
P
r
e
s
e
n
t
5
10
15
20
25
30
35
Haplogale
Stenogale
Proailurus
Pseudaelurus
Hyperailurictis
Nimravides Machairodus
Amphimachairodus
Homotherium Dinobastis
Xenosmilus
Styriofelis
Felinae
(Fig.1)
Para-
machaerodus
Dinofelis
Metailurus
Ginsburgsmilus
Syrtosmilus
Vampyrictis
Afrosmilus
Barbouro
-
felis
Sansanosmilus
Prosansanosmilus
DE
F
G
HI
J
K
L
N
Smilodon Megantereon
M
???
???
???
???
???
???
???
A
B
C
Figure 2.2 Summary of the proposed evolutionary tree of Felidae discussed herein. Thick lines indicate the presence of a
fossil record, thin lines indicate the absence of a fossil record. Labels as in the main text and Table 2.1.
62 Biology and Conservation of Wild Felids
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

Proailurus (with three species, P. lemanensis, P. bour-
bonnensis, and P. major) is a medium-sized cat about
the size of a bobcat, L. rufus. Dentally, it differs from
living cats in the (variable) presence of p1, p2, m2,
and P1, as well as the presence of a small metaconid
and talonid on m1. Overall, the dentition is thus
very similar to that of living felids, but includes
some elements that have been fully reduced in the
modern clade. Further, the auditory bulla of Proai-
lurus has a ventral process of the petrosal promontor-
ium (Hunt 1989, 1998). This process is lost in living
felids. When it was lost in felid evolution has yet to
be established, but it serves to distinguish at least the
modern clade from the basally situated Proailurus.
The geologically youngest Proailurus is from
Laugnac, France, biostratigraphically placed in MN
2b (>20 Ma). In Proailurus we have (as far as it is
known) an essentially modern felid except for a few
minor details of the dentition, auditory bulla, and
postcranium, which has shorter limbs than modern
felids. Coupled with the molecular date for the
divergence of Prionodon and Felidae, this suggests
that there must have been a stem lineage of perhaps
5 Ma in the Early Oligocene leading up to the full
felid morphology. Haplogale and Stenogale are likely
to be members of that lineage (Hunt 1998; Peigne
´
1999), but the details of the process have not been
worked out.
Proailurus is not known with certainty outside Eur-
ope. Hunt (1998) reports the presence of Proailurus
sp. from the Hsanda Gol Formation, Mongolia. How-
ever, Peigne
´(1999) concludes, in our opinion cor-
rectly, that this specimen is better assigned to the
Barbourofelidae. On the other hand, Hunt (1998)
also describes the skull of a Proailurus-grade felid
from the Ginn Quarry, Nebraska (Late Hemingfor-
dian, c. 17–16.5 Ma). According to Hunt the basicra-
nial structure of the Ginn Quarry felid is more
plesiomorphic than that of European Proailurus.
This suggests that phylogenetic diversification in
Felidae had begun already in the Early Miocene and
that North American ‘Pseudaelurus’ (see below) may
have evolved from a Proailurus-grade ancestor rather
than from a migration of early Pseudaelurus into
North America. If so, felids may have migrated
into North America as early as the beginning of the
Hemingfordian (c. 19 Ma), along with a number of
other carnivoran taxa (Qiu 2003).
The next felids to evolve belong to the Pseudaelurus
complex (Fig. 2.2, letter B; Fig. 2.5). This is a group of
Table 2.1 The internal nodes of Fig. 2.2: content, place of origin, and age.
Letter Content Continent
Age, Ma
(approximate)
A Felidae sensu stricto Europe 27
B Pseudaelurine radiation and later
felids
Eurasia 22
C Felinae (radiation of extant felids) Eurasia 14–13
DNimravides North America 14
E Machairodontinae Europe 14–13
FAmphimachairodus lineage Eurasia 10
G Homotheriini Eurasia 6
H Derived Homotheriini Africa 5
I North American Homotheriini North America 4–3
JParamachaerodus lineage Eurasia 11–10
KParamachaerodus and derivates Eurasia 9
L Smilodontini Eurasia, Africa, and North
America
5–4
M Metailurini Eurasia 9–8
N Barbourofelidae Eurasia and Africa 32
Phylogeny and evolution of cats (Felidae) 63
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

species with representatives in Europe, Arabia, Asia,
and North America. The interrelationships of the
species included in Pseudaelurus and the relationship
of this genus (or genera) to the radiations of the
subfamilies Felinae (conical-toothed cats) and Ma-
chairodontinae (sabretooths) are a major challenge
to felid palaeontology. Pseudaelurus is clearly a grade
rather than a monophyletic clade, and this complex
includes the ancestors of all subsequent felids. A
number of generic names are available for parts of
this complex, including Styriofelis, Hyperailurictis,
Miopanthera, Schizailurus, and Pseudaelurus itself. We
will consider the validity and applicability of these in
the discussion below. A fuller knowledge of the inter-
relationships within this group would go a long way
towards an understanding of the evolutionary pat-
terns of the Felidae.
Pseudaelurus is first recorded from Wintershof-
West in Germany (MN 3, 20–18 Ma; Dehm 1950).
Hence, it does not overlap stratigraphically with
Proailurus in Europe. Several reviews of Pseudaelurus
have been published in the past decades (Heizmann
1973; Ginsburg 1983; Rothwell 2003) and we refer to
them for a fuller discussion of evolutionary details.
Four species of Pseudaelurus are known from Eur-
ope. In the order of increasing size they are: P. turn-
auensis (¼P. transitorius), P. lorteti, P. romieviensis, and
P. quadridentatus (type species of the genus). They
range in size from a modern wildcat to a lynx or
small puma. Differences between them, apart from
Figure 2.3 The skull of Proailurus lemanensis, MNHN SG
3509 (holotype) from Saint-Ge
´rand-le-Puy, France, in ventral
view. The anterior and posterior halves do not meet. (Photo
courtesy of Stephane Peigne
´.)
Figure 2.4 Artist’s reconstruction of Proailurus lemanensis, the first cat. (Illustration courtesy of Mauricio Anto
´n.)
64 Biology and Conservation of Wild Felids
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

size, are minute (Heizmann 1973). The first species to
appear is the smallest, P. turnauensis (Dehm 1950).
However, all three remaining species appear in MN 4
(18–17 Ma). This indicates a rapid radiation of the
Pseudaelurus grade, suggesting a monophyletic origin
of at least European Pseudaelurus from a single spe-
cies of Proailurus. P. lorteti and P. romieviensis become
extinct at the end of the Middle Miocene (c. 11.6
Ma), but P. quadridentatus and P. turnauensis survive
into the Late Miocene (MN 9, c. 11.2–9.5 Ma). They
thus overlap stratigraphically with the earliest docu-
mented Machairodontinae (Miomachairodus pseudai-
luroides from Turkey; Schmidt-Kittler 1976; Viranta
and Werdelin 2003) (Fig. 2.2, letter E).
Pseudaelurus is poorly known from Asia, possibly
due to a relative dearth of Middle Miocene localities
on the continent. Two Chinese species are known.
Cao et al. (1990) describe P. guangheensis from Gansu
and Wang et al. (1998) describe P. cuspidatus from
Xinjiang. In addition, Qiu and Gu (1996) describe
material referred to P. lorteti. All this material is
Middle Miocene in age. What the relationship is
between the Chinese and European species has not
been determined, nor has their relationship to the
North American radiation of the grade.
The fossil record of Pseudaelurus in North America
was recently reviewed by Rothwell (2003). There are
five valid species: P. validus (stratigraphic range
c. 17.5–16.5 Ma), P. skinneri (c. 17.5–17.1 Ma), P. intre-
pidus (c. 17.1–13.3 Ma), P. stouti (c. 15.2–12.7 Ma),
and P. marshi (c. 16.4–12.7 Ma). Thus, Pseudaelurus
appears later in North America and goes extinct
sooner there than in Europe. This, and the cladistic
analysis of Rothwell (2003), in which the three youn-
ger species (P. intrepidus, P. stouti, and P. marshi) form
a clade with the two older species (P. validus and
P. skinneri) as outgroups, are consistent with a single
origin for North American Pseudaelurus.
Finally, a single record of P. t u r n a u e n s i s has been
reported from Saudi Arabia (Thomas et al. 1982) in
deposits now considered to be of MN 5 age (17.0–15.2
Ma). Material from Africa previously referred to P. a fr i -
canus (Andrews 1914) is now referred to Afrosmilus,a
barbourofelid (see Morales et al. [2001] and see below).
Figure 2.5 Artist’s reconstruction of Styriofelis lorteti, a member of the stem lineage leading to the extant Felidae, together
with the flying squirrel Petaurista sp. (Illustration courtesy of Mauricio Anto
´n.)
Phylogeny and evolution of cats (Felidae) 65
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

The endemic North American genus Nimravides
undoubtedly originated from one of the above-men-
tioned North American species of Pseudaelurus (Ba-
skin 1981; Beaumont 1990), probably P. intrepidus or
P. marshi, which both have a prominent chin, also
seen in Nimravides (Fig. 2.2, letter D). Nimravides
differs from its putative ancestors only in relatively
minor features: it has a more prominent chin, more
elongated, serrated canines, a more reduced P4 pro-
tocone, and more developed P4 ectoparastyle. These
are all features pointing towards a sabre-toothed
morphology, not dissimilar to that seen in M. pseu-
dailuroides and Machairodus aphanistus (see below),
but evolved in parallel. Four species of Nimravides
are known: N. thinobates (c. 11.0–9.6 Ma), N. pediono-
mus (c. 12.0–11.5 Ma), N. hibbardi (c. 7.0–6.4 Ma),
and N. galiani (c. 11.6–10.7 Ma). Near the end of the
Miocene, Nimravides became extinct, apparently
without leaving descendant lineages. A North Amer-
ican felid of uncertain affinities that may possibly
belong here is Pratifelis martini from the Late Mio-
cene (c. 7–6 Ma) of Kansas (Hibbard 1934). This spe-
cies has a distinctively enlarged m1 talonid and does
not fit comfortably into any of the larger felid
lineages.
Sabretooths
The further evolution of Felidae beyond the Pseudae-
lurus grade begins with M. pseudailuroides (Fig. 2.2,
letter E). This taxon, which is at present known only
from Turkey (Schmidt-Kittler 1976; Viranta and Wer-
delin 2003), has cheek teeth that are very similar to
those of P. quadridentatus, but the upper canines are
more flattened and have small crenulations on the
mesial and distal faces that are not present in Pseu-
daelurus spp. (Schmidt-Kittler 1976, figs. 114a, 1c, 2,
and 3, plate 5). In an important contribution,
Schmidt-Kittler (1976) discusses the relationship be-
tween M. pseudailuroides and the Pseudaelurus-grade
and how the morphological transition may have
occurred. However, he does not pinpoint any specific
relationships between taxa, nor does he extend his
discussion to conical-toothed cats. M. pseudailuroides
is at present known only from MN 7/8 and MN 9
(c. 12.5–9.5 Ma). The taxonomic status of the species
and genus has been discussed several times. Beau-
mont (1978) made Miomachairodus a subgenus of
Machairodus, and included Machairodus robinsoni
from the early Late Miocene (c. MN 9) of Tunisia
(Kurte
´n 1976) in the subgenus. On the other hand,
Ginsburg et al. (1981) synonymized M. pseudailur-
oides with M. aphanistus, type species of the genus
Machairodus. Morlo (1997) followed this, but sug-
gested that M. robinsoni in that case be considered a
separate genus. This discussion is far from settled,
but at the very least shows that these forms grade
into one another. Another early form about which
there is taxonomic disagreement is M. alberdiae from
MN 9 of Spain. Ginsburg (1999) considers this to be
the most primitive Machairodus, but Morlo (1997)
synonymizes it with M. aphanistus.
M. aphanistus was described by Kaup (1833) and
was the first Miocene felid to be named. Its cranio-
dental morphology was recently reviewed in detail
(Anto
´net al. 2004). These authors found that the
functional morphology of the killing bite in M. apha-
nistus, and characters related to this behaviour, were
considerably more primitive than in later machairo-
donts from the Eurasian Late Miocene. They con-
cluded that Machairodus should be restricted in
content to Vallesian (c. 11.2–9.0 Ma) forms, while
Turolian (c. 9.0–5.3 Ma) forms should be referred to
Amphimachairodus (Fig. 2.2, letter F). Morlo and Seme-
nov (2004) objected to this procedure, arguing that
the evolution from Machairodus to Amphimachairodus
was gradual and mosaic and that the two could not be
generically distinct. However, making the distinction
is taxonomically useful and in line with a trend in
recent years of trying to restrict the usage of Machair-
odus to something other than a waste-basket taxon for
any or all Miocene sabretooths (Beaumont 1978;
Ginsburg et al. 1981; Ginsburg 1999).
Some time in the Vallesian, Machairodus probably
migrated to North America, where it gave rise to
M. coloradensis (c. 9.0–5.3 Ma). This is a fairly
generalized species, similar to M. aphanistus.Itis
possible, if unlikely, that it evolved from the North
American Nimravides. This would require extensive
parallelism with Machairodus. The possibility has
been noted before, however, and the generic name
Heterofelis (Cook 1922) is available for this taxon.
The next stage in the evolution of the machairo-
dont lineage is the genus Amphimachairodus (Fig. 2.2,
letter G). This genus includes a number of closely
66 Biology and Conservation of Wild Felids
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

related species that morphologically lead up to the
Plio-Pleistocene tribe Homotheriini (Fig. 2.2, letter
H), which includes the genera Homotherium, Dino-
bastis, and Xenosmilus. Amphimachairodus includes
the species A. giganteus (Eurasia; c. 9–5.3 Ma), A.
kurteni (Kazakhstan; c. 7.1–5.3 Ma), A. kabir (Chad
and Libya; c. 7–5.5 Ma), and possibly A. irtyschensis
(Russia; c. 7.1–5.3 Ma), though the latter may be a
synonym of A. giganteus. Closely related is also Loko-
tunjailurus emageritus (Werdelin 2003b; c. 7.4–5.5
Ma), which lacks a number of the derived cranial
features of Amphimachairodus, but is dentally the
most derived of the group. A. giganteus is, as the
name implies, characterized by very large size, ex-
tremely long upper canines and a derived mastoid
region relative to that of Machairodus, implying mod-
ifications to the killing bite. The mastoid region is
further evolved in M. kurteni and M. kabir, but has not
yet reached the condition seen in Homotherium. Den-
tally, the upper incisor arcade is modified and the
cheek dentition progressively simplified, with reduc-
tion of p3/P3, complete loss of the m1 talonid, and
nearly complete loss of the P4 protocone. The denti-
tion of L. emageritus is very close to that of primitive
Homotherium, but the skull and skeleton of the for-
mer preclude it from the direct ancestry of that genus
(Werdelin 2003b). L. emageritus has an extremely
enlarged dew claw (absolutely and relative to the
other claws) on the manus and this feature appears
to be present also in Homotherium (Ballesio 1963).
The evolution of Machairodus and Amphimachair-
odus is paralleled in the sabretooth group by the
evolution of the genus Paramachaerodus (Fig. 2.2,
letters J and K). At least two and possibly as many
as four species of this genus are known: P. ogygius
(c. 9–7 Ma), P. orientalis (c. 8–6 Ma), P. indicus (age
uncertain), and P. maximiliani (c. 7–5.3 Ma) (Salesa
et al. 2003). The latter two may be synonymous, with
each other and with P. orientalis. Paramachaerodus is
much smaller than Machairodus and (especially) Am-
phimachairodus (Paramachaerodus is leopard, rather
than lion-sized or larger in the case of Amphimachair-
odus). Clearly, this genus and its larger relatives were
dividing up the prey-spectrum by size, though the de-
tails of this are not yet understood. New material from
the early Late Miocene of Spain is doing much to
clarify the taxonomic, functional, and ecological
relationships between these Miocene sabretooths
(Anto
´net al. 2004; Salesa et al. 2005).
A further lineage that is likely to at least in part
belong among the sabretooths, despite lacking the
typical craniodental attributes of this functional
grade, is the tribe Metailurini (Fig. 2.6). This tribe as
generally conceived includes the larger genus Dino-
felis (Fig. 2.6a), with at least ten species (Werdelin
and Lewis 2001), Metailurus (Fig. 2.6b), with at least
four species, and Stenailurus, with one species
(though the latter may be a synonym of Metailurus).
Dinofelis is in many ways convergent on Panthera,
but its evolution is not straightforward convergence.
Instead, various species of Dinofelis are more or
less pantherine-like, while the oldest and youngest
species are the most sabretooth-like. The Metailurini
is essentially a waste-basket for taxa that show
some sabretooth features but can not be placed
in either the Machairodus or the Paramachaerodus
lineages. It is not clear that Dinofelis and Metailurus
are closely related, nor what their respective an-
tecedents are. Nor is it clear, although it seems
likely, that Metailurus is a member of the subfamily
Figure 2.6 (a) Skull of Dinofelis petteri, KNM ER 2612
(holotype), Tulu Bor member, Koobi Fora Formation, Kenya; in
left lateral view. (b) Skull of Metailurus parvulus PIU M3835,
Locality 108, Baode Province, China; in left lateral view.
Phylogeny and evolution of cats (Felidae) 67
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

Machairodontinae (sabretooth cats). Dinofelis, how-
ever, shares several traits with derived sabretooths
and can confidently be placed in this subfamily
(Werdelin and Lewis 2001). Both of these genera
originate in the Miocene and survive into the Plio-
Pleistocene; Metailurus is mainly a Miocene genus,
while Dinofelis has its main radiation in the Pliocene.
The Plio-Pleistocene sees the appearance of the
two derived sabretooth tribes, Homotheriini and
Smilodontini (Fig. 2.2, letters H and L; Fig. 2.7). The
Homotheriini includes the genera Dinobastis (with at
least one species, D. serus) and Xenosmilus (with one
species, X. hodsonae) from North America and Homo-
therium (Fig. 2.7b) (with several species, including
H. crenatidens and H. problematicum) from Eurasia
and Africa. The relationships between these genera
will be discussed below. The Smilodontini includes
two genera: Megantereon (with at least five species:
M. cultridens, M. whitei, M. hesperus, M. falconeri, and
M. ekidoit) from Africa, Eurasia, and North America;
and Smilodon (with three species: S. gracilis, S. fatalis
[Fig. 2.7a], and S. populator) from North, Central, and
South America.
Differences between Homotheriini and Smilodon-
tini are substantial, both craniodentally and postcra-
nially. The Homotheriini have relatively short,
mediolaterally narrow upper canines with large cre-
nulations on the anterior and posterior edges; their
postcranial skeleton shows some adaptations to a
cursorial lifestyle (except in Xenosmilus), with long,
slender limbs and forequarters that are massive but
not hyperdeveloped. The cheek dentition of Homo-
theriini is dominated by very large carnassials,
which especially in Homotherium become larger in
later forms, with the p4 also usurped into the cutting
blade. The Smilodontini have very long, broad upper
canines with minute serrations (lost in Megantereon).
Their skeleton is very robust and the forequarters
extremely massive. The cheek dentition is reduced,
but the carnassials are not elongated to the extent
seen in Homotheriini.
Conical-toothed cats
The conical-toothed cats, subfamily Felinae, com-
prise the common ancestor of all living cats and all
of its descendants (Fig. 2.8). As the name implies,
conical-toothed cats differ from sabretooths in hav-
ing a more rounded canine cross-section. They are
also united by a few other features, such as the rela-
tively long lower canine. The interrelationships of
the living members of this subfamily were discussed
above. Their fossil history is much less well known
than that of the sabre-toothed cats. This could be
for three reasons: (1) they were predominantly
adapted to environments in which fossilization is
less likely than in the environments inhabited by
sabre-toothed cats (i.e. the poor fossil record reflects
a taphonomic bias; species that today occur in habi-
tats in which fossilization potential can be consid-
ered fair [e.g. cheetahs and lynx], have a reasonably
good fossil record, while species that today inhabit
Figure 2.7 (a) Skull (cast) of Smilodon fatalis from Rancho
La Brea, California, United States; in left lateral view. (b)
Skull (cast) of Homotherium sp., unknown locality, China; in
left lateral view.
68 Biology and Conservation of Wild Felids
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

tropical, wet forests [e.g. golden cats and clouded
leopards] tend to have a very poor fossil record); (2)
they were less common in the past than sabre-
toothed cats (i.e. the poor fossil record reflects a
true pattern that is an outcome of a consideration
of intra-familial competition between sabre-toothed
and conical-toothed cats); (3) they are more similar
to each other in hard-tissue morphology than sabre-
toothed cats (i.e. the poor fossil record reflects a bias
in investigator perception; there are great similarities
between all conical-toothed cats in, for example
mandibular morphology, a region in which sabre-
toothed cats exhibit a number of diagnostic differ-
ences). All three of these possibilities may be true to
some extent. Finally, the poor fossil record of coni-
cal-toothed cats may also reflect the interests of re-
searchers. Sabretooth cats are large, spectacular, and
to some extent mysterious, at least as far as their
feeding behaviour is concerned. Conical-toothed
cats are often small, nondescript and closely similar
to living forms that are comparatively well known
ecologically and functionally. Hence, the former re-
ceive far more attention in the palaeontological lit-
erature than the latter.
Only one researcher, Helmut Hemmer, has fo-
cused almost exclusively on the fossil record of coni-
cal-toothed cats, and it is thus from his work (e.g.
Hemmer 1974, 1976; Hemmer et al. 2001, 2004) that
most of the information on the fossil record of this
group is to be gleaned. In the following section, the
fossil record of conical-toothed cats will be outlined,
following the scheme of eight major lineages as
found in the molecular phylogeny (Fig. 2.1). Focus
will be on the earliest members of each lineage and/
or species.
Some early conical-toothed cats cannot with con-
fidence be included in any of the eight lineages.
These include the first ‘Felis’, ‘F.’ attica, known from
MN 11–MN 13 (c. 9.0–5.3 Ma) in western Eurasia.
This species is a little larger than a wildcat. In mor-
phology it is very similar to smaller species of Pseu-
daelurus, but it has a dentition that is reduced
beyond the Pseudaelurus grade. It is noteworthy
that the stratigraphic range of ‘F.’attica is younger
than the estimated age of the base of the radiation of
extant Felidae (Fig. 2.1), so that it may belong within
that radiation rather than to the stem lineage. The
same is true of ‘F.’ christoli, another primitive cat,
known from MN 13–MN 14 (c. 7.1–4.2 Ma) of Spain
and France. In addition, there are significant collec-
tions of Late Miocene small cats from China that
remain undescribed. This material may answer
some questions regarding the early evolution of
extant cats.
The clade with by far the best fossil record is the
Panthera lineage. Despite this, it is also the clade
with the longest ghost lineage (cladistically inferred
lineage undocumented by fossils). According to mol-
ecular data ( Johnson et al. 2006b) this lineage split
off from the Felidae stem lineage about 10.8 Ma.
However, the oldest fossils unequivocally assigned
to the lineage are no older than 3.8 Ma (Barry 1987;
Werdelin and Dehghani, in press), leaving a ghost
lineage that is nearly twice as long as the documen-
ted lineage. The earliest fossil Panthera from Laetoli
belong to two species: a lion-sized one and a leopard-
sized one. They have been suggested to belong to the
Figure 2.8 Skulls of extant Felidae in left lateral view: (a)
Lion, Panthera leo; (b) Eurasian lynx, Lynx lynx; (c) Domestic
cat, Felis catus.
Phylogeny and evolution of cats (Felidae) 69
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

extant species (Turner 1990), but in fact differ from
them morphologically (Werdelin and Dehghani, in
press). The molecular dates suggest that they may
belong to the stem lineage of these species and
there is nothing in the fossils that would suggest
otherwise.
The first definite lions are from Olduvai, Bed 1
(<2 Ma), which is also in line with the molecular
data. The subsequent fossil history of lions is well
known, with dispersal out of Africa across Eurasia
and into North (and possibly South) America. These
developments have been discussed by numerous
people (Vereshchagin 1971; Hemmer 1974; Burger
et al. 2004; Yamaguchi et al. 2004a). It is not until
the middle Pleistocene that lions significantly ex-
tend their range outside Africa, and by about 500 ka
they are found throughout Europe and parts of Asia
north and east of the Black and Caspian Seas. By 300
ka their range had extended to encompass most of
northern and eastern Asia except for the south-east
and southern China, possibly due to competition
with tigers (although this cannot be verified in the
fossil record). At about this time, lions probably
crossed the Bering Strait into North America, where
they are known from Illinoian (<310 ka) and later
deposits. In the Sangamonian, after the retreat of
the Illinoian glaciers, lions could spread further
into North America and, arguably, also northern
South America. Lions became extinct in the Americas
and large parts of Asia at the end of the latest glacia-
tion. Further range contraction occurred in historic
times.
Lion taxonomy has long been controversial. Some
authorities place all fossil lions in the modern spe-
cies, P. leo, while others recognize a number of ex-
tinct species, for example P. spelaea, the cave lion,
and P. atrox, the North American lion. Burger et al.
(2004) analysed mtDNA cytochrome bsequences of
some cave lions and found them to form a mono-
phyletic clade distinct from living lions. Until more
data from a broader range of fossil lions have been
studied, the question of whether lions conform bet-
ter to a one-species or a multiple-species model must
remain open.
The other members of the Panthera lineage are less
well known in the fossil record and some aspects of
their evolution are at present controversial. The char-
acters linking fossils with extant species are often of
uncertain value and the material commonly limited.
The snow leopard, P. uncia, and clouded leopard,
N. nebulosa, are, for example, only known from
isolated fossil teeth, and it is doubtful whether this
is sufficient for specific attribution.
The jaguar, P. onca, has been traced back to the
‘European jaguar’, P. gombaszoegensis, which is con-
sidered by some to be a subspecies of the extant
species (Hemmer et al. 2001). If P. toscana can be
included in this species, as suggested by Hemmer, it
is first known from the latest Pliocene and survived
into the Middle Pleistocene. During this time it was
mainly distributed across western Eurasia.
The earliest leopards, P. pardus, are known from
Africa. As noted, Laetoli (c. 3.8–3.4 Ma) includes a
leopard-sized pantherine. Hemmer et al. (2004) have
suggested that these remains should be referred to
the Puma lineage, but the fossils provide no support
for this hypothesis (Werdelin and Dehghani, in
press). The oldest unequivocal leopards in Africa are
from about 2 Ma, and the first leopards appear in
Eurasia about 1 million years later.
Tiger remains are known from the Lower Pleisto-
cene of South-east Asia (Kurte
´n, 1962). However, the
oldest member of the tiger lineage is generally con-
sidered to be P. palaeosinensis (Fig. 2.9) from (proba-
bly) Upper Pliocene sediments in northern China
(Zdansky 1924). However, renewed study (Christian-
sen 2008) indicates that its specific relationship to
tigers is tenuous at best. Metrically, the specimen is
not particularly close to any extant Panthera.Itis
generally agreed that fossil tigers have not been
Figure 2.9 Skull of Panthera palaeosinensis, PIU M3654
(holotype), one of the earliest fossil Panthera, in left lateral
view.
70 Biology and Conservation of Wild Felids
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

recorded outside Asia (but see Herrington 1987;
Groiss 1996).
The bay cat lineage is not known with certainty in
the fossil record. The Caracal lineage is represented
in the fossil record by specimens dating back c. 4 Ma.
These specimens group into two distinct size classes,
large and small. Given the molecular ages of these
lineages, these may represent members of the cara-
cal/golden cat stem lineage and serval stem lineage,
respectively. Whether any or all of the fossils, which
are known from a number of sites in eastern and
southern Africa, are conspecific with the extant
forms is not determinable on the basis of the avail-
able material, which consists mainly of isolated teeth
and fragmentary jaws. An intriguing recent sugges-
tion is that ‘Felis’ issiodorensis, a species generally
referred to the genus Lynx (Werdelin 1981) should
instead be referred to Caracal (Morales et al. 2003b).
This conclusion is based on the observation that the
metric analyses of Werdelin (1981) showed that spe-
cimens identified as belonging to L. issiodorensis were
more similar to specimens of Caracal than to speci-
mens of Lynx. This possibility deserves further study,
but it is well to remember that it is just as likely that
the similarities between Caracal and L. issiodorensis
are shared ancestral characters.
The fossil record of the ocelot lineage is relatively
poor. This record has recently been reviewed by Sey-
mour (1999) with updates by Prevosti (2006). The
South American record of the group is limited, and
with the exception of some remains of Leopardus co-
locolo from Argentina in sediments dating as far back
as c. 0.5–1 Ma, and the enigmatic ‘Felis’vorohuensis of
about the same age, all records are latest Pleistocene in
age. North American fossils unequivocally referable to
this lineage are also from the Late Pleistocene (Werde-
lin 1985). The inferred age of the radiation of the
extant taxa at c. 2.9 Ma (Fig. 2.1) is younger than
previous estimates and compatible with a radiation
from a single immigration event into South America
(Werdelin 1989). However, this leaves a long ghost
lineage back to the reconstructed age of the node
leading to this group at c. 8.0 Ma. A number of
North American taxa have been proposed at one
time or another as members of this ghost lineage,
including ‘F.’lacustris,‘F. ’rexroadensis,‘F.’longignathus,
and ‘F.’proterolyncis (e.g. Werdelin 1985; Seymour
1999). The first of these is likely to belong to the
Puma lineage, but the relationships of the others are
unclear. They may belong to the Lynx or ocelot
lineages, or be on the backbone of the phylogeny
between them. The earliest members of several of
these taxa are Late Miocene (c. 7–6 Ma) in age.
The short phylogenetic distance between the ocelot
and Lynx lineages may explain why several taxa men-
tioned above could be assigned to either. The genus
Lynx is well represented in the fossil record, both in
Eurasia and NorthAmerica (Werdelin 1981). In light of
the above, it is likely that the earliestfossil membersof
the lineage are Late Miocene in age. The earliest record
of unequivocal Lynx in the fossil record has been con-
sidered to be L. issiodorensis from the Pliocene and
Pleistocene of western Europe (but see the opinion of
Morales et al. [2003a], as discussed above). This species
is not, however, found on the African continent as
previously suggested (Hendey 1974; Werdelin 1981).
The only record of the genus on that continent is the
Pleistocene L. thomasi from Morocco (Geraads 1980).
The Puma lineage has a long, if uneven, fossil re-
cord. The oldest fossils unequivocally belonging to
this lineage are specimens referred to Acinonyx sp.
from Laetoli (c. 3.8–3.4 Ma) (Barry 1987; Werdelin
and Dehghani, in press). These specimens are about
the size of the modern species but differ slightly in
morphology. The cheetah subsequently has a contin-
uous though sparse fossil record in Africa. The genus
Acinonyx has a long history in Eurasia. The ‘giant’
species A. pardinensis appeared in western Europe a
little over 3 Ma. This form is also found in China (as
A. pleistocaenicus) and India (as A. brachygnathus). It
was about the size of a small lion, though consider-
ably lighter. In most other respects it displayed typi-
cal characters of Acinonyx, though the skull does not
show the extreme vaulting seen in A. jubatus. During
the later Pliocene there is a marked size reduction
in Eurasian cheetahs, leading Thenius (1953) to
describe the younger form as a separate species,
A. intermedius. However, some Pleistocene specimens
are as large as the Pliocene ones and we agree with
Viret (1954) and Kurte
´n (1968) that the difference
probably does not warrant specific separation. The
Eurasian cheetah became extinct in the early Middle
Pleistocene. The North American ‘cheetah’, Miraci-
nonyx, with two species, M. inexpectatus and M. stu-
deri (Adams 1979; Van Valkenburgh et al. 1990), is
not the sister taxon to Acinonyx (Barnett et al. 2005).
Phylogeny and evolution of cats (Felidae) 71
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

Instead, it apparently evolved its cheetah-like fea-
tures independently, from puma-like ancestors. The
oldest members of this lineage are c. 2.5 Ma. How-
ever, the oldest ‘F.’lacustris is somewhat older than
this. An interesting specimen of about the same age is
the Felis sp. of Gustafson (1978) from the Blancan of
Oregon, which may also belong to this lineage. The
presence of Puma in Europe has also been suggested,
in the form of P. pardoides (Hemmer et al. 2004). The
oldest of this material is of Pliocene age and may be
the oldest material of Puma on record. The suggestion
that Puma is present at Laetoli is hardly tenable,
however (Werdelin and Dehghani, in press). The old-
est fossil jaguarundi is less than 0.5 Ma.
The leopard cat lineage is very poorly known in the
fossil record. A few fossils probably pertaining to this
lineage and possibly to Prionailurus bengalensis have
been found in Middle Pleistocene sites in South-east
Asia (Hemmer 1976). In addition, fossils tentatively
referred to O. manul have been recorded from Kamyk,
Poland (Kurte
´n 1968). These may be more than 1 Ma.
The fossil record of the domestic cat lineage is not
poor, but much of it is hidden beneath the general
designation of Felis sp., since the species are all but
indistinguishable on the basis of incomplete remains.
The oldest ‘Felis sp.’ that definitely belongs to this
lineage is from Kanapoi, Kenya, dated to >4Ma(Wer-
delin 2003a). If the molecular dates are correct, this
material belongs to a member of the stem lineage of
Felis. Further specimens belonging to this lineage
occur intermittently in the African fossil record. A
species of some interest that may be the oldest mem-
ber of the F. silvestris group is F. lunensis from Europe.
This species goes back at least to the Early Pleistocene
and possibly to the Late Pliocene. Specimens referable
to F. chaus have been found in Holocene strata of Java
(outside the modern range of the species; Hemmer
1976). No specimens definitely referable to F. nigripes
or F. margarita have been found in the fossil record.
Barbourofelidae
Finally, we must touch upon the family (or subfamily)
Barbourofelidae (Fig. 2.2, letter N), which consistsof a
number of derived sabre-toothed forms (though not
all may be sabre-toothed—see below). Traditionally,
they have been seen as Neogene members of the
Nimravidae, a group that itself has been the subject
of much phylogenetic discussion. The nimravids were
once known as ‘paleo-felids’ because of their felid-like
craniodental morphology. They are known from the
Late Eocene to Late Oligocene of North America and
Europe and include genera such as Nimravus, Hoplo-
phoneus,andEusmilus. Studies of basicranial morphol-
ogy have, however, clearly shown that nimravids are
not felids (Neff 1983; Hunt 1987). They are therefore
placed in the family Nimravidae. In its original con-
ception, Nimravidae also included the barbourofelids,
Miocene sabretooths with representatives both in
North America and Europe (Schultz et al. 1970).
These, however, have a basicranial morphology, in-
cluding an ossified bulla, that differs from those in
both Nimravidae and Felidae. Therefore, Morales et al.
(2001) proposed removing them from the Nimravidae
and placing them as the subfamily Barbourofelinae
within the Felidae. This proposal was amended by
Morlo et al. (2004), who proposed raising Barbourofe-
linae to full family status as the Barbourofelidae,
which is the path followed here. The Nimravidae are
likely to be basal Carnivora, while the Barbourofelidae
are either the sister-group to Felidae or the sister-
group to other Aeluroidea (Fig. 2.2, letter N). Because
of their phylogenetic and ecomorphological close-
ness to Felidae, their fossil record is outlined here.
In Africa, the likely centre of origin of Barbourofe-
lidae, the family is known from a number of genera
(Morales et al. 2001; Morlo et al. 2004). Afrosmilus has
two east African species, A. africanus (Fig. 2.10) and
A. turkanae, both c. 18–17 Ma. Ginsburgsmilus, the
most primitive member of the family, has a single
Figure 2.10 Left horizontal mandibular ramus of
Afrosmilus turkanae, KNM MO 15929, Moruorot, Kenya, a
barbourofelid. Note the well-developed metaconid at the
posterior end of the tooth—a diagnostic difference
between Barbourofelidae and Felidae (Morlo et al. 2004).
72 Biology and Conservation of Wild Felids
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

east African species, G. napakensis (c. 20.5–17 Ma).
Syrtosmilus with one species, S. syrtensis (c. 19–15
Ma), and Vampyrictis with one, V. vipera (c. 12.5–9.5
Ma), are North African representatives of the family.
In Europe, Barbourofelidae is known from several
genera, Prosansanosmilus, with two species, P. peregri-
nus (MN 4, c. 18–17 Ma) and P. eggeri (MN 5, c. 17–15.2
Ma), Sansanosmilus with two species, S. palmidens (Fig.
2.11) (MN 5–MN 7/8, c. 17–11.2 Ma) and S. jourdani
(MN 6–MN 9, c. 15.2–9.5 Ma), and Afrosmilus with
one European species, A. hispanicus (MN 5, c. 17–15.2
Ma). These show a temporal progression towards lar-
ger and more sabretooth forms, though they are gen-
erally less extreme in their adaptations than the North
American Barbourofelis spp. Sansanosmilus is also
known from the Middle Miocene of China, though
it is less common there than in Europe.
In North America, the Barbourofelidae consists of
the single genus Barbourofelis, with five species, B.
fricki (c. 10 Ma), B. loveorum (c. 11–9.8 Ma), B. morrisi
(c. 11.5 Ma), B. osborni (c. 11.5 Ma), and B. whitfordi
(c. 12–11.5 Ma). They are all extreme sabretooth eco-
morphs, with long sabres, large mental flanges and
short, stout limbs (where known).
Finally, two species from southern Africa must be
mentioned, Diamantofelis ferox (the size of a small
puma) and Namafelis minor (lynx-sized) (Morales
et al. 1998, 2003a). Both are from the late Early–
earliest Middle Miocene of Arrisdrift, Namibia (c.
17–15.2 Ma). These species are not, as far as is
known, sabre-toothed in morphology, as neither
has a squared-off symphyseal region, but they do
share other mandibular and dental features with
species of Afrosmilus. D. ferox has a short and deep
mandible, while that of N. minor is longer and more
slender. The oldest true felid from Africa is a small
specimen from Songhor, Kenya (c. 18–17 Ma), prob-
ably referable to Pseudaelurus sensu lato.Thus,since
Felidaeisrareornon-existentinAfricaatthistime,
whereas Barbourofelidae is known from a number
of sites and regions, and given the morphological
similarities between them, it should at least be con-
sidered whether the Arrisdrift species might be ‘con-
ical-toothed’ barbourofelids.
The Barbourofelidae was a relatively short-lived
group (c. 20.5–9.5 Ma), within which the vast major-
ity of species were specialized sabretooths. Their ex-
tinction in the early Late Miocene may be tied to the
spread of sabre-toothed Felidae at this time.
Discussion
In this section, we will attempt to draw some con-
clusions from the review above. We advocate the use
of generic names for fossils that maximizes the num-
ber of monophyletic taxa by splitting up at least
those genus-level groups that are obviously para- or
polyphyletic. In so doing we hope to create a more
consistent framework for future studies. Unfortu-
nately, many of the assertions made in the following
discussion are at present untested, though we hope
that it will be possible to test them in the future. We
aim to erect a series of hypotheses to establish the
basic level of understanding of felid evolution, that
of interrelationships. When the interrelationships of
fossil felids have been better established, a founda-
tion for the understanding of the ecological, bio-
geographical, and functional patterns of felid
evolution will have been laid, in much the same
way as the current phylogeny of living felids pro-
vides such a foundation for study of their radiation.
We will also point out areas where we know too
little, which is especially true of the fossil record of
the living felids, which at present has not much to
contribute to an understanding of the modern
Figure 2.11 Artist’s reconstruction of the head of
Sansanosmilus palmidens, a barbourofelid. (Illustration
courtesy of Mauricio Anto
´n.)
Phylogeny and evolution of cats (Felidae) 73
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

radiation. Johnson et al. (2006b) estimate that fossil
representation in the modern cat radiation is about
24%, leaving large areas unknown (cf. Fig. 2.1). Fig.
2.2 provides a graphical summary of the discussion
in the following section.
Early cats
The origins of the family Felidae are relatively uncon-
troversial, though that may merely be because the
gap between the earliest unequivocal felids and their
ancestors among Carnivoramorpha is relatively sub-
stantial. Thus, Proailurus is unquestionably a felid
and Stenogale and Haplogale are likely to belong to
this family as well. All of these genera are undoubt-
edly closer to crown group Felidae than is the extant
sister taxon, Prionodon. Aspects of this early evolu-
tion are covered by Hunt (1998) and Peigne
´(1999)
and require no further elaboration here.
The subsequent radiation of Felidae in the Early–
Middle Miocene is far more complex, however. The
genus Pseudaelurus comprises 11 named species, 4
from Europe and Arabia, 2 from China, and 5 from
North America. Unanswered questions surrounding
this radiation include: Does Pseudaelurus have a sin-
gle origin? What are the interrelationships of the
European species to each other? What is the relation-
ship between Chinese and European species? From
which species did North American Pseudaelurus orig-
inate? Which species of Pseudaelurus belong to which
lineages of later, more derived felids?
Answers to some of these questions have been
proposed in the past, whereas some have rarely
been discussed, if at all. An example of the latter is
the first question raised above: Does Pseudaelurus
have a single origin? This has tacitly been assumed
in discussions of felid evolution in the past. How-
ever, the data in favour of this hypothesis are largely
circumstantial. The presumed ancestor, Proailurus,
has a limited geographic distribution and Pseudae-
lurus from Europe is older than Pseudaelurus on
other continents, arguing for a single origin and
subsequent dispersal. There is, on the other hand,
no phylogenetic framework in which this has been
demonstrated to be the most parsimonious hypoth-
esis. It is certainly also possible that different species
of Proailurus or related genera gave rise to different
species of Pseudaelurus, rendering the latter polyphy-
letic. The Ginn Quarry felid discussed above (Hunt
1998) makes such a scenario more plausible. At pres-
ent, there does not seem to be any way to resolve
this issue definitively and the monophyletic origin
of Pseudaelurus is assumed here as a working
hypothesis.
On the other hand, Pseudaelurus is undoubtedly
paraphyletic, with different species groups giving
rise to different descendant taxa. The paraphyletic
nature of Pseudaelurus has been recognized for a
long time, though perhaps the first to do so explicitly
was Kretzoi (1929b), and this was also implicitly
acknowledged by Viret (1951) before being elabor-
ated on by Beaumont (1964, 1978). Beaumont
(1964), like Kretzoi before him, split Pseudaelurus
into a number of genera at the bases of several
subsequent radiations. Though Beaumont (1978) re-
duced these to subgenera, his Figure 2 remains the
fullest envisioning of felid evolution to this day. If we
ignore the subgenera, he split Pseudaelurus into three
genera: Pseudaelurus (Gervais 1850, type species
P. quadridentatus), Schizailurus (Viret 1951, type
species P. lorteti), and Hyperailurictis (Kretzoi 1929b,
type species P. intrepidus). The first-mentioned in-
cludes only the type species, while the second
includes P. turnauensis in addition to P. lorteti. The
third includes all North American species of Pseudae-
lurus listed above. However, it should be noted that
Schizailurus is an objective junior synonym of Mio-
panthera Kretzoi (1938) (based on the same type spe-
cies), and this, in turn is a subjective junior synonym
of Styriofelis Kretzoi (1929a; type species F. turnauen-
sis). Thus, the latter name is the senior valid synonym
and is used here.
Beaumont (1978) places Pseudaelurus at the base of
the radiation of sabre-toothed cats, Styriofelis at the
base of the radiation of conical-toothed cats, and
Hyperailurictis at the base of the North American
radiation, as well as the radiation of ‘intermediate’
forms such as Metailurus, Stenailurus, and Dinofelis.
The radiation of these three genera takes place at
letter B in Fig. 2.2. The evidence for this scenario is
not particularly strong, as it is based mainly on the
somewhat more sabretooth-like characteristics of
P. quadridentatus, as opposed to the clearly conical-
toothed features of Styriofelis lorteti and Styriofelis
turnauensis. Although the generic separation between
74 Biology and Conservation of Wild Felids
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

these taxa has generally not been considered in reviews
of Pseudaelurus, (Heizmann 1973; Ginsburg 1983), the
separation has been implicitly acknowledged by several
other workers (e.g. Morlo 1997). The status of the North
American pseudaelurines (Hyperailurictis)asadistinct,
generic-level clade is somewhat more secure, as these
species are relatively derived, very similar to each other,
and also very similar to their presumed descendant
Nimravides.
This scenario provides possible answers to several
of the questions posed above, apart from the ques-
tion of the relationship of Pseudaelurus to later, more
derived felids. Among European ‘Pseudaelurus’, S. lor-
teti and S. turnauensis are closely related and more
distant from P. quadridentatus (the status of P. romie-
viensis is unclear). The North American Hyperailurictis
did not evolve from any of them, though it may be
related to one or both of the Chinese species. It is
more likely, however, that Hyperailurictis descended
from a felid similar to the Ginn Quarry felid de-
scribed by Hunt (1998). The wholly unanswered
question is the relationship between European and
Chinese pseudaelurines.
A reasonable consensus, however, is that Styriofelis
gave rise to the radiation of modern cats (Fig. 2.2,
letter C). Aside from S. lorteti and S. turnauensis,no
species definitely belonging to the stem lineage are
known (but see Felis attica and see below).
In North America there is little doubt that Hyper-
ailurictis gave rise to Nimravides (Fig. 2.2, letter D).
Whether the former is mono- or paraphyletic is not
known at this time. If it is paraphyletic, further no-
menclatural complications may arise, but these need
not concern us here. Nimravides seems to have gone
extinct without leaving descendants, though it is
just possible that M. coloradensis evolved from this
genus rather than being an immigrant from Eurasia.
The close similarity between M. coloradensis and the
Eurasian early Late Miocene M. aphanistus argues
against this, however.
The relationship between Hyperailurictis and Dino-
felis, Metailurus and Stenailurus is far less well estab-
lished and is not followed here. These genera are
usually grouped together as the Metailurini, though
the monophyly of this tribe has not been satisfac-
torily demonstrated. This is clearly an Old World
group, with the evolution of Metailurus centred in
Eurasia and that of Dinofelis in Africa. This presents
some biogeographic problems for an origin from
Hyperailurictis in North America as suggested by
Beaumont (1964, 1978). It is very tempting instead
to associate this group with the Chinese pseudaelur-
ines, though these are so poorly known that this
remains pure speculation at present. Here we will
consider the Metailurini to belong to the sabretooth
cats (but see below), and thus a part of the radiation
at letter E of Fig. 2.2.
Upper Miocene to Pleistocene cats
Pseudaelurus, sensu stricto, gave rise to the radiation of
sabretooth cats that first appeared in the late Middle
Miocene of Eurasia (and possibly Africa) and spread
across the world in the Late Miocene (Fig. 2.2, letter
E). There has been much controversy surrounding
sabretooths (subfamily Machairodontinae) and con-
siderable confusion regarding taxonomy and the
allocation of specimens ever since Cuvier (1824)
placed the first sabretooth specimens in the genus
Ursus. Numerous genera and species have been
named over the years and the course of evolution
of the group has been poorly understood. Part of the
problem has been the focus on Smilodon, a late and
highly derived sabretooth, as the exemplar species in
discussions of the functional morphology and evo-
lution of the group (e.g. Bohlin 1940; Simpson 1941;
Miller 1969; Akersten 1985). However, considerable
progress in understanding these issues has come in
recent years with the study of the excellently pre-
served material from the carnivore trap site of Batal-
lones-1 in the Cerro de Batallones, Spain (e.g. Anto
´n
et al. 2004; Salesa et al. 2005). These studies show
that the functional morphology of sabretooths was
not uniform across taxa, evolved over time, and is
compatible with a gradual origin from Pseudaelurus-
grade forms.
Nevertheless, numerous questions regarding the
systematics and evolution of sabretooth cats remain.
Some of these are: What is the relationship of Dino-
felis and Metailurus to Machairodontinae? What are
the evolutionary patterns within the paraphyletic
Amphimachairodus group? What is the relationship
between Homotherium and Dinobastis? How did the
Smilodontini evolve and which taxa are their
Phylogeny and evolution of cats (Felidae) 75
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

ancestors? What did sabretooths feed on and how?
How and why did sabretooths become extinct?
The relationship of Dinofelis and Metailurus to Ma-
chairodontinae (and to each other) has always been
controversial. Some authors, (e.g. Beaumont 1978;
Werdelin and Lewis 2001), have considered them
to be members of the Machairodontinae with slight
to moderate sabretooth adaptations, while others
(Kretzoi 1929b; Hendey 1974) have considered them
to be conical-toothed cats with a tendency to develop
sabretooth adaptations. The main feature they share
with Machairodontinae is a reduced lower canine rela-
tive to the upper canine. Dinofelis further shares with
Machairodontinae a deep groove or pit supero-medial
to the trochlear notch of the ulna (Werdelin and Lewis
2001). This feature seems not to be present in Metai-
lurus (Roussiakis et al. 2006). Thus, it appears likely
that Dinofelis belongs in the Machairodontinae, but
the position of Metailurus is equivocal. This also, of
course, makes the relationship between the two
genera uncertain. Thus, the position of this group at
letter J of Fig. 2.2 is problematic, as is the placement,
even existence, of the node at letter M.
Amphimachairodus is clearly paraphyletic, as Homo-
therium evolved from within this species group. This
is reflected in the intermediate position of letter G
(Fig. 2.2), between letter F where Amphimachairodus
splits off from a similarly paraphyletic Machairodus,
and letter H, at the base of the monophyletic Homo-
theriini. What is not clear is exactly which species
gave rise to Homotheriini (Fig. 2.2, letter H). L. ema-
geritus from Kenya has a more derived dentition than
any species currently assigned to Amphimachairodus,
but is too primitive in other respects and too derived
in a few to be the ancestral taxon. Of the species of
Amphimachairodus, A. kurteni seems the most derived
dentally, but A. kabir (if the material from Sahabi
belongs there; cf. Sardella and Werdelin 2007) has
the most derived mastoid region. Whichever of these
(or some as yet unknown taxon) is ancestral to Homo-
therium, Amphimachairodus as presently conceived
becomes paraphyletic. To resolve this issue, the de-
tailed relationships of Amphimachairodus spp. need
to be better understood.
The relationship between Homotherium and Dino-
bastis (and Xenosmilus) is particularly interesting
(Fig. 2.2, letter I). Traditionally, they are synony-
mized in the genus Homotherium (Turner and Anto
´n
1997). However, early North American homother-
iines such as that from the Delmont Local Fauna,
South Dakota (Martin and Harksen 1974) (c. 2.9–2.6
Ma) differ considerably from contemporary forms in
Eurasia (see, e.g., Ficcarelli 1979), suggesting a long,
separate evolution. In addition, Homotherium and
Dinobastis differ in a number of aspects of their mor-
phology. As an example, the upper canine of Dino-
bastis is smaller than that of Homotherium in
specimens of approximately equal skull size (Werde-
lin and Sardella 2006, plate 1, fig. 1). This is an area
that deserves further in-depth study.
Regardless of which species in the Amphimachair-
odus group is closest to Homotherium, it is nearly
universally acknowledged that there is, broadly con-
ceived, an ancestor–descendant relationship be-
tween the two genera. However, the origins of the
other major Plio-Pleistocene sabretooth lineage, the
Smilodontini (Fig. 2.2, letter L), is much less clear.
This group consists of the genera Megantereon and
Smilodon, which share features such as reduced or
absent serrations on the teeth and extremely long
and relatively mediolaterally broad upper canines
compared to Homotheriini (the latter probably a
plesiomorphic feature). It is tempting to associate
them with the other Miocene sabretooth lineage,
Paramachaerodus (Turner and Anto
´n 1997) (Fig. 2.2,
letter K), but the morphological distance between
that genus and Plio-Pleistocene Smilodontini is con-
siderable and the hypothesized relationship is not
based on any clear synapomorphies. Another ques-
tion germane to this issue is the difference between
Smilodontini and Homotheriini: why is it there and
what does it mean for the functional morphology
and ecology of the respective groups? One answer
would be that the former were closed-habitat taxa
and the latter open-habitat taxa, but can such a sim-
plistic view be maintained? Martin (1980) and Mar-
tin et al. (2000) discuss some of these questions, but
more research needs to be done on the functional
differences between Homotheriini and Smilodon-
tini, and in particular on the latest Miocene species
of Paramachaerodus (P. orientalis and P. maximiliani),
to understand their ecology and feeding behaviour,
and whether these can be directly related to those of
Smilodontini.
The extinction of Homotheriini and Smilodontini
occurs at different times on different continents. In
76 Biology and Conservation of Wild Felids
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

Africa, both Homotherium and Megantereon became
extinct some time before 1.4 Ma (with Dinofelis lin-
gering on another 500 ka). In Europe, Homotherium
became extinct at c. 0.5 Ma (the recent record of a
Late Pleistocene Homotherium from North Sea sedi-
ments [Reumer et al. 2003] needs to be corroborated
by further material before its implications can be
fully assessed) and Megantereon at c. 1 Ma. In North
America, on the other hand, both tribes survive into
the latest Pleistocene, with the last occurrence of
Dinobastis from Friesenhahn Cave (Texas) at
c. 11,000 BP and the last occurrence of Smilodon
from Rancho La Brea (California) at c. 13,000 BP.
We don’t fully understand why these dates differ so
much between continents. The differences may re-
flect the different first appearance datums on each
continent of advanced hominid competitors in suffi-
cient numbers to affect the populations of sabre-
tooths, through direct or indirect competition for
resources. Or they could be the result of major faunal
changes on each continent brought about by human
interference, climatic change, or a combination of
the two. In building and testing these scenarios, it
is also important to consider the conical-toothed
cats and their impact on their sabretooth competi-
tors, for example, the relatively rapid range expan-
sion of lions from Africa through Eurasia during the
Middle–Late Pleistocene (Yamaguchi et al. 2004a),
understanding of which has been hampered by the
poor fossil record of the conical-toothed cats.
Possibly no subject in mammal palaeontology has
been more debated than that of sabretooth feeding
adaptations. How did they use their canines? What
did they feed on? What was their killing behaviour
like? Questions like these have been posed and an-
swered numerous times since sabretooths were first
discovered (see Kitchener et al., Chapter 3, this vol-
ume). To answer these questions, it is important to
realize that this ecomorphology is not restricted to
felids and their carnivoran relatives among nimra-
vids and barbourofelids. The package (with varia-
tions) is also present in some creodonts, an extinct
order of mammals that lived from the Paleocene to
the Miocene (genera Apataelurus and Machaeroides,
Early–Middle Eocene of North America), in marsupials
(genus Thylacosmilus; Miocene–Early Pleistocene of
South America) and in various groups of synapsid
‘reptiles’ of the Late Palaeozoic, for example, Gorgo-
nopsia (Kemp 2004). Despite this, it is among felids,
nimravids, and barbourofelids that the adaptation ap-
pears to have been most successful. Recent work on
early felid sabretooths (Salesa et al. 2003; Anto
´net al.
2004; Salesa et al. 2005) has begun to close the func-
tional gap between sabre-toothed and conical-toothed
cats. This and other lines of evidence, such as the
meandering evolutionary history of Dinofelis from
more sabretooth to less sabretooth and back (Werdelin
and Lewis 2001), suggest that the ecomorphology of
the feeding apparatus in felids is more of a continuum
than a dichotomy. The implications of this for under-
standing the ecology of sabretooths and competition
between sabretooths and conical-toothed cats are in
need of detailed investigation.
One possible implication of the feeding apparatus
of sabre-toothed and conical-toothed cats being on a
continuum is that there may have been more direct
competition between the two groups than previous-
ly thought. Previous models tend to emphasize the
difference, with sabretooths specializing in larger
prey than similar-sized conical-toothed cats. How-
ever, more recent analyses suggest that perhaps the
two groups focused on very similar prey. In Africa,
sabretooths are fairly common fossils and conical-
toothed cats rare until around the time when the
number of fossils of sabretooths decreases (Werdelin
and Lewis 2005). This can be explained if sabretooths
were dominant in the most commonly sampled
habitats and competitively excluded conical-toothed
cats. Support for such an idea can be found at Laetoli.
This site (or at least the Laetolil Member, Upper Beds)
is unique among eastern African sites in not being
near a large body of standing water. It is also unique
among sites in having a large number of fossils of
conical-toothed cats and very few fossils of sabre-
tooths. Further research on competition between
sabretooths and conical-toothed cats is needed, as is
research on the competitive structure of the carni-
vore guild as a whole.
The single most important issue impeding an
increased understanding of the evolution of coni-
cal-toothed cats is the extensive ghost lineage be-
tween the oldest fossil members of the Panthera
lineage and the common ancestor of all Felinae.
Two explanations for this gap in the fossil record
immediately spring to mind: a poor fossil record in
the earliest Pliocene and the possibility that the
Phylogeny and evolution of cats (Felidae) 77
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

Panthera lineage (and the Felinae as a whole) evolved
in an environment that is not conducive to the pro-
cess of fossilization. Both of these factors are un-
doubtedly in play, but it is hard to escape the
impression that pantherines are present in the fossil
record prior to 4 Ma, but that they are misidentified
for as yet unknown reasons. To either identify these
fossils or explain why they have not been found is
the most pressing issue in felid palaeontology and
evolution and without progress here it will not be
possible to move towards a fuller reconciliation of
the fossil record with the molecular evidence for felid
evolution as presented by Johnson et al. (2006b).
The position of Barbourofelidae is, of course, very
uncertain, since there is no consensus at present on
how closely related it is to the Felidae. Here, we have
opted for the view that it split off from the stem
lineage leading to Felidae after Prionodon but before
the evolution of Proailurus (Fig. 2.2, letter N). This
leaves an extensive barbourofelid stem lineage that is
at present entirely unknown.
Evolutionary patterns
The availability of a phylogeny of extant Felidae
makes it possible to consider evolutionary patterns
within the family in the absence of fossils. Such
studies have been attempted in the past (Ortolani
and Caro 1996; Werdelin and Olsson 1997; Ortolani
1999; Mattern and McLennan 2000), but given that
the current phylogeny (Fig. 2.1) is fully resolved and,
we believe, better corroborated than older hypoth-
eses, this work is worth reconsidering. Further, since
the current hypothesis is based on molecular data, it
is possible to study morphological character evolu-
tion without the need to discuss possible circularity
in the results. Some such uses of the phylogeny were
presented by Johnson et al. (2006b) and O’Brien and
Johnson (2007) (intercontinental migrations, ghost
lineage analysis), and we will only briefly present two
further examples of the sort of work that can and
should be done on felid evolution using the phylog-
eny as a baseline. For other examples based on previ-
ously proposed phylogenetic hypotheses, see in
particular Mattern and McLennan (2000).
Werdelin and Olsson (1997) presented a phylogen-
etic study of coat patterns in Felidae using a selection
of then-current phylogenetic hypotheses as the base-
line. Their conclusion was that ‘most transforma-
tions of coat pattern originate from the flecked
pattern, which we consider to be primitive for the
Felidae as a whole’ (Werdelin and Olsson 1997,
p. 399). The current phylogeny has some substantial
differences from the phylogenies used in that study,
so the question arises whether the conclusions hold
up. Fig. 2.12 shows coat pattern mapped on the
current phylogeny. The data are identical to those
in Werdelin and Olsson (1997) except for P. tigris,
which has been recoded from vertical stripes to ro-
settes, as we believe that what appear to be vertical
stripes in the tiger’s coat in reality are enormously
vertically elongated rosettes. This is indicated
through examination of various coat pattern anoma-
lies in tigers and can be more simply seen by holding
up an image of a tiger pelt nearly parallel to one’s line
of sight. One difference from the previous results is
immediately obvious: under the current phylogeny,
the primitive coat pattern for Felidae as a whole is
large blotches. This coat pattern is present in only
two genera: Neofelis, clouded leopards and Pardofelis,
marbled cat. Both are basal within their clades, and
these clades are basal within the family and hence
the primitive condition is reconstructed as large
blotches. Above the node leading to Pardofelis, how-
ever, flecks are primitive as they were in the previous
study. If we consider the number and direction of the
state changes in the cladogram (Fig. 2.13), we can
also see that changes to and from flecks are still the
dominant transformations, though not quite as
dominant as previously thought. Thus, the new phy-
logeny corroborates the main thrust of the results of
Werdelin and Olsson (1997), but also leads to some
modifications of specific parts of their conclusions.
In a second demonstration of possible phylogenetic
reconstructions, we mapped habitat (as open or
closed), activity pattern (diurnal or nocturnal), and
pupil shape (slit-like or rounded in the contracted
state) in all felids. The data are partly from Mattern
and McLennan (2000) and partly original. Many spe-
cies occur in both open and closed habitats and the
mapping reflects this, not showing any clear phyloge-
netic associations of open- or closed-habitat specialists
(Fig. 2.14), although the Panthera and domestic cat
lineages are dominated by open-habitat taxa and
have this habitat reconstructed as primitive for the
78 Biology and Conservation of Wild Felids
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

respective clades. Likewise, there are no clear phyloge-
netic patterns underlying activity patterns in modern
felids (mapping not shown). Rounded pupils, on the
other hand, only occur in three clades, the Panthera-
lineage, where all species except the two Neofelis (A.
Kitchener, personal communication) have rounded
pupils, the Puma-lineage, where all three species have
rounded pupils, and the leopard cat lineage, where the
single species O. manul has rounded pupils.
The question of the occurrence and causes behind
slit-like or rounded pupils has been intermittently
discussed in the literature without a consensus
being reached (see Kitchener et al., Chapter 3, this
volume, for a discussion of some recent research).
One suggestion that has been considered is that
slit-like pupils allow the pupil to be more completely
closed than rounded pupils (Walls 1942). This would
suggest that the former would be more useful in the
Prionodon linsang
Neofelis nebulosa
Panthera tigris
P. uncia
P. pardus
P. leo
P. onca
Pardofelis marmorata
P. badia
P. temmincki
Leptailurus serval
Caracal caracal
C. aurata
Leopardus pardalis
L. wiedii
L. colocolo
L. jacobita
L. tigrinus
L. geoffroyi
L. guigna
Lynx rufus
L. canadensis
L. pardinus
L. lynx
Acinonyx jubatus
Puma concolor
P. yagouaroundi
Felis chaus
F. nigripes
F. silvestris
F. margarita
Otocolobus manul
Prionailurus rubiginosu
s
P. planiceps
P. bengalensis
P. viverrinus
Large blotches
Flecks
Rosettes
Small blotches
Stripes
Uniform
Polymorphic
Figure 2.12 Coat patterns (as labelled) mapped on the phylogeny of extant Felidae.
Phylogeny and evolution of cats (Felidae) 79
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

brighter light of day: that is, slit-like pupils should be
preferentially present in diurnal species. However, a
comparison between the patterns does not corrobor-
ate this idea (not shown). There seems to be no
correlation at all between pupil shape and activity
pattern. However, if we compare habitat and pupil
shape (Fig. 2.14), we find that with the exception of
Puma yagouaroundi, which, if the fossil record of this
clade is taken into account, must be considered sec-
ondarily adapted to closed habitats, rounded pupils
never occur in closed-habitat specialist species. All
the taxa with rounded pupils are either open-habitat
species or occur in a variety of habitats. Further, all
three nodes where there is a change from slit-like to
rounded pupils are also nodes where there is a shift
from closed-to open-habitat preference. What this
means in functional terms is beyond the scope of
this chapter, but the results point to a fruitful avenue
of research. These very tentative results must be cor-
roborated by more in-depth study and statistical test-
ing. More generally, phylogenetically based studies
such as the ones discussed above can direct future
research and provide tests of functional hypotheses
that could otherwise not be investigated due to a lack
of independent data. The existence of a well-corro-
borated phylogeny such as that in Fig. 2.1 is a pow-
erful tool for future research on felid evolution.
Final words
This chapter presents one possible scenario for the
evolution and interrelationships of cats. Some of this
work, such as that which has led to the phylogeny of
Johnson et al. (2006b), is strongly corroborated by
and based on considerable amounts of data. The
fossil record of Felidae is uneven. Some groups,
such as parts of the Machairodontinae, have a fairly
extensive fossil record, while others, such as the line-
age leading to the extant radiation, are much more
1
0–1
1
3
1–2
2–4
0–1
0–1
0, 2
Figure 2.13 The coat pattern transformations implied by the mapping in Fig. 2.12. The majority of transformations
involve flecks (centre pattern). Thus, clockwise from bottom, there are three transformations from flecks to uniform,
one transformation from large blotches to flecks, two to four transformations from flecks to stripes, zero or two (no
reconstruction allows for one) transformations from striped to flecks, one to two transformations from flecks to small
blotches, and zero to one transformations from small blotches to flecks. Remaining reconstructed transformations are
zero to one transformation from large blotches to uniform, one transformation from large blotches to rosettes, and
zero or one transformations from small blotches to stripes. No other transformations are allowed by the phylogeny of extant
Felidae.
80 Biology and Conservation of Wild Felids
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

poorly documented. In no case, however, can the
fossil record be said to be adequate, either in quantity
or quality. Nor can the fossil record of Felidae be said
to have been adequately studied. Some areas, such as
the functional morphology of sabretooths, have
been investigated over and over, while others, such
as the stem lineage of modern cats, have been rela-
tively neglected. Overall, the phylogeny and evolu-
tion of fossil Felidae have been neglected in favour of
studies of their functional morphology and ecology.
Given the limited resources available for this work,
this is understandable, as the latter topics have prov-
en more tractable and have yielded interesting and
significant results. But if our understanding of the
group is to progress, we must try to address such
pressing issues as the fossil record of living cats, the
origins of Smilodontini, and the relationship of Bar-
bourofelidae to Felidae. This will require extending
the work of Johnson et al. (2006b) into the realm of
fossils, by comparing the fossil record with the re-
sults obtained from the phylogeny of extant cats on
aspects such as continental migration (O’Brien and
Johnson 2007), to see if the timing of intercontinen-
tal migrations of fossil cat groups can be matched up
with those postulated for the extant cats based on
phylogeny and geology.
It must be understood that developing a phylo-
geny, or even the simpler task of testing some aspect
of the scenario developed herein, requires more than
a superficial glance at the record and doing a phylo-
genetic analysis of the first few characters that come
to mind. It will require developing new characters
and looking at the fossil record in new ways. If the
fossil record and phylogeny of extant Felidae can be
better integrated, we can expect to develop a signifi-
cantly better understanding of the evolution of this
Prionodon linsang
Neofelis nebulosa
Panthera tigris
P. uncia
P. pardus
P. leo
P. onca
Pardofelis marmorata
P. badia
P. temmincki
Leptailurus serval
Caracal caracal
C. aurata
Leopardus pardalis
L. wiedii
L. colocolo
L. jacobita
L. tigrinus
L. geoffroyi
L. guigna
Lynx rufus
L. canadensis
L. pardinus
L. lynx
Acinonyx jubatus
Puma concolor
P. yagouaroundi
Felis chaus
F. nigripes
F. silvestris
F. margarita
Otocolobus manul
Prionailurus rubiginosus
P. planiceps
P. bengalensis
P. viverrinus
Slit-like pupils
Round pupils
Open habitats
Closed habitats
Figure 2.14 Habitat type preference and pupil shape mapped on the phylogeny of extant Felidae.
Phylogeny and evolution of cats (Felidae) 81
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi

fascinating group and its conditions for existing,
thereby not only enhancing current knowledge, but
also building a better platform for the conservation
of the many endangered species of Felidae today.
Acknowledgements
We would like to thank David Macdonald for
inviting us to contribute to this volume, as well as
for his encouragement and a careful reading of the
manuscript. We thank Mauricio Anto
´n for letting us
use his reconstructions of extinct felids, Ste
´phane
Peigne
´for the photographs of Proailurus, Mikael Ax-
elsson for photographs of felid skulls, Mats Wedin for
help with MacClade, and Margaret Lewis and Susan
Cheyne for reading and commenting on the manu-
script. Lars Werdelin acknowledges support from the
Swedish Research Council. We thank Andrew Lover-
idge, Blaire van Valkenburgh, Andrew Kitchener, and
an anonymous reviewer for their thorough readings
that helped us clarify and correct parts of the text,
making the whole far more accessible to a broad
audience than it otherwise would have been.
82 Biology and Conservation of Wild Felids
OUP CORRECTED PROOF – FINAL, 1/5/2010, SPi