sign in sign up
<<
Home , Aeluroidea, Barbourofelidae, Dinictis, Eupleridae, Felis attica, Hemigalinae

Florida State University Libraries

Electronic Theses, Treatises and Dissertations The Graduate School

2007 Phylogeny and Character Change in the Feloid Jill A. (Jill Alexandra) Holliday

Follow this and additional works at the FSU Digital Library. For more information, please contact [email protected] THE FLORIDA STATE UNIVERSITY

COLLEGE OF ARTS AND SCIENCES

PHYLOGENY AND CHARACTER CHANGE IN THE FELOID CARNIVORA

By

JILL A. HOLLIDAY

A Dissertation submitted to the Department in Biological Science in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Degree Awarded: Spring Semester, 2007

The members of the Committee approve the dissertation of Jill A. Holliday defended on March 21, 2007.

______Scott Steppan Professor Directing Dissertation

______William Parker Outside Committee Member

______Gregory Erickson Committee Member

______Joseph Travis Committee Member

______David Swofford Committee Member

Approved:

______Tim Moerland, Chair, Department of Biological Science

______Dean, College of Arts and Sciences

The office of Graduate Studies has verified and approved the above named committee members.

ii

ACKNOWLEDGEMENTS

I would like to express sincere appreciation to my committee for ongoing discussion and support throughout the course of this project: S. Steppan, G. Erickson, J. Travis, D. Swofford and B. Parker. In addition, I thank J. Albright, J. Burns, M. Reno, K. Rowe and L. VandeVrede for technical discussion and assistance, and I thank J. J. Flynn, L. Werdelin, G. Wesley-Hunt, and A. Goswami for their unselfish encouragement and helpful suggestions regarding this project. I gratefully acknowledge all of those institutions and individuals who provided tissues and samples for study, including the American Museum of Natural History; The Brookfield Zoo; Carnegie Museum of Natural History; Louisiana State Museum; Texas Technical University; L. Heaney, B. Stanley, and S. Goodman, the Field Museum; Candace McCaffery and David Reed, Florida Museum of Natural History; J. Dragoo, Museum of Southwestern Biology; C. Conroy, Museum of Comparative Zoology, UC Berkeley; C. Matthee, Ellerman Museum, South ; Jerry Hooker and Daphne Hills, the British Museum of Natural History; The National Museum of Natural History, Paris; Harvard Museum of Comparative Zoology, Boston; The National Museum of Natural History, Washington, D.C.; the Page Museum, Los Angeles; the Los Angeles County Museum; Museum of Vertebrate Zoology, Berkeley; and the Swedish Museum of Natural History, Stockholm. A variety of funding sources contributed to this study, and support was received from the following: Florida State University Dissertation Research Grant, Florida State University Bennison Memorial Scholarship, Sigma Xi Grants in Aid of Research, The Society of Systematic Biologists, The American Society of Mammalogists, the American Museum of Natural History Collections Study Grant, the American Museum of Natural History Theodore Roosevelt Scholarship, and NSF DDIG # 050-8848 and NSF DEB # 0108450 to Scott Steppan.

iii

TABLE OF CONTENTS

List of Tables...... v

List of Figures ...... vi

Abstract ...... vii

INTRODUCTION...... 1

1. OF HYPERCARNIVORY: THE EFFECT OF SPECIALIZATION ON SUBSEQUENT MORPHOLOGICAL AND TAXONOMIC DIVERSIFICATION ...... 3

Methods...... 5 Results ...... 12 Discussion ...... 13

2. EVOLUTION IN CARNIVORA: IDENTIFYING A MACROEVOLUTIONARY RATCHET...... 27

Methods...... 30 Results ...... 33 Discussion ...... 35 Conclusions ...... 37

3. PHYLOGENY OF THE FELOID CARNIVORA: BALANCED TAXON SAMPLING AND CONCATENATED NUCLEAR GENES PROVIDES RESOLUTION AT DEEPER NODES...... 47

Materials and Methods ...... 49 Results ...... 51 Discussion ...... 52 Conclusions ...... 55

4. A COMBINED EVIDENCE PHYLOGENY OF THE FELOID CARNIVORA: COMPREHENSIVE ANALYSIS OF AND MOLECULAR DATA...... 61

Materials and Methods ...... 64 Results ...... 73 Discussion ...... 75 Conclusions ...... 78

CONCLUSION ...... 87

APPENDICES...... 88

REFERENCES...... 125

BIOGRAPHICAL SKETCH...... 140

iv

LIST OF TABLES

CHAPTER 1.

1. Taxonomic diversity...... 17 2. Disparity ...... 18 3. Frequency of change ...... 19 4. Degree of specialization ...... 20

CHAPTER 2.

5. Generalist forward and reverse changes...... 39 6. Frequency of change for RBL without basal...... 40 7. Frequency of change for RBL with basal...... 41 8. Frequency of change for the complex without basal branches...... 42 9. Frequency of change for the hypercarnivore complex with basal branches...... 43 10. Summary of p values for Relative Blade Length...... 44 11. Summary of p values for the hypercarnivore complex...... 45

CHAPTER 3.

12. PCR primer sequences for genes used in this study ...... 56 13. PCR programs...... 57

v

LIST OF FIGURES

CHAPTER 1.

1. indicating increasing specialization ...... 21 2. Representative cranial and dental measurements ...... 22 3. Average disparity ...... 23 4. Average frequency of change...... 24 5. Position of taxa in morphospace based on degree of specialization ...... 25 6. Disparity for different degrees of specialization to hypercarnivory ...... 26

CHAPTER 2.

7. Method for calculating frequency of change...... 46

CHAPTER 3.

8. Cladogram illustrating current understanding of relationships within Feloidea ...... 58 9. Phylogram of the feloid Carnivora...... 59 10. Maximum likelihood phylograms for each gene ...... 60

CHAPTER 4.

11. Bayesian phylogeny for recent taxa only...... 80 12. Constrained analysis of morphological data. Recent taxa only...... 81 13. Constrained analysis of morphological data. Taxa at least 50% complete...... 82 14. Constrained analysis of morphological data. Taxa at least 20% complete...... 83 15. Constrained analysis of the ...... 84 16. Constrained analysis of the (Herpestidae/)/Hyaenidae ...... 85 17. Constrained analysis of the ...... 86

vi

ABSTRACT

This study presents the results of dissertation research performed by Jill Holliday, and includes a study of the effects of specialization to a dental/dietary morphotype, that of the hypercarnivore, on subsequent morphological and taxonomic diversification. This work demonstrates that morphological specialization to hypercarnivory does limit subsequent morphological diversity (disparity) and also reduces frequency of morphological change over the course of a lineage, but has no effect on subsequent taxonomic diversity. Additionally a test of the possibility of biased morphological character evolution as taxa evolve a more hypercarnivorous phenotype indicates that specialists are not subject to strong directional selection, but are instead unable to reverse to a more generalized condition or even move into alternative open niche space, which strongly implies the effects of a functional constraint. To provide a more robust and detailed phylogeny than is currently available, a molecular and morphological dataset was compiled for the feloid Carnivora, a group in which hypercarnivory has evolved at least three times. The molecular dataset is composed of three nuclear and one mitochondrial gene and represents over 5000 base pairs of data for 39 ingroup taxa. This dataset was analyzed under maximum likelihood and Bayesian models, and represents the most robust, thoroughly sampled phylogeny yet available for the feloid Carnivora. However, since the focus of this research is to evaluate character evolution, 103 morphological characters were added to the dataset so that fossil material, particularly ancestral feloids, could also be placed in a phylogenetic context. This combined evidence phylogeny was analyzed using Bayesian inference and with parsimony using a backbone constraint tree based on the previously obtained molecular phylogeny. Both the molecular-only and the combined evidence phylogenies significantly clarify relationships among the feloid families, and also establish the early pattern of divergences within this group. This study establishes that the monotypic family Nandiinidae is the sister to all other extant feloids, while the family Viverridae is the sister to a comprised of (Felidae)((Hyaenidae)(Herpestidae/Eupleridae))). The combined evidence tree also allows placement of a number of feloid that occupy key positions at the base of the radiation of extant families. Thus, the is an early member of Hyaenidae(Herpestidae/Eupleridae), while Plioviverrops is an early herpestid. Finally, the primitive feloids , Paleoprionodon, and Haplogale do not position within any of any of the extant families, and instead comprise a paraphyletic grouping at the base of the feloid tree. With these results, the combined evidence phylogeny will enable a more accurate designation of character polarities and estimation of ancestral conditions, which will in turn facilitate additional research in these groups. These phylogenies are the best supported, most thoroughly sampled trees yet produced for the feloid Carnivora, and represent a significant contribution to feloid phylogenetics.

vii

INTRODUCTION

The mammalian order Carnivora is a diverse group of mostly carnivorous that share the synapomorphy of the carnassial pair, the first upper premolar and the fourth lower molar, which are modified for the slicing of meat. Despite this shared characteristic, the carnassials do vary depending on diet, and Carnivora includes taxa whose diets range from frugivory to omnivory to hypercarnivory. As the diet changes, the evolves as well, and modifications to the dentition are particularly apparent in durophagous taxa such as the , with enlarged premolars, and also in , whose adaptations to a high meat diet has led them to elongate the carnassial shearing blade and reduce or even lose the post-carnassial molars. Hypercarnivores are a particularly interesting group because this morphotype has evolved repeatedly within Carnivora, including taxa as diverse as Cryptoprocta ferox, a euplerid from Madagascar, and the genus Mustela within the family , the modern felids () and also within several extinct groups of and canids. A particularly noticeable example of hypercarnivory is the sabertoothed , an extinct clade of -like taxa whose and dentitions were highly specialized. It should be emphasized, however, that the characters of interest are not the canines (the sabers of sabertoothed taxa) but the carnassial complex used for meat-shearing. Extreme phenotypic specialization to specific ecological or dietary niches has been suggested to cause significant limitation on subsequent morphological character evolution, likely as either a result of strong stabilizing selection to maintain that specialization or, alternatively, due to strong directional selection towards ever-increasing specialization. Additionally, it has been demonstrated by several workers that when certain morphological features are lost —either as a result of specialization or due to some other process —it is particularly difficult, if not impossible, to regain those structures. This is commonly known Dollo’s Law, and in a simple sense describes the inability to regain (re-evolve) a complex structure. In an analysis of morphological evolution in canids and other groups Van Valkenburgh (1991) suggested that extreme specialization to hypercarnivory may act to limit subsequent morphological evolution in with this phenotype. Chapter 1 of this study directly tests this hypothesis through the use of replicated sister group comparisons, evaluating the effects of specialization to hypercarnivory on subsequent morphological diversity for six sets of hypercarnivores and their sister taxa. Through use of multivariate statistics that evaluate occupied morphospace, and character mapping methods that are used to calculate changes in character states on a phylogeny, it is demonstrated that hypercarnivory does have a significant effect on subsequent morphological but not taxonomic diversity. Given these findings, the underlying mechanisms that may cause this limitation are evaluated in Chapter 2. Possibilities include the competing hypotheses of stabilizing selection, directional selection, and constraint. As with the disparity analyses, replicated sister group comparisons and character mapping methods are employed. This time, however, rates of morphological change towards and away from specialized character states are compared both within and between hypercarnivore clades and their sister groups. Results indicate that rates of forward change are not different between hypercarnivores and their sister taxa, but rates of change away from the specialization are very much reduced in hypercarnivores, and, in conjunction with a consideration of the ecologies of the taxa, strongly suggests the presence of a functional constraint. The obvious next step in understanding the evolution and effects of hypercarnivory is to trace the evolution of the condition from root to tip, and to seek ecological or morphological correlates that may bias taxa towards this phenotype. However, such analyses are restricted due to a lack of well-sampled, robust phylogenies for the clades of interest. Under these circumstances, but with the goal of obtaining a comprehensive phylogeny for as many groups as possible,

1

Chapters 3 and 4 focus on the phylogenetic relationships between the feloid Carnivorans, represented by the families Felidae (cats), Hyaenidae (hyaenas), Viverridae ( and genets), Eupleridae (Malagasy ), Herpestidae () and the monotypic family Nandiinidae, the African palm . Feloids exhibit a variety of ecological and dietary habits, including frugivorous, omnivorous, durophagous, and hypercarnivorous taxa, and fossil material, particularly for the Felidae and Hyaenidae, is relatively common. Because three of the six constituent families have independently evolved hypercarnivory, phylogenetic analysis of this clade, in a framework that includes both recent and fossil material, offers the best potential information gain relative to relative to the data that must be collected to obtain a comprehensive phylogeny. In Chapter 3, molecular data for 39 of feloid representing all of the constituent families was obtained for the nuclear genes RAG1, GHR, and c-myc. Cytochrome b data was available on Genbank, and this was added to the dataset. Phylogenies were analyzed in both maximum likelihood and Bayesian frameworks, and the signal from each of the genes was evaluated. This phylogeny helps to establish the pattern of diversification between the four feloid families, and consequently helps to clarify character polarities as well. It is also the most thoroughly sampled, well-supported phylogeny available for the recent feloid Carnivorans, and will be useful for many workers studying ecological and evolutionary patterns. Although molecular analysis of the recent feloids provides valuable information on its own, the question of the taxonomic position of early feloids and the more basal members of each of the four families must still be addressed. Inclusion of these taxa is key in reconstructing ancestral character states and, consequently, in establishing the primitive morphological condition(s) from which hypercarnivory may have evolved. In Chapter 4, 103 morphological characters, representing data from over 1000 fossil and recent specimens and over 183 operational taxonomic units, were added to the molecular data matrix for a combined evidence analysis of both recent and fossil taxa. Data was analyzed using a constraint tree (the previously obtained molecular phylogeny) as well as a combined Bayesian analysis of both molecular and morphological data. The effects of missing data and of homoplasious signal were taken into account and data was analyzed in subsets that included family level analyses and also subsets that removed taxa with specific amounts of missing data (e.g. taxa less than 20% complete, taxa less than 50% complete). The results of the combined evidence analyses confirmed placement for some stem taxa, but also helped to establish the phylogenetic position of an equal number of questionable species. These analyses show that Viverridae is paraphyletic, that Herpestides is a member of the Hyaenid(Euplerid/Herpestid clade) and Plioviverrops is a herpestid. As anticipated, these results help to establish character polarities, and improve ancestral state estimations. The information obtained from these phylogenies will vastly improve future study of character evolution, while the trees themselves represent a significant advance in available data for this clade.

2

CHAPTER 1 EVOLUTION OF HYPERCARNIVORY: THE EFFECT OF SPECIALIZATION ON SUBSEQUENT MORPHOLOGICAL AND TAXONOMIC DIVERSIFICATION

Understanding the effect of specialization on subsequent taxonomic and morphological evolution is fundamentally important to understanding the tempo and mode of evolution and the role of adaptation in macroevolution (e.g. Futuyma and Moreno 1988, Janz et al. 2001). However, there is little consensus as to how specialization affects diversity: does it act to increase or decrease rates of cladogenesis? How does specialization affect probability of ? Does it constrain further adaptation? Certainly, much attention has been given to the possibility that particular specializations may promote taxonomic diversification, that is, a morphological or behavioral specialization may act as a key innovation, leading to an increase in rates of cladogenesis (Liem 1973; Mitter et al. 1988; Farrell et al. 1991; Hodges and Arnold 1995; de Queiroz 1999; Dodd et al. 1999), but empirical studies have produced contrasting results. Some workers find that specialization increases taxonomic diversity (e.g. Liem 1973; Mitter et al. 1988; Farrell et al. 1991; Hodges and Arnold 1995; de Queiroz 1998; Dodd et al. 1999), some report the opposite pattern (e.g. Price and Carr 2000), and others find no effect at all (e.g. Wiegmann et al. 1993; de Queiroz 1999; Janz et al. 2001). Despite numerous studies of the effect of specialization on taxonomic diversification, studies of its effects on subsequent morphological diversity (disparity) are few. In a theoretical context, a number of workers have suggested that possession of certain morphological character states may reduce the ability to attain certain other states (Lauder 1981; Smith et al. 1985; Emerson 1988; Futuyma and Moreno 1988; Werdelin 1996; Donoghue and Ree 2000; Wagner and Schwenk 2000), implying that the subsequent evolutionary trajectories of some specialized taxa may be limited. At its extreme, therefore, specialization could act as a dead end (Moran 1988, Janz et al. 2001), limiting morphological diversification and potentially reducing rates of cladogenesis or, alternatively, increasing extinction rates as specialized taxa reduce their ability to adapt to changing conditions. However, few studies have directly identified and tested effects of specialization on subsequent phenotypic change (but see Liem 1973; Moran 1988; Warheit et al. 1999). I determined the effect of dental and cranial specialization to a meat-only diet, hypercarnivory, on subsequent morphological and taxonomic diversity in mammalian carnivores. I used an explicitly phylogenetic approach and applied it to repeated convergences on hypercarnivory, increasing statistical power by evaluating results from multiple sister groups. I specifically tested the hypotheses that (1) hypercarnivores have lower taxonomic and dental morphological diversity than do their sister groups and (2) increasing specialization leads to lower morphological diversity. To test these hypotheses, I quantified and compared diversity between hypercarnivore clades and their primitively nonhypercarnivorous sister groups. The advantage of using sister groups is that both groups (when their stem lineages are included) have by definition had equal time to diversify. I used both species counts and the methods of Slowinski and Guyer (1993) to assess taxonomic diversity. I compared morphological diversities (disparities) by comparing variances of factor scores obtained from principal-components analysis (Foote 1993; Wills et al. 1994) and by comparing the differences in average frequency of character change between categories (Sanderson 1993). Finally, I assigned “degrees of specialization” to hypercarnivores using discriminant function analysis and compared the disparity values of different levels of specialization.

3

The Order Carnivora

The order Carnivora is composed of 13 extant and two extinct families of meat-eating mammals. The diagnostic character for Carnivora is the carnassial pair, the fourth upper premolar and first lower molar, which in this group have been modified as shearing blades for effective slicing of meat. Although the shearing carnassials are a synapomorphy for this group, members of Carnivora, hereafter called carnivorans, have diversified to occupy a wide range of ecological niches, and include highly carnivorous clades such as cats (Felidae) and and (Mustelidae), generalists like the and (), like the (Herpestidae), like the (Ursidae) and (), and strict such as the . Variation in ecology is strongly reflected in the dentition (VanValkenburgh 1989), so a more omnivorous/frugivorous diet is accompanied by a relative increase in grinding surfaces while a more highly carnivorous diet is reflected by a relative decrease in grinding surfaces and an increase in shearing edges. Because of the tight correlation between dentition and ecology, dental characters can be used effectively to infer aspects of the diet or ecological niche. Van Valkenburgh (1988, 1989) showed that variables including relative blade length, canine-tooth shape, premolar size and shape, and grinding area of the lower molars distinguished between dietary types in extant carnivores. She compared guild compositions of carnivoran communities, concluding that each guild comprised a broadly similar set of morphotypes occupying a limited number of ecological niches (Van Valkenburgh 1988, 1989). There is thus a substantial overlap in certain regions of morphospace (Crusafont-Pairo and Truyols-Santonja 1956; Radinsky 1982; Van Valkenburgh 1988, 1989) resulting from convergence on similar ecomorphological types, including meat- specialists, -crackers/, omnivores, and generalists (Van Valkenburgh 1988; Werdelin 1996). Such iterative evolution produces natural replicates and is conducive for comparative study. Of the recognized carnivoran ecomorphs, the niche of the meat specialist, or hypercarnivore, is associated with a diet comprising more than 70% meat, in contrast to the generalist (Van Valkenburgh 1988, 1989), which may eat 50–60% meat with vegetable matter and invertebrates making up the remainder of the diet. Ecological specialization to hypercarnivory is associated morphologically with specific changes in the and dentition that include a relative lengthening of the shearing edges, composed of the trigon of the upper fourth premolar and the trigonid of the lower first molar, and reduction or loss of the postcarnassial dentition, the second and third lower molars and first and second upper molars, teeth used for chewing or grinding food (Van Valkenburgh 1989; Hunt 1998). The facial portion of the skull frequently shortens as well, an alteration thought related to maintaining high bite force (Van Valkenburgh and Ruff 1987; Radinsky 1981a,b; Biknevicius and Van Valkenburgh 1996). While the absence of dietary data for many fossil taxa suggests that the term “hypercarnivore-morph” may be more appropriate, in this paper those taxa with morphological characteristics consistent with a hypercarnivorous diet will be called “hypercarnivores.” Figure 1A illustrates a generalized carnivoran with a “typical” tooth formula; individual cusps are labeled. Figure 1B–D illustrates hypercarnivorous modifications in order of increasing specialization. Certain extant and extinct members of such diverse lineages as mustelids (weasels and ), viverrids (civets and genets), canids (dogs and foxes), hyaenids (hyaenas), amphicyonids (extinct -dogs), and ursids (bears) have all evolved phenotypes characteristic of hypercarnivory (Van Valkenburgh 1991; Biknevicius and Van Valkenburgh 1996; Werdelin 1996), although the most extreme cases appear to be in the families Felidae (cats) and Nimravidae (extinct, noncat saber-tooths). Taxa trending toward hypercarnivory are frequently referred to as “cat-like” (Martin 1989; Hunt 1996, 1998; Baskin 1998), a descriptor that reflects the distinctive adaptations of felids for this niche. In a study of evolution of hypercarnivory in the family Canidae, Van Valkenburgh (1991) commented on the low apparent variability in the cranial and dental morphologies of felids and

4

nimravids relative to canids, and suggested that this was possibly due to the extreme specialization to hypercarnivory in the former two groups. Lack of variation in felid craniodental characteristics has been noted qualitatively by many authors (e.g. Radinsky 1981a,b; Flynn et al. 1988), but few have attempted to quantify this phenomenon or to ascertain cause and effect. Quantitative evaluations of felid diversity have been generally limited to within family or genus (e.g. Glass and Martin 1978; Werdelin 1983; Kieser and Groeneveld 1991; O’Regan 2002) or, if among families, have dealt primarily with relative positioning of groups in morphospace (e.g. Glass and Martin 1978; Radinsky 1981a,b, 1982; Werdelin 1983; Van Valkenburgh 1991) or body-size correlates (Van Valkenburgh 1990; Gittleman and Purvis 1998; Gardezi and da Silva 1999). There are, of course, a variety of reasons why any particular clade might not exhibit certain morphologies, including lack of genetic variation, functional constraint, stabilizing selection, or competition (Smith et al. 1985; Brooks and McLennan 1991). Additional causes may include intrinsically low rates of evolution or a recent rapid radiation (Schluter 2000), either of which might suggest a pattern of constraint but actually reflect a lack of time. Another possibility is sampling bias, where alternative morphotypes may exist but occur in geographic areas where sampling is rare or nonexistent. Any study of the evolution of a character therefore benefits from the inclusion of as many different groups as possible that have evolved the trait of interest. I evaluated six separate clades of hypercarnivorous taxa for which phylogenies are available in the literature. By including a variety of groups, I was able to mitigate the effects of phylogeny and thus evaluate the effects of the specialization itself. Furthermore, because different hypercarnivorous taxa exhibit varying degrees of specialization, I could also assess the effects of increasing specialization on character change and morphological diversity.

METHODS

Definition Because identification of a hypercarnivorous taxon is partly a subjective decision, opinions may vary between workers. In a broad sense, the designation “hypercarnivore” has been used to describe taxa that have increased the slicing component of the dentition relative to the grinding component (Van Valkenburgh 1991). However, taxa that fit this overall categorization can be further subdivided based on relative robustness (widening) of the premolars (Van Valkenburgh1991). In combination with the features of elongated blade, reduced postcarnassial teeth, and a shortened face, normal sized or narrowed premolars produces the “cat-like” phenotype. In contrast, relative broadening of the premolars appears to be an alternate trajectory that leads not to a truly cat-like condition but towards more hyaena-like (bone-cracking) characteristics (Van Valkenburgh 1991). While end members of these groupings (e.g. felids vs. hyaenines) are easily differentiated, gradations between groups can be subtle, as can be the difference between a generalist with hypercarnivorous tendencies and a hypercarnivore (e.g. between some ancestors and descendants). To reduce the necessity for arbitrary judgments, for this study I established a minimum definition of a hypercarnivore in order to more objectively differentiate between cat-like, hyaena-like and transitional forms. The following combination of characteristics was therefore considered minimally diagnostic when evaluating putatively hypercarnivorous taxa for inclusion in this study: trigonid not less than 70% of the length of the m1, width of the 4th lower premolar not greater than 60% of its length, entoconid and hypoconid unequal. Given the above qualifications, it is important to note that hypercarnivorous clades in this study were recognized on the basis of being basally hypercarnivorous. Therefore, if the first two branches for a given clade were hypercarnivorous, the entire group was considered so, because hypercarnivory was established as the ancestral condition. Any evolution of the phenotype subsequent to that ancestral condition was then considered part of the total diversity of that clade.

5

Taxa and Phylogenies Hypercarnivorous taxa were initially identified through literature searches for taxa described as hypercarnivorous or highly predaceous. Diet (where known) and dental formula were also taken into account. Besides all members of the families Felidae and Nimravidae, taxa described by various workers as hypercarnivorous include the hyaenid genera Chasmaporthetes, Hyaenictis, and Lycyaena (Werdelin and Solounias 1991), members of the Simpsonian subfamily (now recognized as paraphyletic), especially the genus Mustela (Ewer 1973; King 1989), the viverrid Cryptoprocta ferox (Wozencraft 1984, 1989; Werdelin 1996), and hesperocyonine canids of the genera Enhydrocyon, Ectopocynus, and Parenhydrocyon (Van Valkenburgh 1991; Wang 1994). Additional taxa identified include the “paleomustelids” Megalictis and Oligobunis (Baskin 1998), as well as certain borophagine canids including Euoplocyon, Epicyon, Osteoborus, and Borophagus (Wang et al. 1999), the amphicyonids Daphoenictis (Martin 1989), Temnocyon, and Mammocyon (Van Valkenburgh 1991, 1999), and the ursids Hemicyon johnhenryi (Van Valkenburgh 1991), Cephalogale, and Phoberocyon (Van Valkenburgh 1999). A literature search for species-level, character-based phylogenies for these taxa and their sister groups produced mixed results. In some cases, character-based phylogenies are not available and these taxa were consequently excluded from study (e.g. amphicyonids, ursids, and paleomustelids). In those cases where multiple phylogenies were available, I critically examined the possibilities and chose the better-supported tree based on criteria that included use of a data matrix, type and amount of evidence (molecular vs. morphological, kind and number of morphological or molecular characters, appropriateness of gene or genes used), and congruence with alternative phylogenies. Sister groups were identified from available higher-level analyses, however, in several cases (e.g. Mustela, Felidae), there is significant disagreement regarding the appropriate sister taxon. When a definitive sister group could not be determined, analyses were repeated with several alternate sister groups. Two groups that contain hypercarnivorous taxa, the borophagine canids and the “paleomustelids,” were not included in diversity comparisons but were included in degree-of-specialization analyses. Borophagines, which trend strongly toward a bone-crushing phenotype, did not meet the working definition of hypercarnivores and were consequently excluded from sister group comparisons. However, the distinctive bone-crushing modifications of the group were useful in determining relative positioning in morphospace and in assigning degrees of specialization for other taxa being evaluated. In addition, there is a substantial range of variation within the “bone-crushing” specialization of borophagines, and although sampling was incomplete, the placement of specific taxa such as the derived Euoplocyon was of general interest. Paleomustelids did not meet the criteria set out for phylogenies (none available was based on a data matrix), but because these taxa have been repeatedly described as very specialized to the hypercarnivore niche (Baskin 1998), their position in morphospace relative to other hypercarnivores was of interest. Six sets of sister groups did meet the criteria for inclusion in this study, in that all met the minimum requirements for a hypercarnivore and a species-level phylogeny was available. These sister group sets are the clades Felidae/Hyaenidae, Mustela/--Pteronura-- Enhydra--Amblonyx-, Philotrox-Sunkahetanka-Enhydrocyon/Cynodesmus, Cryptoprocta/-, Chasmaporthetes-Lycyaena-Hyaenictis/Palinhyaena-Ikelohyaena- Belbus-Leecyaena(Hyaena)-Parahyaena-Hyaena--Pachycrocuta-Adcrocuta-Crocuta (hereafter designated Palinhyaena-Crocuta [Werdelin and Solounias 1991]), and Nimravidae/. Phylogenies used are shown in Appendix A.

6

Sister Groups

Felidae/Hyaenidae Relationships among the feliform families—Felidae, Hyaenidae, Viverridae, and Herpestidae—have been notoriously difficult to ascertain. The hypercarnivorous family Felidae is most commonly recognized as sister to the family Hyaenidae (Wozencraft 1989; Wyss and Flynn 1993; Bininda-Emonds et al. 1999), although other workers have found support for a sister group relationship with the //generalist group Viverridae (Hunt 1987; Hunt and Tedford 1993) or all other feliforms (Flynn and Nedbal 1998). Use of Hyaenidae may be considered a relatively conservative comparison, since Hyaenidae has less diversity than either of the alternative sister groups (Flynn et al. 1988). However, it should be noted that a hypercarnivorous clade (Chasmaporthetes-Lycyaena-Hyaenictis) is nested within the hyaenids as well. Because of the lack of consensus regarding the appropriate sister taxon, I performed two comparisons for Felidae, one against Hyaenidae and one against Viverridae. Within Felidae, the relationships among the various genera and species have likewise been problematic. The phylogeny I used is a composite based on Mattern and McLennan (2000) for crown-group felids and Neff (1982), Turner and Anton (1997), and Martin (1998a) for machairodontines. General consensus for the ancestry of felids leads from , which exhibits a trenchant talonid and three lower molars, to , which has a much reduced talonid and a reduced m2, to the clade comprising + , which has lost the m2 as well as the talonid and has devoted the entire lower carnassial to slicing (see Ginsburg 1983; Hunt 1996, 1998; Martin 1998b). The species-level phylogeny for Hyaenidae is from Werdelin and Solounias (1991).

Chasmaporthetes-Lycyaena-Hyaenictis/Palinhyaena-Crocuta Within Hyaenidae, tendencies toward increased carnivory first appear in generalist forms such as Ictitherium, Thalassictis, Hyaenotherium, and Hyaenictitherium. Recognizably hypercarnivorous taxa are present in the clade comprised of the genera Chasmaporthetes- Lycyaena-Hyaenictis. These taxa have been described as cursorial and somewhat “cat-like” (Werdelin and Solounias 1991). Their sister clade paralleled the development of certain of these hypercarnivorous characteristics in the evolution of a trenchant heel and loss of the m2 in some taxa (Werdelin and Solounias 1991, 1996), but members of the Palinhyaena-Crocuta clade trend strongly toward bone-cracking modifications (e.g. premolar width >60% of premolar length), and extant hyaenids in this group are known to occupy a scavenging/bone-cracking niche (Ewer 1973; Nowak 1999) in contrast to the apparently highly predaceous tendencies of Chasmaporthetes- Lycyaena-Hyaenictis. I follow Werdelin and Solounias (1991) in recognizing these groups as monophyletic with distinctly different ecologies, acknowledging that the taxa in the Palinhyaena- Crocuta clade have become specialists in their own right.

Philotrox-Sunkahetanka-Enhydrocyon/Cynodesmus Within the hesperocyonine canids, the clade identified as hypercarnivorous comprises the genera Enhydrocyon, Philotrox, and Sunkahetanka. The four species of Enhydrocyon are clearly hypercarnivorous relative to earlier hesperocyonines—Enhydrocyon crassidens has been described as the most derived hesperocyonine for this niche. However, Philotrox and Sunkahetanka also exhibit modifications characteristic of hypercarnivory, along with other characteristics consistent with bone-crushing habits. In Wang's (1994) cladogram, Sunkahetanka and Philotrox are successive outgroups to Enhydrocyon, thus establishing hypercarnivory as the basal condition for this clade. I therefore included these taxa in a hypercarnivorous clade and contrast them to a sister group composed of the two-species nonhypercarnivorous genus Cynodesmus. Although the immediate outgroup to this Philotrox-Sunkahetanka- Enhydrocyon/Cynodesmus clade, Mesocyon, has also been described by some workers (Van Valkenburgh 1991; Wang 1994) as hypercarnivorous, its tendencies are very slight, and it does

7

not meet the criterion: Mesocyon retains a strongly basined talonid and well-developed postcarnassial teeth (m2 and m3) and has a relative blade length of less than 70%.

Cryptoprocta/Eupleres-Fossa Cryptoprocta is a monotypic genus in the family Viverridae (Wozencraft 1984) whose phylogenetic affinities have been problematic. There is ongoing work regarding the relationships of viverrids (e.g. Veron and Catzeflis 1993; Veron 1995; Veron and Heard 1999), however, these phylogenies have limited taxon and character sampling and are not adequate for these purposes. The most robust phylogeny for viverrids presently available, and that based on the widest sampling, is that of Wozencraft (1984), wherein the two taxa Eupleres goudotii and Fossa fossana are the sister group to Cryptoprocta ferox.

Mustela/Galictis-Ictonyx-Pteronura-Lontra-Enhydra-Lutra-Amblonyx-Aonyx. Mustela is a highly predaceous genus of small carnivores in the family Mustelidae (Ewer 1973; King 1989). Despite a great deal of recent attention that has included both molecular and morphological analyses (Bryant et al. 1993; Masuda and Yoshida 1994; Dragoo and Honeycutt 1997; Koepfli and Wayne 1998), intrafamilial relationships remain contentious. As a result, I used several alternate phylogenies (Bryant et al. 1993; Dragoo and Honeycutt 1997; Koepfli and Wayne in press) and evaluated results for each group in turn.

Nimravidae/Aeluroidea In contrast to the large number of phylogenetic hypotheses proposed for extant Mustelidae, the extinct saber-toothed family, Nimravidae, is relatively impoverished. Of the phylogenies available, only those of Bryant (1996) for Nimravinae and Geraads and Gulec (1997) for Barbourofelinae meet the criteria for inclusion. I grafted these nonoverlapping phylogenies together to create a single composite tree for use in this study. The relationship of Nimravidae to other carnivoran taxa is also not well established, and various workers place nimravids basal to all of Carnivora (Neff 1983), Canidae (Flynn et al. 1988), sister to Felidae (Martin 1980) and (Baskin 1981). Bryant (1991) and Wyss and Flynn (1993) evaluated the various hypotheses and attempted to obtain a better resolution by incorporating additional evidence. Both concluded that the best (if weakly) supported hypothesis is that Nimravidae is sister to the aeluroid carnivorans, an opinion followed here.

Taxonomic Diversity

Several methods are available for comparison of taxonomic diversity and determination of whether rates of cladogenesis have been affected by a given trait. The simplest is a binomial sign test (Sokal and Rohlf 1994), which allows direct comparison of species diversity between sister groups. Species are counted and the group (hypercarnivorous vs. nonhypercarnivorous) with more species receives a plus sign; the group with fewer species receives a minus. The numbers of signs across all groups under study are then contrasted under a null hypothesis of no significant difference. A more complex alternative is that set forth by Slowinski and Guyer (1993). Their method, which incorporates a model of random speciation and extinction, uses ’s combined probability test (Sokal and Rohlf 1994) to determine whether certain traits have caused significantly higher diversity. This approach has been applied in evaluations of species diversity for both plants (Hodges and Arnold 1995; Dodd et al. 1999; Smith 2001) and (Gardezi and da Silva 1999). I obtained species counts from the primary literature (see Table 1 for references) for all six sister group pairs under study and applied both tests against the null hypothesis that specialization to hypercarnivory had no effect on the subsequent diversification of an affected clade. The most common alternative hypothesis is that specialization to hypercarnivory, as with

8

many other kinds of resource specialization, should reduce subsequent cladogenesis. However, this may not be the case. Using matrix representation to create supertrees for all extant carnivoran taxa, Bininda-Emonds et al. (1999) tested for adaptive radiations under the methods of Nee et al. (1995). Contrary to the argument that specialization should negatively affect cladogenesis, their results suggested that both Mustela and Felidae may have undergone more speciation events than would be expected by chance (Bininda-Emonds et al. 1999), although, for Felidae at least, this result may be an artifact of a high rate of extinction in the sister group (Hyaenidae). Because the effects of specialization on taxonomic diversity are uncertain, I performed one-tailed tests in the direction of decreased species diversity and then repeated the tests in the direction of increased species diversity.

Morphological Diversity

Data Collection Cranial and dental material in the collections of The American Museum of Natural History, the Field Museum, the Florida Museum of Natural History, and the Natural History Museum, London, was measured for thisstudy. Appendix B lists species name, collection number, and museum for each specimen. Measurements were taken to the nearest 0.01 mm with digital calipers. Where only a single specimen was available for a species, measurements were repeated 2–3 times, and the mean taken of those measurements. Where multiple specimens were available, 2–4 specimens were measured, and those measurements used to obtain a mean for the species. Where information regarding sex was available and species were known to be sexually dimorphic (e.g. Mustela, some felids), only males were included in the data set.

Measurements I measured 290 specimens for the following distances (Fig. 2): jaw length (JL), tooth- row length (TRL), zygomatic-arch width (ZAW), length of P4 (lP4), length of M1 (lM1), length of M2 (lM2), length of p3 (lp3), width of p3 (wp3), length of p4 (lp4), width of p4 (wp4), length of m1 (lm1), width of m1 (wm1), length of m2 (lm2), length of m3 (lm3), trigonid length (m1 trigonid), height of ascending ramus (HAR), distance from condyloid process to coronoid process (MAT), distance from condyle to front of masseteric fossa (MFL), and distance from carnassial notch to condyle (COM1). The variables JL, TRL, ZAW, MAT, MFL, and COM1 are from Radinsky (1981a,b). For hesperocyonine canids, I used the published data of Wang (1994). The following eight shape or proportion variables were derived from the original measurements. These variables were used for both disparity analyses and character mapping, and are based in part on those of Van Valkenburgh (1988, 1989): relative blade length (RBL, defined as trigonid/lm1), grinding surface length relative to tooth-row length (GSL/TRL, GSL defined as lp3+lp4+lm1+lm2+lm3–trigonid), shape of the m1 (m1 shape, wm1/lm1), shape of the p4 (p4 shape, wp4/lp4), shape of the p3 (p3 shape, wp3/lp3), ratio of lM1 to lP4 (M1/P4), m1 length relative to tooth-row length (lm1/TRL), and grinding surface length relative to m1 length (lm1/GSL). Of these, the variables RBL and p4 shape have been shown previously to differentiate effectively between dietary types in carnivorans (Van Valkenburgh 1988, 1989). Van Valkenburgh (1988, 1989) also used an area measure, TGA (total grinding area), which I have modified to a linear measure, GSL. Like TGA, GSL denotes the relative amount of grinding surfaces in the tooth row. Relative blade length is generally used as an indicator of highly carnivorous taxa, since hypercarnivores increase the size of the trigonid relative to the length of the entire m1. Shape of the p3 and p4 are both indicative of bone in the diet, since wider premolars indicate a more durophagous dietary niche in taxa that have reduced the postcarnassial dentition (Van Valkenburgh 1988, 1989; Werdelin 1989; Werdelin and Solounias 1991). As mentioned above, widening of the premolars in conjunction with elongation of the shearing blade can be considered an alternate trajectory for hypercarnivorous specialization. M1/P4 is a partial

9

measure of the postcarnassial dentition and therefore an indicator of the relative amount of postcarnassial surface area; m1 shape, estimated by dividing the width of the m1 by its length, is representative of an emphasis on slicing as the tooth narrows. GSL/TRL indicates the proportion of the teeth in the jaw not devoted to slicing; GSL is standardized by the upper tooth row because carnivoran jaw lengths are independent of skull length and neither is a reliable standard at a level higher than family (Van Valkenburgh 1990). lm1/GSL likewise indicates the proportion of postcanine tooth surface area taken up by the lower carnassial. lm1/TRL is essentially a measure of the length of the face standardized to body size (as indicated by m1 length; Gingerich 1974; Van Valkenburgh 1988; Werdelin and Solounias 1991): the face tends to shorten as taxa become more highly carnivorous and the position of the carnassials is altered to maintain high bite force (Radinsky 1981a,b; Van Valkenburgh and Ruff 1987; Biknevicius and Van Valkenburgh 1996). In addition to the above eight characters, the following 19 shape or proportion measures were used for character mapping but not multivariate analyses: P4 shape, blade/GSL, p3/m1, p4/m1, m1/JL, MFL/JL, MAT/HAR, MFL/COM1, m1/MAT, TRL/ZAW, ZAW/JL, m1/ZAW, (M1+M2)/P4, MAM/HAR, and (discrete characters) shape of the m1 talonid basin (Van Valkenburgh 1988), position of carnassial (Van Valkenburgh 1988), position of P4 protocone, shape of protocone, and presence or absence of the m2.

Data Preparation: Missing Data A substantial portion of these data comes from fossil material, so some proportion will be missing in most groups (total original missing cells ranged from 50% in hyaenids to 28% in Felidae). Because some statistical methods require that all cells contain values, however, I treated missing data as follows: For disparity analyses involving PCA, individual taxa with >50% missing data were excluded from analysis. Missing data in the remaining taxa were handled either by replacement with the group mean or by replacement using regression based on another highly correlated character. Replacement values for individual measurements were calculated based on family or generic level means and regression. Where correlation analysis did not indicate any good correlate for a particular variable, missing data for that measure were of necessity replaced by the overall mean. Because results from separate analyses by the two methods did not differ substantially, I report only those based on replacement by regression. For analyses involving character mapping, taxa with missing values were excluded for that character.

Variance

Distributions of all variables were evaluated prior to analyses, and transformations were performed as appropriate. To compare disparities between hypercarnivores and their sister groups, I performed principal-components analysis (PCA) based on correlation matrices of the above-listed variables and analyzed differences in variance for the categories “hypercarnivore” and “sister” in each of the six sets of sister taxa. To obtain disparity values, I calculated the total variance for each member of each hypercarnivore–sister group pair by determining the variance of its factor scores (eigenvalues) for the first five eigenvectors. Five factors were generally sufficient to explain >90% of the variation in a set of data. Values thus obtained were then scaled by the amount of variance explained by each vector, and scaled values summed to create a single composite disparity score for each member of the set. Each of the six pairs of sister groups was analyzed separately because of a concern that strong loadings for particular variables in some pairs of sister groups could unduly influence the apparent variance in other groups, masking taxon-specific variation (see also Warheit et al. 1999). However, categorical disparity results did not change when all data were analyzed in a single large set, although individual results for sister pairs did change slightly. Variance was chosen as a representative disparity measure because it is relatively insensitive to sampling (Foote 1997). Total values for each taxon set for each category were then contrasted by Wilcoxon Rank Sums against a null model of no difference.

10

Character Mapping

An alternative measure of evolutionary rate, frequency of character change, is calculated as the number of independent derivations of any given character state divided by the number of branches on a tree (Sanderson 1993). Up to 27 characters were mapped onto phylogenetic trees for groups under study, and frequency of change for each character was determined for each sister clade. Frequency of change was averaged over all characters for each member of the sister group pairs. Incomplete sampling (either missing taxa or missing characters) prevented inclusion of all 27 characters in some groups. In these situations, the average was calculated from the characters available. Results by category (hypercarnivore, sister) were pooled and evaluated with Wilcoxon Rank Sums.

Degree of Specialization

As noted above, hypercarnivorous taxa can be subdivided generally into “cat-like” and “hyaena-like” morphotypes based on robustness of the premolars. However, the stages of evolution of a hypercarnivorous phenotype are not discrete, but can better be viewed as grades on a continuum. Thus, I combined all hypercarnivorous taxa into a single data set and evaluated the position of each specimen in a single, “hypercarnivore” morphospace. Initial data exploration consisted of examination of two and three-dimensional graphs of combinations of original variables as well as plots of PC factor scores for the combined data set based on the eight variables listed in Figure 2. Distributions of all variables were examined prior to analysis and transformations performed as necessary. Because the intent was to assign each specimen to a specific degree of specialization within hypercanivores overall, some sense of relative position in space of each specimen was necessary. Thus, the following were designated as “reference” taxa and assigned levels of specialization a priori: Proailurus (= level 3), Pseudaelurus (= level 4), (= level 5), and Hyaena (= level 7). These taxa have distinct specializations (hyaena = bone) or known degrees of development relative to each other (Proailurus→Pseudaelurus→Felis) (Radinsky 1982; Ginsburg 1983; Hunt 1998), and were used as identifiers to provide reference (positional) information for the remaining taxa. This enabled us to determine the appropriate number of levels and to assign the remaining specimens to levels based on clustering and spacing around these reference taxa. Hyaenas (level 7) were included as indicators of taxa with bone-eating tendencies, thus allowing us to distinguish between cat-like and hyaena-like morphs. However, this numeric designation is only an identifier and is not meant to imply a position along a continuum of change; this level was not included in analyses of disparity based on degree of specialization. After assigning each specimen to a level, taxon assignments and the original eight variables were entered into a discriminant function analysis (DFA). Unlike PCA, which is an objective method used to identify that combination of variables that explain the maximal amount of variation along successive orthogonal axes, DFA is used to find the combination of variables that most clearly distinguishes between previously defined categories (in this case, degrees of specialization), so that the probability of misclassification when placing individuals into categories is minimized (Dillon and Goldstein 1984). The accuracy and stability of the classifications is assessed using jack-knifing. Because levels were being compared to each other, it was important that they be as well-supported as possible. A small number of taxa could not be unequivocally assigned to a specific degree of specialization, which made their a priori assignments for DFA necessarily somewhat arbitrary. Since the functions used by DFA to discriminate between categories is based on the membership in those categories, I performed successive iterations of DFA and made adjustments to taxon assignments until I could obtain high

11

accuracy upon resampling while still maintaining correspondence with the natural divisions observed in PC plots. The final percentage of correct assignments was 98.6% for fit based on original assignments and 96% for cross-validated data.

Disparity Based on Degree of Specialization

Methods for determining disparity values based on degree of specialization are identical to those set out for comparison of sister group sets, but rather than comparing the variances of the members of a pair of sister groups, I performed a single PCA including all hypercarnivorous taxa and then determined scaled variance for each level of degree of specialization. Disparity values thus obtained were plotted against degree of specialization to determine whether degree of specialization and disparity were correlated.

RESULTS

Taxonomic Diversity

Species counts for the six sets of hypercarnivores are shown in Table 1. There is no significant difference in taxonomic diversity between hypercarnivores and their sister groups using either a sign test or the method of Slowinski and Guyer (1993). Most sister group sets were roughly equivalent, with hypercarnivorous Felidae and hesperocyonine canids exhibiting slightly greater taxonomic diversity and Viverridae and Mustela slightly lower diversity. Species counts for the hypercarnivorous Chasmaporthetes-Lycyaena-Hyaenictis and the bone-cracking Palinhyaena-Crocuta were equal. The only hypercarnivorous taxon that showed any notable difference in taxonomic diversity relative to its sister group was Nimravidae, a group whose relatively basal position in the carnivoran phylogeny places it as sister to all of Aeluroidea. Setting aeluroid diversity equal to one (the opposite extreme for sister group comparisons) did not alter these results.

Morphological Diversity

Variance Hypercarnivores show significantly lower morphological diversity than do their sister groups (p < 0.01, Wilcoxon Rank Sums; Table 2, Fig. 3), and these results are robust to perturbations of the various data sets (e.g. different included variables, data transformations, inclusion or exclusion of questionable taxa, alternative sister groups). Method of replacement of missing values also had no effect on the results of the analyses. Exclusion of Proteles cristatus, the highly derived , from sister group analyses for felids and hyaenids did not affect the significance of the results overall, although it did result in roughly equivalent disparity values for felids and hyaenids. Comparisons of felids with an alternate sister clade (viverrids), also did not affect our results: felids are relatively lower and viverrids relatively higher in disparity when compared to each other. Exclusion of both felids and nimravids and their sister groups from the pooled values also did not affect the results, which remained significant at p < 0.02. Nimravidae, the saber-toothed noncat family, was the only group of the six that showed disparity equivalent to or higher than that of its sister taxon.

Frequency of Change Average frequency of change was calculated for six sets of hypercarnivore/sister pairs. I found that the two categories, hypercarnivore vs. sister, were significantly different for the two groups (p < 0.037, Wilcoxon Rank Sums). Of the six, all hypercarnivore clades except Nimravidae showed lower frequency of change relative to their sister group. Congruent with

12

comparisons of variance, the family Nimravidae exhibited a higher frequency of change relative to its sister taxon; this value was the second highest frequency of change of any clade evaluated (Table 3, Fig. 4).

Degree of Specialization Six categories of specialization were identified for hypercarnivorous taxa, ranging from the somewhat specialized hesperocyonine canid genera Philotrox-Sunkahetanka-Enhydrocyon through highly specialized saber-toothed taxa. A two-dimensional plot based on the shape variables of RBL and GSL to TRL indicates placement of key specimens and is coded by degree of specialization (Fig. 5) rather than taxon. Although the assignment of taxa to a particular degree of specialization was, as much as possible, phylogeny free, taxa did tend to fall strongly into groupings consistent with phylogeny. Thus, hesperocyonine canids comprised all but one member of level 1, while mustelids made up all members of level 2. Levels 3 and 4 were more diverse and included saber-toothed taxa, viverrids, felids, and hyaenids, but level 5 was composed entirely of members of Felidae (both Felinae and Machairodontinae), and level 6 was made up of saber-tooths from both Felidae and Nimravidae. Assignments of particular species to degrees of specialization are shown in Table 4.

Disparity by Degree of Specialization Disparity values obtained by summing the variance of the first three factor scores for each degree of specialization show a distinct positive correlation between variance and increasing specialization. However, this correlation becomes nonsignificant as successive factor scores are added to the sum (Fig. 6). Level 6, composed exclusively of saber-toothed taxa, is noteworthy in that disparity is higher than that seen in any of the other specialist groups, and this result does not change even when level 6 is partitioned by nimravids or machairodontines (felid saber-tooths). Nimravids in level 6 score particularly high for disparity values, although their disparity remains comparatively low relative to those of “generalist” families also evaluated (Viverridae and Mustelidae).

DISCUSSION

These results show that morphological specialization to hypercarnivory strongly affects morphological but not taxonomic diversity. Discordance between levels of taxonomic and morphological diversification has been addressed previously by several workers (e.g. Foote 1993; Roy and Foote 1997; Eble 2000) and, depending on the direction of the difference, may be explained as a result of diffusion through morphospace, morphospace packing, or selective extinction. In this case, the lack of an effect on species number is likely a result of continued subdivision of the available resources, possibly by body size (see, e.g., Dayan et al. 1989, 1990), even as the structure of the feeding apparatus is maintained. On a larger scale, however, it is surprising that subsequent evolution has not produced a greater diversity of form in the time since the specialization appeared. The hypercarnivore clades in this study are identified under a criterion of being basally hypercarnivorous; it is certainly not a requirement for the clade as a whole. Their lower morphological diversity is also not a result of lack of time: Pseudaelurus evolved approximately 20 Ma, and modern felids appeared ~16 Ma (Radinsky 1982). In the same time period, the sister group, Hyaenidae, diversified greatly, producing forms as varied as insectivore/omnivores, generalists, bone-cracker/scavengers, and its own version of the hypercarnivore (Werdelin and Solounias 1991, 1996). Interestingly, the presence of hypercarnivores within the basally non-hypercarnivorous hyaenid clade appears to have had only a small effect on overall morphological diversity within this group, although the alternative sister taxon, Viverridae, exhibits even higher disparity relative to felids. The clade of hypercarnivorous hyaenids, Chasmaporthetes-Lycyaena-Hyaenictis, is known from at least the late (Berta

13

1998), but although taxonomically it diversified equally with the Palinhyaena-Crocuta clade, all indications are that it deviated little from the meat-specialist morphotype. In contrast, although the Palinhyaena-Crocuta clade evolved specialists in its own right (bone-crackers), the brown and striped hyaenas are arguably somewhat omnivorous in known habits (Ewer 1973; Van Valkenburgh 1989; Nowak 1999), and certainly the specialization for ingestion of bone does not appear to limit morphological disparity in the variables included in this study. Disparity for the single species of Cryptoprocta was calculated from six specimens, two of which are subspecies, but is still substantially lower than that of its sister clade Eupleres-Fossa: the latter taxa are highly divergent in phenotype relative to each other. In the case of Mustela, members of this genus occupy a highly predaceous niche that, with the exception of the extinct sea , Mustela macrodon (Estes 1989), and Mustela vison, which eats fish, crabs, etc. (Ewer 1973), appears to have been retained over the past 10–15 myr. Mustela’s putative sister groups include the and various generalist taxa. The noncat saber-toothed family, Nimravidae, had approximately 30 myr to differentiate and all remained hypercarnivorous, while the last group evaluated, hypercarnivorous canids of the genera Philotrox, Sunkahetanka, and Enhydrocyon, exhibit disparity lower than that observed between two members of the single sister genus Cynodesmus. Results from variance comparisons are supported by all comparisons of average frequency of character change, an important finding since these approaches capture distinctly different views of morphological change. Whereas disparity based on variance reflects occupied morphospace around a group mean, frequency of change has utility in assessing evolutionary flexibility or rates of change (see, e.g., McShea 2001), as represented by the number of changes in a given character state relative to the number of opportunities (branches) on the tree (Sanderson 1993). Note that, for this study, this measure was used explicitly to evaluate the frequency of any state change in any direction rather than a comparison of forward changes to reversals or stasis, a topic that will be addressed in a subsequent paper. Thus, not only do hypercarnivorous taxa occupy less morphospace overall than do their sister groups, they also appear to move from state to state less frequently within that space. This suggests that once taxa achieve the hypercarnivorous morphotype, they are effectively limited in their subsequent evolution. This consistent reduction in variability in five out of six clades evaluated strongly suggests the presence of a functional constraint, and this pattern is made even more interesting because of the sharp contrast with saber-toothed nimravids, a group whose high values for disparity and frequency of change suggest the possibility of an escape from such a constraint. In a combined morphospace, where nimravids consistently show high disparity relative to other taxa, the variables with the largest loadings on the first two principal-components axes are m1/TRL and shape m1 on axis 1, and GSL/TRL and RBL on axis 2. Thus, nimravids are more variable in precisely those hypercarnivore characters of the greatest importance: relative size and shape of the carnassial (axis 1) and the proportion of the total tooth row used for slicing (axis 2). However, one of the more interesting results is that the two nimravid clades, Nimravinae and Barbourofelinae, do not exhibit the same patterns of variation in this combined principal- components space. In the analysis described above, variance for barbourofelines was highest on the first axis, while variance for nimravines was highest on the second. This difference is intriguing, especially in light of suggestions that Nimravidae is paraphyletic (Neff 1983; Morales et al. 2001; Morlo et al. in press). Recognizing that, in most groups, hypercarnivory does strongly affect subsequent morphological flexibility, the idea that increasing specialization within the hypercarnivorous niche should be accompanied by decreasing disparity has intuitive appeal. The lack of correlation between the two is therefore an unexpected result, although it may be an artifact of the method used to assign degree of specialization. As noted previously, there were several taxa that did not fit clearly into particular categories. Such taxa were thus outliers in any grouping, and consequently exerted relatively greater influence on disparity of the group in which they were

14

placed. While this had little effect on the ability of DFA to accurately classify taxa (at worst, 78- 80% were still correctly reclassified), it did lead to low confidence in the fine scale pattern of disparity between degrees of specialization, even when the categorical functions (the groups) themselves were well supported by resampling. One pattern did not change, however: level six, composed entirely of highly derived saber-toothed taxa, exhibited discontinuously high disparity levels regardless of how the groups were partitioned. This finding is consistent with the unexpectedly high values observed for nimravids for measures of variance and average frequency of change. Indeed, diversity in saber-tooths was commented on by Radinsky (1982), who noted the unexpected positioning of Eusmilus within canid morphospace on the first axis in his own analysis. If a hypothesis of functional constraint on characters related to the carnassial feeding apparatus can be accepted, then it follows that strong selection on some other characteristics, most obviously the canines, may have overridden this constraint in saber-toothed groups. Based on the results presented here, there does seem to be a point as taxa evolve towards hypercarnivory where morphological flexibility becomes substantially curtailed relative to some earlier stage. In a framework of morphological change along a continuum, however, it is also apparent that there must still be alternative trajectories available to hypercarnivores at some stage in their evolution. Hypercarnivorous taxa who enhance the bone-cracking or crushing portion of their dentition (e.g. hyaenas of the Palinhyaena-Crocuta clade, borophagines) appear to retain some measure of flexibility, although bone-crackers were not evaluated in this data set and the lower number of known occurrences make this aspect difficult to assess. Taxa who enhance the canines (saber-tooths) likewise appear to exhibit entirely different patterns of diversity relative to other cat-like hypercarnivores, as though by becoming saber-toothed they have escaped the cat phenotype. It is worth noting, however, that no hypercarnivores appear able to easily reverse to a more generalized condition—in fact, for the phylogenies used in this study, there are no known instances of the “cat-like” phenotype reversing to a generalist or omnivorous/insectivorous condition or even moving into a bone-cracker/ niche. When degree of specialization to hypercarnivory is mapped onto the phylogenies for these groups (results not shown), movement away from a more specialized toward a less specialized condition is also an extremely rare occurrence (one mustelid, one nimravid), suggesting that there is a strong directionality to change for this phenotype. Dollo’s law, which addresses the idea of irreversibility in evolution, states that a structure, once lost, cannot be regained. Felids have received a certain amount of attention in this regard (e.g. Werdelin 1987, Russell et al. 1995) due to their extreme and, excepting , persistent reduction in the dental formula. An alternative explanation for the lower disparity observed for hypercarnivorous taxa, then, may be that it is merely a consequence of simplification via loss and reduction of compound structures: structures that do not exist will not vary. Further, if lost structures cannot be regained, then the only possible direction of change will be toward continued loss. However, this explanation alone is unsatisfactory: there is no reason why a taxon with a reduced dental formula should be less variable in the structures that remain. Additionally, not all of the taxa recognized as morphologically hypercarnivorous exhibit the very extreme specializations of felids. Hypercarnivory is recognized on the basis of a set of proportions: relative lengthening of the carnassial blade, relative shortening of the face, and relative reduction of the postcarnassial tooth row. Thus, while proportions of the skull and dentition may alter, the original structures may not be lost at all, and would thus remain available for selection to act upon in any given direction. To add to this, because sister group comparisons evaluate differences between groups since a common ancestor, phenotypic change such as continued loss or reduction of structures over the course of the lineage will be recognized as variation within the taxon. Finally, and most importantly, the data sets used herein to calculate disparity include a number of non-dental variables. Because the diversity is measured by the total variation in all included characters, the observed disparity values cannot be considered merely a result of simplification of

15

the dental formula. Rather, they are likely a result of both loss of structures (leading to no variation) and of lack of variation in the remaining cranio-dental measures as well. As mentioned, the most obvious explanation for lower morphological diversity in multiple clades of hypercarnivores is functional constraint. A consideration of the known ecologies and behaviors of the taxa involved, however, suggests that this explanation may be an oversimplification. The taxa in this study vary in both size and degree of specialization; prey type (and hence killing method) ranges from the birds and lizards of small cats and mustelids to the larger ungulates favored by big cats. Given this potential diversity in killing mode, it is difficult to imagine that the same functional forces are working at all levels of the size range, especially forces that are so consistently strong that they retard or prevent subsequent modification. The evolutionarily stable systems proposed by Wagner and Schwenk (2000) seem more in line, in the sense of a complex of characters that becomes more and more tightly integrated as selection acts to improve functionality of the system as a whole. Wagner and Schwenk suggested that this is brought about by “internal selection” on the relationships between characters, rather than selection directly on particular characteristics, and the patterns observed here certainly appear to follow this premise. If one views the feeding apparatus as a tightly integrated functional complex from which deviation is unlikely, new questions arise: once taxa begin to trend toward hypercarnivory, is subsequent phenotypic change biased toward this niche? Is there a "point of no return” after which reversal or modification is not possible? Clearly escape is difficult, and in this case the taxa that do not exhibit lower disparity have apparently done so by evolving a very extreme alternative specialization. Is it possible that taxa can only move from a less to a more specialized condition? Mapping of degree of specialization onto phylogenies appears to indicate just such a progression, although more detailed study is needed to determine the generality of this phenomenon. More robust phylogenies for machairodontines and nimravids are sorely needed, as is a better understanding of interrelationships between aeluroids as a whole, especially fossil taxa. Such phylogenies would be of great benefit in ascertaining the most likely patterns of diversification in relation to the carnivoran feeding complex and evolution of hypercarnivorous specialization.

16

TABLE 1. Taxonomic diversity. Taxonomic diversity of hypercarnivores relative to their sister groups. References for species counts are as follows: Hesperocyoninae: Wang 1994; Mustelidae, Viverridae and Herpestidae: Nowak 1999. Hyaenidae: Werdelin and Solounias 1991; Felidae: Nowak 1999, Berta and Galiano 1983, Martin 1998 Ginsburg 1983, Turner 1997, Berta 1987, Hemmer 1978, Glass and Martin 1978, Werdelin 1985, Hunt Jr 1996; Nimravidae: Bryant 1996, Geraads and Gulec 1997.

Hypercarnivore No. of species Sister-group No. of species

Felidae 86 Hyaenidae 69 Hyaenidae:Chasmaporthetes/ 15 Hyaenidae: 15 Lycyaena/Hyaenictis Crocuta/Hyaena Nimravidae 24 Felidae/Hyaenidae/ 228 Viverridae/Herpestidae Canidae: Hesperocyoninae 6 Canidae: Hesperocyoninae 2 Enhydrocyon/Philotrox/ Cynodesmus Sunkahetanka Viverridae: Cryptoprocta 1 Viverridae: Eupleres/Fossa 2 Mustelidae: Mustela 17 Mustelidae: Enhydra/Lutra/ 26 Ictonyx/Galictis ______

17

TABLE 2. Disparity. Disparity values obtained for each set of sister groups. Disparity was calculated as the sum of the scaled variance of the first five factor scores obtained from PCA.

Hypercarnivore Disparity Sister-group Disparity Felidae 46.20 Hyaenidae 116.66 Hyaenidae: Chasmaporthetes/ 44.91 Hyaenidae: 109.95 Lycyaena/Hyaenictis Crocuta/Hyaena Nimravidae 90.65 Felidae/Hyaenidae/ 79.01 Viverridae/Herpestidae Canidae: Hesperocyoninae 84.53 Canidae: Hesperocyoninae 180.56 Enhydrocyon/Philotrox/ Cynodesmus Sunkahetanka Viverridae: Cryptoprocta ferox 21.47 Viverridae: Eupleres/Fossa 107.64 Mustelidae: Mustela 36.87 Mustelidae: Enhydra/Lutra/ 117.60 Ictonyx/Galictis

18

TABLE 3. Frequency of change. Average frequency of change obtained for each set of sister groups, calculated as the number of independent derivations of a character state/number of nodes on the phylogeny.

Average Average Hypercarnivore frequency of change Sister-group frequency of change

Felidae 0.1370 Hyaenidae 0.1841

Hyaenidae: 0.1206 Hyaenidae 0.1637

Chasmaporthetes/Lycyaena Palinhyaena/Crocuta Nimravidae 0.2289 Felidae/Hyaenidae/ 0.1838 Viverridae/Herpestidae Canidae: Hesperocyoninae 0.0714 Canidae: Hesperocyoninae 0.1786

Enhydrocyon/Philotrox/Sunkahetanka Cynodesmus Viverridae: Cryptoprocta 0 Viverridae: Eupleres/Fossa 0.3182 Mustelidae: Mustela 0.1607 Mustelidae: Galictis/Ictonyx/ 0.2195 Pteronura/Lontra/Enhydra/ Lutra/Amblonyx/Aonyx ______

19

TABLE 4: Degree of specialization. Degree of specialization was assigned on the basis of evaluation of location in principal components space and a priori designations where genera or families did not vary.

LEVEL 1 LEVEL 2 LEVEL 3 LEVEL 4 LEVEL 5 LEVEL 6

Mustela putorius Mustela altaica felina Pseudaelurus Felinae

Enhydrocyon basilatus Mustela felipei Lycyaena galiani Nimravides catacopis hodsonae

Enhydrocyon crassidens Mustela frenata Proailurus Cryptoprocta ferox

Enhydrocyon pahinsintewakpa Mustela kathiah Eusmilus

Hoplophoneus Philotrox condoni Mustela nigripes Dinictis cyclops occidentalis

Sunkahetanka geringiensis Mustela nivalis oharrai

Hoplophoneus Mustela sibirica primaevus

Mustela vison Nanosmilus kurteni

Nimravus brachyops

Nimravus gomphodus

20 FIGURE 1. Dentitions indicating increasing specialization. A. A generalist dentition. The talonid is basined, with the hypoconid and entoconid cusps roughly equal in size. Note that the m2 and m3 are unreduced. B. Dentition trending toward hypercarnivory. The shearing blade is slightly elongate, while the hypoconid and entoconid are unequal in size (hypoconid is larger). The m2 and m3 are somewhat reduced in size. C. Trenchant talonid. The shearing blade is elongate, and the hypoconid is enlarged and medial, while the entoconid is completely reduced. The m2 and m3 are reduced. D. Note the absence of a talonid, including loss of the hypoconid and entoconid. The shearing blade extends the entire length of the m1, the m2 is reduced or absent, and the m3 is absent.

21 FIGURE 2. Representative cranial and dental measurements. Representative cranial and dental measurements used in morphological analyses. JL, jaw length; TRL, tooth-row length; ZAW, zygomatic-arch width, lM1, length of M1; trigonid, trigonid length; HAR, height of ascending ramus; MAT, distance from condyloid process to coronoid process; MFL, distance from condyle to front of masseteric fossa; COM1, distance from carnassial notch to condyle.

22 200

150

100

50

Disparity based on variance 0 Hypercarnivore Sister-group

FIGURE 3. Average disparity. Average disparity for six clades of hypercarnivorous taxa and their respective sister groups. Disparity is the sum of the scaled variances of the first five factor scores obtained from PCA of the eight variables representative of skull and dental morphology illustrated in Figure 2. Wilxocon Rank Sums p < 0.01.

23 0.4

0.3

0.2

0.1

0.0 Average frequency of change Average

-0.1 Hypercarnivore Sister-group

FIGURE 4. Average frequency of change. Average frequency of change for six clades of hypercarnivorous taxa and their sister groups. Nimravids are the uppermost symbol in the left column. Wilcoxon Rank Sums p < 0.04.

24 DEGREE 1 r 2 p GSL to TRL to GSL 3 ù 4 ò 5 £ 6 •

Relative blade length

FIGURE 5. Position of taxa in morphospace based on degree of specialization. Plot of proportional variables indicating positioning of taxa in morphospace when coded by degree of specialization. Relative blade length is calculated as the length of the trigonid (shearing blade) relative to the length of the entire m1 and reflects the amount of meat in the diet. GSL/TRL is a measure of tooth surfaces not devoted to slicing relative to the length of the face. Degrees 1 and 2 represent taxa that are relatively less specialized (e.g., mustelids and hesperocyonine canids). Degree 6 represents taxa that are relatively more specialized and is composed exclusively of felid and nimravid sabertoothed taxa.

25 150 = sum of first 5 factor scores x = sum of first 4 factor scores 125 = sum of first 3 factor scores

100

75 Variance 50

25 x

0 123456 Category

FIGURE 6. Disparity for different degrees of specialization to hypercarnivory. Disparity for different degrees of specialization to hypercarnivory, ranging from less (1) to more (6) specialized. Degree 1 is composed of hesperocyonine canids and one mustelid. Degree 2 is exclusively members of Mustela. Degree 3 is composed of Proailurus, Cryptoprocta, hyaenids, and some felids and nimravids. Degree 4 includes Pseudaelurus and some felids and nimravids. Degree 5 is exclusively felines and some machairodontines. Degree 6 is composed of machairodontine and nimravid sabertooths. Note the discontinuously high variance of degree 6 relative to degrees 1-5.

26

CHAPTER 2 EVOLUTION IN CARNIVORA: IDENTIFYING A MACROEVOLUTIONARY RATCHET

To understand the role of adaptation in generating macroevolutionary patterns, it is necessary to understand whether and in what ways specific features of the phenotype affect subsequent phenotypic diversification. This area has been much debated by both past and present workers, who considered whether certain morphologies might be “channeled” (e.g. Gould 1984; Emerson 1988) to appear once a specific starting morphology was attained. Less radically, a number of workers have suggested that possession of certain morphological character states may reduce the ability to attain certain other character states (Lauder 1981; Maynard-Smith et al. 1985; Emerson 1988; Futuyma and Moreno 1988; Werdelin 1996; Wagner and Schwenk 2000; Donoghue and Ree 2000; Wagner and Mueller 2002; Porter and Crandal 2003; Van Valkenburgh et al. 2004), implying that, in some cases, taxa may be limited in their subsequent evolutionary trajectories. Both morphological channeling and a limitation on specific character states fall into the realm of a character change bias, where certain states are more likely to appear than others (Donoghue and Ree 2000; Sanderson 1993). Despite ongoing theoretical debate, however, there has been relatively little empirical exploration of the possibility of bias or directionality in morphological character change and this area remains poorly understood (Arthur 2001, 2004; Schluter et al. 2004). Much of the available work evaluating questions of bias in character evolution has been performed in a functional context, assessing whether certain starting morphologies act to limit the specific kinds of phenotypes that are subsequently attained (Emerson 1988; Richardson and Chipman 2003). Such functional limitations, if they are shown to exist, will necessarily result in a bias in the appearance of certain phenotypes. Additionally, recent years have seen increased interest in the study of morphological integration (Olson and Miller 1958; Pigliucci and Preston 2004; Polly 2005; Goswami 2006) and, in particular, Evolutionarily Stable Systems (ESS) (Wagner and Schwenk 2000). Morphological integration studies are predicated on the presence of strong biases in (or against) character state transitions, but are more frequently discussed in the context of constraint (usually functional). Recent work in the area of integration has also begun to tackle the role that embryological development plays in phenotypic evolution, particularly in the sense of non-independence of character sets and potential biases in character transformations (Donoghue and Ree 2000). Finally, irreversible evolution, often discussed in the context of “Dollo’s Law,” is nothing more than an extreme bias against reversals: a character, once lost, cannot be regained.

Character Evolution and Specialization Theory

For many years, there has been a general sense among workers in ecology and evolution that ecological or phenotypic specialists are very unlikely to revert to a more generalized condition; a significant amount of theoretical discussion addresses the possibility that specialists may be subject to strong stabilizing selection to maintain their particular niche, either through habitat tracking or simply due to lower fitness for variants that fall outside the basic (fundamental) niche (Holt and Gaines 1992; Losos and Irschick 1994). Other workers suggest that specialists may in fact be influenced by biased or directional selection toward an increasingly specialized morphotype as a result of increasingly fine niche partitioning among specialized forms (Van Valkenburgh 1991; Losos and Irschick 1994; Nosil 2002; Van Valkenburgh et al. 2004). Because a theoretical framework is already in place, the evolution of specialization is an obvious choice for detailed testing of questions of bias or directional trends. Unfortunately, to

27

date, empirical studies that evaluate specialization and how it evolves have produced highly equivocal results (Futuyma and Moreno 1988). Furthermore, tests of bias are not only difficult to interpret, but detecting even general patterns is problematic due to varying research scales and methodological approaches (Richardson and Chipman 2003). Typically, research into character evolution (and specialization in particular) can be divided into two main areas: the question of total irreversibility (Bull and Charnov 1985; Emerson 1988; Moran 1988) and the ease with which certain character states may be gained or lost (Sanderson 1993; Wiens 1999; McShea and Venit 2002). Clearly, these two areas are not necessarily mutually exclusive: an extreme bias in favor of character state gain over loss would be interpreted as irreversibility. Regardless, studies of either type are often qualitative in nature and their power is accordingly weak: researchers frequently seek to identify a change or reversal to test an “always or never” (or mostly versus seldom) hypothesis (Siddall et al. 1993; Rouse 2000; Omland and Lanyon 2000) and few studies provide quantified transition frequencies despite their potential importance (but see McShea 2001; see also McShea 2001; McShea and Venit 2002 for the importance of considering frequencies). The need for quantified transition rates is especially evident in irreversibility studies where, without a frequency value, a single instance of a reversal supports a null, despite the recognition that a strong bias would certainly be biologically relevant and in many cases a more interesting finding. At present, attempts to identify and quantify biases in transition rates remain uncommon (but see, e.g., Jansen 1992; Rouse 1999; McShea 2001; Bokma 2002; McShea and Venit 2002; Geeta 2003), although an increasing number of methods allow quantification of character transitions in both a statistical and non-statistical framework (Harvey and Pagel 1991; Sanderson 1993; Hansen 1996). Furthermore, much of the available empirical research to date that explicitly assesses differences in transition rates (aka gain:loss bias) has focused on broader–scale transitions, such as increasing complexity or changes in a macroevolutionary hierarchy (e.g. McShea 2001; McShea and Venit 2002), ecological niche changes (Geeta 2003) or patterns of (Omland 1997). Only a handful of studies consider specific morphologies or the role particular characters or character complexes play with respect to subsequent adaptive change (Emerson 1988; Richardson and Chipman 2003; Holliday and Steppan 2004). Understanding potential biases in character evolution is of fundamental importance to workers in fields as diverse as ecology, conservation, systematics, and evolutionary biology, since a bias or limitation in character change can – and will – have a significant effect on the ability of a taxon to survive and adapt. If character change is biased in a particular direction, this has implications for patterns of ecological interactions, including competitive ability and guild or community composition. The goal of this paper is to present a general approach to testing for a gain:loss bias using a variation of the method presented by Sanderson (1993). I provide an example from Carnivora, where I use replicated sister group comparisons to assess the possibility of a bias in character change relative to a specific dental morphology, that of the hypercarnivore.

What is a Hypercarnivore?

In the mammalian order Carnivora, a hypercarnivore is an ecological and morphological specialist to a diet composed primarily or exclusively of flesh. Morphologically, hypercarnivores are readily identifiable on the basis of a well-described suite of phenotypic features, including relative loss or reduction of the post-carnassial molars, relative elongation of the carnassial shearing blade, and relative shortening for the face (Hunt 1998; Van Valkenburgh 1988; see also Holliday and Steppan 2004). Hypercarnivory has evolved at least eleven times within the Carnivora, and in extant and extinct members of such diverse carnivoran lineages as felids (cats), viverrids (civets and genets), hyaenids (hyaenas), nimravids (extinct non-cat sabertooths), canids (dogs and foxes) and mustelids (weasels, otters and stoats). Guild formation in carnivorans is known to follow a fairly repetitive pattern, producing ecomorphs that include generalists,

28

scavenger/bone-crushers and meat-specialists (hypercarnivores) (Van Valkenburgh 1988; Werdelin 1996). Such iterative evolution allows for natural replicates and is conducive to study of broad-scale patterns.

Previous Findings

Holliday and Steppan (2004) quantified the broad-scale effects of specialization to hypercarnivory on subsequent character change, and showed that not only felids, but hypercarnivores as a group are reduced in their morphological diversity relative to sister taxa. Holliday and Steppan (2004) compared the variance of factor scores (Foote 1993; Wills et al. 1994) from a principal components analysis of six sets of hypercarnivore clades and their sister groups and found that hypercarnivores occupy relatively less morphospace relative to their sister taxa. They also applied the method of Sanderson (1993), mapping morphological characters onto phylogenies and comparing average rates of character change between hypercarnivores and their sister groups. These “frequency of change” measures indicated that hypercarnivores change character state less often on a given phylogeny (Holliday and Steppan 2004). Together with the finding of lower variance in hypercarnivores, fewer character state transitions suggests that some cause may be acting to limit change within hypercarnivores. However, noting that rate differences exist does not provide information on the underlying mechanisms: do hypercarnivores exhibit less change overall (i.e. lower mutation rates/genetic variability or stabilizing selection) or is the difference due to higher rates of change in one direction (directional evolution due to strong selection) and/or reduced rates of change in another (constraint)? In carnivorans, increasing amounts of meat in the diet can be related to increasing length of the carnivoran shearing blade on the carnassial tooth (Van Valkenburgh 1988, 1991). This character, relative blade length (RBL) is defined as the length of the carnassial shearing blade, or trigonid, relative to the length of the entire lower first molar and is a crucial feature in understanding dental evolution within Carnivora (Van Valkenburgh 1988, 1989, 1991; Holliday and Steppan 2004; Dayan and Simberloff 2005). In describing possible selective forces that would shape the evolution of the shearing blade in carnivorans, Van Valkenburgh (1991) proposed that strong competition for resources (e.g., meat), even among littermates, should result in directional selection for the evolution of a longer, more efficient, slicing blade. Thus, the first question to be tested is whether there is a difference in rates of change for the morphological character RBL. But because hypercarnivory is not defined solely by the character RBL, but is instead made up of a combination of several characteristics, a second, more complete measure of bias is also required: are there differences in rates of change for the hypercarnivore complex? To answer these questions, I evaluate the competing hypotheses of stasis (stabilizing selection, e.g., Holt and Gaines 1992; Losos and Irschick 1994) versus directional selection toward increasing specialization (e.g. Van Valkenburgh 1991; Losos and Irschick 1994; Nosil 2002; Van Valkenburgh et al. 2004) versus a limitation on reversals to a more generalized condition (constraint). When characters are polarized so that the extreme “hypercarnivore” phenotype represents an end state, then stasis will be observed as no difference in rates of forward to reverse change. Directionality, if present, should be observed as a higher rate of forward change (relatively more gains than losses), and constraint would be indicated by a reduced rate of reverse change (relatively fewer losses than gains).

Quantifying Gain:Loss Bias

As with any comparative study, in those cases where the causes of morphological rate differences are explicitly addressed, comparisons between sister groups or some other closely comparable taxon (Holliday and Steppan 2004; Nosil 2002; Warheit et al. 1999) is a preferred approach. There are presently several methodological approaches that may be used to quantify

29

transition rates in morphological characters, including the parsimony based method of Sanderson (1993) and a Markov based model (Harvey and Pagel 1991; Pagel 1994). Because the data used in this study have recognized limitations (including missing data and no branch length information), I chose to apply Sanderson’s method to mitigate additional complicating assumptions and the risk of overanalysis. Sanderson’s method is based on information directly available on the phylogeny (number of branches) and characters (number of changes); the principle of parsimony and its assumptions are well-established. A value for frequency of character change (forward changes, reverse changes) is thus based solely on the number of changes in a given character relative to the number of times that character could have possibly changed on a tree (McShea and Venit 2002; Sanderson 1993). While the accuracy of this method depends on both the underlying topology and on accurate ancestral state reconstructions, either (or both) of which may be subject to significant uncertainty, issues surrounding phylogeny estimation and ancestral state reconstruction have been discussed at length by other workers (Pagel 1994; Schluter et al 1997; Cunningham et al. 1998; Mooers and Schluter 1999; Omland 1999; Pagel 1999; Martins 2000) and are not the focus of this paper. Rather, I present a variation and extension of Sanderson’s (1993) approach for testing gain:loss bias using mammalian hypercarnivores (Carnivora: Mammalia) as a study group.

METHODS

Sanderson (1993) presented a method for testing for the presence of a bias in rates of forward change versus rates of reverse change for binary characters. In its simplest form, this method entails merely counting the number of character state changes on a given phylogenetic tree and then dividing those changes by the number of times the character could have possibly changed, as represented by the branches on the tree (Fig. 7). Sanderson’s method anticipates information on branch length, however, in this case branch length information is unavailable and I therefore assume a punctuated model of evolution, where change occurs only during speciation. The calculation (# changes/# opportunities for change) provides a frequency of change metric (i.e. rate of gain of a character versus rate of loss of a character) that can then be compared to evaluate whether there is a bias. Here, I extend Sanderson’s approach, which utilized only binary characters, to include multistate characters. I also use replicated sister group comparisons to consider rates of forward and reverse change for multiple clades of hypercarnivore ecomorphs, comparing rates of forward change to rates of reverse change within clades and comparing rates of forward change between hypercarnivore clades and their non-hypercarnivorous sister groups. Rates of reversal are compared in the same way. Use of replicated sister group comparisons allows me to test for patterns that may be applicable across categorical designations (e.g. hypercarnivores as a group), and enables me to consider possible mechanisms that may be responsible for differences in patterns of change on a macroevolutionary scale. Statistical analyses of the paired data were performed using Wilcoxon signed-ranks test (Sokal and Rohlf 1998). This modified approach represents a novel application of Sanderson’s (1993) method, and is also one of the first studies to use sister group comparisons to quantify and explicitly test for gain:loss bias in a vertebrate group (Richardson and Chipman 2003). Previously, Holliday and Steppan (2004) evaluated overall frequency of change for hypercarnivores relative to their sister clades. This work compared six sets of sister groups and applied Sanderson’s method to obtain a rate of character change for each clade. Because the purpose of that study was to quantify overall patterns of morphological diversity rather than to assess bias, rates of change (= frequency of change) were evaluated as total amounts of change within a given morphospace. Thus, any character state change in any direction was considered a change. Unlike that study, the work presented here explicitly calculates forward changes and reverse changes separately, and compares forward: reverse change both within hypercarnivore clades (testing for a gain:loss bias within hypercarnivores) and between hypercarnivores and their

30

sister groups (testing rates of forward change in hypercarnivores relative to their sister groups, testing rates of reverse change in hypercarnivores and their sister groups).

Multistate Characters

As presented, Sanderson’s (1993) method considers differences in rates of gain versus rates of loss for a single binary character. Here, I have extended the method to include multistate characters. The primary difference in calculating frequency of change for binary versus multistate characters involves terminal character codes. For example, under Sanderson’s method, a character with state zero cannot experience a reversal because it cannot go to a lower state. Likewise, a character in state one cannot experience a new forward change because, in a binary system, it is already at the highest state (Sanderson 1993). When counting change, then, forward changes (0-1) are possible only for branches at state 0. Likewise, reverse changes (1-0) are possible only for branches at state one. Incorporating multistate characters is fairly straightforward: regardless of the number of character states, if characters are ordered and polarized so that 0 represents the least hypercarnivorous condition, change is calculated in essentially the same way as for binary characters. Thus, a character at state 0 cannot reverse - but all other character states may - and its branch is not counted when calculating reversals. Likewise, a character in the terminal (highest) state cannot move forward - all other states can - and that branch is not counted when calculating forward change (Fig. 7).

Data

Morphological characters were chosen based on their functional significance for hypercarnivorous taxa and are a specific subset of characters previously described in Holliday and Steppan (2004). Relative blade length (RBL) is a known indicator of the amount of meat in the diet (Van Valkenburgh 1988). Additional characters used in this study include length of the trigonid (the carnassial shearing blade) relative to the remaining tooth surfaces (blade/GSL), shape of the lower 4th premolar (p4 shape, width of the lower 4th premolar divided by its length), ratio of the length of the upper 4th premolar relative to the length of the upper first molar (P4/M1), length of the upper first molar relative to tooth row length (M1/TRL), and total grinding surfaces relative to the length of the lower carnassial tooth (GSL/m1). All of the characters described represent features that are known to modify as lineages move toward the hypercarnivore phenotype. To ensure that character codings were consistent across all families, quantitative data from all taxa was pooled prior to coding each character. Values were then ordered from smallest to largest, segment coded, and discretized (Simon 1983; see also Chappill 1989). Segment coding was used instead of gap coding because the pooled data were continuously distributed (i.e., there were no gaps). Characters were polarized so that lower states indicate lower levels of carnivory; higher states represent changes toward hypercarnivory. Most characters were divided into between 4 and 6 states, and as much as possible segment sizes were chosen to reflect the biological relevance of particular phenotypes. For example, while a 2 state character based on arbitrary segment sizes of 0-.50 and .51-1.0 could very likely mask information about intermediate phenotypes, 3 segments of 0-.35, .35-.70, and .71-1.0 are much more informative. Missing data were not replaced; individual species or taxa for which data was not available were excluded for that character. Characters were mapped onto phylogenies using both ACCTRANS and DELTRANS optimizations. For each clade, character changes were counted for each individual character and then divided by available branches. Any change was counted as a single step, regardless of the number of states in between (for example, a change in state from 1-2 or from 1-3 was still a single step).

31

After frequencies of forward and reverse change were calculated for each character and for each clade under study, individual frequencies over all six characters (per clade) were combined to obtain an average rate of forward change and an average rate of reverse change. These averages represent the “hypercarnivore complex, ” i.e. the characters that are expected to change as unit as taxa become hypercarnivorous. In limited cases (Mustela, Enhydrocyon, Lycyaena), missing character data or incomplete taxon sampling prevented inclusion of one or two individual characters for that clade. In those cases, the average was calculated based on the remaining four or five characters. Thus, for every group, a set of metrics was produced that provided information on forward and reverse change both on an individual character basis and as an average rate of change overall. Within and between group comparisons were performed for the character RBL and for the average of all characters (the hypercarnivore complex). RBL was the only character fully assessed on an individual basis because this character is considered the most indicative of transformations toward a high meat diet and therefore the most likely to show an individual bias.

Phylogenies: Ingroups

The following hypercarnivorous taxa were included in this study: Felidae (cats), Nimravidae (saber-toothed non-cats), the mustelid genus Mustela (weasels), the herpestid Cryptoprocta ferox (the Malagasy fossa), a clade of hypercarnivorous hyaenids (Chasmaporthetes-Lycyaena-Hyaenictis), and a clade of early hesperocyonine canids (Enhydrocyon-Philotrox-Sunkahetanka). In keeping with Holliday and Steppan (2004), sister group comparisons (hypercarnivore/sister group) included Felidae/Viverridae (civets and genets), Nimravidae/Aeluroidea (felids, hyaenids, viverrids and herpestids), Mustela/Enhydra-Lutra (the otters), Cryptoprocta/Eupleres-Fossa (the falanouc and the ), Chasmaporthetes- Lycyaena-Hyaenictis/-Crocuta (bone-cracking , including fossil forms), and Enhydrocyon-Philotrox-Sunkahetanka/Cynodesmus (a small to mid sized, generalist canid). Phylogenies used in Holliday and Steppan (2004) were drawn from the literature, and included both molecular and morphological treatments. In some cases, complete and/or robust phylogenies were not available and it was necessary to build composite trees based on partial trees. Relevant details of taxa, their phylogenetic hypotheses, and justifications for inclusion are presented in Holliday and Steppan (2004). While it is recognized that ongoing data collection and analysis will almost certainly change hypotheses of relationships, the topologies used by Holliday and Steppan (2004) were retained here in order to maintain consistency between the two studies.

Phylogenies: Outgroups/non-specialists

To establish baseline (or expected) rates of character change, rates of change for four “generalist” families and subfamilies were obtained. These unspecialized taxa, treated as additional outgroups for comparison, include Canidae (dogs and foxes), Herpestidae (mongooses), Hesperocyoninae (a primitive canid subfamily), and Mustelidae (weasels, otters and stoats). It should be noted that several of these groups (Mustelidae, canids) do have hypercarnivorous taxa (and their sister groups) nested within them. However, since on whole these larger clades can be considered generalist or non-specialist, the more inclusive clade was evaluated as a “generalist” taxon under the justification that any individual biases in rates of change due to presence of a hypercarnivore/sister clade would be balanced by the inclusion of the remainder of the family or subfamily as long as that group is broadly unspecialized.

Within Group Comparisons

32

To understand the transition rates of the characters and character complexes under study, outgroup and sister group comparisons were used to establish an “expected” pattern of forward versus reverse change (Schwenk and Wagner 2004). In these analyses, I calculated rates of forward and reverse change for clades of broadly generalized (non-specialized) carnivorans and then compared the two metrics (forward:reverse) to determine whether unspecialized groups exhibit any detectable difference in their rate of forward change relative to their rate of reverse change. This provided a baseline comparison for subsequent analyses of hypercarnivores and their sister groups. Rates of forward and reverse change were then calculated and compared within the sister groups of hypercarnivore clades and finally for hypercarnivores as well.

Between Group Comparisons

One limitation of within group comparisons is that, even when a rate difference exists, without an outgroup or a sister group for comparison one cannot determine whether such a difference is due to a higher rate of forward change (directional selection) or a lower rate of reverse change (constraint). To address this difficulty, I used sister group comparisons to compare rates of forward change between hypercarnivores and their sister groups. I also compared relative rates of reverse change for hypercarnivores versus their sister groups.

Methodological Issues

Because the sister-group method introduces several variations of Sanderson’s original approach, certain methodological issues must be addressed. The most striking of these is intrinsic to the use of paired sister groups, and involves the way in which branches and changes are counted. Sanderson’s (1993) description of his method states that change should be counted “from the root” – but he did not address the question of paired clades. Depending on how the root is defined (as the basal portion of an individual clade or the root of the entire sister group pair; see Fig. 7), a branch may be added or lost during rate calculations. In practice, inclusion or exclusion of this single branch should not significantly affect the frequency of change values unless a clade is extremely small. However, precisely this circumstance arose during rate calculations for the sister group set composed of Cryptoprocta/Eupleres-Fossa. In these analyses, the hypercarnivorous herpestid genus Cryptoprocta is represented by a single species, and inclusion of this individual taxon without including the shared basal branch leaves the frequency of change metric undefined ((2 * 1 – 2) / 1 = 0), effectively eliminating it as a data point. For this reason, in comparisons labeled “without basal branch,” Cryptoprocta and its sister group are excluded. At the same time, exclusion of a valid hypercarnivorous taxon or comparison could bias the results, as well as reduce already low statistical power by permitting only five sets of comparisons. All analyses were therefore performed both with and without the shared root of the sister group pairs (with and without the basal branch), and I was thus able to include Cryptoprocta and its putative sister group (Eupleres/Fossa) in the computations (all computations labeled w/basal). For all analyses that include this basal branch, the root state is established using outgroup comparison.

RESULTS

All analyses were performed using Wilcoxon Ranked Pairs, a nonparametric test for comparison of paired data that performs especially well when sample sizes are small (Sokal and Rohlf 1998). Data from both ACCTRANS and DELTRANS optimizations are provided in Tables 5-9. Tables 10 and 11 are summaries of p values for different comparisons.

33

Relative Blade Length: without Basal Branches (5 comparisons)

Within group comparisons There was no significant difference between rates of forward and reverse change for generalized clades (Tab. 9). Differences in rates of forward and reverse change for sister groups showed only marginal significance under ACCTRANS (p < .07, Tab. 2) and no significant difference under DELTRANS (p = .27, Tab. 2). There was a significant difference between rates of forward and reverse change for hypercarnivore clades under either optimization (p < 0.04, Tab. 2). It is worth noting that, for within-group comparisons, the differences in rates of forward and reverse change for the sister groups of hypercarnivores and for hypercarnivores were in opposite directions, with hypercarnivores themselves exhibiting comparatively lower rates of reversal (or relatively higher rates of forward change) and their sister groups exhibiting comparatively higher rates of reversal (or relatively lower rates of forward change). This suggests that different and potentially opposing selective forces may be influencing the evolution of relative blade length of both sets of taxa.

Sister group comparisons Rates of forward change for hypercarnivores versus their sister groups and rates of reversal for hypercarnivores versus their sister groups showed no significant difference in relative rates of forward change (Tab. 10) but a significant decrease in the relative rate of reversal for hypercarnivores as compared to their sister groups under either ACCTRANS or DELTRANS optimizations. (p < 0.04, Tab. 10).

Relative Blade Length: with Basal Branches (6 comparisons)

Within group comparisons When basal branches were included in comparisons, relative rates of forward:reverse change for hypercarnivores were all significantly different (p < 0.05, Tab. 7). The sister groups of hypercarnivores were significantly different under ACCTRANS (p < 0.05, Tab. 7) but not under DELTRANS (P = .138, Tab. 7) and, as in the analyses that did not include basal branches, the directions of the difference for hypercarnivores and their sister taxa were in opposite directions.

Sister group comparisons Rates of forward change for hypercarnivores and their sister groups and rates of reversal for hypercarnivores and their sister groups showed no significant difference in relative rates of forward change (Tab. 7) but a significant decrease in the relative rate of reversal for hypercarnivores as compared to their sister groups under either optimization. (p < 0.03, Tab. 3).

Hypercarnivore Morphotype without Basal Branches (5 comparisons)

Within group comparisons Because the evolution of the hypercarnivore morphotype involves changes in not just a single character but in a suite of characters that (presumably) interact, it was important to test the evolution of the entire character complex. Results for analyses of the hypercarnivore complex were very similar to those obtained for RBL. There was no significant difference between rates of forward and reverse change for generalized clades (Tab. 5). There was also no significant difference between forward and reverse rates for sister groups of hypercarnivore clades (Tab. 8). There was a significant difference between rates of forward and reverse change for hypercarnivore clades under ACCTRANS (p < 0.03) and DELTRANS (p < 0.05, Tab. 8), and rates of forward change were higher than rates of reverse change.

34

Sister group comparisons Relative rates of forward change for hypercarnivores versus their sister groups and relative rates of reverse change for hypercarnivores versus their sister groups showed no significant difference in the rate of forward change for the two groups (Tab. 8) but a clearly decreased rate of reversal for hypercarnivores relative to their sister taxa (p = 0.08 under ACCTRANS, p < 0.05 under DELTRANS, Tab. 8).

Hypercarnivore Morphotype with Basal Branches (6 comparisons)

Within group comparisons For the character complex, there was no significant difference between rates of forward and reverse change for sister groups of hypercarnivore clades (Tab. 9). There was a significant difference (p < 0.03, p< 0.05) between rates of forward and reverse change for hypercarnivore clades under ACCTRANS and DELTRANS, respectively (Tab. 9), with forward changes outnumbering reverse changes.

Sister group comparisons Comparison of relative rates of forward change for hypercarnivores and their sister groups and rates of reverse change for hypercarnivores and their sister groups showed no significant difference in the rate of forward change for the two groups (Tab. 9) but a significantly decreased rate of reversal for hypercarnivores relative to their sister taxa (p < 0.05, Tab. 9).

DISCUSSION

Given the availability of methods for evaluating character evolution and quantifying transition frequencies, identification of broad patterns in character evolution (e.g, trends or unusual levels of disparity in a particular clade) is only a first, necessary step toward addressing more specific questions of character bias and causality. Clearly, the presence of a constraint or limitation on character change, or strong selection for some feature or phenotype, can have significant effects on the directions and rates of morphological evolution and/or species diversification. It follows that these effects, when strong enough, will produce recognizable macroevolutionary patterns. Once these patterns are identified, though, the focus should shift to questions about mechanisms. For example, given an observed difference in morphological diversity for a specific taxon, is that difference a result of a higher rate of morphological change in one clade (strong selection for a particular kind of change), or a reduced rate in another (a constraint or limitation against a specific change)? How large (quantitatively) is the difference in rates? How strong is the pattern? Previously, Holliday and Steppan (2004) showed that hypercarnivores occupy less morphospace and exhibit lower rates of character change relative to their sister taxa. To better understand the underlying causes of this disparity, I used sister group and outgroup comparisons to explicitly evaluate rates of forward and reverse change. As the results show, hypercarnivores have a significant difference in their rates of forward change relative to rates of reverse change for both relative blade length and for the hypercarnivore complex (Tables 10, 11). Further, comparison with sister groups indicates that this is due to a decreased rate of reversal, not an increased rate of forward change. Potential explanations for the lower overall frequency of change in hypercarnivores relative to their sister taxa included stabilizing selection, directional selection, or constraint. By breaking down overall frequency of change into rates of forward change and reverse change, however, it becomes clear that neither stabilizing selection nor strong directional selection can explain the observed patterns.

35

Stabilizing selection, which would maintain the phenotype “as is” without significant shifts towards or away from the optimum, is rejected because there is no apparent limitation on forward change for hypercarnivorous taxa. Strong directional selection is also not supported: although forward changes do occur at a higher rate than reversals within hypercarnivores, rates of forward change are not significantly different between hypercarnivores and their sister taxa. Instead, it appears that hypercarnivores are “ratcheting” forward into increasingly specialized morphospace, with no limitation on change in a forward direction but with significant limitations on reversals. This contrasts with the situation in generalists and sister taxa, where the rate of forward and reverse change is not significantly different and forward shifts are roughly balanced by reversals. Theoretical arguments also contradict directional selection as a causal mechanism. Whether directional selection can produce a measurable trend is not at question in this study. However, whether directional selection can persist over extremely long time scales (e.g., hundreds of thousands or millions of years) is a less likely possibility (Sheldon 1996), and it is unlikely (and in fact almost certainly not the case) that hypercarnivores would have been continuously exposed to the same selective factors through their entire history. Over the history of any clade, it is reasonable to anticipate that there will be periods where various perturbing influences, including competition, climate change, or other ecological and environmental factors, will alter selection regimes and, hence, the direction of phenotypic evolution (Sheldon 1996). During such times, when the shape of the adaptive landscape itself is varying, some changes in the observed phenotype should also result (Sheldon 1996). This is a key point, since Van Valkenburgh et al. (2004) made an important observation regarding hypercarnivorous canids: these taxa do not exhibit adaptive change when their environments change (i.e., do not exhibit shifts in phenotype when selection is relaxed). Instead, they have a tendency to go extinct (Van Valkenburgh et al. 2004). While these workers’ results can only be applied to the taxa studied (canids), both their findings and those presented here support the possibility that hypercarnivores as a group are strongly limited in their ability to respond to environmental/ecological change, at least in the sense of reversing to a more generalized condition. It is the lack of reversals to a more generalized condition, rather than directional selection towards the specialization, that appears to influence the evolution of hypercarnivory Now that the “how” has been established (low rate of reversal), the question that follows is, of course, “Why?” Why can hypercarnivorous taxa not reverse or shift their phenotype as selective pressures change? What processes might be acting to bring about this constraint? Given the present data, several possibilities remain, including genetic or developmental limitations, and functional constraint. The generally accepted definition of developmental constraint is “a bias or limitation on the production of variant phenotypes” (Maynard-Smith et al. 1985), and is typically viewed as the result of selection on gene interactions and gene products during embryological development (Arthur 2001; Fusco 2001; Salazar-Ciudad 2006). Since developmental constraint is directly predicated on the viability of specific gene combinations (Maynard-Smith et al. 1985; Fusco 2001; Schwenk and Wagner 2001; Salazar-Ciudad 2006), a selective bias in the production of certain phenotypes could feasibly occur for any number or reasons (Arthur 2001, 2004; Fusco 2001). Having said that, it is recognized that developmental constraint will be difficult to identify, much less reject, without significant detailed study (Arthur 2001; Fusco 2001; Schluter et al. 2004). In this case, the use of replicated sister-group comparisons is the best available evidence against such a constraint: it is unlikely (although by no means impossible) that the same developmental bias against reversals would appear repeatedly in different taxa (but see, e.g., Schluter et al. 2004). In contrast to developmental constraint, functional constraint affects variants of the phenotype from the standpoint of phenotypic construction or integration, and maintains the ability

36

of a character complex to function effectively as a unit by eliminating less well-suited variants of that whole (Wagner and Schwenk 2000; Schwenk and Wagner 2001). Further, theoretical predictions involving morphological integration, arguably the most extreme version of functional constraint, suggest that the functionality of a particular suite of characters is affected not just by the existence of a complex of correlated (and functionally related characters), but by the specific interactions among the characters (Olson and Miller 1958; Schwenk and Wagner 2001; Schwenk and Wagner 2004; see also Eble 2004). In fact, an integrated unit may lose its ability to function at all if an individual character varies outside a particular range (Wagner and Schwenk 2000), and such integration would allow for only limited phenotypic change in any character at any time (Wagner and Schwenk 2000). At the same time, however, any change that acts to improve the fit of the characters to the functional unit (i.e., forward change toward increasing integration, or in this context, increasing specialization) would be selected for, while reverse changes in any one characteristic, which would decrease the integration of the characters and hence negatively affect functionality, would be strongly selected against (Bull and Charnov 1985; Wagner and Schwenk 2000). In the case of hypercarnivores, then, change in individual characters in the direction of increased adaptation to meat-eating would be reinforced - or at least not selected against -, while reverse change, particularly for individual characters, would exhibit a very low rate of occurrence. This is precisely the pattern the current data exhibits.

CONCLUSIONS

Studies to date have evaluated the evolution of hypercarnivory at steadily increasing levels of detail. The patterns described here are clearly the result of some constraint – either developmental or functional or both. Determining which of these possibilities is the most likely – and whether there is a “point of no return” in hypercarnivore morphospace, after which reversals cannot occur - will require even more fine-scale study, including a better understanding of character correlations and development within Carnivora. Studies with the specific intent of quantifying types and amounts (degrees) of character correlations and levels of morphological integration would be ideal, and some work has already begun in this area (Meiri et al. 2005; Goswami 2006). Improvements in quantitative methods should allow more specific testing of pattern/process hypotheses by enabling the comparison of developmental data for various species to the evolution of character complexes over time and across phylogenies (Steppan 1997b; Marroig and Cheverud 2001). Furthermore, recent methodological advances in evaluating and comparing the structure of phenotypic variance-covariance matrices at multiple phylogenetic levels (Steppan 1997a, 1997b; Marroig and Cheverud 2001; Baker and Wilkinson 2003) may offer significant insights into the myriad factors that affect evolutionary patterns. Hypercarnivores have a significant difference in their rates of forward change to reverse change for both relative blade length and for the hypercarnivore complex (Tables 10, 11). Comparison with sister groups indicates that this is due to a decreased rate of reversal, not an increased rate of forward change. These findings are little affected by character optimization (ACCTRANS or DELTRANS) or minor modifications of the approach (inclusion or exclusion of basal branches) although inclusion of basal branches (which allows inclusion of Cryptoprocta ferox and its sister group) generally strengthens the findings, almost certainly because of the concomitant increase in statistical power. Thus, instead of support for the prediction of directional evolution toward a longer shearing blade or a more specialized hypercarnivore phenotype, I instead found a limitation on reversals to a more primitive condition. These results indicate that the evolution of the hypercarnivore specialist morphotype proceeds very much as an evolutionary ratchet, allowing taxa to passively move forward into an increasingly specialized morphospace

37

but actively inhibiting reversals to a more primitive (= generalized) condition. The finding that hypercarnivorous taxa are able to move only in a forward direction (and at no significantly higher rate of forward change), leads to the inescapable conclusion that these taxa are under the influence of a strong constraint. The approach described here, which combines the frequency of change method of Sanderson (1993) with the comparative tool of replicated sister group comparisons, is a promising new way to handle old macroevolutionary questions – in this case, to test for directionality in character evolution, or gain:loss bias. It should be emphasized that the results presented in this paper must be interpreted cautiously; as with any character evolution study, new phylogenies can potentially alter these results (Pagel and Harvey 1988). Despite these issues, many additional applications and extensions of this methodology are readily apparent: sister group comparisons, character comparisons, comparison of ecomorphs or time series data. Further, as noted by Sanderson (1993), power may be increased by including sets of characters that, as with hypercarnivores, are all expected to exhibit the same behavior, while rigor may be increased by incorporating maximum likelihood estimates or temporal information about branch lengths. In its most basic form, however, the method requires only a phylogeny, making it accessible to workers in a variety of fields.

38

TABLE 5. Generalist forward and reverse change. Rates of forward and reverse change for generalist groups for relative blade length and the hypercarnivore complex under both ACCTRANS and DELTRANS optimizations. There is no significant difference between rates of forward change and rates of reversal under either optimization.

ACCTRANS DELTRANS

Relative Blade Length Forward Reverse Forward Reverse

Mustelidae .15 .11 .13 .09 Herpestidae .08 -- .08 -- Hesperocyoninae .14 .02 .14 .02 Canidae .11 .06 .11 .06 No significant difference No significant difference

Hypercarnivore Complex Forward Reverse Forward Reverse

Mustelidae .06 .09 .07 .09 Herpestidae .10 .08 .08 .07 Hesperocyoninae .08 .02 .08 .02 Canidae .09 .06 .10 .05 No significant difference No significant difference

39

TABLE 6. Frequency of change for RBL without basal. Frequency of change values for relative blade length for hypercarnivores and their sister groups under both ACCTRANS and DELTRANS optimizations. Calculations performed without basal branches.

ACCTRANS Clade Hypercarnivores Sister Groups Forward Reverse Forward Reverse 0.10 0 0.50 0.50 Enhydrocyon Cynodesmus Felidae 0.60 0.08 Viverridae 0.08 0.12 0.17 0 0 0.33 Mustela Enhydra/Lutra 0.19 0.10 0.08 0.18 Chasmaporthetes Hyena/Crocuta Nimravidae 0.18 0.08 Aeluroidea 0.10 0.11

DELTRANS Clade Hypercarnivores Sister Groups Forward Reverse Forward Reverse 0.10 0 0.50 0.50 Enhydrocyon Cynodesmus Felidae 0.40 0.06 Viverridae 0.03 0.32 0.17 0 0 0.33 Mustela Enhydra/Lutra 0.04 0 0.14 0.11 Chasmaporthetes Hyena/Crocuta Nimravidae 0.19 0 Aeluroidea 0.10 0.11

40

TABLE 7. Frequency of change for RBL with basal. Frequency of change values for relative blade length for hypercarnivores and their sister groups under both ACCTRANS and DELTRANS optimizations. Calculations performed with basal branches.

ACCTRANS Clade Hypercarnivores Sister Groups Forward Reverse Forward Reverse

Enhydrocyon 0.09 0 Cynodesmus 0.11 0.06 Felidae 0.50 0.08 Viverridae 0.07 0.08

Mustela 0.16 0 Enhydra/Lutra 0.09 0.19

Chasmaporthetes 0.18 0.09 Hyena/Crocuta 0.12 0.07 Nimravidae 0.17 0.07 Aeluroidea 0.06 0.06

Cryptoprocta 0 0 Eupleres/Fossa 0 0.28

DELTRANS Clade Hypercarnivores Sister Groups Forward Reverse Forward Reverse

Enhydrocyon 0.09 0 Cynodesmus 0.33 0.33 Felidae 0.40 0.06 Viverridae 0.09 0.17

Mustela 0.16 0 Enhydra/Lutra 0 0.29

Chasmaporthetes 0.33 0 Hyena/Crocuta 0.13 0.10 Nimravidae 0.19 0 Aeluroidea 0.10 0.11

Cryptoprocta 0 0 Eupleres/Fossa 0 0.33

41

TABLE 8. Frequency of change for hypercarnivore complex without basal. Frequency of change values for the hypercarnivore complex for hypercarnivores and their sister groups under both ACCTRANS and DELTRANS optimizations. Calculations performed without basal branches.

ACCTRANS Clade Hypercarnivores Sister Groups Forward Reverse Forward Reverse

Enhydrocyon 0.09 0.03 Cynodesmus 0.25 0.13 Felidae 0.13 0.04 Viverridae 0.08 0.12

Mustela 0.09 0.03 Enhydra/Lutra 0.10 0.22

Chasmaporthetes 0.11 0.02 Hyena/Crocuta 0.14 0.08 Nimravidae 0.14 0.10 Aeluroidea 0.06 0.06

DELTRANS Clade Hypercarnivores Sister Groups Forward Reverse Forward Reverse

Enhydrocyon 0.09 0 Cynodesmus 0.25 0.13 Felidae 0.11 0.02 Viverridae 0.07 0.11

Mustela 0.05 0.06 Enhydra/Lutra 0.10 0.16

Chasmaporthetes 0.15 0 Hyena/Crocuta 0.18 0.04 Nimravidae 0.14 0.04 Aeluroidea 0.06 0.06

42

TABLE 9. Frequency of change for hypercarnivore complex with basal. Frequency of change values for the hypercarnivore complex for hypercarnivores and their sister groups under both ACCTRANS and DELTRANS optimizations. Calculations performed with basal branches.

ACCTRANS Clade Hypercarnivores Sister Groups Forward Reverse Forward Reverse

Enhydrocyon 0.12 0.02 Cynodesmus 0.11 0.06 Felidae 0.11 0.04 Viverridae 0.07 0.08

Mustela 0.08 0.03 Enhydra/Lutra 0.09 0.19

Chasmaporthetes 0.08 0.02 Hyena/Crocuta 0.12 0.07 Nimravidae 0.13 0.09 Aeluroidea 0.06 0.06

Cryptoprocta 0.33 0 Eupleres/Fossa 0 0.28

DELTRANS Clade Hypercarnivores Sister Groups Forward Reverse Forward Reverse

Enhydrocyon 0.12 0 Cynodesmus 0.11 0.06 Felidae 0.11 0.02 Viverridae 0.09 0.06

Mustela 0.04 0.05 Enhydra/Lutra 0.08 0.14

Chasmaporthetes 0.10 0 Hyena/Crocuta 0.14 0.03 Nimravidae 0.14 0.03 Aeluroidea 0.06 0.06

Cryptoprocta 0.25 0 Eupleres/Fossa 0 0.33

43

TABLE 10. Summary of p values for Relative Blade Length. The table shows p values obtained from comparisons of ACCTRANS and DELTRANS optimizations as well as results obtained with and without the basal branches. The first column indicates the type of comparison; the top row whether comparisons were within taxon or between sister groups. The column labeled Hypercarnivores FOR:REV shows p values obtained from comparisons of relative rates of forward change to relative rates of reversal within hypercarnivore clades; the column labeled Sister FOR:REV indicates p values obtained from comparisons of relative rates of forward change to relative rates of reversal within the sister groups of hypercarnivores; The column labeled FORWARD indicates the p values obtained from sister group comparisons (hypercarnivores versus sister groups) for rates of forward change; the column labeled REVERSE shows p values from sister group comparisons for rates of reverse change.

WITHIN TAXON SISTER GROUP COMPARISONS

FOR:REV FOR:REV FORWARD REVERSE Hypercarnivores Sister

ACCTRANS .042 .043 .345 .027 with basal ACCTRANS .042 .068 .345 .043 without basal DELTRANS .043 .138 .345 .027 with basal DELTRANS .043 .273 .500 .043 without basal

44

TABLE 11. Summary of p values for the hypercarnivore complex. Values for the hypercarnivore complex are calculated as the average of values for the following characters: RBL, p4 shape, blade/GSL, P4/M1, M1/TRL, GSL/m1. Characters are explained in the text. Table format follows that of 10.

WITHIN TAXON SISTER GROUP COMPARISONS

FOR:REV FOR:REV FORWARD REVERSE Hypercarnivores Sister

ACCTRANS .028 .463 .173 .046 with basal ACCTRANS .043 .893 .893 .080 without basal DELTRANS .046 .917 .463 .028 with basal DELTRANS .080 .686 .686 .043 without basal

45 Clade A Clade B

0 A without basal B without basal 1 2 A and B with basal

FIGURE 7. Method for calculating frequency of change. Method for calculating frequency of change for multistate characters for a hypothetical sister group pair. Characters are polarized and ordered, with 0 being the least specialized and increasing specialization indicated by increasing numbered states. Any forward change counts as one step, even if a state is skipped (0-1 or 0-2 both = 1 forward step). Reversals are calculated in the same way (1-0 or 2-0 both = 1 reverse step). The total number of forward or reverse changes is divided by the number of branches that could possibly change (= opportunities for change). Branches already in state 0 cannot reverse further and are excluded when counting branches for reversals. Likewise, branches already in a highest terminal state cannot experience additional forward change and are excluded when counting branches for forward changes. The value obtained from these calculations is the frequency of change, and these values can be averaged over multiple characters for an average frequency of change. Calculations: Clade A: Without basal. There are 5 forward changes and 1 reverse change. There are 26 branches that can possibly move forward (branches in state 2 are excluded). There are 18 branches that can possibly reverse (branches in state 0 are excluded). Clade A. With basal. This adds one more branch to the calculations of forward change and no branches to the calculations of reverse change (this branch is already at state 0). Clade B. Without basal. There are 2 forward changes and no reverse changes. There are 22 branches that could possibly move forward and 4 branches that could possibly reverse. Clade B. With basal. This adds one more branch that could move forward and no additional branches that could reverse. Frequency of forward change: A w/out basal = .20; A with basal = .19; B without basal = .09; B with basal = .09. Frequency of reverse change: A w/out basal = .08; A with basal = .08; B without basal = 0; B with basal = 0.

46

CHAPTER 3 PHYLOGENY OF THE FELOID CARNIVORA: BALANCED TAXON SAMPLING AND CONCATENATED NUCLEAR GENES PROVIDES RESOLUTION AT DEEPER NODES

The order Carnivora (Mammalia) is a diverse group of 13 living and two extinct families that includes taxa such as the (seals and ), canids (dogs), ursids (bears), felids (cats), and hyaenids (hyaenas). Since their common ancestor, carnivorans have diversified widely into a variety of ecological niches and adaptive morphotypes, including durophagous, insectivorous, omnivorous, and even strictly herbivorous taxa such as the Giant Panda, an ursid. Carnivorans also exhibit high rates of guild structuring and phenotypic convergence (Van Valkenburgh 1988, 1989), and similar morphotypes have appeared repeatedly during the course of this clade’s history, including hypercarnivorous, osteophagous, and saber-toothed forms (Van Valkenburgh 1988, 1989, 1991, Werdelin 1996; Holliday and Steppan 2004). The initial carnivoran radiation occurred approximately 45 million years ago (mya) (Wesley-Hunt and Flynn 2005), and proceeded in two waves: an - radiation that produced the caniform (-like) families such as Canidae, Ursidae, and Mustelidae, and a later radiation during the early Miocene of that led to the feliform (cat-like) clades, including Felidae (cats), Hyaenidae (hyaenas), Viverridae (civets and genets), Eupleridae (Malagasy carnivorans), Herpestidae (mongooses), and Nandiniidae. Within the superfamily Feloidea, there are 113 living species, including 37 species in Felidae, four species within Hyaenidae, 34 species in Viverridae, 33 species in Herpestidae, six species in Eupleriidae and one in Nandiniidae (Wozencraft 2005). The feloids are a particularly intriguing group of carnivorans because they exhibit not only a variety of phenotypes but also significant variation in their behaviors and ecologies. For example, the solitary aardwolf, a hyaenid, is an -specialist with vestigial teeth, while its close relative, Crocuta crocuta, the spotted hyaena, is better known for its predatory and bone-crushing abilities. Within Felidae, social systems range from the male-dominated pride of leo (), to the solitary P. pardus, (), which has been suggested to be P. leo’s sister taxon (Janczewski et al 1995, Jae-Heup et al. 2001). In herpestids, mongoose species range in their behaviors from highly predaceous, solitary types to highly social colonies that exhibit cooperative care such as guarding and babysitting. This same level of diversity also characterizes ecological niche and habitat preferences: feloid habitats include high mountains (Panthera uncia, the ), deserts (Felis libyca and F. margarita), forests (P. onca, P. pardus and many of the viverrids), and savannah (hyaenas and many mongooses). The phenotypic variety observed within the feloid carnivorans serves to make this group of particular interest to workers who study character evolution and ecological or behavioral correlates (see, e.g. Hunt 1989, 1991, 1998; Werdelin 1996; Mattern and McClennan 2000, Gaubert et al 2004; Holliday and Steppan 2004; Veron et al 2004). Such studies, however, require a robust phylogenetic hypothesis in order to effectively understand patterns of change. To date, there is no well-resolved, comprehensive phylogeny of the Feloidea as a whole, and this limits efforts to understand patterns of taxonomic diversification and to test evolutionary hypotheses. In order to address this need, molecular information was used to obtain a well-sampled, robust phylogeny of the feloid Carnivora.

47

History of Phylogenetic Study

For many decades, there has been diligent effort to elucidate relationships both within and between the members of Feloidea, and careful, comprehensive analyses, both morphological and molecular, have been produced for nearly every family within this group (e.g. Wozencraft 1984; Werdelin and Soulinias 1991; Salles 1992; Mattern and McClennan 2000; Johnson et al 2006; Koepfli et al. 2006). Despite this history of work, however, there has until now been little success in elucidating relationships within the feloids as a whole, particularly in relation to the deeper nodes of the phylogeny (e.g., between families and major subclades). These nodes are of especial importance because they establish the identity of the most basal taxon and deepest divergence(s) – and hence the likely primitive condition – necessary for evaluating patterns of subsequent evolutionary change. There are several likely reasons for the historical difficulty in establishing phylogenetic relationships within the Feloidea. First, convergence in commonly used morphological characters, particularly in the skull and dentition, in combination with other, highly derived or autapomorphic characters, has led to much confusion regarding primitive vs. derived character states. Molecular data has proven problematic as well, since rapidly evolving mitochondrial genes appear to be ineffective for several feloid lineages due to saturation (Gaubert et al. 2003; Gaubert et al. 2004; Koepfli et al. 2006, but see Masuda et al. 1996; Yu and Zhang 2005 for felids), while there is also evidence for incomplete lineage sorting in other clades (see, e.g., Masuda et al. 1996; Johnson and O’Brien 1997; Johnson et al. 2006, for felids). Taxon sampling has also been problematic within the feloids. Since the most thorough analyses have focused only on a single family or subfamily within the Feloidea (e.g. Wozencraft 1984; Gaubert et al. 2003; Gaubert et al. 2004; Veron et al. 2004; Yu and Zhang 2005; Johnson et al. 2006; Koepfli et al. 2006), both taxon and gene sampling for Feloidea as a coherent whole has, until very recently, been extremely limited (but see Flynn et al. 2005). Instead, high profile clades such as felids have received the bulk of the attention from researchers (Salles 1992; Johnson and O’Brien 1997; Mattern and McClennan 2000; Yu and Zhang 2005; Johnson et al. 2006), while the small, rare herpestids have only recently begun to generate interest from molecular systematists (see, e.g., Veron et al. 2004; Perez et al. 2006). Viverrids, a diverse and charismatic group of small to medium sized carnivorans, have been periodically included in molecular analyses, but these analyses typically consider members of a particular genus or geographical area (Gaubert et al. 2003; Gaubert and Veron 2003; Gaubert et al. 2004; Gaubert et al. 2006), and few researchers have attempted to evaluate molecular relationships within the entire family Viverridae (but see Gaubert and Cordeiro-Estrala (2006) for ). The need for more comprehensive analyses has been underscored in the past several years as recent molecular work has indicated that Nandinia binotata, the palm civet, may be sister to all remaining extant feloids, while Cryptoprocta ferox (the Malagasy fossa), Fossa fossana (the falanouc), and Eupleres goudottii (the Malagasy civet), previously considered viverrids (Wozencraft 1984) all appear to be part of an endemic Malagasy herpestid radiation (Yoder et al. 2003; Veron et al. 2004; Flynn et al. 2005; Gaubert and Cordeiro-Estrela 2006).

Hyaenidae/Eupleridae/Herpestidae Relationships

In a review of the then-available data for feloid phylogenetics, Flynn et al. (1988), noted that virtually every possible combination of families has been proposed – and supported - at one time or another. Nearly 20 years later, the situation has still not been fully resolved: Flynn et al. (2005) performed a multigene molecular analysis of the feloid Carnivora and established only that Hyaenidae and (Eupleridae/Herpestidae) are sister taxa, a conclusion that has been reached in several recent molecular studies (see, e.g. Gaubert and Veron 2003; Yoder et al. 2003; Yu and

48

Zhang 2004; Flynn et al. 2005; Gaubert and Cordeiro-Estrela 2006; Koepfli et al. 2006). However, regarding the relationships between Felidae/Viverridae/(Hyaenidae(Eupleridae/Herpestidae)), Flynn concluded that, at present, the best way to present the initial feloid divergences is as a polytomy (Figure 8). Thus, despite ongoing efforts to elucidate feloid relationships, neither careful morphological studies nor molecular data have provided the desired clarification for this group. While several recent nuclear + mitochondrial multigene datasets have indicated that Felidae may be sister to a clade containing ((Hyaenidae(Eupleridae/Herpestidae))Viverridae) (Yoder et al. 2003; Flynn et al. 2005; Koepfli et al. 2006), other datasets suggest that it is Viverridae that is sister to ((Hyaenidae(Eupleridae/Herpestidae))Felidae) (Gaubert et al. 2003; Flynn et al. 2005). Given these issues, it is clear that establishing the order and patterns of speciation and morphological diversification within and among these families will not be easily solved. In the past, many phylogenetic studies that attempted to achieve broad taxonomic coverage did so at the expense of taxonomic depth (see, e.g., Wyss and Flynn 1993; Flynn and Nedbal 1998; Koepfli et al. 2006 for carnivorans). Thus, the general approach for performing higher-level molecular analyses (within Feloidea or other groups) had been to include a few representative members of each clade of interest and then to assess relationships between those representatives (Zwickl and Hillis 2002). There are many valid reasons to use “exemplars” – tissue availability and cost are obvious considerations. However, there is a growing recognition that more extensive sampling is important in both lower and higher-level analyses (Maddison and Knowles 2006, Zwickl and Hillis 2002, Rannala et al. 1998). Recognizing the previous difficulties with evaluating feloid relationships, a key aspect of the design of this study was to sample as extensively and as evenly as possible throughout the feloid families, and to include three unlinked nuclear genes (RAG-1, GHR and c-myc) as well as mitochondrial (cyt b) sequence data. This study samples 45% of all species of felid (17/38), 50% of hyaenids (2/4), 28% of herpestids (11/39), and 24% of viverrids (9/35), as well as all members of Eupleridae and Nandiinidae. This sampling represents over 50% of all extant genera (26/45 total), and the tree obtained represents the most robust, thoroughly sampled phylogeny yet available for the feloid Carnivora. Here I present and describe the results of a phylogenetic analyses of the Feloidea. This study addresses specific ongoing questions regarding the affinities of particular taxa, including Nandinia, Cryptoprocta, and, in particular, the phylogenetic relationships between the families Felidae/Hyaenidae/Eupleridae/Herpestidae/Viverridae. The phylogeny will be a valuable resource for workers who wish to test ecological and/or evolutionary hypotheses, and will allow better estimates of rates of evolution, tests of evolutionary flexibility, or study of ecomorphological patterns.

MATERIALS AND METHODS

Taxon Sampling

This phylogeny includes 39 members of the superfamily Feloidea and five outgroup taxa. Tissue samples included muscle, hide, or blood samples, were obtained from zoos and museums in and Africa. Voucher information is listed in Appendix C. Some sequences were obtained from GenBank, and their accession numbers are also shown in Appendix C.

Genes

The nuclear gene RAG-1 is over 3000 base pairs in length and exhibits different rates of evolution in different regions of the gene. I sequenced 2162 base pairs (bp), including the first 1000 bp, known as the divergent region (Fugmann et al. 2000; Steppan et al. 2004b) as well as

49

the more commonly sequenced but relatively conserved middle region. RAG-1 has relatively uniform substitution properties that have been demonstrated to maximize phylogenetic signal (G. Naylor, pers. comm.), and has been used with success in a variety of organisms, including carnivorans, rodents, bats, and birds (Barker et al. 2001; Ericson et al. 2002; Teeling et al. 2002; Steppan et al. 2004a; Steppan et al. 2004b; Koepfli et al. 2006). In addition to RAG-1, I sequenced 900 bp of c-myc, a proto-oncogene that has been successful both at the family level and interordinally for rodents and other vertebrates (Miyamoto et al. 2000; Barker et al. 2001; Ericson et al. 2001; Harshman et al. 2003; Steppan et al. 2004a, Steppan et al. 2004b). Finally, a 934 bp exon of the nuclear gene GHR (growth hormone receptor) was sequenced. This exon has also been used successfully within feloids in previous studies (Koepfli et al. 2006). GHR has a uniform base composition and low among site rate variation that provides strong phylogenetic signal in mammals; recent work in rodents indicates that GHR evolves at a rate comparable to that of the 5’ end of RAG-1 (Steppan et al. 2004b). Because the nuclear genes in this dataset were selected for their ability to resolve deeper nodes and not for optimal resolution at the tips, I chose to include a more rapidly evolving, mitochondrial gene in the dataset as well. Cytochrome b data was available on GenBank for the majority of the ingroup taxa, and sequences for this gene were downloaded and analyzed in the combined dataset and as an additional data partition. Sources of cytochrome b are listed in Appendix C.

Extraction and Sequencing

Total genomic DNA was extracted from frozen or ethanol-preserved tissue samples by PCI (phenol/chloroform/isopropanol)/CI (chloroform/isopropanol) “hot” extraction (Sambrook et al., 1989). For PCR, one of two protocols was used. The first included a standard 24 ul PCR mix comprised of dH20, dNTPs, MgCl, DMSO, Amplitaq gold DNA polymerase, and 1-2ul of DNA or, alternatively, I utilized 22.5 ul of PCR Supermix (Invitrogen, Carlsbad, CA) with 0.4 ul of each primer and 1ml of template. Successful amplification was achieved in many cases using general mammalian primers. Where necessary, carnivore specific primers were designed. Primers used for each of the genes are listed in Table 12. For most genes, a generic PCR program set to an amplification temperature of 58 degrees was successful in amplifying the segments of interest. In specific instances, these programs were modified slightly to optimize results. Negative (no DNA) controls were included in every reaction to reveal possible contamination of reagents. PCR products were visualized on an agarose gel with ethidium bromide, and successful amplifications were prepared directly by enzymatic digestion with Exo-SAP-IT (USP, Cleveland). PCR products were sequenced at the Florida State University sequencing lab (Tallahassee, Florida) using an Applied Biosystems 3130X1 Genetic Analyzer with Capillary Electrophoresis. Despite varying the protocol (e.g., changes in temperature and MgCl concentrations), some taxa could not be successfully amplified. These included the following: for RAG-1, region 2: Felis silvestris, Chrotogale owstoni, Genetta tigrina and Genetta sp. For c-myc: Ichneumia, Arctictis , and hermaphroditus. Felis concolor did not amplify for RAG-1 or c-myc, but I was able to obtain a successful amplification for GHR. For RAG-1 and GHR, sequence editing and alignments were unambiguous and were performed using the program Sequencher 4.7. Additional sequence data for these genes was available on GenBank for the following taxa: the red , vulpes; the , latrans; the American , americanus; and the , Cuon alpinus. Alignment of the third nuclear gene, c-myc, was also largely unambiguous, but did contain two small (5bp, 8bp) insertions for three species of herpestid, the Malagasy genera Mungotictis (narrow striped mongoose), Galidia (ring-tailed mongoose) and (broad-striped mongoose). C-myc

50

sequence was available on GenBank for Ursus americanus. Sequences were downloaded from GenBank for the cyt b gene for all ingroup taxa excepting the felids Felis viverrina (), rufus (), F. manul (Pallas’ cat), F. margarita (), F. temmincki (golden cat), F. (caracal). F. yagouaroundi (the jagouraoundi),and Panthera onca (the ). The caniform taxa Vulpes vulpes, Canis latrans, Ursus americanus and Cuon alpinus were designated as outgroups for the cyt b analysis.

Phylogenetic analysis

Parsimony and maximum likelihood (ML) analysis of these data sets were performed with the program PAUP* (Swofford 2004) and data were analyzed as individual genes and in combination (all genes). Each dataset was analyzed under parsimony (heuristic search algorithm, tree bisection-reconnection (TBR), and 500 maximum trees, with all characters receiving equal weight) and also under a ML model selected via the program Modeltest (Posada and Crandall 1998), which suggests a best fit model using either the likelihood ratio test (Huelsenbeck and Crandall 1997) or the Akaike information criterion (Akaike 1974). In most cases the model indicated by Modeltest was identical under either criterion; where disagreement occurred, the more complex model was applied (Posada and Crandall 1998). Specific model settings are provided in Appendix D, and all ML analyses were performed under heuristic search conditions, with tree bisection-reconnection branch swapping and 10 random addition replicates. A Bayesian analysis was conducted on the combined data set with MrBayes (Huelsenbeck and Ronquist 2001) using the GTR + I + gamma model. Each gene was designated as a partition (subset) and model parameters for each subset were estimated separately (“unlinked”) in the analysis; each subset had its own rate multiplier. The analysis assumed flat priors, and included two sets of four MCMC chains for 3,000,000 generations, sampling trees every 100 generations with a heating parameter of 0.2. Based on graphical output from AWTY (Wilgenbusch et al 2004), and the number of generations before the 20 most variable bipartitions became stable, I applied a 33% burn-in, discarding the output for the first 1,000,000 generations. A majority rule consensus of the remaining 20,000 trees was obtained in PAUP* (Swofford 2004). All datasets and topologies were compared for bootstrap (500 replicates) and Bayesian support values.

RESULTS

Combined analysis of all genes

The Bayesian and ML analyses of the combined dataset produced an identical phylogeny (Figure 9). Parsimony and ML bootstrap values and Bayesian support are indicated in the figure. Hyaenidae and Herpestidae are sister taxa, while the genus is paraphyletic. The family Herpestidae, however, is supported with 100% bootstrap values and can be divided into a herpestine radiation consisting of (Herpestes, , Ichneumia), with Suricata and as successive sister taxa, and an endemic Malagasy radiation that is comprised of (Mungotictis, Galidictis) Galidia) with Cryptoprocta and Fossa as successive sister taxa. Hence, Cryptoprocta ferox, along with Fossa fossana, are part of a Malagasy herpestid radiation. Crocuta and Proteles, representative of the Hyaenidae in this dataset, are consistently monophyletic with 100% support values in all analyses. In most analyses, the Felidae are sister to the (Herpestidae/Hyaenidae) clade. Within Felidae, I found strong (100%) support for a (pantherine) lineage comprised of ((Panthera uncia)(P. pardus/P. leo)(P. tigris/P. onca)). The remaining, smaller cats also formed a clade with 100% support. The sister group relationship between Panthera leo and P. pardus was

51

recovered in nearly all analyses at 100% bootstrap, while the sister group relationship between P. tigris and P. onca received intermediate levels of support, depending on the gene used. Additionally, I recovered a strongly supported sister group relationship between Felis silvestris and F. margarita (100%), the Lynx lineage (100%), and a moderately well-supported Felis viverrina, , and F. concolor clade. The last feloid group of interest, Viverridae, consistently forms a monophyletic group. Within the viverrids, Paguma and Paradoxurus are sister taxa with consistently strong bootstrap support. Arctictis and Chrotogale form successive sister groups to the former taxa, and these topologies are again supported with support values over 70%. tangalunga and Civettictis are sister taxa, and the Genetta lineage is monophyletic; a sister group relationship between (Viverra tangalunga/Civettictis) and Genetta is indicated in several partitions and in the combined dataset but bootstrap support is low. Viverridae are shown to be sister to the (Felidae)(Hyaenidae/Eupleridae/Herpestidae) clade with moderate to good bootstrap support. Excepting cyt b under parsimony, all analyses show that Nandinia binotata is sister to the remaining feloids.

Gene-specific results

The parsimony and ML bootstrap analyses of RAG-1 did not differ from the results obtained in the combined analysis described above (Fig. 10A). For c-myc, many nodes in Felidae were not recovered. However, parsimony and ML bootstrap analyses for the c-myc partition did support several groupings at >70%, including (Acinonyx/Felis yagouaroundi), although this was the only resolved relationship within the small cats. Within the big cats, only Panthera tigris and P. uncia were recovered as sister taxa, and bootstrap support was low at 57% parsimony, 66% ML. All other relationships were recovered as described for the combined dataset, although bootstrap support in the c-myc partition relative to the combined analysis was slightly reduced for many of the shallower nodes (Fig. 10B). For GHR, an unexpected felid/viverrid sister group relationship was recovered with 80% parsimony and 74% ML bootstrap support. Other relationships were as per the combined analysis, excepting a poorly supported big cat clade (69% parsimony, 69% ML bootstrap) that GHR did not resolve and instead collapsed to a polytomy. The Lynx lineage and the sister group relationship between Felis margarita and F. silvestris were recovered in the GHR partition, but the remaining small cats were polytomous (Fig. 10C). In contrast to the consistent and congruent topologies obtained using the nuclear genes, the results of analysis of the cyt b data partition are characterized by incongruent topologies and unexpected taxonomic placements, as well as generally poor bootstrap support. For the cyt b partition, ML bootstrap and parsimony bootstrap recovered Fossa fossana as sister to the Galidine mongoose radiation (69% for ML and less than 50% support under parsimony), with Cryptoprocta ferox as sister to that clade. Felis silvestris was placed as sister to the big cats with 29% parsimony and 56% ML bootstrap support (Fig. 10D).

DISCUSSION

Much of the feloid phylogenetic research produced over the past decade has focused on improving data and taxon sampling (Johnson and O’Brien 1997; Masuda et al. 1996; Gaubert et al. 2004; Veron et al. 2004; Flynn et al. 2005; Yu and Zhang 2005; Gaubert and Cordeiro-Estrela 2006; Koepfli et al. 2006; Perez et al. 2006). This includes increasing the number of sampled genes and taxa, including nuclear genes where saturation is recognized for mitochondrial genes, and improved modeling of molecular evolution in analyses. Some workers have chosen to specifically evaluate different model parameters or weighting schemes in order to control for

52

known issues (see, e.g. Gaubert et al. 2004; Veron et al. 2004). Additionally, workers considering subfamily or genus level relationships (Gaubert et al. 2003; Gaubert et al. 2004; Veron et al. 2004), as well as those working at higher levels, have recognized the importance of increased taxonomic representation and have incorporated this into their analyses (e.g., Flynn et al. 2005). Particularly in the case of “problematic” groups, where there have been repeated attempts to elucidate a phylogeny (e.g., Fishbein et al. 2001 (Saxifragiles); Buckley et al. 2006 (cicadas); Flynn et al. 2005 (Carnivora); Poe and Chubb 2004 (Aves); Schwarz et al. 2006 (); Weisrock et al. 2006 (salamanders)), the question of more data (taxon or character) becomes extremely relevant. Maddison and Knowles (2006) recently pointed out that while taxon sampling is definitely important, balanced taxon sampling is even more so, at least in a set of recent simulation studies (Maddison and Knowles 2006). This study takes just such a balanced approach, so that, depite a relatively modest increase in number of species relative to other, recent studies (Flynn et al. 2005), sampling is much more evenly distributed throughout the families. The result is a better-resolved, better-supported phylogeny than has yet been produced. A comparison of topologies, branch lengths, and bootstrap support between the four genes in the dataset illustrates not only the stability of specific nodes across phylogenies, but also highlights areas of incongruence and suggests possible reasons why some groupings may not be well-supported. Thus, an analysis based on the mitochondrial gene cyt b, which is expected to be less informative at deeper nodes due to its higher rate of evolution and the risk of saturation ((see Gaubert et al. 2003; Gaubert et al. 2004; Koepfli et al. 2006), indicates a trichotomy at the primary divergence (feloids less Nandinia) and receives only 53% ML bootstrap support at that node. Saturation in this gene is a likely cause of the unexpected topologies obtained under both parsimony and ML, which include the placement of Fossa fossana as sister to the galidictine mongooses and of Nandinia as sister to (Felidae (Hyaenidae(Eupleridae/Herpestidae))). However, since recent felids are the product of a rapid radiation that occurred sometime within the past 10 my (Johnson and O’Brien 1997; Johnson et al. 2006), cyt b has not had enough time to undergo a high proportion of homoplastic changes within the felid lineage, and appears to still be effective at elucidating relationships for this clade (Masuda et al. 1996; Yu and Zhang 2005), despite being much less informative in closely related, but older groups. In contrast to the rapid evolution of cyt b, c-myc and GHR evolve relatively slowly, and as a result may accumulate too few changes to be useful in a separate (individual) analysis. Having said that, analysis of GHR produces a moderately supported ((felid/viverrid)((herpestid/euplerid)/hyaenid))) grouping (78% parsimony bootstrap, 68% ML). Most of the deeper divergences in these families are generally well supported, with bootstrap values ranging from 80-100%. However, the more shallow divergences, as expected with a slowly evolving gene, are either unresolved (polytomy) or have very low bootstrap values (Fig. 10C). C-myc, like GHR, was less helpful at the terminal branches and, particularly within Felidae, returned low bootstrap support for all clades. The monophyly of Felidae, however, was upheld, as was that of the clades Viverridae, Herpestidae, Eupleridae, and Hyaenidae. Viverridae was recovered as the sister to the remaining feloids, but was supported with only 48% bootstrap; likewise, ((Herpestidae/Eupleridae)/Hyaenidae), a clade that has been recovered in nearly every molecular analysis in recent years (Yoder et al. 2003; Flynn et al. 2005; Koepfli et al. 2006), received only 56% parsimony bootstrap support. Additionally, while a clade comprised of (Felidae (Hyaenidae/(Eupleridae/Herpestidae))) was recovered, bootstrap support was extremely low (29%) (Fig. 10B). While maximum likelihood also recovered this topology, resampling (ML bootstrap) produced a three-way polytomy, indicating low support for any one grouping. Given the rate of evolution of c-myc (approximately the same rate as GHR, exon portion slightly slower but intron slightly faster than RAG-1 [Steppan et al 2004a]), it is probable that this relatively slower rate of evolution provides too few informative characters for this level of analysis. The third nuclear gene, RAG-1, seems to be the most informative for this level of analysis, and produced the best-resolved and best supported phylogeny of the four individual

53

genes. RAG-1 appears to evolve slowly enough to keep the terminal branches relatively short, but rapidly enough that the internal branches accrue mutations that will inform the deeper divergences. It should be noted that RAG-1 also includes more sampled base pairs (and thus informative sites) than the other genes in this analysis. Support values for the RAG-1 data are uniformly high, and RAG-1 appears to be particularly useful in elucidating deeper divergences within feloids as a whole. Overall, parsimony bootstrap support was good, although parsimony analysis indicated sister-group relationships between (Felidae/Viverridae) (56% support) and (Hyaenidae(Eupleridae/Herpestidae)) (99% support). ML contrasted with parsimony in only very minor ways at the terminal portions of the phylogenies, but under likelihood Viverridae were placed as the earliest diverging of the feloids, while felids were sister group to (Hyaenidae(Eupleridae/Herpestidae)) (68% ML bootstrap support). Bearing this in mind, the RAG-1 topology and branch lengths also serve to highlight the speed of the diversification of the feloid families, and this gene suggests a radiation rapid enough to very nearly approximate a true hard polytomy (Fig. 10A). Finally, analyzing these genes in combination offers the best-supported topology. When all of the genes are concatenated, parsimony and, in particular, ML bootstrap results are high (90-100%) for nearly all of the deeper nodes in the tree. Interestingly, a parsimony analysis of the combined data recovers a topology that places Felidae as the sister to the (Viverridae((Herpestidae/Eupleridae)Hyaenidae)) clades. However, given the high bootstrap values elsewhere on the tree, the low (47%) parsimony bootstrap support at the critical branch (that leading to (Viverridae((Herpestidae/Eupleridae)Hyaenidae)) becomes difficult to overlook. In contrast, both ML and Bayesian analysis indicate that Viverridae is the sister group to the remaining feloids (less Nandinia) and this grouping is supported by 77% ML bootstrap and 0.89 posterior probabilities. As noted in other analyses, the branch leading to (Felidae (Hyaenidae(Eupleridae/Herpestidae))) is extremely short (Fig.9).

Polytomies and Rapid Radiations

Recently, Flynn et al. (2006) suggested that, given the history of previous work and the information available at present, the basal divergence of Feloidea is best described as polytomy. A hard polytomy is identified as the three or more new lineages deriving simultaneously from a single ancestral lineage (DeSalle et al. 1994; Slowinski 2001; Poe and Chubb 2004). A soft polytomy, on the other hand, is simply the appearance of a polytomy, when in reality there is not enough information (in the form of data, either characters or taxa) to identify which lineage diverged first (Maddison 1989; Poe and Chubb 2004). It can be difficult to discriminate between the two (Braun and Kimball 2001; Poe and Chubb 2004), however, some general criteria have been identified: in a hard polytomy, there is no congruence between gene trees and no observable “central tendency” or majority signal (Maddison 1997; Poe and Chubb 2004). Gene trees are essentially randomized relative to the pertinent node (Slowinski 2001; Poe and Chubb 2004), and branches on all gene trees are consistently short (or nonexistent) at the node in question (Slowinski 2001; Poe and Chubb 2004). In contrast, a soft polytomy should begin to exhibit resolution given additional data, and a consensus should eventually emerge from multiple gene trees (Maddison 1997; Poe and Chubb 2004). Further, there will be resolution at the node in question, even if the branch is extremely short (Slowinski 2001, but see Poe and Chubb 2004 for zero length branches). The analyses presented here indicate that, despite suggestions of polytomy, the earliest divergence is simply represented by a very short branch, as indicated by the partitioned data and also the combined evidence tree. However, although the feloid radiation may not represent a hard polytomy in the sense of a truly simultaneous divergence, what is clear is that the diversification of this clade was extremely rapid (see also Flynn et al. 2006), and consequently there was only a

54

very short time between speciation events for informative changes to accumulate in the relevant genes. It is a paradox of genetic evolution that the genes that might evolve rapidly enough to accrue information at the level of these rapid, deep divergences, are also the genes that are likely exhibit saturation before they reach that level (Maddison and Knowles 2006). This study was specifically designed to correct for these issues, and included multiple genes evolving at different rates as well as detailed, balanced taxonomic sampling across all families, an approach that enabled me to establish the pattern of the initial divergences more clearly than in any previous studies.

CONCLUSIONS

The trees presented here represent the most thoroughly sampled, robust phylogenetic analyses yet available for the recent feloid carnivorans. Importantly, this work provides the first solid evidence that Viverridae is sister to a clade comprised of (Felidae(Hyaenidae(Eupleridae/Herpestidae))). I further confirm that Nandinia is sister to the remaining feloids, and that Cryptoprocta and Fossa are part of an endemic Malagasy radiation. Now that the pattern of divergence among the families has been clearly established, this new phylogeny will enable workers to assign character polarity more accurately, which will significantly improve the quality of research for workers seeking to understand character evolution or to test ecomorphological, behavioral, or biogeographical hypotheses.

55

TABLE 12. PCR primer sequences for genes used in this study. Target Gene Primer name Primer type Nucleotide sequence 5’-3’

RAG-1(1) S211 External GGGTGMGATCYTTTGAAA S212 External CVGTYCTGTACATCTTRTGRTA S213 Internal CAGCTYAGCAARAAACTMAAAAC S214 Internal ATCTGGCAGGAGATGGATTTCACAA RAG-1(2) S71 External TGGCTTCTGGTTATGGAGTGGA S118 Internal GAAGACATCTTGGAAGGCAT S119 Internal GAAGGGACCATTCAGGTAGTC S155 External TCCCTGCTTCCCTRCYGACCTG C-myc S91 External CCMAAGACYCAGCCAAGGTTGTGAGGT GHR S92 External RRAGCCTCATTAAGTCTTAGGTAAGAA S192 External GGRAARTTRGAGGAGGTGAACACMATCTT S193 Internal TTCTAYARYGATGACTCYTGGGT S194 Internal GTAAGGCTTTCTGTGGTGATRTAA S196 External CTACTGCATGATTTTGTTCAGTTGGTCTGTGCTCAC

56

TABLE 13. PCR programs. Generic PCR programs used for each gene or gene segment. Programs were modified as needed to optimize amplification results.

Temperature Time

RAG1 Lid 105 degrees 1 94 15:00 2 94 0:45 3 65.7 0:45 4 72 1:30 5 Go to 2, Repeat 40 6 72 6:00 7 Hold 4.0

C-myc Lid 105 degrees 1 94 5:00 2 94 0:35 3 58 0:35 4 72 1:00 5 Go to 2, Repeat 40 6 72 6:00 7 Hold 4.0

GHR Lid 105 degrees 1 94 5:00 2 94 0:30 3 50 0:30 4 72 1:30 5 Go to 2, Repeat 35 6 72 6:00 7 Hold 4.0

57 Outgroup

Nandinia binotata

Herpestidae

Hyaenidae

Felidae

Viverridae

FIGURE 8. Cladogram illustrating current understanding of relationships within Feloidea. A cladogram illustrating the current state of understanding regarding the basal divergences within Feloidea. Redrawn from Flynn et al. 2006.

58 Felis viverrina 93/- /59 Felis concolor

64/- /- Acinonyx jubatus

*/99/* Lynx canadensis 70/- /- Lynx rufus 81 /- /- Felis temmincki 87/ - / - Felis caracal Felidae */82/80 Felis yagouaroundi

*/99/* Felis margarita 91 /- /- Felis silvestris Felis manul */*/* 95/57/- Panthera tigris 96/-/- Panthera onca */65/65 Panthera leo */86/83 Panthera pardus */*/* Panthera uncia Neofelis nebulosa

77/-/- Herpestes sanguineus 89/77/- 95/90/73 Galerella pulverulenta */*/99 Ichneumia */*/* Herpestes javanicus Herpestidae Helogale parvula Suricata suricatta */*/* 100/83/74 Mungotictis decemlineata */*/* Galidictis fasciata 92/77/69 Galidia elegans * */*/* */*/* Cryptoprocta ferox

*/*/95 Hyaenidae Fossa fossana

*/*/* Crocuta crocuta Proteles cristatus

*/94/* Paguma larvata */*/* Paradoxurus hermphrotitus Viverridae */*/* */98/92 Arctictis binturong Chrotogale owstoni */*/* */*/* Viverra tangalunga Civettictis civetta 52/-/- */*/* Genetta sp. Genetta tigrina Nandinia binotata Canis latrans Cuon alpinus Vulpes vulpes Ictonyx striatus Ursus americanus

FIGURE 9. Phylogram of the feloid Carnivora. Combined evidence phylogram of the feloid Carnivora derived from concatenated data from three nuclear (RAG-1, c-myc and GHR) and one mitochondrial (cytochrome b) genes. Numbers above the line are Bayesian posterior probabilities/ maximum likelihood bootstrap/maximum parsimony bootstrap values based on 500 bootstrap replicates. An asterisk (*) indicates 100% support. A "-" indicates support values less than 50%.

59 A. 88 B. 70 Acinonyx jubatus Felis viverrina Felis yagouaroundi Acinonyx jubatus 99 Felis viverrina Lynx canadensis Lynx canadensis Lynx rufus Felis temmincki Lynx rufus 84 73 Felis caracal Felis manul Felis yagouaroundi Felis margarita 98 Felis margarita Felis silvestris Felis silvestris Felis manul Felis temmincki 100 100 Panthera tigris Felis caracal Panthera onca Panthera tigris 66 Panthera leo Panthera uncia Panthera pardus Panthera uncia Neofelis nebulosa 69 Neofelis nebulosa Panthera leo 75 Herpestes javanicus Panthera pardus Galerella pulverulenta Panthera onca 99 Herpestes sanguineus Ichneumia albicauda Herpestes sanguineus 92 76 Suricata suricatta Herpestes javanicus Helogale parvula 65 Galerella pulverulenta 100 Mungotictis decemlineata 100 Helogale parvula 100 Galidictis fasciata Suricata suricatta 99 Galidia elegans 99 97 100 Cryptoprocta ferox Galidia elegans Fossa fossana 100 Mungotictis decemlineata Crocuta crocuta 100 73 99 Galidictis fasciata Proteles cristatus 80 Cryptoprocta ferox 80 Paguma larvata 100 Paradoxurus hermaphroditus Fossa fossana 90 Crocuta crocuta Arctictis binturong 100 Chrotogale owstoni Proteles cristatus Viverra tangalunga 100 82 Viverra tangalunga Civettictis civetta 100 100 Civettictis civetta 100 Genetta sp. Genetta tigrina 79 Genetta sp.

Nandinia binotata Paguma larvata Canis latrans Chrotogale owstowni Cuon alpinus 51 Nandinia binotata Vulpes vulpes Ursus americanus Ictonyx striatus Ictonyx striatus Cuon alpinus 0.005 substitutions/site 0.005 substitutions/site

Lynx canadensis 56 Herpestes sanguineus Lynx rufus 72 C. D. 78 Galerella pulverulenta 95 Felis margarita Felis silvestris Ichneumia albicauda Felis viverrina 100 Herpestes javanicus 90 Acinonyx jubatus Helogale parvula Felis manul 69 62 Felis temmincki Suricata suricatta Felis caracal Fossa fossana Felis yagouaroundi 64 97 Mungotictis decemlineata 83 Felis concolor 99 Galidictis fasciata Neofelis nebulosa Panthera tigris 94 80 Galidia elegans Panthera leo Cryptoprocta ferox Panthera pardus 69 Crocuta crocuta 100 74 Panthera onca Proteles cristatus Panthera uncia 97 Paguma larvata Panthera tigris 97 Paradoxurus hermaphroditus Panthera pardus Arctictis binturong Panthera leo 100 Viverra tangalunga 97 Civettictis civetta Panthera uncia 99 56 100 Genetta sp. Neofelis nebulosa 54 Genetta tigrina Felis silvestris Chrotogale owstoni 68 100 Lynx canadensis Ichneumia albicauda 61 Acinonyx jubatus Herpestes javanicus 53 99 Galerella pulverulenta 61 Felis concolor 80 Herpestes sanguineus Paguma larvata 98 Suricata suricatta 97 99 Paradoxurus hermaphroditus Helogale parvula 81 Arctictis binturong 88 62 64 Mungotictis decemlineata 97 Galidictis fasciata Chrotogale owstoni Galidia elegans 75 Viverra tangalunga 82 Fossa fossana 90 78 Civettictis civetta Cryptoprocta ferox Genetta sp. 100 Crocuta crocuta 100 Proteles cristatus Genetta tigrina Nandinia binotata Nandinia binotata Vulpes vulpes Ictonyx striatus Canis latrans Cuon alpinus Ursus americanus Ictonyx striatus Vulpes vulpes Ursus americanus Canis latrans 0.005 substitutions/site 0.1 substitutions/site

Figure 10. Maximum likelihood phylograms for each gene. Maximum likelihood phylograms for each gene. A. RAG-1. B. c-myc. C. GHR. D. cytochrome b.

60

CHAPTER 4 A COMBINED EVIDENCE PHYLOGENY OF THE FELOID CARNIVORA: COMPREHENSIVE ANALYSIS OF FOSSIL AND MOLECULAR DATA

The order Carnivora (Mammalia) is a diverse group of 13 living and two extinct families (Wozencraft 2005), and includes groups such as the pinnipeds (seals and walruses), canids (dogs), ursids (bears), felids (cats), and hyaenids (hyaenas) as well as two extinct families, the Amphicyonidae (bear-dogs) and the Nimravidae, a clade of non-felid sabertoothed taxa. All carnivorans share the synapomorphy of the carnassial pair, comprised of a modified 4th upper premolar and 1st lower molar that have been adapted for slicing flesh. This is a diagnostic feature, since, relative to other, closely related taxa such as creodonts, carnivorans have limited the shearing aspect of the dentition to only these teeth (Flynn et al. 1988). Despite this specialized dental form, however, carnivorans have diversified widely since their common ancestor, and fossil and extant forms include durophagous, insectivorous, omnivorous and even strictly herbivorous taxa such as melanoleuca, the giant panda. Carnivorans also exhibit high rates of guild structuring and convergence in ecological niche (Van Valkenburgh 1988, 1989), and similar morphotypes have appeared repeatedly during the course of this clade’s history, including hypercarnivorous, osteophagous, and saber-toothed forms (Van Valkenburgh 1988, 1989, 1991; Werdelin 1996; Holliday and Steppan 2004). The initial carnivoran radiation occurred approximately 45 million years ago (mya) (Wesley-Hunt and Flynn 2005), and proceeded in two waves: an Eocene-Oligocene radiation that produced the caniform (dog-like) families Canidae, Ursidae, Mustelidae, Procyonidae, and Amphicyonidae as well as the extinct Nimravidae and a later radiation during the early Miocene of Europe that led to the feliform (cat-like) clades Felidae, Hyaenidae, Viverridae (civets and genets), Herpestidae (mongooses), Eupleridae (Malagasy carnivorans) and the monotypic family Nandiinidae. It is this second group of cat-like carnivorans, together known as the Feloidea (=Aeluroidea of Flower, 1869), which is the focus of this study. Here, I follow Wyss and Flynn (1993) in viewing as feloid any taxon that is a member of the clade represented by the most recent common ancestor of the extant feloid families and its descendants. The superfamily Feloidea contains 113 living species, including 37 species in Felidae, four species within Hyaenidae, 34 species in Viverridae, 33 species in Herpestidae, six species in Eupleridae and one in Nandiniidae (Wozencraft 2005). The Eurasian and American fossil record also indicates the presence of numerous and diverse feloid species dating from the late Miocene (around 20 million years ago) to recent times. Furthermore, fossils that represent more primitive members of this clade (e.g. Stenoplesictis, Paleoprionodon) have been identified in early Miocene deposits of Europe (St. Gerand) as well as in formations that date to the late Oligocene (Quercy [Peigne and de Bonis, 1999]; Hsanda Gol in Mongolia [Dashzeveg 1996, Hunt 1998]) and possibly the Eocene (Stenoplesictis [Dashzeveg 1996, Hunt 1998]). This indicates that, although the major radiation of the feloids occurred during the Miocene, this taxon was present in Europe or at a much earlier date. The fossil record thus contains both taxa that can be reliably assigned to extant families as well as geologically older and more primitive specimens that, by virtue of their more plesiomorphic character states, are less easily assigned to any specific family, and may, in fact, be ancestral (or close) to the entire group. Despite relatively abundant fossil material, particularly for felids and hyaenids, the task of interpreting historical patterns of diversification and morphological evolution in feloids has presented great difficulty for comparative anatomists and, in particular, taxonomists. Comparative morphologists and systematists who have studied the morphology of this clade have recognized for decades (Wozencraft 1984, 1989; Hunt 1989, 1998; Werdelin and Soulinias 1991; Wyss and Flynn 1993; Gaubert et al. 2004) that many feloids are highly autapomorphic, or else possess a mixture of plesiomorphic and autapomorphic characteristics that can make their true affinities difficult to discern (e.g., dentition of Eupleres goudotii, ear region in Nandinia binotata). To add

61

to this, feloids also exhibit high levels of convergence in specific morphological characters, particularly those of the skull and dentition (Van Valkenburgh 1988, 1989; Werdelin 1996; Hunt 1998; Gaubert et al. 2004; Holliday and Steppan 2004). In particular, feloids appear to have a strong tendency towards evolution of the hypercarnivore, or meat specialist morphotype: hypercarnivory has evolved at least three times independently within the six constituent families, and recognition that the many primitive feloids were also relatively highly carnivorous (Hunt 1998) carries strong implications for hypotheses of parallel evolution and biased morphological character change. In combination with their morphological characteristics, the varied and distinctive ecologies and behaviors of feloids serve to make this group of particular interest to workers who study character evolution and ecological or behavioral correlates (see, e.g. Hunt 1988, 1989, 1998; Werdelin 1996; Gaubert et al 2004; Holliday and Steppan 2004; Veron et al. 2004). Such studies, however, require a robust phylogenetic hypothesis in order to effectively understand patterns of change. In particular, the inclusion of fossils in a phylogenetic analysis can have a significant effect on interpretations of character evolution, since fossil data helps to polarize characters, break up long branches, and preserve combinations of character states that are no longer found in extant taxa (Gauthier et al. 1988; Donoghue et al. 1989; Wyss and Flynn 1993). However, despite a long history of work on the phylogenetics of exant feloids (e.g. Flower 1869; Flower and Lydekker 1891; Gregory and Hellman 1939; Hemmer 1978; Wozencraft 1984; Salles 1992; Gaubert et al. 2003; Gaubert et al. 2004; Veron et al. 2004; Perez et al. 2006; Johnson et al 2006; Koepfli et al. 2006; Gaubert and Cordeiro-Estrala 2006), and extant carnivorans more broadly (Yoder et al 2003; Flynn et al 2005), there is still only a limited understanding of the phylogenetic relationships of the feloids within a broader perspective. More specifically, there have been few studies that evaluate feloid or carnivoran phylogeny in the context of fossil data (but see, Werdelin and Soulinias 1991; Wyss and Flynn 1993), a lack that can severely limit efforts to understand character evolution and patterns of diversification and to test specific evolutionary hypotheses.

History of Systematic Study

For many decades, there has been diligent effort to elucidate relationships both within and between the members of Feloidea, and careful, comprehensive analyses, both morphological and molecular, have been produced for nearly every family within this clade (e.g. Flower 1869; Flower and Lydekker 1891; Gregory and Hellman 1939; Hemmer 1978; Wozencraft 1984; Hunt 1987; Werdelin and Soulinias 1991; Salles 1992; Mattern and McClennan 2000; Johnson et al 2006; Koepfli et al. 2006 to name only a few). Despite this history of work, however, there has been little success in obtaining a robust, well-sampled phylogeny for the feloids as a whole, particularly in relation to the deeper nodes of the phylogeny (e.g., between families and major subclades). These nodes are of especial importance because they establish the identity of the most basal taxon and deepest divergence(s) – and hence the likely primitive condition – for evaluating subsequent change. There are several likely reasons for the historical difficulty in establishing relationships within the Feloidea. First, convergence in commonly used morphological characters, as well as highly derived taxa with autapomorphic character states, has led to much confusion regarding primitive vs. derived character states and patterns of morphological diversification. Molecular data has proven problematic as well, since rapidly evolving mitochondrial genes, often used by molecular systematists to elucidate relationships between species within families, appear to be of limited utility for several lineages due to saturation (Gaubert et al. 2003; Gaubert et al. 2004; Koepfli et al. 2006 [viverrids], but see Masuda et al. 1996; Yu and Zhang 2005 for felids). More slowly evolving nuclear genes, typically used to understand deeper divergences, have also produced incongruent results depending on the gene or genes used (Yoder et al. 2003; Flynn et al.

62

2005; Holliday and Steppan in prep). Incongruent topologies at deeper nodes is a phenomenon that is typically observed when there has been a rapid radiation early in the history of a clade (Fishbein 2001; Fishbein and Soltis 2004). In the case of Feloidea, there were actually several rapid radiations, including one that occurred around 30-35 my (Johnson et al 2006) and represents the initial radiation of the major families Felidae, Viverridae, Hyaenidae and Herpestidae, again during the late Miocene radiation of Felidae, and, most recently, within the six species felid genus Panthera, a clade whose radiation has been dated at only 2-3 my (Yu and Zhang 2005; Johnson et al. 2006). In addition to character data, taxon sampling has also been problematic. Since the most thorough analyses have focused only on a single family or subfamily within the Feloidea (e.g. Wozencraft 1984 (Viverridae); Werdelin and Soulinias 1991 (hyaenids); Veron and Heard 1999 (Viverridae); Gaubert et al. 2002, 2003, 2004 (genets); Veron et al. 2004 (Viverridae); Yu and Zhang 2005 (Panthera); Gaubert and Cordeiro-Estrala 2006 (Viverrinae); Johnson et al. 2006 (Felidae); Koepfli et al. 2006 (Hyaenidae)), both taxon and gene sampling for Feloidea as a coherent whole has, until very recently, been extremely limited (but see Flynn et al. 2005; Holliday and Steppan in prep.). Instead, high profile clades such as felids have received the bulk of the attention from researchers (Salles 1992; Johnson and O’Brien 1997; Mattern and McClennan 2000; Yu and Zhang 2005; Johnson et al. 2006), while herpestids have only recently begun to generate interest from molecular systematists (see, e.g., Veron et al. 2004; Perez et al. 2006). Viverrids, a diverse and charismatic group of small to medium sized carnivorans, have recently been the focus of a number of molecular and morphological studies, but these analyses typically consider members of a particular genus or geographical area (Gaubert et al. 2003; Gaubert and Veron 2003; Gaubert et al. 2004; Gaubert and Cordeiro-Estrala 2006), and few researchers have attempted to evaluate molecular relationships within the entire family Viverridae (but see Gaubert and Cordeiro-Estrala (2006) for Viverrinae), despite ongoing questions regarding the Viverridae’s taxonomic status as a monophyletic group (Flynn et al. 1988; Veron 1995; Flynn and Nedbal 1998; Gaubert and Veron 2003; Yoder et al. 2003; Koepfli et al. 2006).

Among-family Relationships

In a review of the then-available data for feloid phylogenetics, Flynn et al. (1988), noted that virtually every possible combination of families has been proposed –and supported— at one time or another. Although there have been continued efforts to elucidate feloid relationships since that time, molecular data as well as additional morphological study have not provided the desired clarity. Based on key morphological features, including the loss of the alisphenoid canal, presence of an a2 arterial shunt, and reduction of the hallux (see also Flynn et al. 1988; Wozencraft 1989; Wyss and Flynn 1993), many previous workers had argued in support of a felid/hyaenid sister group (Wozencraft 1984; Hunt 1987; Wyss and Flynn 1993). However, substantial molecular evidence for a sister group relationship between Hyaenidae and Herpestidae has begun to accumulate over the past several years (Flynn and Nedbal 1998; Gaubert and Veron 2003; Yoder et al. 2003; Yu and Zhang 2004; Flynn et al. 2005; Gaubert and Cordeiro-Estrela 2006; Koepfli et al. 2006). Further, while several recent nuclear + mitochondrial multigene datasets have indicated that Felidae may be sister to a clade containing ((Hyaenidae/Eupleridae/Herpestidae)Viverridae)) (Yoder et al. 2003; Flynn et al. 2005; Koepfli et al. 2006), other datasets suggest that Viverridae is sister to ((Hyaenidae/Eupleridae/Herpestidae)Felidae)) (Gaubert et al. 2003; Flynn et al. 2005). In a recent, multigene study by Flynn et al. (2005), the problem of among-family relationships was perceived as so nearly intractable that the authors suggested that, at present, the best way to present the initial divergences within the feloids is as a polytomy. However, a study including three nuclear and one mitochondrial gene by Holliday and Steppan (in prep), did resolve the diversification of the families, and showed that Nandinia is sister of the remaining extant feloids,

63

while Viverridae (sensu stricto) is sister to a clade comprised of Felidae + ((Herpestidae/Eupleridae)/Hyaenidae).

Objectives

Here, I add 103 morphological characters to the molecular data set of Holliday and Steppan, and an additional 25 extant and over 100 extinct species to the analysis. By including both morphological and molecular data, I am able to make use of the greater number of characters available in molecular data for recent taxa to help stabilize the major relationships for recent and fossil taxa. The data are analyzed in the following ways: 1) the molecular phylogeny is used as a backbone constraint tree for parsimony analysis of the morphological data and 2) molecular and morphological data are analyzed under a Bayesian mixed model, which allows data to be partitioned and modeled by datatype. The goal of this study is to obtain a well-sampled, comprehensive phylogeny of the feloid Carnivora, including recent and fossil material, in order to obtain the detailed information on relationships that is necessary to better understand character evolution within this clade.

MATERIALS AND METHODS

Morphological data

Taxa Sixty crown group species, representing members of all extant families of feloid, were included in the morphological dataset. As much as possible given the availability of specimens, sampling for recent species was balanced between male and female specimens. I included only adult specimens in the dataset, although dental characters for some juveniles were included if the adult teeth had erupted. Adult status was assessed on the basis of fully erupted dentition, tooth wear, and whether or not suture fusion had occurred. Juveniles typically have unfused cranial sutures indicating that skull growth is incomplete. Skulls that exhibited damage, either through injury before death or damage after death, were excluded except for comparative/confirmatory purposes. Sexually dimorphic taxa were coded as polymorphic for those characters. Over 100 fossil species, representing over 500 specimens in 66 genera, were included in the fossil data matrix. Unlike recent taxa, damage to fossil material is the rule rather than the exception. Damaged, adult specimens were thus included in the dataset, but only characters that did not exhibit visible damage or distortion were coded for the matrix. Since species identifications on fossil material are not always conclusive, and because small sample sizes limited the ability to detect sexual dimorphism, all questionable material was included in the data set as separate operational taxonomic units (OTUs) subject to confirmatory analysis of their taxonomic identification; combination of data from conspecific fossil taxa was performed subsequent to this confirmation. Taxa and museum identifications are shown in Appendices 1 and 2. The following descriptions briefly describe crown group and fossil taxa included in these analyses:

Felidae (Fischer 1817) Although felids are considered to have originated in the late-Oligocene with Proailurus (Hunt 1989; Turner and Anton 1997), the primary radiation of this clade was mid-Miocene (Werdelin 1996), and included both conical-toothed (subfamily Felinae) and sabertoothed (subfamily Machairodontinae) taxa. The conical toothed felids first appear in the fossil record approximately 18 million years ago, during the early Miocene of Europe, and immigrated to

64

North America soon after, during the latest early Miocene, approximately 17 my (Martin 1980; Janis et al 1998). The machairodontines appear relatively soon after, at approximately 15 my, and radiated to include at least eight different genera with varying levels of sabertooth adaptation (Turner and Anton 1997). Felids are defined on the basis of their extreme dental reduction and hypercarnivorous phenotype, and the usual felid dental formula is 3-1-2-0 (upper) and 3-1-2-1 (lower) (Nowak 1999). The earliest felids, however, did possess additional teeth, and the first true “cat”, Proailurus, had a full complement of premolars (pm1-pm4) as well as a small m2 while its immediate successors, members of the the genus Pseudaelurus, retained only pm2-pm4. All felids have highly sectorial premolars and carnassials (P4/m1), but post-carnassial molars, when they exist, are vestigial. Along with the recent felids, I include the following fossil material in this study:

Proailurus: Multiple species of the genus Proailurus first appeared in the late Oligocene/Early Miocene fossil record of (Quercy, St. Gerand) (Hunt 1989, 1991, 1998; Turner and Anton 1997; Rothwell 2003). Members of this genus are relatively small (house-cat to lynx, approximately 8-30 lbs) sized feloids, with comparatively longer rostrums and a viverrid-like body form (Martin 1980). Proailurines had been unknown outside of European fossil localities until recently, when a specimen was found in in deposits dating to the mid-Miocene (16 my) (Hunt 1989, 1998). I examined eight specimens of Proailurus lemanensis, primarily maxillary and dental remains. Character data for these specimens ranged from 14-73% complete.

Pseudaelurus: Pseudaelurus is viewed as the transitional form between Proailurus and the more modern felines/machairodontines, and multiple species have been named and described from the early Miocene of Europe, Asia and North America (Ginsburg 1983; Turner and Anton 1997; Rothwell 2001). Pseudaelurus retains a 2nd premolar but has lost the p1 and m2. It also retains the alisphenoid canal, which has been lost in all modern cats (Rothwell 2003). The oldest North American specimens have been dated to approximately 18-20 my (Turner and Anton 1997; Rothwell 2001), with a temporal range extending to the mid-Miocene, approximately 12 my (Rothwell 2001). The temporal range of the European forms is similar but extends slightly earlier and later, from around 20 my to 11 my (Rothwell 2001). Species within the genus Pseudaelurus may have given rise to the sabertoothed machairodontines (i.e. the French Pseudaelurus quadridentatus) as well as the conical toothed felids (i.e. Pseudaelurus transitorius and Pseudaelurus lorteti) (Turner and Anton 1997), however, Pseudaelurus has also been referred to as a “wastebasket” taxon, with nearly all early and middle Miocene felid fossils referred to this genus (Rothwell 2001). Materials examined ranged from 16-72% complete for this dataset, and included 30 specimens representing at least six different species.

Felinae: The subfamily Felinae denotes the conical-toothed members of the felid family, and includes all living cats. Similarity in many features of the skull and dentition, and near complete homogeneity in the dental formula, has made it very difficult to determine the precise systematics and taxonomic designations within this group (Martin 1980; Werdelin 1983, 1985; Rothwell 2003). Depending on the author, Felinae may be divided into anywhere from 3-17 distinct genera, but generally such divisions include the “small cats” in the genus Felis, the “big cats” in the genus Panthera, and the , Acinonyx. The genus Felis is also generally understood to include Felis () concolor, the American puma. Neofelis nebulosa, the , is also typically placed in a separate genus, despite its position at the base of the pantherine radiation. Fossil material for Panthera and Felis is relatively abundant in the fossil record; Felis dates from the late Miocene of Europe and Asia (Rothwell 2001).

Felis attica: This mid-Miocene (7-12 my), Eurasian cat is considered to be near the base of the modern felid radiation (Werdelin 1996; Turner and Anton 1997), and has been suggested as

65

possibly ancestral to all other species of Felis (Turner and Anton 1997). It differs slightly from modern Felis, however, in that it has a slightly longer skull (Hunt 2001). I examined 11 specimens of Felis attica for this study, and the matrix includes data obtained from complete skulls as well as fragmentary dental material.

Felis daggetti and Felis hawveri are Rancholabrean (Pleistocene) pumas from western North America. I examined a single, highly fragmentary representative of each species; each was only about 15% complete.

Panthera atrox: The Late Pleistocene lion Panthera atrox is popularly known as the American Cave Lion, although authors have differed on whether this is a subspecies of Panthera leo, closely related to Panthera tigris or a distinct species (Hemmer 1978; Turner and Anton 1997; Sotnikova and Nikolskiy 2006). I examined three specimens of P. atrox, all of which were approximately 65-70% complete.

Machairodontinae: The subfamily Machairodontinae denotes the saber-toothed members of Felidae, and includes taxa such as , , Smilodon, as well as less derived sabertooths such as (Turner and Anton 1997). Machairodonts first appeared in the Eurasian fossil record approximately 13-15 my, during the late-Miocene/early Pleistocene and subsequently radiated into North and (Turner and Anton 1997).

Dinofelis: Dinofelis was a primitive sabertooth with a large, stocky build and heavy forelimbs, a morphology that suggests adaptations to dense forests (Turner and Anton 1997). Specimens ascribed to Dinofelis have been found throughout Africa, Eurasia, and North America; North American remains have been dated to about 5 my (late Miocene) (Turner and Anton 1997). I examined five specimens of Dinofelis, representing at least two species. Material included in the matrix ranged from 35-63% complete and included both skull and dental remains.

Metailurus: Metailurus was a Eurasian cat present during the end of the Miocene (about 8 my) (Turner and Anton 1997). It had very slightly developed sabertooth characteristics (it has been described by [Turner and Anton 1997] as somewhere between a sabertooth and a conical toothed cat) but was slightly more derived in the direction of sabertoothy relative to Dinofelis. It is believed that it might have hunted in trees, and may have had a habitat similar to (Turner and Anton 1997). I examined seven specimens of Metailurus, representing both recognized species (M. major and M. parvulus). Material ranged from 12-79% complete.

Nimravides is another taxon that has been described as having incipient sabertooth features (Turner and Anton 1997). This long-limbed, puma to lion sized cat first appeared in North America at the end of the middle Miocene (Janis et al. 1998). I examined four damaged specimens of Nimravides, representing both skull and dental material, and was able to complete 20-38% of the data matrix.

Homotherium: Unlike Nimravides, Dinofelis and Metailurus, Homotherium was a relatively derived sabertoothed genus, with elongate forelimbs and a short tail that, along with its large size (about the size of a lion) has suggested to some workers that it may have been a specialist predator on young mammoths. Homotherium is known from the Miocene of Europe, Asia, and Africa, and apparently immigrated to North America around the end of the Miocene, where it persisted until about 0.5 my. Data from two species of Homotherium were included in the analysis, one a cast with 65% of the characters of interest present, and the other specimen limited to 13%. Each of these individuals was coded and included separately in the dataset.

66

Smilodon: Smilodon was a highly derived sabertooth with powerful forelimbs, and may have hunted cooperatively in prides like the lion. Three recognized species of Smilodon existed in North and South America, with famous fossil remains collected from the La Brea Tar Pits in California. This Pleistocene taxon first appears in the fossil record around 2.5 my, and Smilodon did not go extinct until around 10,000 years ago (Turner and Anton 1997). I examined 14 specimens of Smilodon. Material included in the data matrix ranged from 13-67% complete.

Hyaenidae (Gray 1821) Despite the paucity of recent species in the Hyaenidae, fossil evidence shows that there was a high level of diversity during the Miocene (over 100 species), and specimens that can be reliably identified as belonging to the Hyaenidae begin to appear in the fossil record approximately 17 million years ago, during the late Miocene (Werdelin and Soulinas 2001). The earliest hyaenids share many plesiomorphic characters with both viverrids and herpestids, and it has been suggested by some authors (Gregory and Hellman 1939; Hunt 1987) that some or all of the Hyaenidae were an offshoot of the early viverrid or herpestid stock. However, Hunt, Jr. (1991) stated that the “hyaenid” bulla type is recognizable in mid-Miocene deposits in East Asia, which suggests that, despite the European late Miocene fossil record, stem taxa may have begun diversifying as early as the Oligocene (Hunt 1991). Despite this, the conventional view is that primitive hyaenas very likely originated in western Europe as small, semi-arboreal, omnivorous/insectivorous types (Werdelin and Soulinias 1991), and, from these early taxa, larger, more generalized, meat-eating forms evolved and dispersed into other parts of Eurasia and into Africa (Werdelin 1996), then later into North America (Chasmaporthetes) during the (Janis et al. 1998; Turner and Anton 1997). Crown group Hyaenidae include Hyaena brunnea, Hyaena hyena, Crocuta crocuta and Proteles cristatus (Wozencraft 2005). The clade formed by Hyaena and Crocuta is most readily recognized on the basis of their adaptations to and their large size, while Proteles, the 4th extant member of Hyaenidae, exhibits a highly autapomorphic dentition related to its diet of . Features of the basicranium and skull, however, place Proteles definitively within the hyaena family, a position that has been unequivocally supported by molecular results (Koepfli et al 2006, Holliday and Steppan in prep). In addition to the crown group hyaenids, I include the following extinct species in this analysis.

Protictitherium: Protictitherium is a primitive hyaena and appears to have been a semiarboreal, insectivore/omnivore morph subsisting on a diet of small mammals, birds, and (Werdelin and Soulinias 1996). It possessed a generalized, civet-like dention that included a full set of premolars and molars (Werdelin and Soulinias 1996). This dataset includes only a single, highly fragmentary (10% complete) cast of the lower jaw of a specimen of Protictitherium.

Plioviverrops: Plioviverrops is arguably the most primitive of the hyaenids and is in many ways very similar to the viverrids. Ecologically, Plioviverrops appears to have been a mongoose-like, insectivore/omnivore type (Werdelin and Soulinias 1996), with a dentition that appears to have been specialized for insectivory, including reduction of the slicing teeth and an increase in the height of the cusps on the molars and premolars (Werdelin and Soulinias 1996). I examined two specimens of Plioviverrops orbigny, both including skull and mandibular material.

Tungurictis: One of the most primitive hyaenids, Tungurictis has been classified variously as both a hyaenid and a viverrid (Colbert 1939; Hunt 1991). Although dentally it appears highly similar to the civets, recent study of both the auditory bulla and the dentition indicate that it possesses a distinctly hyaenid bullar structure (Hunt 1989; Werdelin and Soulinas 1991). I examined two specimens of Tungurictis, a skull representing the species Tungurictis spocki and an as yet undescribed specimen that included both the skull and dentaries. I was able to complete 40% of

67

the matrix based on the T. spocki specimen, while the undescribed Tungurictis offered material for 30%.

Hyaenictitherium, Ictitherium and Thalassictis – These hyaenas were all relatively generalized both in terms of locomotion and also diet (Werdelin and Soulinas 1991). They exhibit a comparatively unspecialized carnivorous dentition much like canids, and appear to have been generalized meat/bone eaters (Werdelin and Soulinias 1996). Taxonomic identification and phylogenetic placement of members of these species has been difficult due to changing concepts of each genus and the phenotypic similarities between these relatively primitive taxa (Werdelin and Soulinias 1991). I examined 19 specimens of Thalassictis, representing the skulls and jaws of two different species. Material ranged from 10-40% complete. A further 21 specimens of Ictitherium viverrinum were evaluated, and this material was slightly over 54% complete. Lastly, I evaluated three species Hyaenotherium wongii, and this material ranged from 63% to 90% complete.

Palinhyaena reperta: Palinhyaena was one of the earliest hyaenas to show widening of the premolars, a characteristic that indicates adaptations towards bone consumption (as in modern hyaenas). Palinhyaena thus appears to be a primitive member of the “bone-cracking” lineage (Werdelin and Soulinias 1991, 1996). I examined five specimens of Palinhyaena reperta, representing skull and dental material; completeness of the data matrix ranged from 23-85%.

Adcrocuta eximia: This monotypic genus was large, bone-cracking hyaena with a stocky build and short, robust limbs that is very likely the sister taxon to Crocuta crucuta (Werdelin and Soulinias 1990). Unlike Crocuta, however, Adcrocuta does not exhibit that taxon’s adaptations for pursuit (Werdelin and Soulinias 1996) and instead appears to have filled a more truly scavenging/osteophagous niche (Werdelin and Soulinias 1996). I examined four specimens of Adcrocuta eximia, and material ranged from 20-45% complete.

Lycyaena: This taxon belonged to a clade that also included the genera Chasmaporthetes and Hyaenictis. The morphology of these taxa indicates that they evolved in a different direction relative to the specialized meat/bone eating lineage represented by Palinhyaena-Crocuta, and instead specialized as pursuit predators (Werdelin and Soulinias 1996). Members of this clade thus exhibited more cursorial adaptations and a highly sectorial dentition more developed for meat-eating than bone crushing (Werdelin and Soulinas 1991, 1996). I examined 19 specimens of Lycyaena, representing skull and dental remains of three different species. Completeness of the material for the data matrix ranged from 28-62%.

Chasmaporthetes: Like Lycyaena, Chasmaporthetes was more adapted for a meat- specialist diet than for crushing bone. Species of Chasmaporthetes included the long- legged, North American “” hyaena, the only hyaena known to have immigrated to the New World (Janis et al. 1998; Werdelin and Soulinas 1996; Turner and Anton 1997). I examined nine specimens of Chasmaporthethes, representing both skull and mandibular material, but was able to complete only 25% of the character matrix.

Viverridae (Gray 1821) Crown group Viverridae is a diverse clade that includes over 30 living species divided into approximately 19 genera in four different subfamilies (Wozencraft 2005). Viverrids range in size from 6 – 20 kgs, and exhibit a variety of dental and habitat specializations. Their geographic distribution is likewise relatively broad, and includes both African and Asian/Indonesian forms. Viverridae has traditionally been treated as a “wastebasket” taxon (Hunt 1989; Werdelin 1996),

68

with some workers choosing to place aberrant or unusual taxa into Viverridae rather than erecting a new family (Hunt 1989). Consequently, the taxonomic history of the viverrids has been one of reinterpretation, synonomies, and shifting of clades in and out of the family. Unsurprisingly, most molecular studies indicate that Viverridae is a paraphyletic grouping. Most recently, the “viverrid” Nandinia binotata has been shown to be the sister to the remaining extant feloids (Yoder et al 2003; Koepfli et al 2006; Flynn et al 1995; Holliday and Steppan, in prep) and has been recognized as the monotypic family Nandiniidae (Wozencraft 2005), while the Malagasy carnivorans Cryptoprocta ferox, Eupleres goudotti, and Fossa fossana, long considered closely related to viverrids (Wozencraft 1989), are more closely related to herpestids, and belong in the herpestid sister-family Eupleridae (Yoder et al 2003; Wozencraft 2005, Koepfli et al. 2006; Holliday and Steppan, in prep). Fossil material that can be unequivocally assigned to Viverridae proper is rare, and what is available is highly fragmentary (Hunt 1989). The situation is further complicated because of the preference of many viverrids for forested habitats where fossilization is less likely. This study does not include any known viverrid fossil material, but does incorporate several “stem” feloids that may or not belong within the Viverridae proper (Stenoplesictis, Paleoprionodon).

Herpestidae (Bonaparte 1845) Crown group Herpestidae includes over 33 living species that can be separated into 14 genera and two main divisions: the social mongooses (e.g. Herpestes) and the highly predaceous Indian mongooses such as Cynictis (Nowak 1999). The modern geographic distribution of mongooses is mainly Indonesian, but includes taxa in Europe and Africa as well (Nowak 1999). Their small size and habitat preference ( to forest, as well as some burrowing forms) has contributed to a poor knowledge of fossil taxa. The earliest putative herpestids are dated to the early Mocene and early middle Miocene of Europe (Savage and Russell 1983), as well as the middle Miocene of Africa (Schmidt-Kittler 1987). Clearly modern forms from the Late Miocene of Pakistan (Asia) are dated to approximately 7-9 my (Hunt 1989; Peigne et al. 2005). Herpestidae was long treated as a subfamily within the Viverridae (Flower and Lydekker, 1891, Simpson 1945), despite a lack of synapomorphies to unite the two clades. However, Gregory and Hellman (1939) and Wozencraft (1984, 1989) considered the two groups to be distinctly separate families, and most recent workers have followed Wozencraft’s lead (Wozencraft 2005). In addition to the crown group herpestids, I include eight specimens of the extinct herpestid genus Kichechia, dated to the mid-Miocene. In my dataset, this African taxon is represented entirely by partial maxillary and mandibular remains, and despite examination of multiple specimens I was only able to complete 25% of the matrix.

Eupleridae (Chenu 1850) The family Eupleridae is composed of the Malagasy carnivoran taxa in the genera Galidia, Galidictis, , Cryptoprocta, Fossa and Eupleres. Galidia, Galidictis, Salanoia have traditionally been treated as members of the Herpestidae (Nowak 1999), but recent molecular phylogenetic analyses (Yoder et at. 2003, Flynn et al. 2005, Holliday and Steppan in prep) have established the presence of an endemic Malagasy radiation, and these taxa, along with Cryptoprocta, Fossa and Eupleres, instead to form a clade that is sister to the herpestids. The latter three genera, Cryptoprocta, Fossa and Eupleres, have consistently recognized as problematic in their taxonomic affinities, although likely most closely related to each other: Cryptoprocta, which shares significant morphological (dental) similarities to the early felids, has been variously placed with the Felidae, Viverridae and Herpestidae, while Fossa and Eupleres, long-recognized as sister taxa, were placed in Viverridae despite the recognition that they may constitute a separate group (Wozencraft 1989, Wozencraft 2005). There is no fossil material yet known for Eupleridae prior to the Pleistocene.

69

Nandiinidae (Pocock 1929) In a series of detailed morphological analyses of the auditory bulla of recent and fossil feloids, Hunt (1987, 1989, 1998) showed that Nandinia binotata, the , has a highly plesiomorphic auditory bulla relative to other feloids. From this, Hunt argued that Nandinia should be removed from Viverridae and instead designated as a monotypic family. Molecular studies over the past decade have supported this position, and analyses by Yoder et al. (2003), Flynn et al. (2005), and Holliday and Steppan (in prep) confirm that Nandinia is sister to all exant feloids.

Feloidea (unassigned)

Haplogale: The genus Haplogale, although originally placed in the primitive felid genus Proailurus (Teilhard 1915; Hunt 2001), is a long-snouted, early feloid with similarities to viverrids (Hunt 2001). Hunt (1998) describes it as sharing dental similarities with PalaeoPrionodon but with larger teeth. Haplogale has been recorded in the Oligocene Quercy deposits of France (Hunt 1998). I examined 13 partial specimens of Haplogale media, including and a single partial cranium, and completeness of the material ranged from 10-30%.

Stenogale is a small, short-snouted, cat-like taxon from Oligocene (30-25 my) deposits of Europe and Asia (Savage and Russell 1983; Dashzeveg 1996; Hunt 1998). Stenogale was initially thought to represent a primitive mustelid (Savage and Russell 1983), but Hunt (1989) showed that the auditory region is distinctly feloid and this genus in fact shares many features with the basicranium of Proailurus (Hunt 1989; Hunt 2001). Stenogale’s true phylogenetic position has consistently been problematic, however, and it has been suggested as either a sister to Felidae or possibly ancestral to Proailurus lemanensis (Hunt 2001). I examined 19 specimens of Stenogale, representing multiple species. Material was almost exclusively comprised of lower jaws and dentitions, and I was able to complete the matrix for only 14-35% of the data.

Herpestides antiquus: Herpestides is a primitive feloid with a highly carnivorous dentition and a Holarctic distribution (Quercy, St. Gerand) (Hunt 1991). Although fossil material dates it to the early Miocene, approx 20-23 my (Hunt 1991), Hunt (1989) argues that it most likely diverged sometime prior, during the Oligocene. Several authors have suggested a relationship with either herpestids (Viret 1929, as cited in Hunt, Jr. 1991; Petter 1974) or ancestral hyaenids (Hunt, Jr. 1989); Hunt, Jr. (1989) additionally points out that Herpestides appears to be an early feloid that has not yet developed a basicranial pattern typical of one of the living families, and in consequence may be close to the ancestry of both hyaenids and herpestids. However, as noted by Werdelin and Soulinas (1991) many of the features Herpestides shares with the hyaenids are plesiomorphic, and further study by Hunt (1991) has instead suggested that this highly variable taxon may be more closely affiliated with the Viverridae. I examined seven specimens of Herpestides antiquus, including both skull and mandibular fragments, and was able to complete the data matrix for 20-73% of the characters

Paleoprionodon: As the name implies, Paleoprionodon exhibits many morphological features that are shared with the Asiatic in the genus Prionodon (Gregory and Hellman 1939, Hung, Jr. 2001), as well as the African , . Hunt, Jr. (2001) has described Paleoprionodon’s basicranial region as highly similar to that of Prionodon, and noted that it may represent a reasonable transitional form towards more specialized viverrids. Unlike Prionodon, however, which has an Asian distribution (Nepal southward to Java), Paleoprionodon fossil evidence indicates a more Holarctic distribution (Quercy and Hsanda-Gol) (Hunt 1989, Hunt 2001, Dashzeveg 1996). Paleoprionodon is thus more likely a stem taxon rather than a representative member of any particular family (Hunt Jr, 2001), although Hunt, Jr. (2001) makes

70

an excellent case for Paleoprionodon as a potential ancestor of a clade that includes the viverrids Poiana, Prionodon, and the genets. Paleoprionodon often co-occurs in deposits with the feloid genus Stenoplesictis, and the two have similarities in both temporal and geographic range. However, Paleoprionodon is described as being slightly more highly carnivorous relative to Stenoplesictis. I examined nine specimens of Paleoprionodon, representing two species. Material was 15-53% complete for the character matrix, and included skulls with relatively intact basicrania as well as mandibular remains.

Stenoplesictis (Filhol 1880): Stenoplesictis is an Oligocene (30-25 my) feloid with clearly primitive characteristics, and members of the genus are typically viewed as a good stem feloids (Hunt 1998); Hunt (1998) has argued that the basicranium is similar to that which might be expected in an ancestral hyaenid. Remains have been recovered in the Quercy deposits of France (Hunt 1998), and possible remains in Hsanda-Gol of Asia (Dashzeveg 1996; Peigne and de Bonis 1999), indicating that Stenoplesictis had a holarctic distribution. From this material, Peigne and de Bonis (1999) have suggested that Stenoplesictis may have originated in Asia. Stenoplesictis cayluxi, the type species, was described by Filhol as being similar to Proailurus but with smaller teeth, and further commented that “the distinction from Proailurus is difficult” (as discussed in Peigne and de Bonis, 1999). A second, slightly younger and larger species, Stenoplesictis crocheti, is known from the late Oligocene of Europe but not Asia (Peigne and de Bonis 1999), and may represent a later immigrant from the same lineage. A further three Asian species (Alag Tsab, Ergilin Dzd, and Hsanda Gol formations) have been identified and described by Dashzeveg (1996), but their generic designation has been questioned by Peigne and de Bonis (1999) and Hunt (2001) on the basis of their morphology, although both authors do concede that these specimens very likely represent a distinct feloid lineage. The new species described by Dashzeveg (1996) are all geologically older (Eocene-Early Oligocene) and are based on only a few fragmentary specimens. I examined 52 specimens of Stenoplesictis, representing the species Stenoplesictis cayluxi and S. crocheti. The available material was composed almost exclusively of dental and mandibular remains, with a few maxillary fragments, and I was able to complete only 10-23% of the character matrix.

Characters

Morphological data This dataset included 103 morphological characters, including 24 skull, 23 jaw, 18 maxillary, and 38 from the basicranial region. Morphological characters included those that were adapted from other phylogenetic studies (Neff 1982, Wozencraft 1984, Van Valkenburgh et al. 1990, Wyss and Flynn 1993, Veron 1995, Hunt, Jr. 1998, Gaubert et al. 2002) as well as characters derived from study of museum specimens; character states were scored from direct study of museum material. Characters were chosen specifically to reflect features identifiable on fossil material, and consequently do not include many other, potentially useful features of the soft anatomy. All characters were analyzed as unordered, an approach that requires the fewest assumptions about character evolution. Descriptions of characters, character states, and citations (where appropriate) are detailed in Appendix E. The data matrix is shown in Appendix F.

Molecular data The molecular dataset included 39 ingroup taxa and was comprised of over 5000 base pairs from three nuclear genes (c-myc, RAG-1 and GHR), and one mitochondrial gene (cyt b). The three nuclear genes were expressly chosen for their utility in mammals and their anticipated ability to resolve the deeper divergences within the Feloidea (see Holliday and Steppan, in prep). These genes evolve at varying rates and are thus informative at different levels of the phylogeny.

71

RAG-1 is over 2000 base pairs (bp) in length, and contains both a divergent and a conserved region that evolve at different rates. C-myc is 900 bp in length, and includes an intron region that also evolves at a slightly higher rate. GHR, growth hormone receptor, is approximately 934 bp and evolves at a rate approximately equal to the 5’ end of RAG-1. Holliday and Steppan (in prep) give details of data collection and phylogenetic analysis of these genes. For the current study, the aligned molecular data matrix is utilized in two ways. First, it was analyzed in combination with the morphological data in both parsimony and Bayesian analyses. Secondly, the maximum likelihood phylogeny obtained from analysis of this molecular data matrix alone (Holliday and Steppan, in prep) was used as a backbone constraint tree for subsequent parsimony analysis of the morphological dataset (Fig. 9).

Analysis

Combined data Molecular information was combined with the morphological dataset in two different ways: as additional data in an unconstrained, combined analysis, and as a backbone constraint tree, where the phylogeny obtained from previous maximum likelihood analysis of the molecular dataset was incorporated into a constrained parsimony analysis of the morphological data. The unconstrained, combined data was analyzed using both parsimony and Bayesian frameworks For parsimony analysis of the combined (unconstrained) dataset, I evaluated the effects of different levels of taxonomic inclusion on the resulting topology. Data subsets were based on the amount of missing data for each taxon, and included recent taxa only, taxa where at least 50% of the data matrix was completed, taxa where at least 20% of the data matrix was completed, and all taxa, regardless of the amount of missing data. Data was analyzed in PAUP* (Swofford 2004) with a heuristic search method, simple addition sequence and 3000 maximum trees. A Bayesian analysis with Mr. Bayes (Huelsenbeck and Ronquist 2002) offers the option of designating a dataset as “mixed”, which allows the application of different models to data subsets, including molecular and morphological data. Such a mixed model enables each of the subsets to each be analyzed under the most appropriate model of evolution for that data type. Each of the four genes was evaluated individually using Mr. Modeltest (Posada and Crandall 1998), which indicated that the most appropriate model for all of the genes was GTR + I + G. Additionally, the morphological dataset for recent taxa was analyzed under a ML model that approximates JC + G. The combined data, with appropriate models, was then analyzed using the program Mr. Bayes (Huelsenbeck and Ronquist 2002) under flat priors (no “favored” hypotheses) and included three heated and one cold chain. Because examination of the 20 most variable bipartition frequencies, analyzed graphically in AWTY (Wilgenbusch et al. 2004), indicated that the bipartition frequencies did not converge until nearly six million generations had passed, the analysis was run for 12 million generations and I applied a 50% burn-in to the resulting trees. I requested tree output every 100 generations and discarded the first 60,000 trees as burn-in. I imported the remaining 60,000 trees into PAUP* (Swofford 2002), from which I obtained a majority rule consensus tree for 68 recent species.

Constrained analyses Recent-only and Recent + fossil morphological data-sets were analyzed under parsimony with a molecular backbone constraint applied. Given the large amount of morphological data, in particular the inclusion of fragmentary fossils, use of a constraint tree was viewed as a way to help provide additional structure to the morphological data without overwhelming the morphological signal with molecular characters. Additionally, a constraint tree allowed me to incorporate a framework in which I had high confidence while still making use of the information from morphological characters. As with the combined but unconstrained data analysis, I obtained

72

phylogenies based on taxon subsets of data for recent taxa only, data that was at least 50% complete, data that was at least 20% complete, and all taxa.

Data subsets

Although it is clear that the clade (Hyaenidae(Eupleridae/Herpestidae)) and the family Felidae are monophyletic, taxonomic and phylogenetic position of a number of species and clades within the Feloidea have presented more difficulty (e.g. monophyly of Viverridae, placement of Prionodon, fossil taxa such as Stenogale and Haplogale) (Flynn et al. 2005). Furthermore, more primitive/older taxa (closer to the root of the tree) are also likely to be more similar to each other due to the slight amount of evolutionary distance between them (Radinsky 1982; Werdelin and Soulinas 1991), a similarity that may be reflected in symplesiomorphic characters that have potential to mislead in phylogenetic analysis. Because it was evident at the outset of this study that plesiomorphic characters, convergence, and missing data were likely to complicate the analysis, I chose to explore the data as fully as possible by partitioning the dataset by taxon and level of completeness of the character matrix. Taxon subsets included Felidae (with putative stem taxa), Herpestidae, Eupleridae and Hyaenidae (together and separately, with putative stem taxa), and Viverridae, with stem taxa. The partitioned data was analyzed under the same backbone constraint topology used for the full dataset.

Rooting

Trees were rooted with caniform outgroups for both the molecular and the morphological datasets. Molecular information for the , Vulpes vulpes, was combined with morphological data for the , . Molecular data for Ictonyx striatus, the (zorilla), was combined with morphological information for the striped , Mephitus mephitus. Both morphological and molecular data were available for the genera Ursus and Canis. Use of composite matrices for outgroups was justified under the recognition that the species that were combined for molecular and morphological information were much more closely related to each other than they were to any other taxa.

RESULTS

Phylogenies

Combined analysis using parsimony, unconstrained In a combined parsimony analysis of molecular plus morphological data for Recent taxa only, the tree obtained was highly similar to a majority rule parsimony consensus tree derived from the original molecular characters alone (tree not shown). Inclusion of less complete morphological (fossil) data in the combined dataset resulted initially in loss of resolution at the tips of the tree. As more taxa, with increasingly fragmentary data were added, however, taxon placement in the tree appeared to become more and more arbitrary. By the time all taxa had been added, including those with only 10-20% of the data matrix, the phylogeny had reached near-randomization, with all families and many genera paraphyletic (tree not shown).

Combined analysis using Bayesian inference Figure 11 shows the results of a Bayesian analysis of the Recent dataset that included 68 taxa (63 ingroup, 5 outgroup). 25 of the taxa had only morphological data available. This analysis yielded a well-supported phylogeny that indicated a sister group relationship between

73

Herpestidae/Eupleridae and Hyaenidae, with Felidae as sister to that clade and Viverridae as sister to the clade comprised of (Felidae((Herpestidae/Eupleridae)/Hyaenidae)). Nandinia binotata was sister to the remaining recent feloids. However, the viverrid genus Prionodon, which was represented by morphological data for two species, was positioned as sister to the herpestid/hyaenid clade. Felidae, while well supported as a monophyletic group, contained a large polytomy; only the Panthera/NeoFelis clade and the small cats received good support. Cryptoprocta ferox was sister to the small Malagasy herpestids (Galidinae), while Eupleres and Fossa were sister to the entire Malagasy herpestid radiation, forming the family Eupleridae). The position of Cryptoprocta, while unexpected, had high Bayesian posterior probabilities. Viverridae (less Prionodon) was supported as an otherwise monophyletic group, but although three distinct clades were recovered, they did not represent any of the traditionally recognized subfamilies of Viverridae (Viverrinae, , and ) and received low to moderate posterior probabilities ranging from 0.54 to 0.74.

Constrained analysis A parsimony analysis of 65 recent taxa (60 ingroup), using the backbone constraint tree, yielded 984 most parsimonious trees of 1070 steps (Fig. 12) and, even with morphological material included, continued to support the monophyly of Felidae, Hyaenidae, Herpestidae, Eupleridae and Viverridae. Furthermore, the genus Poiana was placed within the genets. The genus Prionodon was recovered as sister to this clade and, in contrast to the Bayesian analysis of Recent data, positioned well within Viverridae. As in the Bayesian analysis, Eupleres and Fossa were sister to all Malagasy herpestids including Cryptoprocta, forming the family Eupleridae. Cryptoprocta was sister to the remaining Malagasy herpestids (the subfamily Galidinae). The hyaenids, felids, and the genus Panthera were all monophyletic. The next analysis included all taxa at least 50% complete, and this analysis thus contained 87 ingroup taxa, as well as a good proportion of fossil material. A majority rule consensus of 124 mpts (Fig. 13) showed that all of the major clades had been recovered, but Viverridae and Eupleridae were both paraphyletic. In this tree, Fossa fossana was sister to the entire viverrid clade, while Hyaenotherium wongii and a single specimen of Herpestides antiquus were sister taxa nested within a subclade of the Viverridae. Likewise, a single specimen of Proailurus lemanensis was sister to the euplerid Cryptoprocta ferox. Felidae, although otherwise monophyletic, also had several unexpected groupings: a specimen of Pseudaelurus was placed within a clade formed by Felis concolor and Acinonyx jubatus, while members of the genus Metailurus were positioned variously as sister to all of Felidae, sister to recent felids, and sister to two specimens of Felis concolor. At the same time, the placement of other taxa corroborated the findings of prior workers: the early felid Felis attica, which has been suggested to be ancestral to the entire genus Felis, did position as sister to all other members of that genus. Including taxa at which had completed a minimum of 20% of the data matrix added a number of fragmentary taxa to the dataset and produced a similar result to that observed in the same subset analysis (taxa at least 20% complete) for the unconstrained, combined dataset: there appeared to be a breakdown in the structure of the tree and an increase in incongruent, or even arbitrary, groupings. Figure 14 shows a majority rule consensus of 3000 mpt of 1682 steps. In this topology, Nandinia is nested with other stem feloids such as Paleoprionodon and Haplogale, while the euplerid Fossa fossana, the hyaenid Tungurictis and the early feloid Stenogale are all positioned within the viverrids. Plioviverrops, an early hyaenid genus, is represented by two specimens. One is positioned at the base of a herpestid radiation, while the other is placed at the base of the more derived and crown group hyaenids. Furthermore, one highly fragmentary hyaenid specimen is positioned with the canids, which were originally designated as outgroup taxa. The hyaenidae are thus polyphyletic, and form several additional, distinct clades, including one as sister to Herpestidae/Eupleridae and one as sister to (Hyaenidae(Herpestidae/Eupleridae)). A third and fourth group of hyaenids are joined multiple genera that form successive sister taxa to

74

Felidae, including Pseudaelurus, Proailurus, Stenoplesictis, Paleoprionodon, Herpestides, and Stenogale. The machairodontine lineage is also polyphyletic, with several sabertoothed taxa nested within the Panthera group and other machairodonts, including the derived Smilodon and Homotherium, positioned as successive sister groups to the Felinae. The final constrained analysis of combined data included all Recent and fossil taxa, regardless of the level of completeness. This 177 taxon phylogeny contained a significant number of individuals, particularly those based on more fragmentary taxa, whose position was wholly inconsistent with expectations based on literature and prior analyses and, as with the combined dataset without constraint, appeared to contain multiple arbitrary groupings (figure not shown).

Partitioned data: Taxa The dataset for Felidae included all known felids and then all putative stem taxa as outgroups that were at least 20% complete in the data matrix. Figure 15 represents a majority rule consensus of 3000 trees obtained when Felidae + stem feloids were analyzed under parsimony with a backbone constraint. In this pylogeny, Proailurus and Stenogale are sister taxa, and form a clade that is sister to Pseudaelurus and all other cats. The transitional genus Metailurus, which has been considered ancestral only to the Machairodontine felids, instead seems to share features with not only the machairodonts but also Felis concolor, the puma, Felis silvestris,the European wild cat, and F. attica. Specimens of Metailurus are sister to each of these clades. The Machairodontinae are paraphyletic with a specimen of Smilodon, Nimravides and Dinofelis nested within a paraphyletic Panthera. A majority rule consensus tree derived from 3000 most parsimonious trees obtained from a constrained analysis of the hyaenid/euplerid/herpestid clade is shown in Figure 16. In this analysis, a paraphyletic Hyaenidae was recovered, while the genus Herpestides was sister to the (Hyaenidae(Eupleridae/Herpestidae)). The genus Plioviverrops, generally considered a primitive hyaena, instead was sister to crown group Herpestidae. The early feloid genus Stenoplesictis cayluxi, along with the genus Stenogale coupant, both place as sister to the Eupleridae. The tree shown in Figure 17 is the majority rule consensus of 3000 most parsimonious trees obtained from a constrained parsimony analysis of the viverrid and stem feloid data. Analysis of the Viverridae produced a monophyletic viverrid clade with the exception of a single specimen of Stenogale coupant, which nested within the genets. The genus Genetta is shown to be paraphyletic, and both Poiana and Prionodon are also positioned in this group. The early feloids form a single clade, but there is little structure and it is contains an eight taxon polytomy.

DISCUSSION

Addition of fossil material has long been recommended for morphologically based phylogenetic analysis, since fossil material is useful for breaking up long branches, polarizing characters, and preserving combinations of character states that may no longer exist (Gauthier et al. 1988; Donoghue et al. 1989; Wyss and Flynn 1993). Furthermore, it has been noted repeatedly that, since the usual effect of missing data (e.g. fragmentary fossils) is simply loss of resolution (Huelsenbeck 1991; Wiens and Reeder 1995), the benefits of including such material outweigh the drawbacks (see, e.g. Gauthier et al 1988; Donoghue et al. 1989). However, as more and more fragmentary fossils are included in the dataset, character signal will be lost and, particularly in the case of stem taxa with relatively few derived features, topological positioning can become increasingly equivocal. The value of many shared characters, whether molecular or morphological, is thus underscored in this study. While inclusion of as much character information as possible, both molecular and morphological, is clearly a preferable approach, it is also important to analyze the different data types in the most optimal way possible. It is well documented that choice of the wrong or

75

inappropriate model for molecular data analysis can yield incorrect topologies (Swofford et al. 1996; Sullivan and Swofford 1997), and, in consequence, careful model estimation is crucial to obtaining a robust molecular phylogeny (Swofford et al. 1996). This requirement still applies when analyzing combined morphological and molecular datasets, and parsimony analysis of such datasets without considering the mode of evolution of the molecular data, can, as described earlier, lead to incongruent results. Properly accounting for the requirements of the combined dataset, however, comprises a more complex analytical problem than simple parsimony. This study thus applied the only two currently viable options: 1) a Bayesian analysis of the combined data and 2) a backbone constraint tree based on the molecular dataset.

Bayesian analysis

Bayesian analysis proved to be of good utility in analyzing the recent only dataset, and produced a reasonably well-supported phylogeny that was mostly congruent with the tree obtained from the recent data set using a constraint tree. Only one taxonomic placement differed between the recent-only constrained analysis and the Bayesian analysis: the position of the genus Prionodon as sister to the herpestid/hyaenid clade. Although Prionodon has been traditionally considered a viverrid (Wozencraft 2005), recent analyses (Gaubert and Veron 2003) have suggested that it may instead be sister to the Felidae. However, while the Bayesian results do support the placement of Prionodon outside Viverridae, these data do not support its position as sister to the felids. Having said that, the current dataset is based on an entirely different kind of character information from that of Gaubert and Veron (2003), and I was unable to obtain molecular data for this genus. Thus, I can only conclude that additional molecular analysis, preferably nuclear DNA, is clearly warranted. Viverridae, of course, has long been recognized as highly diverse and not necessarily monophyletic, and it has been suggested several times (Petter 1974; Flynn et al. 1988; Hunt 1989, 2001) that the viverrids may be comprised of multiple different lineages that branched off separately from the early stock. If subsequent research upholds the placement of Prionodon outside the Viverridae, it follows that the taxonomic placement other members of the family may also deserve additional scrutiny.

Constrained analysis

As described previously, inclusion of a backbone constraint tree helped stabilize the morphological data by providing a framework for the morphological dataset, while the morphological data available for the recent taxa on the backbone helped position taxa for which there was no molecular information available. This approach thus has the advantage of retaining the most information from both types of data, although, as noted, including taxa with large amounts of missing data will lead to decreased resolution in the tree. In turn, the topological position of such individuals is likely to become increasingly arbitrary, since there are fewer characters to guide their placement in the phylogeny. Thus, while the recent-only and even the dataset with up to 50% missing data were relatively congruent with expectations based on prior work, at 20% (or less) complete, the signal from the morphological data had become effectively randomized. While there are clearly limitations to the benefits of including fragmentary taxa, however, it is important to note that the signal from different kinds of data (e.g. character regions) is not necessarily equivalent. Thus, while data may mislead due to convergence, they may also offer a rich source of phylogenetically informative characters. In point, this is the goal of morphological phylogenetic studies – to locate the characters that offer the strongest and truest signal. Assuming such characters can be identified, problems of missing data become much less important than evaluating the information that remains. Thus, if all of the fragmentary taxa in a dataset were missing the same kind of data, such as dental characters, but possessed

76

phylogenetically informative basicranial data, then it is possible that there would still be a strong enough signal to place even the highly fragmentary taxa. But is this realistic? The morphological dataset used here included characters drawn from the skull, maxilla, dentition, and basicranium. Separate phylogenetic analyses were performed using data representing each of these regions, and then compared to a robust phylogeny based on molecular data (the constraint tree). Although none of the topologies are perfectly congruent, the trees based on basicranial and skull characters are much more consistent with the molecular phylogeny, while the dental and jaw characters, despite an equally strong character signal, are completely incongruent. Given this information, if the fragmentary material is primarily basicranial or skull data, then it is likely that taxa missing even a large number of characters from the total data set would still be correctly placed on the tree. The better solution for analysis of fossils, then, may not be to exclude fragmentary fossil material on the basis of completeness, but to actively include taxa that retain characters and regions deemed most informative, however incomplete.

Taxon subsets

Because it is recognized that both missing data and homoplasy can lead to weak or misleading signal in phylogenetic analysis, this study also evaluated and analyzed family level datasets in order to mitigate these confounding effects. In the complete tree, I found that Viverridae is sister to the Felidae (Hyaenidae/Eupleridae/Herpestidae clade). However, when fossil taxa such as Paleoprionodon, Stenogale, and Stenoplesictis are included in the dataset, they tend to scatter among the basal members of the different families and it becomes difficult to follow patterns of character change or to establish polarities. It is likely that the inconsistent positions of various taxa within each topology can be ascribed to three primary causes: 1) convergent signal in derived characters, 2) lack of signal due to missing data, and 3) shared, plesiomorphic character states. As to the first issue, one benefit of partitioning the data by family is that it prevents unrelated taxa, drawn together by convergent signals, from distorting the true phylogenetic signal, however weak, that may be present for each monophyletic family or clade. In the felid only analysis, the relationship of Proailurus to Pseudaelurus and of Pseudaelurus to the remainder of the Felidae is confirmed for the first time in a cladistic framework, while the genus Metailurus, which has been considered ancestral only to the Machairodontine felids, instead appears to share features with the machairodonts as well as several members of Felis. Metailurus has been described as being somewhere between the Felinae and Machairodontinae (Turner and Anton 1997), so this is not necessarily a surprising result. Additionally, the genus Stenogale, which has been suggested as a possible ancestral taxon to the Felidae (Hunt, Jr. 1998), is positioned as a sister taxon with Proailurus lemanensis, and this clade is sister to the remaining cats. Despite this, in other analyses, Stenogale does not consistently position near to the Felidae, and its placement at this time should still be considered equivocal. For the hyaenid/euplerid/herpestid analysis (Figure 18), the paraphyletic status of the hyaenid family appears to be the result of a combination of missing data and plesiomorphic character states. Thus, since crown group hyaenids, such as Crocuta and Hyaena, are constrained to be placed as sister to the herpestids on the basis of the molecular analyses, other derived hyaenids, such as Adcrocuta, also fall in this position. The most primitive taxa, however, such as Tungurictis and Plioviverrops, position outside the hyaenid group, with Tungurictis sister to a clade that includes the Eupleridae, and Plioviverrops as sister to the crown group herpestids. These two genera are both less derived and more fragmentary, and consequently appear to lack the synapomorphies that would join them to the remaining hyaenids. One very important finding in the hyaenid/euplerid/herpestid analysis is the position of the Miocene genus Herpestides as sister to both the herpestid/euplerid and the hyaenid radiations. Herpestides has been suggested as

77

a possible ancestor for either hyaenids or herpestids (Hunt 1991, 1998), and these results substantiate that argument. The final clade analyzed as a subset was the Viverridae (Figure 19), a group that has only recently begun to approach monophyly with the removal of several Malagasy carnivorans and Nandinia binotata (Wozencraft 2005). Although there are still questions regarding the affinities of Prionodon, this taxon nests firmly within the genet lineage in this analysis. In fact, Viverridae is paraphyletic only with respect to a single specimen of Stenogale, which also falls out with the genet clade. The lack of fossil material for Viverridae is also underscored in this subset: crown group viverrids form a distinct lineage relative to the early feloids and, unlike the felid and the herpestid/euplerid/hyaenid analyses, there are no transitional forms that to link the recent and fossil taxa.

CONCLUSIONS

The primary motivation for building this combined evidence phylogeny, and for the inclusion of fossil material, was to better understand character evolution within the feloids. As is well known, character mapping and diversity analyses are only as good as the quality of the data they are built on. Thus, attempting to understand character evolution or patterns of speciation/diversification without including available fossil data excludes a valuable resource, and can easily lead to erroneous conclusions. The analyses and trees obtained here are much more comprehensive then any produced so far, and broad patterns can be readily identified even without extended analyses. Thus, examination of the data and phylogenies shows that the primitive condition was that of a small, relatively highly carnivorous taxon with a full complement of teeth (pm1-pm4, m1 and m2) and associated cusps. However, the dentition was also highly developed for slicing, apparently foreshadowing the later appearance of highly carnivorous and hypercarnivorous descendants. As for tracing the evolution of hypercarnivory, the subject of prior work (Holliday and Steppan 2004, Holliday in prep.) and a main impetus of this study, it is clear based on the topologies presented here that hypercarnivory is primitive for neither Feloidea nor for the common ancestor of the derived, hypercanivorous clades. Instead, it appears that the specialist clades all evolved separately from a more generalized ancestor and, confirming expectations, hypercarnivory has evolved independently three times, in Felidae, Hyaenidae, and in the euplerid Cryptoprocta. Of all of these, however, the hypercarnivorous hyaenids are the only taxon with a well-enough developed fossil record to trace the evolution of the condition from generalist to specialist (Tungurictis, Thalassictis through to the Chasmaporthetes/Lycyaena clade). But why is it that there are no transitional forms for the cats and for Cryptoprocta? Proailurus, the first true cat, is already hypercarnivorous when it appears in the Oligocene fossil record (Hunt 2001), although it retains a small pm1 and pm2 and m2. The taxa suggested as ancestral form(s) leading to Proailurus are questionable at best (putatively, Stenogale and Haplogale, although the results of this study offer only weak support for Stenogale and none for Haplogale). Further, the euplerid, Cryptoprocta, is very convergent dentally to Proailurus, but the ancestor of Cryptoprocta is also completely unknown. If the insectivorous/omnivorous Eupleres/Fossa are indeed sister to the remaining Malagasy euplerids (Galidinae + Cryptoprocta), there are likely many missing fossil taxa that are needed to fill the gap.

Fossil biases

Two questions immediately come to mind when considering character evolution and morphological diversification in feloids: first, where are the transitional forms leading to derived taxa such as Cryptoprocta and Proailurus? Second, where are all the fossil viverrids and herpestids? There is substantial fossil evidence for hyaenids and felids during the Miocene, but

78

little for viverrids or herpestids until the late Miocene/Pleistocene. To investigate these questions, a consideration of biogeography and relevant climatic and geologic events that occurred during the feloid radiation may provide some insight. The earliest feloids appeared in Asia during the Oligocene (about 33 my) and included taxa such as Stenoplesictis, Stenogale and Paleoprionodon (Dashzeveg 1996). At about this same time, there was a significant shift in the European faunal composition (Prothero 1994; Dashzeveg 1996; Alroy et al. 2000), during a period known as la Grande Coupure (the Great Cut). This time period was characterized by changes in continental air and water currents that led to increased aridification and cooling (Prothero 1994; Alroy et al. 2000; Hunt 2001). Over the next 10-15 my, tectonic changes on the European and Asian continents led to the uprising of the Alps, and also destroyed the connection between the Asian epicontinental seas and the Tethys, and formed the Paratethys Sea in the western European basin (Prothero 1994; Dashzeveg 1996). In combination, these events could plausibly have enabled the stem feloids in Asia to migrate overland via the now-exposed landmass (Heissig 1979; Dashzeveg 1996; Hunt 2001). It is worth noting that when feloids appear in the European fossil record, they are represented by the same taxa as in Asia: Stenoplesictis and Paleoprionodon (Hunt 2001, Peigne and de Bonis 1999), along with the more derived genera Proailurus and Haplogale (Hunt 2001). Since Asia’s fossil record is largely unexplored, additional, transitional forms (such as ancestors of Proailurus) may still wait for discovery. Once in Western Europe, feloids experienced a rapid radiation (Dashzeveg 1996), and subsequently migrated into Africa, Southern Europe, back across Eastern Europe, Asia, and (in the case of felids) into North and South America (Tedford et al. 1987; Hunt, Jr. 1996; Rothwell 2003). The re-dispersal across Europe has particular relevance for viverrids and herpestids: migrations overland could be expected to produce fossils in many areas, but modern viverrids and many herpestids are preferentially tropical and forest-dwellers – climates and habitats that are not conducive to fossilizaton. Additionally, if any suitable habitats were located near the ocean during the Miocene, those habitats are now under water, covered by the Mediterranean and Indian Oceans. Thus, if it is possible to accept the assumption that no substantial habitat shift has occurred during viverrid diversification, then gaps in the fossil record may be a direct result of the ecology of the taxa involved: the fossils simply do not exist. There are, however, even simpler explanations for the low number of viverrid and herpestid fossils. The first is that fossils are abundant in Europe because people are abundant – the well-known Quercy deposits in France were initially excavated as mines. In contrast, the continent of Africa, which was almost certainly the site of significant taxonomic diversification and probably the earliest herpestids/euplerids, has only recently begun to be the subject of non-hominid fossil exploration; indeed, most fossils that are recovered are still found at hominid sites. Thus, ongoing work in this region may well produce good quality material that will help to fill in the missing data. Here I present the most thoroughly sampled phylogeny yet available for the feloid Carnivora. The relationships between various feloid families have been significantly clarified, while patterns of diversification within those families (e.g. placement of the Machairodontines within the Felidae and the taxonomic placement of early hyaenids) have also been better established. The position of many recent and fossil taxa of questionable affinities, such as Nandinia, Herpestides, Haplogale, Stenogale, and Stenoplesictis, are also much better understood, which helps to establish character polarities as well as improves estimation of the primitive condition for the Feloidea. Identification of potential biases in the fossil record will inform those estimates. This study and the phylogeny it produced will facilitate subsequent research into character evolution, as well as stimulate further testing of evolutionary and ecological hypotheses within the feloid Carnivora.

79 Herpestes sanguineus Galerella pulverulenta 79 Herpestes auropunctatus * Herpestidae 80 A tilax paludinosus * Cynictis penicillata * 91 85 Ichneumia albicauda crassicauda * 64 93 Herpestes javanicus Herpestes ichneumon * 100 Suricata suricatta 91 Helogale parvula mungo *

100 84 Mungotictis decemlineata Eupleridae 100 Galidictis fasciata 98 Galidia elegans Salanoia concolor * 100 100 Cryptoprocta ferox 100 Fossa fossana

Eupleres goudotii * Hyaenidae 80 Hyaena brunnea * 87 99 Hyaena hyena * 100 Crocuta crocuta Proteles cristata 100 Prionodon linsang * Prionodon pardicrolor * 100 Felis margarita 81 Felis silvestris 59 Felis manul Felis yagouaroundi 58 Lynx canadensiss 87 Lynx rufus 97 Felis bengalensis * Felis * Felis viverrina Felidae 85 Acinonyx jubatus Felis temmincki Felis caracal Felis concolor Felis nigripes * Felis geoffroyi * 100 Felis colocolo * Felis pardalis * 100 Panthera tigris 100 100 Panthera onca 98 100 Panthera leo Panthera pardus 100 Panthera uncia Neofelis nebulosa 100 Paguma larvata 96 Paradoxurus hermaphroditus 85 Arctictis binturong 64 Arctogalidia trivirgata * Viverridae 74 Chrotogale owstoni Cynogale bennetti * 100 Genetta sp. 100 75 Genetta servalina 90 Genetta tigrina 98 Genetta maculata * 72 Poiana richardsoni * 100 Viverricula indica * 93 Viverra tangalunga Civettictis civetta Nandinia binotata 100 Canis latrans 100 Cuon alpinus Vulpes vulpes 97 Ictonyx striatus Ursus americanus

FIGURE 11. Bayesian phylogeny for recent taxa only. Taxa represented by only morphological data are indicated by *. Bayesian support values are indicated on the branches; posterior probabilities shown as percents. Nodes with less than 50% posterior probability are collapsed.

80 Felis bengalensis * Felis serval * Felis pardalis * Acinonyx jubatus Felis concolor Felis viverrina

Lynx canadensis Felidae Lynx rufus Felis temmincki Felis caracal Felis geoffroyi * Felis colocolo * Felis yagouaroundi Felis attica * Felis silvestris Felis nigripes * Panthera tigris Panthera onca Panthera leo Panthera pardus Panthera uncia Neofelis nebulosa Herpestes auropunctatus * Cynictis penicillata *

Herpestes sanguineus Herpestidae Galerella pulverulenta Atilax paludinosus * Herpestes javanicus * Herpestes ichneumon * Ichneumia albicauda Mungos mungo * Suricata suricatta Helogale parvula Bdeogale crassicauda *

Galidia elegans Eupleridae Salanoia concolor * Galidictis fasciata Cryptoprocta ferox Fossa fossana Eupleres goudotti * Hyaenidae Crocuta crocuta Hyaena brunnea * Hyaena hyena * Proteles cristatus Genetta sp. Genetta servalina * Poiana richardsoni * Genetta maculata * Prionodon pardicolor * Viverridae Prionodon linsang * Genetta tigrina Viverricula indica * Viverra zibetha * Civettictis civetta Cynogale bennetti * Paguma larvata Paradoxurus hermaphroditus Arctictis binturong Chrotogale owstoni Arctogalidia trivirgata * Nandinia binotata Ictonyx striatus Ursus americanus Vulpes vulpes Canis latrans

FIGURE 12. Constrained analysis of morphological data. Recent taxa only. Majority rule consensus of constrained parsimony analysis of recent taxa only. Taxa represented by only morphological data are indicated by *.

81 Felis bengalensis Felis serval Felis viverrina Felis geoffroyi Felis colocolo Acinonyx jubatus Pseudaelurus validus 61803 AMNH f Felis concolor Felis Lynx rufus X9489 f Felis nigripes Lynx canadensis Lynx rufus Felis temmincki Felispardalis Felis caracal Lynx shansius AMNH f Felis yagouaroundi Felis silvestris Felis attica AMNH f Panthera tigris Panthera Panthera onca Panthera leo Panthera pardus Panthera atrox 579 LACM f Panthera atrox 2900 19 LACM f Panthera atrox f Panthera uncia Metailurus parvulus M3895 f Metailurus major f

Felis concolor X8627 f Machairodontinae Felis concolor f Neofelis nebulosa Metailurus major (composite) SMNH f Dinofelis BC 120 SMNH f Dinofelis CB 20 f Smilodon (composite) PM f Homotherium nestianum 104641 AMNH f Machairodontinae genus indet. f Metailurus sp. 131854 f Herpestes auropunctatus Cynictis penicillata

Herpestes sanguineus Herpestidae Galerella pulverulenta Atilax paludinosus Herpestes ichneumon Ichneumia albicauda Herpestes javanicus Suricata suricatta Bdeogale crassicauda Helogale parvula

Mungos mungo Eupleridae Galidia elegans Galidictis fasciata Salanoia concolor Cryptoprocta ferox Proailurus lemanensis 10931 AMNH f Fossa fossana Eupleres goudotti

Palinhyaena reperta Ch 42 L344 AMNH f Hyaenidae Hyaenotherium wongii (composite) AMNH f Palinhyaena reperta SMNH f Palinhyaena reperta CH L46 AMNH f Lycyaena chaeretis (composite) CH 11 L12 f Crocuta crocuta Hyaena brunnea Hyaena hyena Proteles cristatus Genetta sp. Genetta servalina Prionodon pardicolor Prionodon linsang Genetta tigrina Viverridae Poiana richardsoni Genetta maculata Viverricula indica Viverra zibetha expectata 18725 f Herpestides antiquus (composite) MNHN f Hyaenotherium wongii (composite) f Civettictis civetta Paguma larvata Paradoxurus hermaphroditus Arctictis binturong Arctogalidia trivirgata Chrotogale owstoni Cynogale bennetti Fossa fossana spelaea 9391 BM f Nandinia binotata Ictonyx striatus Ursus americanus Vulpes vulpes Canis latrans

FIGURE 13. Constrained analysis of morphological data. Taxa at least 50% complete. Majority rule consensus of constrained parsimony analysis of taxa at least 50% complete. Fossil material is represented by f.

82 Pseudaelurus sp. 61910 AMNH VP f Pseudaelurus validus 61803 AMNH f Nimravides (composite) f Acinonyx jubatus Dinofelis sp. 50446 AMNH f Felis geoffroyi Felis concolor f Felis concolor X8627 f Felis concolor

Felis bengalensis Felis Felis serval Felis viverrina Felis pardalis Lynx rufus X9489 f Felis nigripes Lynx canadensis Lynx rufus Felis temmincki Felis attica AMNH f Felis caracal Felis colocolo Felis yagouaroundi Lynx shansius AMNH f Felis silvestris Metailurus sp. Ch 45 L399 f Panthera atrox 579 LACM Panthera atrox 2900 19 LACM f Panthera atrox f Panthera Smilodon (composite) f Nimravides sp. HIG 227 1790 AMNH f Panthera onca Panthera tigris Panthera leo Panthera pardus Metailurus parvulus M3895 f Metailurus major f Panthera uncia Neofelis nebulosa Machairodontinae Felis attica PIK 3232 f Dinofelis sp. 50445 AMNH f Metailurus major (composite) SMNH Dinofelis 50453 f Dinofelis BC 120 SMNH f Dinofelis CB 20 f Smilodon (composite) PM f Homotherium nestianum 104641 AMNH f Machairodontinae genus indet. Metailurus sp. 131854 f Pseudaelurus intrepidus 18271 AMNH f Adcrocuta eximia SMNH f Chasmaporthetes lunensis 99787 AMNH f Adcrocuta eximia 30 SMNH f Pseudaelurus quadridentatus MNHN f Proailurus lemanensis 10931 AMNH Proailurus lemanensis SG 3509 MNHN f Stenoplesictis cayluxi (composite 2) MNHN f Stenoplesictis cayluxi (composite 3) MNHN f Paleoprionodon lemandini Qu 9366 MNHN f Herpestides antiquus SG3116 MNHN f Stenogale julieni SG 3210 MNHN f Herpestides antiquus GQ13186 MNHN f Herpestides antiquus (composite) MNHN f Herpestides antiquus SG319 f Tungurictis spocki 26600 AMNH f Thalissictis robusta M8981 BM f Thalissictis robusta M8983 BMf Ichneumia albicauda Herpestidae Mungos mungo Herpestes ichneumon Atilax paludinosus Herpestes sanguineus Galerella pulverulenta Cynictis penicillata Herpestes auropunctatus Herpestes javanicus Suricata suricatta Helogale parvula

Bdeogale crassicauda Eupleridae Galidia elegans Salanoia concolor Galidictis fasciata Cryptoprocta ferox Pseudaelurus marshi Rothwell f Cryptoprocta ferox spelaea 9948 BM f Fossa fossana Eupleres goudotti Kichechia composite f Plioviverrops orbignyi PIK 3016 MNHN f Crocuta crocuta Hyaena brunnea Adcrocuta eximia M8966 BM f

Hyaena hyena Hyaenidae Lycyaena chaeretis 8978 BM f Proteles cristatus Plioviverrops orbignyi Ictitherium viverrinum PIK 3004 MNHN f Palinhyaena reperta Ch 42 L344 AMNH f Hyaenotherium wongii (composite) AMNH f Lycyaena chaeretis (composite) CH 11 L12 AMNH f Palinhyaena reperta CH L46 AMNH f Palinhyaena reperta (composite) SMNH f Palinhyaena reperta M3854 SMNH f Ictitherium viverrinum M3206 SMNH Hyaenotherium wongii (composite) f Genetta maculata Stenogale coupant (composite) MNHN f Poiana richardsoni Genetta sp. Genetta servalina Viverridae Prionodon pardicolor Prionodon linsang Genetta tigrina Tungurictis X. Wang (undescribed) Viverricula indica Civettictis civetta Fossa fossana spelaea 9391 BM f Viverra zibetha expectata 18725 f Paguma larvata Paradoxurus hermaphroditus Arctictis binturong Chrotogale owstoni Cynogale bennetti Arctogalidia trivirgata Haplogale media 9435 MNHN f Paleoprionodon lamandini 9370 MNHN f Nandinia binotata Vulpes vulpes Canis latrans Ursus americanus Lycyaena dubia M3856 f Ictonyx striatus

FIGURE 14. Constrained analysis of morphological data. Taxa at least 20% complete. Majority rule consensus of constrained parsimony analysis of taxa at least 20% complete. Fossil material indicated by f.

83 Felis geoffroyi Feliscolocolo Felis caracal Felis yagouaroundi Lynx shansius AMNH Metailurus parvulus M3895 Metailurus major Felis concolor X8627 Felis concolor Pseudaelurus marshi Rothwell Pseudaelurus validus 61803 AMNH Pseudaelurus sp. 61910 AMNH VP

Nimravides (composite) Felis Acinonyx jubatus Felis concolor Lynx rufus X9489 Felis nigripes Lynx canadensis Lynx rufus Felis manul Felis pardalis Felis bengalensis Felis serval Felis viverrina

Feli smargarita FELIDAE Felis temmincki Felis attica PIK 3232

Machairodontinae genus indet. Machairodontinae Smilodon (composite) PM Homotherium nestianum 104641 AMNH Dinofelis BC 120 SMNH Dinofelis CB 20 Dinofelis piveateau 50453 Metailurus major (composite) SMNH Dinofelis sp. 50445 AMNH Felis silvestris Felis attica AMNH Metailurus sp. Ch 45 L399 Panthera atrox 579 LACM Panthera atrox 2900 19 LACM Panthera atrox

Smilodon (composite) Panthera Nimravides sp. HIG 227 1790 AMNH Metailurus sp. 131854 Dinofelis sp. 50446 AMNH Panthera uncia Panthera tigris Panthera onca Panthera leo Panthera pardus Neofelis nebulosa Pseudaelurus intrepidus (composite) AMNH Early felids Pseudaelurus quadridentatus MNHN Proailurus lemanensis 10931 AMNH Proailurus lemanensis SG 3509 MNHN Stenogale coupant (composite) MNHN Herpestides antiquus (composite) MNHN Herpestides antiquus SG319 Herpestides antiquus SG3116 MNHN Early feloids Stenogale julieni SG3210 MNHN Herpestides antiquus GQ13186 MNHN Paleoprionodon lamandini 9370 MNHN Haplogale media 9435 MNHN Paleoprionodon lemandini Qu 9366 MNHN Stenoplesictis cayluxi (composite 2) MNHN Stenoplesictis cayluxi (composite 3) MNHN Ictonyx striatus Ursus americanus Vulpes vulpes Canis latrans FIGURE 15. Constrained analysis of the Felidae. Constrained analysis of the Felidae + early feloids.

84 Ichneumia albicauda Mungos mungo Herpestes ichneumon

Atilax paludinosus Herpestidae Herpestes sanguineus Galerella pulverulenta Cynicti penicillata Herpestes auropunctatus Herpestes javanicus Suricata suricatta Helogale parvula Bdeogale crassicauda Plioviverrops orbignyi PIK 3016 MNHN Plioviverrops orbignyi Galidia elegans

Salanoia concolor Eupleridae Galidictis fasciata Cryptoprocta ferox Cryptoprocta ferox spelaea 9948 BM Kichechia (composite) Fossa fossana Eupleres goudotti Fossa fossana spelaea 9391 BM Stenoplesictis cayluxi (composite 2) MNHN Stenoplesictis cayluxi (composite 1) MNHN Stenogale coupant (composite) MNHN Tungurictis X. Wang (undescribed) Stenogale julieni SG 3210 MNHN Herpestides antiquus SG 3116 MNHN Hyaena brunnea Lycyaena dubia M3856 Hyaena hyena Crocuta crocuta Adcrocuta eximia 30 SMNH Adcrocuta eximia SMNH Adcrocuta eximia M8966 BM Lycyaena chaeretis 8978 BM Hyaenidae Chasmaporthetes lunensis 99787 AMNH Palinhyaena reperta Ch 42 L344 AMNH Hyaenotherium wongii (composite) AMNH Lycyaena chaeretis various CH 11 L12 AMNH Palinhyaena reperta CH L46 AMNH Palinhyaena reperta (composite) Palinhyaena reperta M3854 SMNH Ictitherium viverrinum M3206 SMNH Hyaenotherium wongii (composite) Ictitherium viverrinum PIK 3004 (MNHN) Thalissictis robusta M8981 BM Thalissictis robusta M8983 BM Tungurictis spocki 26600 AMNH

Herpestides antiquus (composite) MNHN Early feloids Paleoprionodon lemandini Qu 9366 MNHN Herpestides antiquus SG319 Nandinia binotata Haplogale media MNHN 9435 Paleoprionodon lamandini MNHN 9370 Herpestides antiquus MNHN GQ13186 Ictonyx striatus Ursus americanus Vulpes vulpes Canis latrans FIGURE16. Constrained analysis of (Herpestidae/Eupleridae)/Hyaenidae. Constrained analysis of (Herpestidae/Eupleridae)/Hyaenidae and early feloids + Nandinia.

85 Genetta maculata

Stenogale coupant (composite) MNHN

Poiana richardsoni

Genetta sp.

Genetta servalina

Prionodon pardicolor

Prionodon linsang

Genetta tigrina Viverridae

Viverricula indica

Viverra zibetha expectata 1872

Civettictis civetta

Cynogale bennetti

Paguma larvata

Paradoxurus hermaphroditus

Arctictis binturong

Chrotogale owstoni

Arctogalidia trivirgata

Herpestides antiquus SG3116 MNHN

Stenogale julieni SG3210 MNHN

Haplogale media 9435 MNHN

Paleoprionodon lamandini 9370 MNHN Early feloids

Stenoplesictis cayluxi (composite) MNHN

Stenoplesictis cayluxi (composite) MNHN

Nandinia binotata

Herpestides antiquus (composite) MNHN

Herpestides antiquus GQ13186 MNHN

Herpestides antiquus SG319

Paleoprionodon lemandini Qu 9366 MNHN

Ictonyx striatus

Ursus americanus

Vulpes vulpes

Canis latrans

Figure 17. Constrained analysis of the Viverridae. Constrained analysis of the Viverridae and early feloids.

86

CONCLUSION

The focus of this dissertation research was to explore the effects of morphological specialization to a dental/dietary phenotype, that of the hypercarnivore. The results clearly show that phenotypic specialization to hypercarnivory has a significant effect on morphological diversity and character evolution, and that this specialization acts to limit subsequent morphological diversity and both the rate and direction of morphological character change. While hypercarnivore clades are significantly reduced in their rate of reverse change, they do not exhibit a higher rate of change towards increasingly specialization. Passive movement into specialized morphospace is a pattern that is consistent with a functional constraint representing an evolutionarily stable system (Wagner and Schwenk 2000). In an evolutionarily stable system, a morphological change that improves the organization and integration of a character complex as a whole will be selected for, but changes that reduce that integration will be selected against. To explore and extend these findings, Chapters 3 and 4 focus on building new phylogenies, both molecular and morphological, in order to better understand patterns of phenotypic evolution within the feloid Carnivora, a group that includes the families Felidae, Hyaenidae, Viverridae, Herpestidae, Eupleridae and Nandiinidae. The phylogenies obtained show that the family Viverridae is sister to Felidae (Hyaenidae(Herpestidae/Eupleridae)) and Nandiniidae is sister to all other extant feloids. Additionally, the fossil genus Herpestides is an early member of the herpestid/euplerid/hyaenid clade, while the Miocene hyaena Plioviverrops is more accurately placed with the Herpestidae. These new trees will allow more rigorous testing of character and/or environmental correlates and, since they include fossil material, will also allow better estimation of ancestral conditions, help to establish character polarities, and will provide additional rate information. On whole, this research represents a significant contribution to both evolutionary biology and to systematics. Chapters 1 and 2 offer innovative approaches for the study of phenotypic evolution, particularly through the use of replicated sister group comparisons, while Chapters 3 and 4 provide carefully evaluated, good quality phylogenetic data that can be used by workers in fields including paleontology, systematics, evolutionary biology, and comparative biology.

87

APPENDIX A. Hypotheses of phylogenetic relationships for hypercarnivorous clades and their sister groups used to build the composite phylogenies used in Chapters 1 and 2. More detailed, species-level phylogenies were used for frequency of change comparisons and are available upon request. A. Felidae/Hyaenidae. B. Philotrox-Sunkahetanka-Enhydrocyon/Cynodesmus. C. Chasmaporthetes-Lycyaena-Hyaenictis/Palinhyaena-Crocuta. D. Mustela/Galictis-Ictonyx- Pteronura-Lontra-Enhydra-Lutra-Amblonyx-Aonyx. E. Cryptoprocta/Eupleres-Fossa. F. Nimravidae/Herpestidae-Viverridae-Felidae-Hyaenidae.

88 A. C. Proctictitherium Lycyaena Plioviverrops Hyaenictis Proteles Tongxinictis Chasmaporthetes Lycyaena Palinhyaena Hyaenictis Chasmaporthetes Ikelohyaena Palinhyaena Belbus Ikelohyaena Hyaenid "sp. E" Belbus Leecyaena (Hyaena) Leecyaena (Hyaena) Parahyaena Parahyaena Pliocrocuta Pachycrocuta Hyaena Crocuta Pliocrocuta Adcrocuta Hyaena Pachycrocta Hyaenid "sp. E". Crocuta Hyaenictitherium Miohyaenotherium Adcrocuta Hyaenotherium Thalassictis Ictitherium Tungurictis D. Proailurus Mustela Pseudaelurus Enhydra Homotherium Paramachairodus Lutra Ictonyx Smilodon Nimravides Galictis Dinofelis Felis Panthera Acinonyx E. Lynx Eupleres Fossa Cryptoprocta

B.

Cynodesmus thooides Cynodesmus martini F. Philotrox condoni Viverridae Sunkahetanka geringensis Felidae Enhydrocyon basilatus Hyaenidae Enhydrocyon pahinsintewakpa Herpestidae Enhydrocyon crassidens Nimravidae Enhydrocyon stenocephalus

89

APPENDIX B. Specimens used to calculate morphological diversity in Chapter 1. Museum abbreviations are as follows: The American Museum of Natural History (AMNH); The Frick Collection, The American Museum of Natural History (F:AM); British Museum of Natural History (BMNH); The Field Museum of Natural History (FMNH); Florida Museum of Natural History (UF); Museum of Comparative Zoology, Harvard University (MCZ), Texas Memorial Museum (TMM). Specimens designated P, PM, UM, UT and UC are currently housed at the Field Museum.

90

Felidae: Acinonyx: BMNH 16573; Dinobastus serus: UF 22908, UF 22909; Felis brachygnatha: BMNH16537; Felis amnicola: UF 1933, UF 19351, UF 19352; Felis aurata: AMNH 51998, AMNH 51994; Felis badia: FMNH 8378; Felis bengalensis: FMNH 62894; Felis chaus: FMNH 105559; Felis colo colo: AMNH 189394, AMNH 16695, FMNH 43291; Felis rexroadensis: UF 25067, UF 58308; Felis serval: AMNH 34767, AMNH 205151; Felis viverrina: FMNH 105562; Homotherium serum: UF 24992; Hoplophoneus occidentalis: AMNH 102394; Machaerodus aphanistus: BMNH 8975; Machaerodus cultridens: BMNH 49967A; Machairodus palanderi: F:AM 50476, F:AM 50478; Megantereon cultridens: AMNH 105446; Megantereon falconeri: BMNH 16350, BMNH 16557; Megantereon hesperus: UF 22890; Metailurus: F:AM China L-604, F:AM 95294; Nimravides: AMNH 25206, UF 24471, UF 24479, F:AM 61855, F:AM 104044; Nimravides catacopis: Kan 45-99, Kan 93-60; Nimravides galiani: UF 24462; Panthera leo: UF 10643, UF 10645; Panthera onca: UF 14765, UF 14766, UF 23685; Panthera paleonca: TMM 31181-192, TMM 31181-193; Panthera pardus: AMNH 35522; Paramachairodus ogygia: BMNH 1574; Paramachaerodus orientalis: BMNH 8959; Proaelurus lemanansis: BMNH 1646, AMNH 105065, AMNH 101931, AMNH 107658; Proailurus medius: BMNH 9636, BMNH 9640; Pseudaelurus: AMNH 18007, AMNH 27318, AMNH 27446, AMNH 27447, AMNH 61938, AMNH 62129, AMNH 62190, AMNH 62192, F:AM 61925, AMNH 27451-A , BMNH 9633; Pseudaelurus marshi: F:AM 27453, F:AM 27457; Pseudailurus intermedius: BMNH 2375; Smilodon californicus: UF 167140, UF 167141; Smilodon floridanus: UF 22704, UF 22705 Smilodon gracilis: UF 81700; Xenosmilus hodsonae: UF 60000 Nimravidae: Barbourofelis: P15811; Barbourofelis fricki: AMNH 103202, AMNH 108193, F:AM 61982; Barbourofelis lovei: UF 24447, UF 24429, UF 36858, UF 37052; Barbourofelis morrisi: AMNH 25201, F:AM 79999; Barbourofelis whitfordi: AMNH C38A-210, F:AM 69454, F:AM 69455; Dinictis cyclops: AMNH 6937; Dinictis felina: AMNH 38805 P12004 PM 21039; Dinictis: UF 155216, UF 207947; Eusmilus cerebralis: AMNH 6941; Eusmilus: F:AM 99259, F:AM 98189; Hoplophoneus: F:AM 69344, UC 1754; Hoplophoneus primaevus latidens: UM 420, UM 701; Hoplophoneus oharrai AMNH 27798, AMNH 82911; Hoplophoneus oreodontis: AMNH 9764; Nanosmilus kurteni: UF 207943; Nimravus: F:AM 62151; Nimravus sectator: AMNH 12882; Nimravus brachyops: AMNH 6930; Nimravus gomphodus: AMNH 6935; Pogonodon: AMNH 1403, F:AM 69369; AMNH 1398; Pogonodon platycopis: AMNH 6938; Sansanosmilus: AMNH 26608; Vampyrictis vipera: T 3335 Hyaenidae: Acrocuta eximia: AMNH 26372, BMNH 8971, M8968, M901; Chasmaporthetes AMNH 99788; Chasmaporthetes exilelus: AMNH 26369; Chasmaporthetes lunensisi: F:AM China 94 B-1046, F:AM China 96B 1054, AMNH 10261, AMNH 26955; Chasmaporthetes ossifragus: AMNH 108691, AMNH 95208; Crocuta crocuta: AMNH 187771, FMNH 98952 UF 5665; Hyaena brunnea: FMNH 34584; Hyaena hyaena dubbah: FMNH 140216; Ictitherium pannonicum: BMNH 8983; Ictitherium Robustum: BMNH 8987, BMNH 8988; Ictitherium viverrinum: F:AM China (G) - L100; Leecyaena bosei: BMNH 1554, BMNH 37133, BMNH 578; Lycyaena chaeretis: F:AM China 26-B47, F:AM China 52- L495, F:AM China 56-L560; Lycyaena choeretis: BMNH 8978, BMNH 8979a; Lycyaena crusafonti: AMNH 108175, AMNH 116120; Pachycrocuta perrieri: AMNH 27756, F:AM 107766, F:AM 107767, AMNH 27757; Palhyaena reperta: F:AM China 42-L338, F:AM China 51-L443; Pliocrocuta: BMNH 16565; Plioviverrops: AMNH 99607; Proteles cristatus: FMNH 127833; Thalassictis hyaenoides: F:AM China 14-L344, F:AM China 14-L35; Thalassictis wangii: AMNH 20555, AMNH 20586; Tungurictis spocki: AMNH 26600, AMNH 26610; Viverridae: Arctogalidia trivirgata stigmatica: FMNH 68709; Chrotogale owstoni: FMNH 41597; Cryptoprocta ferox: AMNH 30035, FMNH 161707, FMNH 161793, FMNH 33950, FMNH 5655; Cryptoprocta ferox spelaea: BMNH 9949; Eupleres goudotii: FMNH 30492, AMNH 188211; Fossa fossa: AMNH 188209, AMNH 188210, FMNH 85196; Genetta genetta senegalesis: FMNH 140213; Genetta maculata: FMNH 153697; Nandinia binotata: FMNH 25306; Prionodon linsang: FMNH 8371; Viverra zibetina picta: FMNH 75883; Viverricula indica babistae: FMNH 75815, FMNH 75816; Canidae: Hesperocyoninae: Cynodesmus thooides: AMNH 129531; Ectopocynus antiquus: AMNH 63376; Ectopocynus simplicidens: F:AM 25426, F:AM 25431; Enhydrocyon basilatus: AMNH 129549, F:AM 54072; Enhydrocyon crassidens: AMNH 12886, AMNH 27579, AMNH 59574; Enhydrocyon pahinsintewakpa: AMNH 129535; Hesperocyon gregarious: AMNH 9313; Mesocyon coryphaeus: AMNH 6859; Mesocyon temnodon: F:AM 63367; Osbornodon fricki: AMNH 27363; Osbornodon: AMNH 54325;

91

Parenhydrocyon: AMNH 81086; Parenhydrocyon josephi: F:AM 54115; Philotrox condoni: AMNH 32796, F:AM 63383; Prohesperocyon wilsoni: AMNH 12712; Sunkahetanka geringensis: AMNH 96714 Canidae: Borophaginae: Aelurodon taxoides: F:AM 61781; Borophagus diversidens: AMNH 67364; Borophagus secundus: AMNH 61640; Epicyon haydeni: F:AM 61461; Epicyon saevus: F:AM 61432; Euoplocyon praedator: AMNH 18261; Euoplocyon: AMNH 25443, AMNH 27315; Paratomarctos euthos: F:AM 61101; Desmocyon thomasii: AMNH 12874 Mustelidae: Arctonyx collaris: AMNH 57373; Eira barbara: AMNH 128127, AMNH 29597, UF 3194; Enhydra lutra: UF 24196; Galictis cuja: AMNH 33281; Galictis vittata UF 29310; Gulo gulo: AMNH 35054, FMNH 14026, AMNH 169501; Ictonyx: AMNH 165812; Lutra canadensis: UF 24007; Martes americana: UF 13212; Martes cauvina: UF 5642; Martes foina: UF 29046, AMNH 70182; Martes pennanti: UF 23316; Mustela altaica UF 26514; Mustela erminea: UF 3982, UF 1417; Mustela felipei: AMNH 63839, FMNH 86745; Mustela frenata: UF 26144, UF 4779; Mustela kathiah: FMNH 32502; AMNH 150090; Mustela nigripes: AMNH 22894; Mustela nivalis: UF 1418; Mustela putorius: UF 1425, UF 14422; Mustela sibirica: AMNH 114878, F:AM 104392; Mustela vison: UF 4569, UF 4724, UF 8108; Mydaus javanensis: AMNH 102701; Poecilogale albinucha: AMNH 86491; Taxidea taxus: UF 384; Vormala peregusna negans: AMNH 60103 "Paleo" mustelids: Aelurocyon: F:AM 25430; Aelurocyon brevifacies: P12283, P12154 P26051; Brachysypsalis: AMNH 25295, AMNH 25299, AMNH 27307, F:AM 25284, F:AM 25363, AMNH 27424; Oligobunis crassivultus: AMNH 6903, PM 537; Oligobunis darbyi: P 25609; Oligobunis floridanus: MCZ 4064; Paroligobunis frazieri: UF 23928; Plesictis cf. Pygmaeus: M27815; Plesiogulo marshalli: UF 19253; Promartes lepidus: P12155; Zodialestes: F:AM 27599, F:AM 27600; Zodialestes daimonelixensis: P12032

92

APPENDIX C. Identification for tissues and DNA indicating families, species, type of preservation, institution from which the material was obtained, the voucher number for that material, and GenBank accession numbers where available. Where no GenBank number is listed, that gene did not amplify for that taxon. New sequences (indicated by “yes”) will be submitted to GenBank. All extractions were performed on either tissue (muscle/organ) or blood samples. Institutions are as follows: Field Museum of Natural History (FMNH); Carnagie Museum of Natural History (CMNH); Museum of Comparative Zoology, Berkeley (MCZ-B); Ellerman Museum (South Africa) (EM); Brookfield Zoo (BZ); Texas Tech University (TTU); Museum of Southwestern Biology, Division of Genetic Resources (MSB-DGR); Louisiana State University Museum of Natural History (LSU); Florida Museum of Natural History (UF).

93

Specimen name Preservation Institution VoucherRAG1-1 RAG1-2 c-myc GHR cyt-b Acinonyx jubatus frozen LSU M4407yes yes yes yes AY463959 Felis caracal blood BZ 1012yes yes yes yes Felis concolor frozen UF 29826 no no no yes AY5984987 Felis lynx blood BZ B224564yes yes yes yes AY773083 Felis manul blood BZ 900234yes yes yes yes Felis margarita blood BZ 900239yes yes yes yes Felis silvestris ethanol FMNH Chicagoyes no yes yes AY170109 Felis temmincki frozen TTU TK 21991yes yes yes yes Felis viverrina blood BZ 870254yes yes yes yes Felis yagouaroundi frozen LSU M3450yes yes yes yes Lynx canadensis blood BZ 24564yes yes yes yes AY928671 Lynx rufus ethanol MSB-DGR 1816yes yes yes yes Neofelis nebulosa frozen TTU TK 21866yes yes yes yes NC008450 Panthera leo ethanol MCZ-B 193137yes yes yes yes DQ022303 Panthera leo blood BZ 930133yes yes yes yes Panthera onca frozen TTU TK 21990yes yes yes yes Panthera pardus frozen TTU TK 21882yes yes yes yes AY005809 Panthera tigris frozen LSU M4474yes yes yes yes X82301 Panthera uncia frozen LSU M4958yes yes yes yes DQ097339 Galerella pulverulenta ethanol CMNH TM 41495yes yes yes yes AF522330 Galidia elegans frozen FMNH 161921yes yes yes yes AY170099 Galidictis fasciata frozen FMNH 156549yes yes yes yes AY170100 Helogale blood BZ 920200yes yes yes yes AF522339 Herpestes javanicus ethanol MCZ-B 186570yes yes yes yes DQ519070 Herpestes sanguineus frozen FMNH 145232yes yes yes yes AF522332 Ichneumia albicauda frozen FMNH 157991yes yes no yes AF522341 Mungotictis decemlineata frozen FMNH 176128yes yes yes yes AY170094 Suricata suricatta ethanol MSB-DGR 43303yes yes yes yes AY170111 Crocuta crocuta ethanol MSB-DGR 9975yes yes yes yes AY928676 Proteles cristatus blood BZ 950042yes yes yes yes AY928679 Arctictis binturong blood BZ yes yes no yes AY048793 Chrotogale owstoni ethanol MCZ-B 186571yes no yes yes AF125144 Civettictis civetta frozen FMNH 149357yes yes yes yes AF511043 Cryptoprocta ferox frozen FMNH 171889yes yes yes yes AY928681 Fossa fossana frozen FMNH 156648yes yes yes yes AF511062 Genetta tigrina frozen TTU TK 33198yes no no yes AY928672 Nandinia binotata frozen FM 161266yes yes yes yes AF522350 Paguma larvata ethanol MCZ-B 186576yes yes yes yes AF125151 Paradoxurus hermaphroditus ethanol MCZ-B 186573yes yes no yes AF511056 Viverra tangalunga ethanol FM 145742yes yes yes yes AF511045 Genetta sp. DMSO EM CM1yes no yes yes AY241922 Galerella sanguinea ethanol TTU TK 33108yes yes yes yes AF522332 Ictonyx striatus ethanol CMNH SP 7550yes yes yes yes AF498156 Vulpes vulpes n/a GENBANK DQ240715 AY928721 DQ498126 Canis latrans n/a GENBANK DQ240691 DQ240657 DQ480509 Cuon alpinus n/a GENBANK DQ240715 AF519448 DQ240663 Ursus americanus n/a GENBANK DQ240717 DQ205799 X82307 94

APPENDIX D. Specific model settings for each data partition for ML and Bayesian analyses. Model choice was performed using Modeltest (Posada and Crandall, 1998).

95

All data combined RAG-1 c-myc GHR cyt-b

Model: GTR + I + gamma Model: GTR + I + gamma Model: GTR + I + gamma Model: K81uf+G Model: GTR + I + gamma -ln likelihood 31153.808 -ln likelihood 8854.32192 -ln likelihood 3694.91634 -ln likelihood 3598.0115 -ln likelihood Base frequencies: Base frequencies: Base frequencies: Base frequencies: Base frequencies: A 0.277 A 0.25848 A 0.28221 A 0.2666 A 0.37298 C 0.2814 C 0.25667 C 0.24567 C 0.2909 C 0.3792 T 0.2261 T 0.26431 T 0.24611 T 0.2294 T 0.06489 G 0.2154 G 0.22054 G 0.22601 G 0.2131 G 0.18293

Substitution model: Substitution model: Substitution model: Substitution model: Substitution model: Rate matrix Rate matrix Rate matrix Rate matrix Rate matrix A-C 1.8669 A-C 1.70995 A-C 1.24208 A-C 1 A-C 0.186718 A-G 5.4698 A-G 7.32309 A-G 2.931769 A-G 3.14 A-G 10.96808 A-T 0.9046 A-T 0.4165 A-T 0.4456 A-T 0.4891 A-T 0.6894 C-G 0.7591 C-G 1.70698 C-G 1.112042 C-G 0.4891 C-G 0.363887 C-T 16.2605 C-T 12.3771 C-T 4.37062 C-T 3.1493 C-T 14.31992 G-T 1 G-T 1 G-T 1 G-T 1 G-T 1

ASRV ASRV ASRV ASRV ASRV I 0.439 I 0.31413 I 0.425645 I 0 I 0.431145 Gamma 0.4419 Gamma 0.76423 Gamma 0.82058 Gamma 0.3398 Gamma 0.560532

96

APPENDIX E. Characters and description of character states used for morphological phylogenetic analysis.

97

1. Basioccipital mid-ventral ridge State: 0 = absent or very weak State: 1 = present but weakly developed State: 2 = present and strongly (well) developed

2. Basioccipital ventral shape State: 0 = planar State: 1 = concave State: 2 = convex

3. Shape of the basisphenoid State: 0 = basisphenoid triangular State: 1 = cranial basisphenoid indented (e.g. felids)

4. Depression in frontals State: 0 = absent State: 1 = pit (thumbprint) State: 2 = groove

5. Nasal extension (view dorsally, or as straight line in lateral view from furthest extension to orbit) State: 0 = do not extend beyond anterior border of the orbits State: 1 = extend beyond anterior border of the orbits State: 2 = at anterior border of the orbits Character citation: Wozencraft 1984

6. Nasals depressed at midline caudally State: 0 = no State: 1 = yes Character citation: Wozencraft 1984

7. Nasals depressed at midline rostrally State: 0 = no State: 1 = yes

8. Nuchal crest position State: 0 = extends to end of occipital condyles State: 1 = does not reach end of occipital condyles State: 2 = extends beyond end of occipital condyles

9. Orientation of the premaxilla/maxillary suture relative to the upper tooth row State: 0 = sloping, shallow angle State: 1 = steep angle Character citation: Van Valkenburgh et al. 1990

10. Tooth row/palate extension (view in direct dorsal view) State: 0 = does not extend beyond post-orbital processes State: 1 = extends beyond post-orbital process Character citation: Veron 1995

11. Palatal extension (posterior molars) State: 0 = does not extend beyond most posterior molar State: 1 = extends past posterior molars State: 2 = extends to midline of zygomatic arch Character citation: Wozencraft 1984

98

12. Position of anterior edge of P4 relative to anterior root of zygomatic arch State: 0 = posterior to root of zygomatic arch or at root of zygomatic arch State: 1 = anterior to root of zygomatic arch Character citation: Van Valkenburgh et al. 1990

13. Post-canine constriction (rostrum continues straight or bows in slightly post canines) State: 0 = absent State: 1 = present

14. Posterior extension of the nasals relative to the furthest extension of the maxillofrontal suture State: 0 = nasals extend beyond the maxillofrontal suture State: 1 = nasals are at maxillofrontal suture State: 2 = nasals are anterior to the maxillofrontal suture

15. Premaxillary/maxillary suture position in relation to P1 (most posterior extension of suture) State: 0 = does not extend beyond canines State: 1 = extends beyond canines, does not reach 1st premolars State: 2 = reaches premolars Character citation: Gaubert et al. 2002

16. Premaxillary-frontal contact State: 0 = absent State: 1 = variable State: 2 = present Character citation: Gaubert et al. 2002

17. Processes on hamulus (hooks) State: 0 = absent State: 1 = present

18. Angle of processes on hamulus State: 0 = low relative to palate (as in herpestids) State: 1 = steep (as in felids)

19. Rostral extension of the nasals (when viewed laterally) State: 0 = not extended anteriorly, angle 45-70 degrees State: 1 = extended rostrally, creates very steep angle (70-90 degrees) from anterior tip of nasals to incisor arcade

20. Rostrum length State: 0 = very short State: 1 = short State: 2 = long State: 3 = very long Character citation: Gaubert et al. 2002

21. Shape of posterior skull (viewed dorsally) State: 0 = rounded/spherical State: 1 = elongate, ovoid, egg shaped State: 2 = flattened laterally (hyaenas) Character citation: Wozencraft 1984

22. Shape of rostrum State: 0 = elongate State: 1 = shortened Character citation: Wozencraft 1984

99

23. Shape of the basioccipital State: 0 = flat State: 1 = convex State: 2 = concave

24. Shape of the palate State: 0 = rectangular State: 1 = triangular Character citation: Veron 1995

25. Skull shape State: 0 = flattened State: 1 = gently rounded (herpestid-like) State: 2 = strongly domed (cat-like) State: 3 = hyaena like

26. Distinct nasal processes at anterior extension of nasals (extend freely) State: 0 = absent State: 1 = present

27. Shape of rostrum in lateral view State: 0 = straight from frontals to nasals State: 1 = strongly concave State: 2 = strongly convex

28. Hypoglossal and posterior lacerate foramen State: 0 = share foramen, large rounded opening State: 1 = share foramen, opening very small State: 2 = separated, both openings clearly visible by a thin wall or plate (thinner in viverrids) State: 3 = separated, both openings clearly visible, plf angled medially (felids) State: 4 = separated, both openings clearly visible, plf angled dorsally (viverrids)

29. Posterior lacerate foramen bounded by: State: 0 = bounded by entotympanic, basioccipital and paroccipital State: 1 = no paroccipital contribution State: 2 = no basioccipital contribution State: 3 = no entotympanic contribution

30. Medial lacerate foramina (aka internal carotid foramina) State: 0 = covered ventrally bby the bulla State: 1 = partially visible beneath a shelf or protrusion State: 2 = fully exposed Character citation: Wozencraft 1984

31. Medio-ventral surface of the petrosal (within the bulla) State: 0 = smooth State: 1 = distinct ventral petrosal process present Character citation: Wyss and Flynn 1993

32. Orientation of the entotympanic/ectotympanic junction State: 0 = diagonal State: 1 = perpendicular to the midline Character citation: Wozencraft 1984

100

33. Petrosal bone emerges medially between the basioccipital and entotympanic bulla State: 0 = no State: 1 = yes Character citation: Wozencraft 1984

34. Inclination of the tympanic ring State: 0 = horizontal or nearly so State: 1 = vertical or nearly so State: 2 = tilted dorsolateral to ventromedial

35. Lateral expansion of the mastoid State: 0 =absent State: 1 = present State: 2 = very well developed Character citation: Wozencraft 1984

36. Mastoid process (adheres) State: 0 = adheres to bullar wall State: 1 = does not adhere to bullar wall

37. Mastoid process (ventral extension) State: 0 = well developed, extends beyond bulla ventrally State: 1 = well developed, does not extend beyond bulla ventrally State: 2 = weakly developed

38. Mastoid process (shape) State: 0 = flat State: 1 = rugose with space for muscle attachments

39. Connection of paroccipital and mastoid State: 0 = no connection, notched State: 1 = connected by a low ridge State: 2 = no distinct separation

40. Fusion of paroccipitals to bulla State: 0 = well fused State: 1 = weakly attached State: 2 = distinct Character citation: Van Valkenburgh et al. 1990

41. Paroccipital process (development) State: 0 = simple State: 1 = cupped anteriorly around the posterior surface of the bulla State: 2 = extends caudally

42. Paroccipital process (width) State: 0 = narrow, triangular State: 1 = broad, triangular State: 2 = narrow, rounded State: 3 = broad, rounded

43. Paroccipital process (size) State: 0 = wide State: 1 = narrow (relative to the extensive size of the viverrid bulla) Character citation: Wozencraft 1984

101

44. Paroccipital processes (adheres) State: 0 = adheres to bullar wall State: 1 = does not adhere to bullar wall

45. Paroccipital processes (ventral extension) State: 0 = well developed, extends beyond bulla ventrally State: 1 = well developed, does not extend beyond bulla ventrally State: 2 = weakly developed

46. Paroccipital processes (shape) State: 0 = flat State: 1 = rugose with space for muscle attachments

47. Ectotympanic arms of the external auditory meatus form complete ring State: 0 = no State: 1 = yes Character citation: Wozencraft 1984

48. Ectotympanic bone anteromedial process (tubarius processus of Neff 1982) State: 0 = absent or very reduced State: 1 = reduced State: 2 = well-developed Character citation: Gaubert et al. 2002

49. Ectotympanic bone laterally expanded State: 0 = absent or very weak State: 1 = developed Character citation: Gaubert et al. 2002

50. Ectotympanic bone shape State: 0 = little inflated State: 1 = inflated Character citation: Gaubert et al. 2002

51. Ectotympanic contributes to external auditory meatal tube State: 0 = no State: 1 = yes Character citation: Wyss and Flynn 1993

52. Relative position of ectotympanic State: 0 = anterior State: 1 = anterolateral State: 2 = lateral, slightly anterior State: 3 = lateral

53. Relative size of ectotympanic to entotympanic State: 0 = equal State: 1 = 1/3 State: 2 = ¼ State: 3 = 1/5

54. Relative size of EAM to ectotympanic State: 0 = ½ size of ectotympanic State: 1 = smaller

102

55. Boundary of ento and ecto tympanic regions State: 0 = not distinct (boundary smooth) State: 1 = clearly distinct and constricted State: 2 = separate Character citation: Wozencraft 1984

56. Ectotympanic bone medially expanded State: 0 = absent or very weak State: 1 = developed Character citation: Gaubert et al. 2002

57. Entotympanic bulla shape State: 0 = spherical/rounded (herpestids) State: 1 = triangular/pyramidal (e.g. Paguma, hyaenids) State: 2 = rounded, ovoid (felids) State: 3 = ovoid/elongate (Prionodon)

58. Midventral ridge on entotympanic bulla State: 0 = absent State: 1 = present Character citation: Wozencraft 1984

59. Mid-entotympanic foramen State: 0 = absent State: 1 = present

60. Rostral entotympanic State: 0 = not visible State: 1 = projects from anterior-medial corner of the bulla Character citation: Wozencraft 1984

61. Alisphenoid canal (presence/absence) State: 0 = absent State: 1 = present Character citation: Wyss and Flynn 1993

62. Alisphenoid canal relative to pterygoids State: 0 = alisphenoid foramen posterior and behind the pterygoid processes State: 1 = posterior and lateral to the pterygoid processes State: 2 = anterior (and lateral) to the pterygoid processes State: 3 = alisphenoid canal absent Character citation: Veron 1995

63. Ethmoid foramen enlarged State: 0 = small (as in felids) State: 1 = nearly the size of the orbital fissure

64. Foramen ovale and foramen rotundum State: 0 = both positioned ventrally State: 1 = only the foramen ovale is positioned ventrally

103

65. Position of the palatine canal relative to sphenopalatine State: 0 = anterior and ventral to sphenopalatine State: 1 = ventral to sphenopalatine State: 2 = anterior to sphenopalatine State: 3 = posterior to sphenopalatine State: 4 = in same foramen

66. Posterior palatine foramina State: 0 = in maxilla State: 1 = in palatine State: 2 = on suture

67. Posterior palatine foramina relative to dentition State: 0 = anterior to p2 (herpestids, hyaenids) State: 1 = between p2-p3 (viverrids) State: 2 = at level of p4 (felids)

68. Size of the orbital fissure State: 0 =larger than optic canal State: 1 = roughly equivalent to optic canal

69. Size of I3 State: 0 = no increase in size State: 1 = slightly larger than I1 and I2 State: 2 = well developed, larger than I1 and I2

70. Incisor arcade State: 0 = curved State: 1 = straight or very slightly curved

71. Size of M1 State: 0 = large State: 1 = reduced or absent Character citation: Wyss and Flynn 1993

72. M1 parastyle enlarged State: 0 = no State: 1 = reduced but present State: 2 = developed State: 3 = very well developed, enlarged relative to other cusps Character citation: Wozencraft 1984

73. M1 protocone State: 0 = weak or absent State: 1 = moderately developed State: 2 = very well developed

74. Presence or absence of M2 State: 0 = absent State: 1 = present Character citation: Wyss and Flynn 1993

75. M2 protocone State: 0 = weak or absent State: 1 = moderately developed State: 2 = very well developed

104

76. Development of P1 State: 0 = absent State: 1 = present State: 2 = vestigial or very small (non-functional, no cusps) Character citation: Wyss and Flynn 1993

77. P3 lingual cusp State: 0 = absent State: 1 = present, development moderate State: 2 = present, very well developed Character citation: Wyss and Flynn 1993

78. P4 protocone position relative to parastyle State: 0 = medial State: 1 = anterior State: 2 = posterior Character citation: Wozencraft 1984

79. Size of P4 parastyle State: 0 = unreduced State: 1 = reduced

80. Number of talonid cusps on m1 O, 1, 2, 3, 4 etc. Note: a trenchant talonid = 1 cusp

81. Presence/absence of metaconid on m1 State: 0 = absent State: 1 = vestigial State: 2 = weakly developed State: 3 = present, well developed

82. m1 talonid State: 0 = absent State: 1 = present Character citation: Wyss and Flynn 1993

83. Development of m2 State: 0 = present, near size of m1 State: 1 = well developed with cusps, smaller than m1 State: 2 = present, reduced, few or no cusps State: 3 = absent Character citation: Wyss and Flynn 1993; Wozencraft 1984

84. Presence or absence of p1 State: 0 = absent State: 1 = present Character citation: Wyss and Flynn 1993

85. Orientation of trigonid on m1 (bounded by metaconid) State: 0 = non-equilateral triangle State: 1 = equilateral triangle State: 2 = nearly straight line State: 3 = no metaconid

105

86. Number of mental foramen on lateral surface of State: 0 = absent State: 1 = 1 State: 2 = 2 State: 3 = 3

87. Size of angular process State: 0 = poorly developed State: 1 = strongly developed

88. Shape of dentary State: 0 = flat, all parts in straight line State: 1 = curved

89. Shape at rear of dentary State: 0 = no change in curvature State: 1 = sharp break in region of angular/condyloid process (curves downward)

90. Shape of coronoid process State: 0 = sharp (salanoia helogale) State: 1 = hooked (paguma, arctictis, paradoxurus) State: 2 = blunted but squared (other viverrids) State: 3 = cat-like State: 4 = narrow (prionodon) State: 5 = triangular/pyramidal (herpestids)

91. Orientation of coronoid process State: 0 = no angulation State: 1 = angled caudally

92. Development of the glenoid processes State: 0 =no angulation State: 1 = strongly dorso-laterally angled State: 2 = strongly dorso-medially angled

93. Fossa (depression) on the ventral face of the mandibular symphysis State: 0 = absent State: 1 = present

94. m1 talonid loss of cusps State: 0 = strongly basined State: 1 = hypoconid relatively larger, entoconid reduced State: 2 = hypoconid medial State: 3 = hypoconid reduced or small talonid present (no cusps) State: 4 = no talonid Character citation: Hunt, Jr. 1998

95. Relative size of p2 to p 3-4 State: 0 = unreduced State: 1 = reduced State: 2 = absent Character citation: Hunt, Jr. 1998

106

96. m2 trigonid basined State: 0 = strong basin (equilateral triangle) State: 1 = weak basin (open triangle) State: 2 = no basin Character citation: Hunt, Jr. 1998

97. m2 ridge between metaconid and protoconid State: 0 = absent State: 1 = present Character citation: Hunt, Jr. 1998

98. p3 anterior cingulum cusp State: 0 = absent State: 1 = visible but weak State: 2 = well developed Character citation: Hunt, Jr. 1998

99. p4 anterior cingulum cusp State: 0 = absent State: 1 = visible but weak State: 2 = well developed Character citation: Hunt, Jr. 1998

100. p3 posterior accessory cusp State: 0 = absent State: 1 = visible but weak State: 2 = well developed Character citation: Hunt, Jr. 1998

101. p4 posterior accessory cusp State: 0 = absent State: 1 = visible but weak State: 2 = well developed Character citation: Hunt, Jr. 1998

102. p3 posterior cingulum cusp State: 0 = absent State: 1 = visible but weak State: 2 = well developed Character citation: Hunt, Jr. 1998

103. p4 posterior cingulum cusp State: 0 = absent State: 1 = visible but weak State: 2 = well developed Character citation: Hunt, Jr. 1998

107

APPENDIX F. Data matrix for specimens included in this study.

108

Nandinia binotata 1001100?2111?101101201120112?1001?101??321?1??02?11111221111000100000200?00011?0011??010?200???2?102110101122112110020?000?????1010111000000?000?1? ?00?100 Felis viverrina ?011010?1122?101000113000010??1?12121?300000??01?211101011111000012300012?010??00001?102?100???1?11010?0020000?3?00?203001??????2???1212000?0110001 20??000 Felis shansius 100&2?110?11???101100?1110?000?01?12?22???0?0???02?101&21000101100000?1210012?010??1??00?01&2??100???2?11100?0000000?3?00?2???00??????2???0&1211&2 00?0?1???2??0??0?1 Lynx rufus X9489 1121120?1112?10100012200?010?01112?20??0??00??01?1?11000102100010?2011012?010??00000???2?10&10?????11??0?0000???????????????????????????????10?1??0 0?2?????? Lynx canadensis 1020122?11?2?101100?2?100010?0111210??300000??01?2111010101110010121110131000??00001?0?1?10????2?11000?0020000?3?00?203000??????2???121211??0110?21 20??001 Lynx rufus 0100121?1022?101000122000010?01012111??00000??02?2111010101110000123000121010??00001?10??100???1?11010?0000000?3?00?203000??????2???1212001?0110?02 20??001 Acinonyx jubatus 121012??1112?101100?1100??00?01212111??00000??0??201101011110211011110012?0010?10011?101?200???2?11010?0020000?3????103000??????2????2?2110101?0022 20??001 Felis silvestris 1000121?1112?101000023000010?01112111?000000??02?101100010111000012200012?010??10001?101?100???1?11110?0000000?3?00?203001??????2???1212001?0110001 20??001 Felis temmincki 1?00121?1122?101001?120?0010?01012121?300000??02?2111010001100000133000121010??00001??02?000???2?11010?00?0000?3?00?210001??????2???0111001?011000? 20??001 Felis geoffroyi 1000020?1110?101001?22000010?110121?2?300000??0??2111020111110010?1210012?010??00011?102?200???3?11010?0020000?3?00?303001??????2???121200?00010002 21????? Felis nigripes 1??1121?1110?100001?22000010?01111101?000000??02?11210101021000101??110121010??00111?102?100?????11000?0020000?3?00?103000??????2???1212000211 100012???001 Felis attica PIK 3232 ???????????????????????????????????????????????????????????????????????????????????????????????????????????00??3?????????????????????2?2?0&1?? ?????????????? Felis bengalensis 1?0102??1022?101000023101010?11112121?310000??02?1111010101110000123000121010??0000??102?100???1?11010??020000?3?00?213000??????2???1212001?00 1000220??001 Felis colocolo 1100021?1122?101000122000010?11112121?300000??0??211101010110011011?10012?010??10101?111?100???1?11010?0021000?3?00?203001??????2???010100?200 1000220??001 Felis pardalis 1?01121?1022?101001??1?00010?01012121???0000??02?2111010101110000?2300012?010??10101?002?100???2?11010?0020000?3?0??203001??????2???1212001?01 1010220??001 Felis serval 1201121?1022?101000022100?10?11012121??00000??01?2111010101110000?2300012?010??10001?102?100???1?11010?00?0000?3?00?213001??????2???1112001201 1000?20??001 Neofelis nebulosa 1100122?1122?10100110?100101??1?121????00000??02?211100120111?00012?0?01?1000??00111?012?200???2?11010?0000000?3?02??03001??????2???121201?211 1000?20??021

109

Panthera tigris 100?102?1122?10110100??0?111?11212101??00000??01?211101?001102100?1310012?010??10011?10??000???2?11110?0000000?3?00???3000??????2???121200??01 1000220??001 Panthera leo 100012??1022?10110110??00001?01212001??00000??02?2111010100112100?2310012?010??11101?101??00???2?11110??020000?3?00?203000??????2???121200??01 1002220??001 Panthera pardus 10001?2?1122?101101?0??00001?112121???300000??12?211101?00111210012?100121010??00010?102?100???2111110?0000000?3?00?203000??????2???1212000?01 100222???001 Panthera onca 000?1?2?1122?10110100?10?001?01212111??00000??01?201100?00010200011210012?010??10111?101?000???2?11010?0000000?3?00?203000??????2???121200??01 1000220??001 Panthera uncia 1000122?1122?101100211100110?01012112??00000??02?21110?0000101010?1?10012?010??10111??12?200???2?110?0??0?0000?3?00?20?000??????2???121200??01 1002?20??001 Felis concolor 2200111?1122?10110100?20??10?01012022??00000??02?21110??100101000?23100121010??00?01??02?100???2?11000?0000000?3??????3000??????2???12?2001221 ?00??20??001 Felis caracal 22011?1?1122?10100011?100010?01?12121?300000??0??211101000110?1101121?01?1010??00001?101?100???2?11000?0020000?3?00?203000??????2???121200?211 1000220??001 Felis yagouaroundi 002?121?1110?101000122100010?1111?120?300000???2?1011000101111000??200012?010??10101?021?100???2?11110?0020000?3?00?113001??????2???122211?201 1010120??001 Arctictis binturong 1?21020?0111?102101221?21002?1000?011?001100??02?2011211000102101?230101??101???0111?010?001???0?01010?20?132112110030?000?????001021100110000 1211010?1120 Paguma larvata 1100021??11??111110???010002?10011111??1?100??01?102121100?10?101??212011?1011?10111?010?001???2?0021102001520122101???000?????00?031202000101 ?11???200100 Paradoxurus hermaphroditus 1200100??11??111101201010002?10001001?001100??0??10212?100?102001?1202011?1011??0110?110?001??????0?11020014301111012?2000?????001011101000101 ?01221210100 Poiana richardsoni 1121021?0101?00100121?200002?10111100?101101??12?0???201001001010?0101013?0011?00110?030?011???1?11310?11111201??10021?000?????20200222211???0 ?0?2?12??100 Prionodon pardicolor 102?020?1111?1010012?0200102?10111100??00101??0??002021011210?010?0101013?0011?10010?140?001???0?10210?101123012210021?000?????0010011121?00?1 1012111??120 Prionodon linsang 102?220?2111?10?011210100102?10111000??1?100??0??0020?001020?2000?1201013?0011?1001??140?001?????11210?10012?012?10021?000??????0203111211?0?1 ?01?111??1?0 Genetta sp. 11201?2?0112?01?011210210003?10101101?101101??02?002020100100?01010101013?001??00110?130?011???1?1031102?1023011110021?000?????10101111100??00 0011?1201100 Genetta tigrina 1021001?1121?101011101101002?10111101?00110???02?002020100110?0101110101310011?00?1??110?011?????1031111110?20111100212200??????0?011212110200 001?21211120 Genetta maculata 0221001?0121?10101120???0002?10101101??01101??01?00202010010010101010101310011?00110?1?0?011?????103111?111??012110121?000?????101031112110200 1012?12111?0

110

Genetta servalina 11201?2?0112?01?011210210003?10101101?101101??02?002020100100?01010101013?001??00110?130?011???1?1031102?1023011110021?000?????10101111100??00 0011?1201100 Viverricula indica 102002??0021?11201??0101?002?1000?101?10?100??02?00202113010000001?201013?010??0011????0?001???1?0031101010330111101215000?????001001112000?00 011??031?100 Viverra zibetha expectata 18725?2011????12???02&310?20??0???2?10101?0?????101??02?012020100000?000?1202013?0112???01??1?0??001?????03111101143?1111?1???????????0?001101 2?1?2?1????2022???? Arctogalidia 1100021?0?01?0?11211?202?002?1000?11??00?100??02??011201001102000?021101??1012?00110?0?0?001??????0?1101021330111101211000?????00200010100?200 001211110100 Cynogale bennetti 2200020?0122?10110020020?103?11001002?101100??00?1021201000100000??212013?0001?10010?010?001???1?102110102131011?100202000?????00???0212110001 ?01210210130 Chrotogale owstoni 0100101?0121?001101210020003?10101011?20?100??00?002120110010?101?1212013?0011?10111?1???001???0?00211012?033111010030?000?????00203222111??01 ?0111?22?13? Civettictis civetta 12002?1?0121?101101200?01003?10200121??01101??01?1021?10000102100?120?013?0011?01110?1?0??01???1?00?11010?0430111101211000?????001?00112?10?01 0112202201?0 Galidia elegans 1000121?201??101111101?01011?110100?1?21?110??01?1020000102012101100110101000???0111?1?0?011?????10211100003301?100021?000??????11111102110200 001021210110 Galidictis fasciata 1001221?201??011011101001001?11?010?0??1??10??01?10210001020121?1?0010010?000??10111?110?001???2?10311100?0330110001200000?????011001102110200 001021210130 Salanoia concolor 0000121?0000?101110221?01011?11001000??11?11??01?0020000102012111?0010110?0?0??10?10?140?011???1?10311021?123011100121?200?????11100110211?200 001220210130 Cryptoprocta ferox 11000?1?2020?101001211000101?1101?101?01??00??02?1011200001111000???00011?0111?00010?040?001???2?11010?210010013?00??15000?????31???12??120100 ?010?10??00? Cryptoprocta ferox spelaea 9948 BM ?00???????21?????????????????1?0????1&2?2&412?00??02?2011???????01000?011?0????111????1??????001?????????????????????????????????????????????? ???????????????? Cryptopropta ferox spelaea 9996 9949 BM ???????????????????????????????????????????????????????????????????????????????????????????????????????????10?10?0??212000?????31???122200???? ?????????001 Eupleres goudottii 0020102?0001?0011?102020?003?10101012??02100??0??0020000102011010?0?11010?0012?10111?0???011?????00311010?023010001021?200?????110002212110200 10?10122?111 Fossa fossana spelaea BM 9391 ?2012?0&110110?011101?200???03?10001?02?101&2101??01?012020010100201010&1111011&2?0011?0??1????0?0???????00??1?1011?????????????????????????? ????????1?0???0?????? Fossa fossana 0020020??011?1111111?1200003?10101010?102101??01?002000010101201010111010?0?11?00111?0?0?001???0?0031111011?30110100210200?????100001112000?01 ?01200221100 Helogale parvula 11010?2?0011?1120102??021010?11000000??0211???00?1021000102011111?001??10?0012?10111?0?0?011???1?0031110210?33011100210000?????101000102010200 001000220130

111

Suricata suricatta 0000101??102?102101023021101?1100?001??0?111??00?2011201001100111?0011110?0012?10110?1?0?001???1?00?1100110?30111000210000?????101001112000201 ?012?1220110 Ichneumia albicauda 0000002??020?102100213101002?10000011??02111??01?102100010?01111110110010?0112??0111?010?001???1?0031111210230100101215000?????000101112000201 001?00220120 Herpestes sanguineus 021002?00?1?1020111032?1011?11000000??0?111??02?0020?0010201?0111001111010012?00111?010?011???1?10311121?0330111001215000?????1010311121102000 01000221110 Herpestes javanicus 1001002?1011?1120101?3121012?10000001?102111??01?102100000?011?11?0010?10?0012?00111?1???011???1?1031101211?3011111121?000?????101001212000200 001000210120 Herpestes auropunctatus 0121011?1021?102010?03?21011?11000001?102111??02?002000010100?1111001111010012?00111?0?0?011???0??031111210?30111101215000?????1000?1202000200 00210022?110 Herpestes composite ???????????????????????????????????????????????????????????????????????????????????????????????????????????2&31?11?1?12111&2???????01???110&1 111???????????????? Herpestes ichneumon 020022?1021?112011?1?2?111??112000?1??0?111??0??1021000101012111?011001??0112?00111?1?0?001?????1031101101230111100215000?????1010011120002000 01000211120 Galerella pulverulenta 1020102??011?112011203121111?11010000??02111??02?00202001010100111001111010012?00111?0?0?011???2?1031112210330110001200000??????01??1112110200 001000210110 Kichechia composite ???0????????1??011?0?0????1?0??00?????????????????????????????????????????????????????????????00001111001021?100??1???????????0&1?0???10111??? ????????????? Mungos mungo ?101002?20?1?112111?02001012?1000?0?1??0?111??01?1021000101012101?0010010?0112?10111?0?0?001???1?0031110220230100001210000?????10100000200?200 001000220120 Bdeogale crassicauda ?0000?2?002??1021002020?1001?1100?011??02?10??02?1021000002000111?0111010?0012?00111?1?0?011???2?003110?200?30101100215200?????011010112010??0 ?01100220110 Cynictis penicillata 0001002?2011?1020101?3?21011?111010?2??01111??01?0020000102010111?0011110?0012?10110?0?0?001???1?1021112210?20100101215000??????01001212110?00 001?00210130 Atilax paludinosus 1?2000??202??112111?020?1011?110011?1?10??11??02?10210001010021111011111?10112?00111?110?001?????10311121?033011?1112??000?????001111112110200 001020220110 Proailurus lemanensis MNHN ????????????????????????????????????????????????????????????????????????????1??????????????????????????????11012?002325?00?????31???12220&11?? ??????????3100 Proailurus lemanensis MNHN SG 12331 ????????????????????????????????????????????????????????????????????????????1??????????????????????????????11010?102&32?5????????3&41???011200 ?????????????0?? Proailurus lemanensis AMNH 10931 ?20?0?0&1?0022?100001?01?2??01?0&11011021?0321?0?????2?211201111000000????00?0??11????11???0?000?????11110?202111?12212221500??????312030&122 211&2???????????????? Proailurus lemanensis MNHN SG 3509 ?201??????20?1??0?????0????0?0111???1&2?30?1?0?????2?212?211?1????????????????11??0111?????000??????1310?2000??????1????5???????????????????0? ?????0??1?????

112

Proailurus lemanensis ????????????????????????????????????????????????????????????????????????????1?????????????????????0010?1010??????1????5??????????????????????1 ?1?2??1?300? Metailurus sp. 131854 ?0011???1122?000100?11?0??01&2?01211?02?400??0??12?2101011&2100100000?0310012?010??0??????12???????2?01300???2010012100?203000?????2&31003011 200??????0???0??030 Metailurus sp. AMNH Ch 45 L399 ????2????1???111100?110???00?01?12?2???30?00??02?101&21???????00000?2310012?0?0??0???1???1???????1?11210?0000?????????????????????2??????????? ?1???0?20????? Metailurus sp. 26599 AMNH ????????????????????????????????????????????????????????????????????????????0??????????????????????????????????3?0031???????????2???121200&1?? ???????????0?? Metailurus parvulus M3895 0000102?1122?10110012110?010?01012?22?301?00??01?2111112101102010?3010012?010??0??11?102?001???2?11000?0000?????????????????????2?????????1201 ???2220????? Metailurus major SMNH M3841 M3842 ????122?11???101100011?0???0?01?12?0?????0????01?2??????????00010???1?0?????0??00010?1?1?000???2?11000?0000000?3?00?2?2001??????2???021200?2?1 ???2220????1 Metailurus major 010??21?1122?101100?2110??10?01012?12?300??0??01?2011000001100010?1010012?010??0??11?1?2?001??????1000?000000??3?00?213000??????2???122201?1&2 01???2220??121 Pseudaelurus marshi Rothwell ??????????2????????????????1??1???????????????????????????1?1???????????????1????????????????????01010?0?0010013?00??2300??????31???021111???? ?????????001 Pseudaelurus lorteti LGR 1380 1381 MNHN ??????????????????????????????????????????????????????????1?????????????????1??????1??????????????1010???0010?13?0?????????????31????1?1&2?0&1 ???????0??0&1????? Pseudaelurus quadridentatus MNHN ????????????????1?????????????????????????????????????????1?????????????????1??????0&1??????????????1210???0011?13?0?2&3??3????????31????21200 ???????0??1??10? Pseudaelurus intrepidus AMNH 18271 plus ??????????????????????????????????????????????????????????1?????????????????1????????????????????????????0010?13?02?20300??????31???122200???? ?????????00&11 Pseudaelurus (composite) ????????????????????????????????????????????????????????????????????????????1??????????????????????????????10010?00?201100?????2&41&2???011111 ???????????????? Pseudaelurus transitorius MNHN LGR 1373 ??????????????????????????????????????????????????????????1?????????????????1??????????????????????????????10?13?0?????????????31???0&11&211& 200?????????????00? Pseudaelurus sp. AMNH VP 61910 ??????0?1&2????10010?1???0???0??????????????????????????????1?????????????????1?????11?1???000???1??1??0?000010?13?02?203?0??????3&41???11&2? 1&2?1?0?????22????01&21 Pseudaelurus validus AMNH 61803 ?20?????1?22?100100?0&10?0??01?01201?21?31???????2?2?11110&21011?0000?1310012?0?1??1??10?1?2?20????2?11110??00010&1010?00?212&3100?????31&2?? ?011211?1?????222???121 Felis atrox LACM 579 100?121?1122?10210020120?101?01202001?2&400?00??01?210101000010100011111012?010??1??11?121?010???1?11??0?0000???????????????????????????????01 11?00222?????? Felis atrox LACM 2900 19 1000121?1122?10110020110?101?01202001?400000??01?21?1010001101000102100121010??00011?121?010???1?11??0?0000????????????????????????????????101 ?00222??????

113

Panthera atrox bebbi type ?12???????12????111?11????01?0110&110?10100?0???01&2?110100?000101001?101?012?010??10122?221?0000102011??0?0???000?0?00?20100???????2???111101 ???????????????? Felis daggetti ???????????????????????????????????????????????????????????????????????????????????????????????????????????000?0?00?201000??????2???111111???? ???????????? Felis hawveri ???????????????????????????????????????????????????????????????????????????????????????????????????????????00??0?0???010????????2???111111???? ???????????? Felis attica AMNH 01001???1122?1010&10011100??10?01012020????000??01?102100000110?00012200001201???1???1???1???????1&2?11100?0000000?3?00?2?3&400???????2???011 1&200???1?0?0?20??001 Felis concolor X8627 210?111?1122?101100001&2?0?010?01212?21???0000??01?2111010101100000112100121010??10111?122?200???2?11????0000????????????????????????????????? ?????????????? Felis bituminosa 2100010?1122?10100000120?110?01012?211000?00??01?21110&110110101000?01100121010??00011?122?200???2?11??0?000000??3?00?313000??????2???22220010 ?1??0222???021 Felis concolor PM (composite) ???????????????????????????????????????????????????????????????????????????????????????????????????????????00??3?00?304200??????2???2222001121 ???022?????? Herpestides antiquus MNHN (composite) 1201122?1022?01110110102?012?10210?01&2?10?1?0?????2?211??0???0???0?????0?????11?00110?110?001?????0031111&20003301??100212200?????01???121200 &1020&11?0?2?121?100 Herpestides antiquus MNHN GQ13186 ?20?1?????20&1?????????1???????1?01???2?412100??02?1121100101000110?001101?1?112????1??????001??????????????????????????????????????????????0? ????0????????? Herpestides antiquus SG319 ????????????????????????????????????????????????????????????????0?????0????????????????????????????????????33?1??1?02?1&2000?????01???12120002 0&11?0?2?121?100 Herpestides antiquus MNHN SG3144 ????????????????????????????????????????????????????????????????0?????0????????????????????????????????????33?12???021?0&2???????01???111201?? ???????????11? Herpestides antiquus MNHN SG3116 ?2010????020???111???????1?2??001???1&2???0??????????????01?11????0?0???0?????11&2?0??10???0?????????0031102010&133?11&2?1?0?120???????0????0 &11&212000??0???0?1211100 Machairodontinae (genus indet.) 120?120?2?22??01100?010??101?01212022?0&402?00??01?2101000001&210200??0310012?000??10020???0?301???2?01??0?0001000?3?01?203000??????2???010000 ?????0?2?2???020 Smilodon composite ?1????????12???????????????????0????1?10??????0&2??100100?1021??00??0&11??0121?1???????????????????????????????00??0?0??100?1???????2????1?1?1 ???????????????? Smilodon PM (composite) 1200122?10&122?10110000110?100?01212111?3&40&10&1000??01?2001202002101000?2&32100121000??00011?102?001???2?00&1110?0020000?3?01?103011??????2 ????2?2?2?1&20&10?202220&10?01&20 Dinofelis sp. 50446 AMNH ?00?1&2???1?22?101100011?0??00?0&111121?2?????00??02?21010?0110100000?0&1310012?01???????????????????1&2?1???0?000&20???????????????????????? ??????????1???????????? Dinofelis sp. 50445 AMNH ?10?1&2???1122?10110000100??00?01212121??????0??11?2?010??????0?000?????01????0??1??1??1?1?000???2?11100?0000????????????????????????????????? ?1????220?????

114

Homotherium ischyrus ???????????????????????????????????????????????????????????????????????????????????????????????????????????000?0?01?2???11??????2???010100???? ???????????? Homotherium nestianum AMNH 104641 220?111?1122?10111000&11?0??01?0&11212??1?0&310??0??01?2001??????100000????0012?010?????21?1?022?????1?01??0?0020000?3?01??0010???????2???020 102?210???22????001 Hyaena brunnea 111?120?0102?001000202001101?21203012?000100??01?2011202000102101?111011??110??00001?010?001???2?11010?20002?010?00?111000?????11?????12???000 ?010?01?0110 Hyaena hyena 2100110?1012?00111120?010102??021301??001100??01?20112020001011011011001?1010??00001?1?0?101???2?11?10?20002?013?00011?000?????11???12120000?0 1010201?01?? Adcrocuta eximia SMNH 30 ????1???20???1?11?00??12??01??1?1??1??????????????????????????001?2300013?0????0???1?1?0???????2?11300?1&20001001??10?2??????????30???111100?? ?0???0211????? Adcrocuta eximia BM M8966 11????????22????100?01????02?20013?11&2?00??0???01?2?112?00?110?101?12?0011?0????????1?1?0??????????1110?2000????????????????????????????????? ?????0?01????? Adcrocuta eximia SMNH 1????22?10???1????0???22??01??1?1?01???????????????????????????????????????????0???1?1?0???????2?11200?1&200010013?00?2&3???00?????2&30???021 &2200?1&20??1?0211?1??? Adcrocuta eximia BM 8970 ???????????????????????????????????????????????????????????????????????????????????????????????????????????20?13?10?201000?????10???0&121200?? ???????????100 Crocuta crocuta 2211110?2012?101101201200101??121301??000100??01?2011200000102101112100121010??00001?100?101???2?11000??00010013?00?101000?????31???011?001000 1010200??120 Proteles cristatus 0001000?1011?102100??2000002?00203010?001111??01?0020200001102111101101111010???1101?1101?0????1?00000000?100?02&3200320?000?????41???00000001 00101?1100?0?1 Plioviverrops orbignyi MNHN PIK 3016 ?1?0?????1&2??????0?0??0????2??0??????0?1021?1??12?1??12?1??21?0010?001?01???011?????1??????????????03111201033?1100?1???????????0&110010212?1 ??????00??22???? Plioviverrops orbignyi ?00????????2??01100??11????2??0?0???0??02?????02?202120?????00010?0???1?????1??1??1????0?01???????03111101033?1????12?5?00?????00???021201???? ?????023?100 Thalissictis robusta BM M8981 ??????????1&22?1?????1?1????02??00?3?2??????1???02?2?112??0?1?0?010?001?0????????????????????????1&2??031102?0???0?2?10?2?110?00????0????????? ??????????21?100 Thalissictis robusta BM M8983 100??20?112???0110110&1100??02?00?0&13?2??302??1??02?21201?00?0100000??210010?0?11????1????0?0&2????????031102000???????????????????????????? ????0?1?0???121???? Thalissictis robusta BM 8986 ???????????????????????????????????????????????????????????????????????????????????????????????????????????31?11&2110021??00?????10101021&2?02 ????????????010? Thalissictis robusta BM 8989 ???????????????????????????????????????????????????????????????????????????????????????????????????????????33?1??100????????????????0&121&2202 ????????????0??? Thalissictis robusta BM M29480 ???????????????????????????????????????????????????????????????????????????????????????????????????????????12?12?1?02022???????1&21???12?2???? ??????????0110

115

Thalassictis certa BM M5555 ???????????????????????????????????????????????????????????????????????????????????????????????????????????33?1??1?1&22??????????1????000&1100 ???????????????? Palinhyaena reperta AMNH Ch 42 L344 ?1012??11022?1?111120100??03?20213??11??1101??01?2?11?000011?1101?2300012?01???????1?1?????????2?0?31100100320121100?0??00?????0&10000122200?? ?0???0212?0&1?1? Palinhyaena reperta SMNH M37 M38 ???0220??0???0011??20&1101?0?2??0?13?10??001?1??01?0?2??01001100000?3200012?01???0??1??1?0?0????????02010200022&3?12?100112000?????10???011100 0??0???020311120 Palinhyaena reperta AMNH CH L46 210?210?1012?01110020120?002?1&20013?11?????0???01?1120200300110110?3300011?0?10?1?011?1?0?001???1&2?1121101000212&3121101112200?????20100011 200?1?0???02?200110 Palinhyaena reperta SMNH M3854 ???????????????????????????????????????????????????????????????????????????????????????????????????????????230222100111200?????20111011100???? ????????0120 Hyaenotherium wongii AMNH (composite) 210111&20&111022?0011012110&10?003?00013?11?00?101??01?1021200001100101?3300012?010??0??01?120?001???2?0021101000320121&2100210&1200?????0000 0122200??10???0?1112101 Hyaenotherium wongii M28 M29 M30 11&20?100?2112?00110121100?002&3?10203?12????1?1??0??????????????1????????01???1???0??1??1?0?001?????0020101&20002301101?011320??????00010222 201?0?1????21210130 Ictitherium viverrinum MNHN PIK 3004 ?00??????????001101211???1?2?20?03????00?1?1??11?2011??01?110?010?0?1?01?????????????0????????????0311011002&33?11&2?1002??????????00???011202 ??????????21?10? Ictitherium viverrinum SMNH M3206 ?11???1?102???11?112???0???2??0103??2?1021?1??01?11210010011??00??0&1110011?0112????1????0?001??????02&30101&2000???????????????????????????? ??????0?????121???? Lepthyaena sivalensis BM K13 273 ???????????????????????????????????????????????????????????????????????????????????????????????????????????2&32&3?1????0???????????0&1?????1? 2?0????????????1??? Tungurictis spocki AMNH 26600 ?20??????122?111011?012???03?00000?2???021?1??0??2121????????00?0?????01???11??1???1?120???????1&2??031111010????????????????????????????????? ?1???2?121???? Tungurictis X. Wang undescribed ?????????????0000????1?????2??0?0??????????????????????????????????????????????????????????????1?002&31111&201023?110??0?020&2???????0?102??? ????????????1211120 Protictitherium gaillardi ???????????????????????????????????????????????????????????????????????????????????????????????????????????2&34?111??0???????????0&1?101?2?2?? ???????????????? Chasmaporthetes lunensis AMNH 99787 ???????????????????????????????????????????????????????????????????????????????????????????????2?01210?100?10013?00?1??000?????2&30???122201?? ???????????100 Lycyaena dubia M3856 2???????1&21???0021012?100?001??1?0?11?????????????????????????????????????????????????100???????1?11200?1000????????????????????????????????0 ?0???0211????? Lycyaena chaeretis MNHN PIK 3383 3384 ???????????????????????????????????????????????????????????????????????????????????????????????????????????22?1??1001??????????11???122200???? ???????????? Lycyaena chaeretis BM 8979 ???????????????????????????????????????????????????????????????????????????????????????????????????????????11013?10010?200??????0???010220???? ?????????100

116

Lycyaena chaeretis BM 8978 ?20???0???12?0?110????????02?20213?21&2???2?????0??2?212?100010??????????????????????????0???????2??1210?2000????????????????????????????????2 00???0?12?1100 Lycyaena chaeretis AMNH CH 11 L12 ?11?0?1?1022?10110120??0??02?20013?11?????0???01??????0&10&21?110?101?3310010&1?010?????11?120?00????1&2?11210?10002&31?13?1001?1?00?????1&20 ???022200?200???0?12?1100 Lycyaena macrostoma BM 13179 ???????????????????????????????????????????????????????????????????????????????????????????????????????????21&2?12???0???????????1&2????022201 ???????????????? Stenogale coupant MNHN (composite) ???????????????????????????????????????????????????????????????????????????????????????????????????????????12?121100212000?????1&2&310&101111 211????????????1100 Stenogale brevidens ???????????????????????????????????????????????????????????????????????????????????????????????????????????2&31?11?1?1???0???????10???001111?? ?????????????? Stenogale gracilis (composite) ???????????????????????????????????????????????????????????????????????????????????????????????????????????11?1110002?1????????1&2?1&210111111 ???????????????? Stenogale julieni MNHN SG 3210 1201021???20????01??01???????10011??1&2?30?1???????1?211?1&31?10????????????????10&1????11?0?0?0?1??????00&21??1&2011???????????????????????? ???????020????00&111????? Stenogale intermedium MNHN Qu 9082 9083 ????????????????????????????????????????????????????????????????????????????????????????????????????????????1?13?1?0???????????3????121211???? ????????011? Dinofelis piveateau 50453 ????1&2????????1??1??0?100??01??1??21????0???0??01?2001???????00000??3?0012?01???????????????????2?0???0?000&20??????????????????????????????? ???1???????????? Dinofelis BC 120 SMNH ?10???2??120?101100??100??01?01202022?002??0??01?110100010210?000?????013?0????00011??0??000???1?01??0?0020000?3?01?203011??????2???022211?2?1 ?0???????110&1 Dinofelis CB 20 ????1?2??120?10110???1????01??121112??002??0??01?210120010210?000?????013?0????????1???????????2?01000?0020000?3?01?1030&111??????????021200&1 ???1?0????0??110 Nimravides (composite) ?????????????001100?0?0????0?11?12????????????02?1?110?01?11???????????????????0????????????????011000?000010?12?0??200?0??????3?????1?1?1???? ???????????? Nimravides sp. AMNH HIG 315 2306 ???????????????????????????????????????????????????????????????????????????????????????????????????????????010?3?012&3203000??????2???0&122201 ????????????3021 Nimravides sp. AMNH HIG 227 1790 ?2??2????122?10110?2?1?0??01??1202101?????????12?2101???????0?000?????012?0????????????????????2?11300?010&20????????????????????????????????? ?1?1????0????? Haplogale media BM M9640 ???????????????????????????????????????????????????????????????????????????????????????????????????????????12?12?1?2???????????31???222200???? ?????????1?? Haplogale media (composite) ???????????????????????????????????????????????????????????????????????????????????????????????????????????01??1?1?02???????????0????1?1?1???? ???????????? Haplogale media MNHN 9422 ???????????????????????????????????????????????????????????????????????????????????????????????????????????11?12?1?0&2212200?????3????222211?? ???????????100

117

Haplogale media MNHN 9473 ???????????????????????????????????????????????????????????????????????????????????????????????????????????01&2??2?1?0&2212?????????1???1&222 212?????????????100 Haplogale media MNHN 9435 0???100?1????0??0?12??1??0?3??0?1??????????????????????????????????????????????0???0?112???????0??0211??00101??21100&2?????????????111?12211?0 ?????0212&3?2??? Stenoplesictis minor (composite) ?????????????????????????????????????????1?????????????????????????????????????????????????????????????????1&21?11???1???????????2?????1?1?1?? ?????????????? Stenoplesictis BM 4494 ?????????????????????????????????????????1????????????????????????????????????????????????????????0210?200??2?1221?0????????????????222211???? ??????1??1?? Stenoplesictis cayluxi MNHN Qu 9372 9393 ?????????????????????????????????????????1?????????????????????????????????????????????????????????????????22?12?1?02???00??????????121&2200&1 ????????????110? Stenoplesictis cayluxi MNHN (composite) ?????????????????????????????????????????1?????????????????????????????????????????????????????????????????22?1211?02???00?????0&11000121&2200 ????????????1??? Stenoplesictis crocheti MNHN (composite) ?????????????????????????????????????????1?????????????????????????????????????????????????????????????????21&2?1210002???00?????31102121200?? ??????????211? Stenoplesictis cayluxi MNHN (composite 2) ?????????????????????????????????????????1?????????????????????????????????????????????????????????????????22?11&211?02??000?????11000120220?? ??????????110? Stenoplesictis cayluxi MNHN (composite 3) ?????????????????????????????????????????1?????????????????????????????????????????????????????????????????21&2?12110021?0&200?????0&11000121 20&10????????????1110 Paleoprionodon lamandini MNHN 9370 002112??1111?1??0?110010??02?00011?02?2&33?1?1??02?11211?01?2100010?0?0&1?00????11?1???0?11&42?0?????????????2001???????????????????????????? ???00?0???21&21?????? Paleoprionodon lemandini MNHN 2003 1 Qu 9322 (composite) ???????????????????????????????????????????????????????????????????????????????????????????????????????????01??21100212000??????1100111100???? ?????????110 Paleoprionodon lemandini MNHN Qu 9366 ???????????????????????????????????????????????????????????????????????????????????????????????????????????01??21100212200??????11001&221200?? ??????????2100 Paleoprionodon M6007 BM ???????????????????????????????????????????????????????????????????????????????????????????????????????????13?1211?021?0???????31102122200???? ?????????10? Paleoprionodon mutablis (composite) ???????????????????????????????????????????????????????????????????????????????????????????????????????????21?11?1?12??????????10???111111???? ???????????? Paleoprionodon lemanensis Qu 9361 MNHN ???????????????????????????????????????????????????????????????????????????????????????????????????????????01??211?02???????????1101221200???? ????????0???

118

APPENDIX G. Museum specimens examined for morphological phylogenetic analysis. Information shown includes museum, specimen number and locality (where known). BMNH: British Museum of Natural History; AMNH: American Museum of Natural History. F: AM: Frick Collection, the American Museum of Natural History; SMNH: Swedish Natural History Museum; MNMH: National Museum of Natural History; NMNH: National Museum of Natural History (Washington, D.C.) PM: Page Museum: LACM: Los Angeles County Museum; HMCZ- VP: Harvard Museum of Comparative Zoology, Vertebrate Paleontology; UCMP: Univ. California Museum of Paleontology (Berkeley); Localities: SG: St. Gerand, France. Qu: Quercy deposits, France. LG:La Grive, France. PIK: Pikermi, Greece. CH: China.

119

Feloidea: Herpestides antiquus: GQ 13186, SG3116, SG3121, SG3122, SG3144, SG3139, SG3137. (St. Gerand); Palaeoprionodon BMNH M1644; Palaeoprionodon mutablis BMNH M6006, 6007, UCMP 63096, 63097, 63098 (Quercy), P. lemanensis HMCZ-VP 8917, MNMH QU 9348, QU 9370 (Quercy); Stenogale BMNH M9635, MNMH QU 2003-1 (Quercy); Stenogale brevidens UCMP 118329 (St. Gerand), S. coupant MNMH QU 9115, QU 9392, QU 9102, QU 9104, QU 9379, QU 9075, QU 9058, QU 9356, QU 9114, QU 9113 (Quercy); S. gracilis UCMP 93076, 63079 (Quercy); S. intermedia BMNH M4005, MNMH QU 9083, QU 9082 (Quercy); S. julieni SG 3210 (St. Gerand); Stenoplesictis BMNH M1723, M9639 (Phosphorites, Caylux), HMCZ-VP 8932 (Phosphorites); MNMH QU 9376, QU 9359, QU 9360 (Quercy), BMNH M9630, M2361b, M1366 (Quercy); S. cayluxi BMNH M1366, HMCZ-VP 8934, MNMH QU 9111, QU 9372, QU 9393, QU 9473, QU 9388, QU 9544, QU 9099, QU 9387, QU 9096, QU 9399, QU 9098, QU 9419, QU 9378, QU 9112, QU 9382, QU 9381, QU 9398, QU 9055, QU 9411, QU 9060, QU 9355, QU 9375, QU 9375, QU 9377, QU 9363, QU 9053, QU 9702, QU 9415, QU 9103, QU 9362, QU 9101, QU 9106, QU 9380, QU 9407, QU 9091, QU 9051 (Quercy); S. crocheti QU 9373, QU 9401, QU 9402, QU 9400, QU 9095; S. minor BMNH M4494, UCMP 93070, 63092 (Quercy) Felidae: Dinofelis SMNH BC 120, AMNH 50445, 50446, 50453; Felis atrox LACM HC 479, UCMP 14001; F. attica: AMNH CH 45-L417, CH 46-L419, CH 83-L692, 97025, 22331, CH 42-L361, F:AM 95295, CH 76-B876, CH 51-L431, 20558, CH 92-L706; F. bituminosa: PM X8628, X9474; F. concolor: PM X8627, X9626, X9073, X8629, X9566, X9469, X9464, X9470; F. daggetti: UCMP 21572, F. hawveri UCMP 10636; F. shansius AMNH 62-B749, 62-B754, 62-B756, 63-B766A, 63-B766B, 63-B766C, 63- B766D, 63-B766E, 63-B766F, 63-B766G, 63-B766H, 63-B766K, 63-B766N, 63-B766O, 69-B822, 78- B890, 87-8945, 96-B1042, F:AM 101219, F:AM 117364, F:AM 117365, F:AM 117366; Haplogale media MNMH QU 9432, QU 9430, QU 9428, QU 9418, QU 9427, QU 9371, QU 9422 (type), QU 9435 (Quercy); Homotherium nestianum AMNH 104641; Lynx rufus PM X9489, X9491, X9490, X92562, Pit 16; Metailurus UCMP 77532, AMNH 226599, IVPP5679; CH45-L399, CH47-L548; Nimravides UCP 34614, 34512, AMNH HIG227-1790, HIG315-2306; Proailurus lemanensis AMNH 101931, UCMP 118323, BMNH M1645, M1381, MNMH QU 9437, QU 9439, QU 9440 (Quercy), SG 3509, SG 12331 (St. Gerand); P. media HMCZ-VP 8936, 8937, UCMP 63099, 63080, BMNH M9636 (Quercy); Pseudaelurus UCMP 67917, AMNH 61910; P. africanus UCMP 53886, P. edwardsii BMNH M7489, HMCZ-VP 8939, P. intermedius BMNH M2375, P. intrepidus AMNH 18271, 61804, 61841, 61805, 61806; P. lorteti LGR 1381, 1380 (La Grive); P. marshi AMNH 62135, 61808, 61807, 22398, 61846, 22401, 22404; P. martini HMCZ-VP 8603; P. pedionomus UCMP 29186, P. quadridentatus LGR (La Grive); P. thinobates UCMP 34513; P. transitorius LGR 1382, 1382 (La Grive); P. validus AMNH 25476, 61832, 61829, 61833, 61803; Smilodon UCMP 38338, S. californicus UCMP 10210 Herpestidae: Herpestes UCMP 63253, 63254, 63256, 63252, 63258, 63257 (St. Gerand), 43988 (Moutigu- le-Bliu), 89283 (Transvaal); Kichechia zamanae UCMP 77529, 77536, 77530, 77527 (Rusinga Island), AMNH 56443, 56444, 56445, 56446 Hyaenidae: Chasmaporthetes kani AMNH 99787, 49-B536, 99788, 99786, 99785, 99789, 226369, 99784; Adcrocuta eximia BMNH M49673, M8966 (Pikermi); Hyaenotherium wongii PIK 3378 (Pikermi); Ictitherium hipparionoum PIK 3042a, PIK 3012 (Pikermi); I. orbignyi PIK 022a, 3022b (Pikermi); I. robustum BMNH M29480, M8983, M8982, M8981, M8987, M8988, M8986, M8989, PIK 3004, a, 3004b, M13180 (Pikermi); I.sivalensis BMNH K13/273 (Siwaliks); I. viverrinum PIK 3005, 3006, 3007, 3011, 3009, 3377, 3030 (Pikermi); Lycyaena BMNH K13/183 (Siwaliks); Lycyaena chaeretis AMNH CH 56- L560, CH 26-B47, CH 11-L120, CH 69-B848, CH 38-B296, CH 45-L400, CH 51-L434, CH 52-L495, CH 56-L470 (Hipparion beds) BMNH M8978, M8979, M8979A, PIK 3383, PIK 3384 (Pikermi); L. dubia SMNH M3856; L. macrostoma BMNH M15703, M13179 (Siwaliks); Palhyaena reperta AMNH CH 51- L443, L146, F:AM 129667, CH L290; Plioviverrops orbignyi PIK 3016 (Pikermi); Progenetta certa BMNH M5555; P.proava BMNH M12793 (Siwaliks); Proctitherium gaillardi (La Grive); Thalassictis hyaenoides AMNH CH 42-L344, CH 20-L90, CH 90-L753, CH 17-L64; T. wangii AMNH 20554, 20555, 20586, 22878, 23032, CH B-L5, CH 14-L17,, CH 51-L464, CH 57, CH L562, CH 52-L569, F:AM 12966, F:AM L788, F:AM 95-L789, F:AM L789; Tungurictis spocki AMNH 26600 Viverridae: BMNH, Viverra BMNH M3182 (Siwaliks); V. antiqua MNMH (St. Gerand); V. augustidens BMNH M4497; V. zibetha AMNH 18725, 18930, 18726, 96435.

120

APPENDIX H. Extant museum specimens examined for morphological phylogenetic analysis. Information shown includes museum and specimen identification, sex of the specimens (where known), and collection locality. FM: Field Museum of Natural History; NMNH: National Museum of Natural History; HMCZ: Harvard Museum of Comparative Zoology; BMNH: British Museum of Natural History.

121

Felidae: Acinonyx jubatus: FM 39475 (u, Tanzania), 104980 (u, ), 34589 (m, Zimbabwe), 29633 (m, Kenya), 29535 (m, Kenya), 34669 (u, E. Africa), 89918 (u, Kenya); Felis caracal: NMNH 251869 (u, Tanzania), 182218 (u, Kenya), 113457, (u, Africa), 384162 (f, Botswana), 439818 (f, Benin), 47014 (f, Botswana), 51167 (m, unknown), 182310 (m, Kenya), 36853 (m, Botswana), 520686 (m, unknown). HMCZ 12611 (m, Palestine), 20048 (m, SW Africa), 12612 (m, Palestine), 31923 (m, Kenya), 35634 (m, Palestine), 20049 (m, SW Africa) FM 32945 (m, Ethiopia), 95922 (f, Zimbabwe, 105607 (f, Sudan), 135042 (u, Israel); F. badia: FM 8378 (u, Malaysia); F. bengalensis HMCZ 36034 (f, Sumatra), 24846 (m, China), 24843 (m, China), 36768 (m, Borneo). FM 31781 (u, Laos), 32552 (f, Vietnam), 75828 (m, India), 65458 (m, Philippines), 79819 (m, Philippines), 32549 (m, Vietnam), 367224 (m, Laos), 36003 (m, China), 39339 (m, China), 62888 (m, Philippines), 106014 (u, Vietnam), 74326 (f, Philippines), 79819 (m, Philippines), 99353 (m, Malasia), 8377 (u, Malaysia), 32517 (f, China), 39526 (f, China), 39504 (m, China), 62888 (m, Philippines); F. chaus FM 84696 (unknown, Egypt); F. colocolo: FM 43291 (m, Ecuador), 24370 (m, Chile), 68318 (f, Peru), 80994 (m, Peru), 52488 (m, Peru); F. concolor: FM 14902 (u, Florida), 51785 (u, Chile), 19136 (f, Mexico), 51472 (m, Arizona), 94318 (m, Brazil), 21713, (m, Arizona), 15532 (u, ), 30818 (u, Argentina), 83480 (u, Texas); F. geoffroyi: FM 28404 (m, Uraguay), 121290 (m, Zoo; S. America), 24360 (f, Argentina); F. guigna FM 24359 (u, Chile); F. manul FM 60734 (f, Zoo: Asia); F. margarita (FM 197299 (f, Egypt); F. marmorata FM 60358 (u, Thailand); F. nigripes NMNH 395519 (f, unknown), 452464 (f, S. Africa), 381275 (f, S. Africa), 452463 (f, S. Africa), 395838 (m, unknown), 397135 (m, Unknown), HMCZ 33976 (u, S.Africa); F. pardalis FM 88888 (f, Peru), 85503 (m, Peru), 91359 (m, Peru), 52427 (u, ZOO), 184555 (u, Guyana), 14175 (f, Mexico), 14177 (f, mexico), 22471 (m, Bolivia), 22468 (m, Bolivia); F. serval: 90022 (f, Zaire), 18862 (m, Kenya), 27165 (m, Ethiopia), 79415 (m, Sudan), 38192 (u, Botswana), 38191 (u, Botswana); F. silvestris FM 101986 (m, ZOO), 89993 (f, Namibia), 101877 (u, Egypt), 122383 (u, Egypt), 32941 (u, Ethiopia), 27248 (u, Ethiopia), 97861 (u, Iran), 112453 (u, Iran), 93874 (u, Sudan), 93306 (u, Sudan); F. temmincki FM 89919 (m, Vietnam) 31778 (u, Laos), 75826 (m, India), HMCZ 24847 (m, China), 5014 (u, Nepal); F. tigrina FM 84554 (m, Columbia); F. viverrina FM 123070 (f, Zoo), HMCZ 6329 (u, E. India), 30255 (u, Ceylon), 57926 (u, Pakistan), 5292 (u, India); F. weidii: FM 52437 (m, Zoo); F. yagouaroundi FM 69647 (m, Columbia), 88479 (m, Colombia), 22472 (f, Bolivia), 14896 (m, Mexico), 15984 (f, Guatemala), 20439 (m, Brazil); Lynx cancdensis NMNH A21452 (u, Alaska), 75638 (u, Alberta), A21451 (u, Alaska), A44472 (f, British Columbia), A44473 (m, British Columbia), 115984 (m, Manitoba), 110096 (m, Manitba), 81780 (m, Alberta), 75518 (m, Quebec) 99249 (m, Quebec), FM 43111 (u, Alaska), 1318835 (m, Alaska), 138828 (m, Alaska), 138836 (u, Alaska), 138837 (u, Alaska), 43111 (u, Alaska), 16022 (u, Alaska); Lynx rufus FM 7224 (m, Arizona), 81502 (m, California), 46954 (u, Texas), 10879 (f, mexico), 81500 (u, California), 51642 (m, Maine), 1643 (u, Maine), 81501 (u, Maine), 90579 (m, Idaho), 42761 (u, S. Dakota), 156710 (u, ), 123987 (f, Michigan), 123989 (u, Michigan), 123984 (f, Michigan), 129343 (f, Wisconsin), 123982 (f, Michigan), 129343 (f, Wisconsin), 160121 (m, Minnesota), 165363 (m, Minnesota), 129342 (m, Texas) 53041 (u, Texas); Neofelis nebulosa: NMNH 31925 (f, unknown), 575150 (f, unknown), 545387 (f, unknown), 399291 (f, unknown), 198705 (f, Indonesia), 172673 (f, unknown), 580716 (f, unknown), 396639 (m, unknown), A49974 (m, Indonesia), 172674 (m, unknown), 75831 (u, India), 75830 (u, India), 42583 (f, India); Panthera leo: FM 20756 (f, Kenya), 30778 (u, Sudan), 35741 (m, Africa), 75608 (u, Kenya), 20762 (m, Kenya); P. onca FM 70566 (m, Colombia, 17768 (u, Guyana), 21392 (u, Bolivia), 14179 (m, Mexico); P. pardus NMNH 259410 (u, Sichuan), 172662 (u, Shanxi), 184818 (u, Kenya), 162416 (u, Kenya), 239610 (f, Hunan), 161911 (f, Kenya), 173329 (f, India), 216606 (m, Angola), 216607 (m, Angola), 173328 (m, Kashmir), 34592 (m, Botswana), 31793 (u, Laos), 33469 (u, China), 35257 (u, Botswana), 83654 (u, Angola), 22364 (u, Kenya) HMCZ 7957, 32578, 29883, 31929, 39737, 46404, 44283; P. tigris FM 31797 (f, French Indo-China), 31152 (f, India), 31153 (m, India); P. uncia (FM 127297 (f, Zoo), 122235 (m, Zoo). Herpestidae: Atilax paludinosus NMNH 269708 (u, Liberia), 545835 (f, Sierra Leone), 220399 (f, Gabon), 541412 (f, unknown), 344022 (f, S.Africa), 182227 (f, Kenya), 367381 (m, Mozambique), 467666 (m, Ivory Coast), FM 39457 (m, Zaire), 27030 (f, Ethiopia, 93642 (f, Sudan), 98318 (f, Sudan), 9858 (m, Sudan), 17806 (m, Kenya), 27435 (m, Tanzania), 38308 (m, Tanzania); Bdeogale crassicauda NMNH 182281 (f, Kenya), 318111 (m, Kenya), 8383 (f, Malawi), 85974 (m, Kenya), 22615 (u, Tanganyika), 39416 (u, Tamgamoula), 85969 (f, Kenya); Chrotogale owstoni FM 41597 (m, Vietnam); obscurus FM 54410 (f, Ghana);); Cynictis penicillata FM 38366 (m, Botswana), 38376 (m, Botswana), 38379 (f, Botswana), 38373 (f, Botswana), 38378 (f, Botswana), 38374 (m, Botswana), 38382 (m, Botswana), 38372 (m, Botswana), 77513 (f, Namibia), AMNH 169020 (f, Botswana), 169014 (f,

122

Botswana), 81759 (f, S. Africa), 169011 (m, Botswana), 70415 (m, S. Africa), 81757 (m, S. Africa), 8384 (u, S. Africa). 38369 (f, S. Africa); Galerella pulverulenta NMNH 238039 (f, Cape Province), 424625 (f, Cape Province), 469872 (f, Cape Province), 469870 (f, Cape Province), 141519 (f, Cape Province), 469871 (f, Cape Province), 344021 (f, S. Africa), 469869 (m, Cape Province), 452469 (m, Cape Province), 238038 (m, Cape Province), 238037 (m, Cape Province), 469867 (m, Cape Province), 344877 (m, Cape Province), 469868 (m, Cape Province), 452468 (m, unknown); G. sanguinea NMNH 164155 (f, Kenya), 154157 (f, Kenya), 164158 (f, Kenya), 018936 (f, Kenya), 164151 (f, Kenya), 164586 (m, Kenya), 164153 (m, Kenya), 154156 (m, Kenya), 181535 (m, Kenya), 467671 (m, Ivory Coast), 467670 (m, Ivory Coast), 467669 (m, Ivory Coast) 161907 (m, Kenya), FM 85978 (m, Kenya), 38354 (m, Botswana), 77511 (m, Nambia), 73050 (f, Tanzania), 85981 (m, Tanzania), 85982 (m, Kenya), 54411 (f, Ghana), 27246 (m, Ethiopia), HMCZ 16112 (u, unknown), 8314 (u, unknown), 16114 (u, unknown), 46565 (u, unknown), 32332 (u, unknown), 16115 (u, unknown) 8289 (u, unknown), 16113 (u, unknown); Helogale parvula FM 35323 (f, Mozambique), AMNH 165750 (f, Nambia), 81487 (f, Angola), 205149 (f, Tanzania), 55726 (f, Zaire), 179305 (m, Tanzania), 89797 (u, Zambia), 83818 (u, Tanzania), 212994(u, ZOO), FM 83652 (u, Angola), 78113 (u, SW Africa), 35823 (u, Mozambique), 83651 (u, Angola), 83653 (u, Angola), 34201 (u, Kenya), 34200 (u, Tanzania); Herpestes auropunctatus FM 57598 (u, Nepal); 39350 (f, China), 75949 (m, Assam), 104402 (m, unknown), 83089 (f, Nepal), 75805 (m, India), 75801 (m, India), 57587 (u, Nepal), 29183 (u, India), 83092 (f, Pakistan), AMNH 232737 (f, Virgin Is.), 239688 (Trinidad and Tobago), 232980 (m, Virgin Is), 232980 (m, Virgin Is.), 239671 (m, Virgin Is.), 239645 (m, Virgin Is.), 218783 (m, Virgin Is.), 235659 (m, Virgin Is.), 232915 (m, Virgin Is.), 232930 (m, virgin Is.), 218836 (m, virgin Is.), HMCZ: 15933 (u, W. Indies), 57931 (u, Pakistan), 35317 (u, Santo Domingo), 24985 (u, China), 17067 (u, W. Indies), 15934 (u, W. Indies), 8089 (u, Grenada), 63333 (u, Hawaii), 5123 (u, India), 49026 (u, Dominican Republic); H. ichneumon FM 83658 (f, Angola), 83647 (m, Angola), 41339 (f, Tanzania), 27027 (m, Ethiopia), 100653 (f, Egypt), 93879 (f, Sudan), 90469 (f, Egypt), 56314 (f, Sudan), 104716 (u, Sudan); H. javanicus AMNH 101656 (f, unknown), 101657 (f, unknown), 101493 (f, unknown), 102031 (f, unknown), 101733 (f, unknown), 101658 (f, unknown), 101655 (m, unknown), 102221 (m, unknown), 102913 (m, unknown), 55556 (u, unknown), FM 46823 (f, Vietnam), 80390 (f, Vietnam), 31038 (m, Vietnam); Ichneumia albicauda FM 95033 (f, Sri Lanka), 38194 (m, Botswana), 73053 (f, Tanzania), 73052 (f, Tanzania), 85968 (f, Kenya), 17803 (f, Kenya), 17805 (f, Kenya), 85970 (m, Kenya), 26039 (m, Uganda), 6040 (m, Uganda); Mungos mungo FM 149364 (f, Cibitoke Prov.), 14935 (f, Cibitoke Prov.), 127824 (m, Tanzania), 27334 (m, Tanzania), 98355 (m, Sudan), 98356 (m, Sudan), 9837 (m, Sudan), 95272 (m, Zambia), 38357 (m, Botwana), 38359 (f, Botswana), 38307 (m, Tanzania); Suricata suricatta (FM 34348 (m, Botswana), 38349 (m, Botswana), 153721 (m, ZOO), 77514 (m, SW Africa), 38309 (m, Botswana), AMNH 168994 (f, S. Africa), 83645 (f, Botswana), 168990 (f. S. Africa), 81756 (m, S. Africa), 168998 (m. Cape Province), 168997 (m, S. Africa), 169451 (u, S. Africa), 31277 (u, unknown) Hyaenidae: Crocuta crocuta NMNH 015202 (u, S. Africa), 182032 (f, Kenya), 182085 (f, Kenya), 163101 (f, Kenya), 163100 (f, Kenya), 182084 (m, Kenya), 182095 (m, Kenya), 35011 (m, Sudan), 163102 (m, Kenya), 161909 (m, Kenya), FM 73035 (f, Tanzania), 98739 (m, Sudan), 104022 (u, Ethiopia), 127829 (u, Tanzania), HMCZ 13232, 21173, 31931, 23098, 13245, 14558; Hyaena brunnea FM 34585 (f, Botswana), 34584 (m, Botswana), 34586 (u, Botswana), 107290 (u, Egypt), Hyaena hyena NMNH 1820404 (f, Kenya), 182080 (f, Kenya), 182078 (f, Kenya), 182136 (f, Kenya), 182086 (f, Kenya), 182047 (f, Kenya), 182100 (f, Kenya), 182034 (m, Kenya), 182045 (m, Kenya), 182134 (m, Kenya), HMCZ 449, 14909, 27645, 51305, 8200, 51908, 27321; Proteles cristatus NMNH 368501 (f, Botswana), 470163 (f, Batwaswana), 368499 (f, Botswana), 469886 (f, S. Africa), 395686 (f, Unknown), 181495 (f, Kenya), 164837 (f, Kenya), 382515 (m, S. Africa), 164503 (m, Kenya), FM 127833 (f, Tanzania), 95919 (u, Zimbabwe, 95920 (u, Zimbabwe), AMNH 112743 (f, Burma), 102040 (f, Indonesia), 163599 (m, Burma), 112745 (m, Burma) Viverridae: Arctictis binturong: FM 98270 (m, Thailand) 104693 (f, Thailand), 98270 (m, Thailand), 85932 (m, Malaysia), HMCZ 5107 9u, Java), 34303 8u, Philippines), 35594 (u, N. Borneo); Arctogalidia trivirgata: FM 85110 (f, Malaysia), 38003 (f, Laos), 141375 (f, Malaysia), 88597 (f, Malaysia), 68709 (M, Malaysia), 8373 (m, Indonesia), 38004 (m, Fr. Indo-China), 38005 (m, Laos), 98648 (u, Malaysia), 98649 (u, Malaysia). Chrotogale owstoni 32556 (m, Vietnam); Civettictis civetta: FM 38188 (u, Botswana), 73802 (m, Gabon), 104964 (u, Mbomou), 153595 (f, Kenya), 27278 (f, Tanzania), 67002 (f, Sudan), 38186 (u, Botswana); Cynogale benetti FM 100462 (m, Malaysia); Genetta genetta FM 140213; G. maculata NMNH 367367 (f, Mozambique), 367369 (f, Mozambique), 367368 (f, Mozambique), 367362 (f, Mozambique), 367370 (f, Mozambique), 481984 (f, Liberia), 182381 (f, Kenya), 318099 (f, Kenya), 367361 (f,

123

Mozambique), 344961 9f, Kenya), 367360 (m, Mozambique), 182706 (m, Kenya), 182360 (m, Kenya), 182358 (m, Kenya), 182708 (m, Kenya), FM 17541 (f, Kenya), 149359 (f, Burundi), 73045 (f, Serengeti), 34321 (m, Tanzania), 32736 (m, Cameroon), 3864 (m, Maun), 17549 (u, Kenya), 38206 (u, Bechuanaland), HMCZ 42664, 32289, 46377, 8317, 32299, 22618, 3229; G. servalina FM 145230 (m, Uganda), 25306 (m, Cameroons), 145228 8m Uganda), 73795 (f, Gabon), 73800 (u, Gabon), AMNH 36017 (f, Kenya), 36013 (f, Kenya), 51557 (f, Zaire), 51552 (f, Zaire), 51568 9f, Zaire), 51573 (m, Zaire), 51576 (m, Zaire), 51562 (m, Zaire), 20793 (u, ZOO), HMCZ 32489 (u, Cameroons), 14741 (u, Cameroons), 32617 (u, Cameroons), 32305 (u, Uganda); G. tigrina NMNH 351943 (f, Cape Province), 469853 (f, Cape Province), 351942 (f, Cape Province), 469854 (m, Cape Province(, 344017 (m, Cape province), AMNH 169060 (f, S. Africa), HMCZ 1943 (u, S. Africa), 35416 (u, S. Africa), 1944 (u, S. Africa), 1897 (u, S. Africa); Paguma larvata: FM 31150 (u, China), 38019 (u, Laos), 32554 (u, Vietnam), 33630 (u, China), 33629 (u, China), 88308 (u, Sarawak); Paradoxurus hermaphroditus: FM 91249 (f, India), 38000 (m, Laos), 62855 (f, Palawan), 62841 (f, Palawan), 68712 (m, Malaysia), 56498 (m, Mindanao), 95032 (u, Ceylon), AMNH 163603 (f, Burma), 102879 (f, Sumatra), 59993 (m, China), 102878 (m, Sumatra); Prionodon linsang: FM 8371 (f, N. Borneo), 88300 (f, Malaysia), 88606 (m, Malaysia); P. pardicolor (FM 35463 (f, India), 35464 (m, India), 39176 (m, Vietnam), 39175 (m, Vietnam), 75814 (m, Assam); Viverra tangalunga: FM 62864 (f, Philippines), 68704 (f, Malaysia), 32693 (f, Borneo), 85115 (m, Malaysia), 88788 (m, Malaysia), 68707 (m, Malaysia), 87721 (f, Philippines), AMNH 106045 (f, Indonesia), 152882 (f, Indonesia), 106054 (m, Borneo), 10011 (m, Borneo), 103738 (m, Borneo), 106003 (m, Borneo), 242234 (u, Philippines); V. zibetha FM 36743 (f, Scechwan); Viverricula indica FM 171891 (u, Madagascar), 75864 (f, Assam), 83088 (m, Sri Lanka), 36739 (f, Sichuan), 33637 (m, Sichuan), 36532 (f, Vietnam), 31040 (m, Annam), AMNH 60181 (f, China), 59946 (f, China), 84425 (f, China), 59947 (m, China), 60073 (m, China), 59954 (m, China), 58373 (m, Sichuan), 84429 (m, Fujian) Eupleridae: Cryptoprocta ferox NMNH 112841 (f, Madagascar), 319987 (u, Zoo); FM 161707 (f, Madagascar), 5655 (m, Madagascar), 161793 (u, madagascar), 33950 (u, Madagascar), 47494 (u, Madagascar), 45971 (u, Madagascar), 45970 (u, Madagascar), 16455 (u, Madagaascar), 45969 (u, Madagascar; Eupleres goudottii FM 30492 (m, Madagascar), AMNH 100462 9f, Madagascar), 100461 (m, Madagascar), 188211 (m, Madagascar), 100484 (m, Madagascar), 880 (u, unknown), HMCZ 6 specimens (u, Madagascar); Fossa fossana NMNH 521565 (m, unknown), 574898 (m, unknown), 318107 (m, unknown) FM 171892 (f, Madagascar), 85195 (u, Madagascar), 156648 (m, Madagascar); Galidia elegans NMNH 199254 (u, Madagascar), 318104 (f, Madagascar), 341845 (f, Madagascar), 318106 (f, Madagascar), 399274 (m, Madagascar), 318105 (m, Madagascar), 463931 (m, Madagascar), FM 162110 (f, Madagascar), 169717 (f, madagascar) 161921 (m, Madagascar), 156550 (m, Madagascar), 156650 (m, Madagascar), 169674 (f, Madagascar), 51925 (m, Madagascar), 85194 (m, Madagascar), 85873 (m, Madagascar), 170876 (m, Madagascar), HMCZ 45956 (u, Madagascar), 45959 (u, Madagascar), 45951 (u, Madagascar); Galidictis fasciata FM 162111 (m, Madagascar), HMCZ 5116 (u, S. Africa), 28623 (u, S. Africa) NMNH 4419210 (u, unknown), 100479 (u, unknown), FM 156549 (u, Madagascar), 156652 (u, Madagascar), Helogale HMCZ 8294 (u, E. Africa), 23003 (u, Tanganyika), 8318 (u, British E. Africa), 36046 (u, SW Africa), 37315 (u, Transvaal), 23006 (f, Port. East Africa), 23007 (f, Tanganyika), 23005 (m, Port. E. Africa); Salanoia concolor FM 33946 (u, Madagascar), HMCZ 27827 (u, Madagascar) Nandiinidae: Nandinia binotata FM 161266 (m, Tanzania), 53868 (m, W. Africa), 149360 (f, Burundi), 151366 (f, Tanzania), 25303 (f, Cameroon), 83642 (f, Angola), 149361 (f, Burundi), 81601 (f, Angola), 73804 (f, Gabon), HMCZ 31623 (f, Kenya), 32248 (f, Uganda), 17704 (f, Cameroon), 32193 (m, Kenya), 31101 (m, Kenya), 14742 (m, Cameroon), 23192 (m, Cameroon), 17703 (u, Cameroon)

124

REFERENCES

Akaike, H. 1974. A new look at the statistical model identification. IEEE Trans. Autom. Contr. 19: 716-723.

Alroy, J., Koch, P.L. and J.C. Zachos. 2000. Global climate change and North American mammalian evolution. Paleobiology. 259-288

Arthur, W. 2001. Developmental drive: an important determinant of the direction of phenotypic evolution. Evol. and Dev. 3: 271-278.

———. 2004. The effect of development on the direction of evolution: toward a twenty-first century consensus. Evol. and Dev. 6: 282-288.

Baker, R. H., and G. S. Wilkinson. 2003. Phylogenetic analysis of correlation structure in stalk-eyed flies (Diasemopsis, Diopsidae). Evolution 57: 87-103.

Barker, F.K., Barrowclough, G.F. and J.G. Groth. 2001. A phylogenetic hypothesis for passerine birds: taxonomic and biogeographic implications of an analysis of nuclear DNA sequence data. Proc. R. Soc. Lond. 269: 295-309.

Baskin, J.A. 1981. Barbourofelis (Nimravidae) and Nimravides (Felidae) with a description of two new species from the late Miocene of Florida. J. Mamm. 62: 122-139.

———. 1998. Mustelidae. In Janis, C.M., Scott, K.M. and L.L. Jacobs (eds.). Evolution of Tertiary mammals of North America. Cambridge Univ. Press, Cambridge. Pp. 152-173

Berta, A. 1987. The sabercat Smilodon gracilis from Florida and a discussion of its relationships (Mammalia, Felidae, ). Bull. Am. Mus. Nat. Hist. 31: 1-63.

———. 1998. Hyaenidae. In Janis, C.M., Scott, K.M. and L.L. Jacobs (eds.). Evolution of Tertiary mammals of North America. Cambridge Univ. Press, Cambridge. Pp. 243-246

Berta, A., and H. Galiano. 1983. Megantereon hesperus from the late Hemphillian of Florida with remarks on the phylogenetic relationships of machairodonts (Mammalia, Felidae, Machairodontinae). J. Paleo. 57: 892-800.

Biknevicius, A. R., and B. Van Valkenburgh. 1996. Design for killing: craniodental adaptations of predators. In: Gittleman, J.L. (ed.) Carnivore behavior, ecol. and evolution. Ithaca. Pp. 393-428.

Bininda-Emonds, O.R.P., Gittleman, J.L. and A. Purvis. 1999. Building large trees by combining phylogenetic information: a complete phylogeny of the extant Carnivora (Mammalia). Biol. Rev 74: 143-175.

Bokma, F. 2002. A statistical test of unbiased evolution of body size in birds. Evolution 56: 2499- 2504.

125

Braun, E.L., and R.T. Kimball. 2001. Polytomies, the power of phylogenetic inference, and the stochastic nature of molecular evolution: a comment on Walsh et al. (1999). Evolution 55(6): 1261-1263.

Brooks, D. R., and D. A. McLennan. 1991. Phylogeny, Ecol. and behavior. Univ. Chicago Press, Chicago.

Bryant, H. N. 1991. Phylogenetic relationships and systematics of the Nimravidae (Carnivora). J. Mamm. 72: 56-78.

———. 1996. Nimravidae. In D. R. Prothero and R. J. Emry, (eds.). The terrestrial Eocene- Oligocene transition in North America. Cambridge Univ. Press, Cambridge. Pp. 453- 475.

Bryant, H. N., A. P. Russell, and W. D. Fitch. 1993. Phylogenetic relationships within the extant Mustelidae (Carnivora)—appraisal of the cladistic status of the Simpsonian subfamilies. Zool. J. Linn. Soc. 108: 301-334.

Buckley, T.R., Cordeiro, M., Marshall, D.C. and C. Simon. 2006. Differentiating between hypotheses of lineage sorting and introgression in New Zealand alpine Cicadas (Maoricicada Dugdale). Syst. Biol. 55(3): 411-425.

Bull, J. J., and E. L. Charnov. 1985. On irreversible evolution. Evolution 39: 1149-1155.

Chappill, J. A. 1989. Quantitative characters in phylogenetic analysis. Cladistics 5: 217-234.

Colbert, E. 1939. Carnivora of the Tung Gur Formation of Mongolia. Bull. Am. Mus. Nat. Hist. 76: 47-81.

Crusafont-Pairo, M., and J. Truyols-Santonja. 1956. A biometric study of the evolution of fissiped carnivores. Evolution 10: 314-332.

Cunningham, C. W., K. E. Omland, and T. H. Oakley. 1998. Reconstructing ancestral character states: a critical reappraisal. Trends Ecol. and Evol. 13: 361-366.

Dashzeveg, D. 1996. Some carnivorous mammals from the Paleogene of the Eastern Gobi Desert, Mongolia, and the application of Oligocene carnivores to stratigraphic correlation. Am. Mus. Nov. 3179: 1-14.

Dayan, T., Simberloff, D. Tchernov, E. and Y. Yom-Tov. 1989. Inter and intra-specific character displacement in mustelids. Ecol. 70: 1526-1639.

———. 1990. Feline canines. Community wide character displacement among the small cats of Israel. Am. Nat. 136: 39-60.

Dayan, T., and D. Simberloff. 2005. Ecological and community-wide character displacement: the next generation. Ecol. Letters 8: 875-894.

De Queiroz, A. 1998. Interpreting sister group tests of key innovation hypotheses. Syst. Bio. 47: 710-718.

126

———. 1999. Do image-forming eyes promote evolutionary diversification? Evolution 53: 1654- 1664.

DeSalle, R., Absher, R., and G. Amato. 1994. Speciation and phylogenetic resolution. Trends Ecol. Evol. 9: 297-298.

Dillon, W.R. and M. Goldstein. 1984. Multivariate analysis. Wiley, New York.

Dodd, M.E., Silvertown, J., and M.W. Chase. 1999. Phylogenetic analysis of trait evolution and species diversity variation among angiosperm families. Evolution 53: 732-744.

Donoghue, M.J., Doyle, J.A., Gautherier, J., Kluge, A.G and T. Rowe. 1989. The importance of fossils in phylogeny reconstruction. Ann. Rev. of Ecol. and Syst. 20: 431-460.

Donoghue, M. J., and R. H. Ree. 2000. Homoplasy and developmental constraint: a model and an example from plants. Am. Zool. 40: 759-769.

Dragoo, J. W., and R. L. Honeyctt. 1997. Systematics of mustelid-like carnivores. J. Mamm. 78: 426-443.

Eble, G. J. 2000. Contrasting evolutionary flexibility in sister groups: disparity and diversity in Mesozoic atelostomate echinoids. Paleobiology 26: 56-79.

Eble, G. J. 2004. The macroevolution of phenotypic integration. In M. Pigliucci and K. Preston, (eds.). Phenotypic Integration. Oxford Univ. Press, Oxford. Pp. 253-273.

Emerson, S. B. 1988. Testing for historical patterns of change —a case-study with frog pectoral girdles. Paleobiology 14: 174-186.

Ericson, Per G. P., Cristidis, L., Cooper, A., Irestedt, M. Jackson, J., Johansson, U.S., and J.A. Norman. 2002. A Gondwanan origin of passerine birds supported by DNA sequences of the endemic New Zealand wrens. Proc. R. Soc. Lond. B 269: 235-241.

Estes, J.A. 1989. Adaptations for aquatic living by carnivores. In Gittleman, J.L. (ed.). Carnivore behavior, ecol., and evolution. Cornell Univ. Press, Ithaca. Pp 242-282.

Ewer, R.F. 1973. The carnivores. Cornell University Press, Ithaca, New York.

Farrell, B.D., Dussourd, D.E., and C. Nitter. 1991. Escalation of plant defense: do latex and resin canals spur plant diversification. Am. Nat. 138: 881-900.

Fishbein, M. and D.E. Soltis. 2004. Further resolution of the rapid radiation of Saxifragiles (Angiosperms, Eudicots) supported by mixed model Bayesian analysis. Syst. Bot. 29:883-891.

Fishbein, M., Hibsch-Jetter, C., Soltis, D.E., and L. Hufford. 2001. Phylogeny of Saxifragales (Angiosperms, Eudicots): analysis of a rapid, ancient radiation. Syst. Biol. 50(6): 817- 847.

127

Flower, W.H. 1869. On the value of the characters of the base of the cranium in the classification of the Order Carnivora, and on the systematic position of Bassiris and other disputed forms. Proc. Zool. Soc. Lond. 4-37.

Flower, W.H. and Lydekker. 1891. An introduction to the study of mammals, living and extinct. London: A. and C. Black.

Flynn, J.J., Neff, N.A., and R.H. Tedford. 1988. Phylogeny of the Carnivora. In: Benton, M.J. (ed.). The Phylogeny and Classification of the Tetrapods, Vol 2: Mammals. Clarendon Press, Oxford. Pp 73-116.

Flynn, J.J., Finarelli J.A., Zehr, S., Hsu, J., and M.A. Nedbal. 2005. Molecular phylogeny of the Carnivora (Mammalia): assessing the impact of increased sampling on resolving enigmatic relationships. Syst. Biol. 54(2): 317-337.

Flynn, J.J., and M.A. Nedbal. 1998. Phylogeny of the Carnivora (Mammalia): congruence vs. incompatibility among multiple data sets. Mol. Phylogenet. Evol. 9(3): 414-426.

Foote, M. 1993. Discordance and concordance between morphological and taxonomic diversity. Paleobiology 19: 185-204.

———. 1997. The evolution of morphological diversity. Ann. Rev. Ecol. Syst. 28: 129-152.

Fugmann, S.D., Lee, A.I, Shockett, P.E., Villey, I.J. and D.G. Schatz. 2000. The rag proteins and V(D)J recombination: complexes, ends, and transposition. Ann. Rev. Immun.18: 495- 527.

Fusco, G. 2001. How many processes are responsible for phenotypic evolution? Evol. and Dev. 3: 279-286.

Futuyma, D. J., and G. Moreno. 1988. The evolution of ecological specialization. Ann. Rev. Eco. Syst. 19: 207-233.

Gardezi, T. and J. da Silva. 1999. Diversity in relation to body size in mammals: a comparative study. Am. Nat. 153: 110-123.

Gaubert P., and P. Cordeiro-Estrela. 2006. Phylogenetic systematics and tempo of evolution of the Viverrinae (Mammalia, Carnivora,Viverridae) within feliformians: implications for faunal exchanges between Asia and Africa. Mol. Phylogenet. Evol. 41: 266-278.

Gaubert P., Fernandes, C.A., Bruford, M.W., and G. Veron. 2004. Genets (Carnivora, Viverridae) in Africa: an evolutionary synthesis based on cytochrome b sequences and morphological characters. Biol. J. Linn. Soc. 81: 589-610.

Gaubert P., and G. Veron. 2003. Exhaustive sample set among Viverridae reveals the sister-group of felids: the linsangs as a case of extreme morphological convergence within Feliformia. Proc. R. Soc. Lond. B. 270: 2523-2530.

Gaubert P., Tranier, M., Delmas, A., Colyn, M., and G. Veron. 2003. First molecular evidence for reassessing phylogenetic affinities between genets (Genetta) and the enigmatic genet-like

128

taxa Osbornictis, Poiana and Prionodon (Carnivora, Viverridae). Zool. Scripta 33(2): 117-129.

Gaubert, P., Veron, G. and Tranier, M. 2002. Genets and “genet-like” taxa (Carnivora, Viverinae): phylogenetic analysis, systematics and biogeographic implications. Biol. J. Linn. Soc. 134: 317-334.

Gauthier, J., Kluge, A.G. and T. Rowe (1988). Amniote phylogeny and the importance of fossils. Cladistics 4: 105-209.

Geeta, R. 2003. The origin and maintenance of nuclear endosperms: viewing development through a phylogenetic lens. Proc. R. Soc. Lond. B 270:29-35.

Geraads, D. and E. Gulec. 1997. Relationships of Barbourofelis piveteaui (Ozansoy, 1965), a late Miocene nimravid (Carnivora, Mammalia) from central Turkey. J. Vert. Paleo. 17: 380- 375. Gingerich, P.D. 1974. Size variability in living mammals and diagnosis of closely related sympatric fossil species. J. Paleo. 48: 895-903.

Ginsburg, L. Sur les modalites d’evolution du genre Neogene Pseudaelurus Gervais (Felidae, Carnivora, Mammalia). Modalites, Rythmes et Mecanismes de L’evolution biologique 330: 131-136.

Gittleman, J.L. and A. Purvis. 1998. Body size and species-richness in carnivores and primates. Proc. R. Soc. Lond. B, Biol. Sci. 265: 113-119.

Glass, G.E., and L.D. Martin. 1978. Multivariate comparison of some extant and fossil Felidae. Carnivore. 1: 80-87.

Goswami, A. 2006. Morphological integration in the carnivoran skull. Evolution 60: 169-183.

Gould, S. J. 1984. Morphological channeling by structural constraint; convergence in styles of dwarfing and gigantism in Cerion, with a description of two new fossil species and a report on the discovery of the largest Cerion. Paleobiology 10: 172-194.

Gregory, W.K., and M. Hellman. 1939. On the evolution and major classification of the civets (Viverridae) and allied fossil and recent Carnivora: a phylogenetic study of the skull and dentition. Proc. of the Am. Phil. Soc. 81(3): 309-392.

Hansen, T. F., and E. P. Martins. 1996. Translating between microevolutionary process and macroevolutionary patterns: the correlation structure of interspecific data. Evolution 50: 1404-1417.

Harshman, J. Huddleston, C.J., Bollback, J.P., Parsons, T.J., and M.J. Braun. 2003. True and false gavials, a nuclear gene phylogeny of Crocodylia. Syst. Biol. 52(3): 386-402.

Harvey, P. H., and M. Pagel. 1991. The comparative method in evolutionary biology. Oxford Univ. Press, New York.

Hemmer, H. 1978. Fossil history of living Felidae. Carnivore 2: 58-61.

129

Hodges, S.A. and M.L. Arnold. 1995. Spurring plant diversification: are floral nectar spurs a key innovation? Proc. R. Soc. Lond. B, Biol. Sci. 262: 343-348.

Holliday, J.A., and S.J. Steppan. 2004. Evolution of hypercarnivory: the effect of specialization on morphological and taxonomic diversity. Paleobiology 30(1): 108-128.

Holt, R. D., and M. S. Gaines. 1992. Analysis of adaptation in heterogeneous landscapes - implications for the evolution of fundamental niches. Evol. Eco. 6: 433-447.

Huelsenbeck, J. P. 1991. When are fossils better than extant taxa in phylogenetic analysis? Syst. Zool. 40: 458-469.

Huelsenbeck, J.P., and K.A. Crandall. 1997. Phylogeny estimation and hypothesis testing using maximum likelihood. Ann. Rev. Ecol. Syst. 28: 437-466.

Huelsenbeck, J.P., and F. Ronquist. 2001. MrBayes: Bayesian inference of phylogeny. Biometrics 17: 754-755.

Hunt, R.M. Jr. 1987. Evolution of the aeluroid Carnivora: significance of auditory structure in the nimravid cat Dinictis. Am. Mus. Nov. 2886: 1-74.

———. 1989. Evolution of the aeluroid Carnivora: significance of the ventral promontorial process of the petrosal, and the origin of basicranial patterns in the living families. Am. Mus. Nov. 2930: 1-32.

———. 1991. Evolution of the aeluroid Carnivora: viverrid affinities of the Miocene carnivoran Herpestides. Am. Mus. Nov. 3023: 1-34.

———. 1996. Biogeography of the Order Carnivora. In: Gittleman, J.L. (ed.). Carnivore behavior, ecology, and evolution. Cornell Univ. Press, Ithaca. Pp. 485-541.

———. 1998. Evolution of the aeluroid Carnivora: diversity of the earliest aeluroids from Eurasia (Quercy, Hsanda-Gol) and the origin of felids. Am. Mus. Nov. 3252: 1-65.

Hunt, R.M., Jr. and R.H. Tedford. 1993. Phylogenetic relationships within the aeluroid Carniovra and implications of their temporal and geographic distribution. In: F.S. Szalay, J.J. Novacek and M.C. McKenna, (eds.). Mammal phylogeny. Springer, New York. Pp. 53- 73.

Jae-Heup, K. Eizirik, E., O’Brien, S.J. and W.E. Johnson. 2001. Structure and patterns of sequence variation in the mitochondrial DNA control region of the great cats. Mitochondrion 1: 279-292.

Janczewski, D.N., Modi, W.S., Stephens, J.C., and S.J. O’Brien. 1995. Molecular evolution of mitochondrial 12S RNA and cytochrome b sequences in the pantherine lineage of Felidae. Mol. Biol. Evol. 12: 690-707.

Janis, C.M., Scott, K.M. and L.L. Jacobs, (eds.). 1998. Evolution of Tertiary mammals of North America. Cambridge Univ. Press, Cambridge.

130

Janz, N., Nyblom, K., and S. Nylin. 2001. Evolutionary dynamics of host-plant specialization: a case study of the tribe Nymphaline. Evolution 55:783-796.

Janson, C. H. 1992. Measuring evolutionary constraints: a Markov model for phylogenetic transitions among seed dispersal syndromes. Evolution 46:136-158.

Johnson, W.E., and S.J. O’Brien. 1997. Phylogenetic reconstruction of the Felidae using 16S rRNA and NADH-5 mitochondrial genes. J. Mol. Evol. 44(1): S98-S116.

Johnson, W.E., Eizirik, E., Pecon-Slattery, J., Murphy, W.J., Antunes, A., Teeling, E., and S.J. O’Brien. 2006. The Late Miocene radiation of modern Felidae: a genetic assessment. Science 311: 73-77.

Keiser, J.A. and H. Groeneveld. 1991. Fluctuating odontometric asymmetry, morphological variability and genetic monomorphism in the cheetah Acinonyx jubatus. Evolution 45: 1175-1183.

King, C. 1989. The advantages and disadvantages of small size to weasels, Mustela species. In Gittleman, J.L. (ed.). Carnivore behavior, ecol., and evolution. Cornell Univ. Press, Ithaca. Pp 302-334.

Koepfli, K.P., and R.K. Wayne. 1998. Phylogenetic relationships of otters (Carnivora: Mustelidae) based on mitochondrial cytochrome b sequences. J. Zool. 246:401-416.

Koepfli, K., Jenks, S.M., Edizirik, E., Zihirpour, T., Van Valkenburgh, B., and R.K. Wayne. 2006. Molecular systematics of the Hyaenidae, relationships of a relictual lineage resolved by a molecular supermatrix. Mol. Phylogenet. Evol. 38: 603-620.

Lauder, G. V. 1981. Form and function – structural analysis in evolutionary morphology. Paleobiology 7: 430-442.

Liem, K. 1973. Evolutionary strategies and morphological innovations: a cichlid pharyngeal jaws. Syst. Zool. 22: 425-441.

Losos, J. B., and D. J. Irschick. 1994. Adaptation and constraint in the evolution of specialization of Bahamian Anolis lizards. Evolution 48: 1786-1798.

Maddison, W.P. 1989. Reconstructing character evolution on polytomous cladograms. Cladistics 5: 365-377.

———. 1997. Gene trees in species trees. Syst. Biol. 46: 523-535.

Maddison, W.P., and L. Knowles. 2006. Inferring phylogeny despite incomplete lineage sorting. Syst. Biol. 55(1): 21-30.

Marroig, G., and J. M. Cheverud. 2001. A comparison of phenotypic variation and covariation patterns and the role of phylogeny, Ecol., and ontogeny during cranial evolution of new world monkeys. Evolution 55: 2576-2600.

Martin, L.D. 1980. Functional morphology and the evolution of cats. Trans. Neb. Acad. Sci. 8: 141-154.

131

———. 1989. Fossil history of terrestrial Carnivora. In Gittleman, J.L. (ed.). Carnivore behavior, ecol., and evolution. Cornell Univ. Press, Ithaca Pp. 536-568.

———.1998a. Felidae. In Janis, C.M., Scott, K.M. and L.L. Jacobs (eds.). Evolution of Tertiary mammals of North America. Cambridge Univ. Press, Cambridge. Pp. 236-242.

———.1998b. Nimravidae. In Janis, C.M., Scott, K.M. and L.L. Jacobs (eds.). Evolution of Tertiary mammals of North America. Cambridge Univ. Press, Cambridge Pp. 228-235.

Martins, E. P. 2000. Adaptation and the comparative method. Trends Ecol. Evol.15: 296-299.

Masuda, R., Lopez, J.V, Slattery, J.P., Yuhki, N., and S. J. O’Brien. 1996. Molecular phylogeny of mitochondrial cytochrome b and 12s rRNA sequences in the Felidae: and domestic cat lineages. Mol. Phylogenet. Evol. 6(3): 351-365.

Mattern, M.Y., and D.A. McClennan. 2000. Phylogeny and speciation of felids. Cladistics 16: 232-253.

McShea, D. W. 2001. The minor transitions in hierarchical evolution and the question of a directional bias. J. Evo. Biol.14: 502-518.

McShea, D. W., and E. P. Venit. 2002. Testing for bias in the evolution of coloniality: a demonstration in cyclostome bryozoans. Paleobiology 28: 308-327.

Meiri, S., T. Dayan, and D. Simberloff. 2005. Variability and correlations in carnivore crania and dentition. Functional Ecol. 19: 337-343.

Mitter, C., Farrell, B. and B. Wiegman. 1988. The phylogenetic study of adaptive zones – has phytophagy promoted diversification? Am. Nat. 132: 107-128.

Miyamoto, M.M., Porter, C.A., and M. Goodman. 2000. C-myc gene sequences and the phylogeny of bats and other eutherian mammals. Syst. Biol. 49: 501-514.

Mooers, A. O., and D. Schluter. 1999. Reconstructing ancestor states with maximum likelihood: support for one- and two-rate models. Syst. Biol. 48: 623-633.

Morales, J., Salesa, M.J., Pickford, M. and D. Soria. 2001. A new tribe, new genus and two new species of Barbourofelinae (Felidae, Carnivora, Mammalia) from the early Miocene of E. Africa and . Trans. R. Soc. Edinburgh: Earth Sci. 92: 97-102.

Moran, N. A. 1988. The evolution of host-plant alternation in aphids - evidence for specialization as a dead end. Am. Nat. 132: 681-706.

Morlo, M.S., Peigne, S. and D. Nagel. 2004. A new species of Prosansanosmilus: implications for the systematic relationships of the family new rank (Carnivora: Mammalia). Zool. J. Linn. Soc. 140: 43-61

Nee, S., Holmes, E.C., Rambout, A. and P.H. Harvey. 1995. Inferring population history from molecular phylogenies. Phil. Trans. R. Soc. Lond. B. Biol. Sciences 344: 25-31.

132

Neff, N. 1982. The big cats. Harry N. Abrams, New York.

———.1983. The basicranial anatomy of the Nimravidae (Mammalia: Carnivora): character analyses and phylogenetic inferences. Ph.D. dissertation. City University, New York.

Nowak R.M. 1999. Walker’s mammals of the world. VII. Baltimore: Johns Hopkins Univ. Press.

Nosil, P. 2002. Transition rates between specialization and generalization in phytophagous insects. Evolution 56: 1701-1706.

Olson, E. C., and R. L. Miller. 1958. Morphological integration. Univ. of Chicago Press, Chicago.

Omland, K. E. 1997. Examining two standard assumptions of ancestral reconstructions: repeated loss of dichromatism in dabbling ducks (Anatini). Evolution 51: 1636-1646.

———. 1999. The assumptions and challenges of ancestral state reconstructions. Syst. Biol. 48: 604-611.

Omland, K. E., and S. M. Lanyon. 2000. Reconstructing plumage evolution in orioles (Icterus): repeated convergence and reversal in patterns. Evolution 54: 2119-2133.

O’Regan, H.J. 2002. Defining , a multivariate analysis of skull shape in big cats. Mamm. Rev. 32: 58-62.

Pagel, M. 1994. Detecting correlated evolution on phylogenies: a general method for the comparative analysis of discrete characters. Proc. R. Soc. Lond. B 255: 37-45.

———. 1999. The maximum likelihood approach to reconstructing ancestral character states of discrete characters on phylogenies. Syst. Biol. 48: 612-622.

Pagel, M. D., and P. H. Harvey. 1988. Recent developments in the analysis of comparative data. Qu. Rev. Biol. 63: 413-440.

Peigne, S. and L. de Bonis, 1999. The genus Stenoplesictis Filhol (Mammalia, Carnivora) from the Oligocene deposits of the phosphorites of Quercy, France. J. Vert. Paleo. 19(3): 566- 575.

Peigne, S. de Bonis, L., Likius, A., Mackaye, H.T., Vignaud, P. and M. Brunet. 2005. The earliest modern mongoose (Carnivora, Herpestidae) from Africa (late Miocene of Chad). Naturwissenschaften 92: 287-292.

Perez, M., Li, B., Tillier, A., Craud, A., and G. Veron. 2006. Systematic relationships of the bushy-tailed and black-footed mongooses (genus Bdeogale, Herpestidae, Carnivora) based on molecular, chromosomal and morphological evidence. The Authors JZS 44(3): 251-259.

Petter, G. 1974. Rapports phyletiques des viverrides (Carnivores fissipedes). Les formes de Madagascar. Mammalia 38(4): 605-636.

Pigliucci, M., and K. Preston. 2004. Phenotypic integration. Oxford Univ. Press, Oxford.

133

Poe, S., and A. L. Chubb. 2004. Birds in a bush: five genes indicate explosive evolution of avian orders. Evolution 58(2): 404-415.

Polly, P. D. 2005. Development and phenotypic correlations: the evolution of tooth shape in Sorex araneus. Evol. and Dev. 7: 29-41.

Porter, M. L., and K. A. Crandall. 2003. Lost along the way: the significance of evolution in reverse. Trends Ecol. Evol. 18: 541:547.

Posada, D., and K.A. Crandall. 1998. MODELTEST: testing the model of DNA substitution. Bioinformatics 14: 817-818.

Price, P.W., and T.G. Carr. 2000. Comparative ecology of membracids and tenthredinids in a macroevolutionary context. Ecol. Evol. Rev. 2: 645-665.

Prothero, D.R. 1994. The Late Eocene-Oligocene . Ann. Rev. Earth Planet. Sci. 22: 145-65.

Radinsky, L.B. 1981a. Evolution of skull shape in carnivores. 1. Representative modern carnivores. Biol. J. Linn. Soc. 15: 69-388.

———.1981b. Evolution of skull shape in carnivores. 2. Additional modern carnivores. Biol. J. Linn. Soc. 16: 337-355.

———. 1982. Evolution of skull shape in carnivores. 3. The origin and early radiation of the modern carnivore families. Paleobiology. 8(3): 177-195.

Rannala, B., Huelsenbeck, J.P., Yang, Z., and R. Nielsen. 1998. Taxon sampling and the accuracy of large phylogenies. Syst. Biol. 47(43): 702-710.

Richardson, M. K., and A. D. Chipman. 2003. Developmental constraints in a comparative framework: a test case using variations in phalanx number during amniote evolution. J. Exp. Zool. 296B: 8-22.

Rothwell, T. 2001. A partial skeleton of Pseudaelurus (Carnivora: Felidae) from the Nambe Member of the Tesuque Formation, Espanola Basin, New Mexico. Am. Mus. Nov. 3342: 1-31.

———. 2003. Phylogenetic systematics of North American Pseudaelurus (Carnivora: Felidae). Am. Mus. Nov. 3403: 1-64.

Rouse, G. W. 2000. Bias? What bias? The evolution of downstream larval-feeding in animals. Zool. Scripta 29: 213-236.

Roy, K. and M. Foote. 1997. Morphological approaches to measuring biodiversity. Trends Ecol. Evol. 12: 277-281.

Russell, A.P., Bryant, H.N., Powell, G.L. and R. Laroiya. 1995. Scaling relationships within the maxillary tooth row of the Felidae, and the absence of the 2nd upper premolar in Lynx. J. Zool. 236: 161-182.

134

Salazar-Ciudad, I. 2006. Developmental constraints vs. variational properties: how pattern formation can help to understand evolution and development. J. Exp. Zool. 306B: 107- 125.

Salles, L.O. 1992. Felid phylogenetics: extant taxa and skull morphology (Felidae, Aeluroidea). Am. Mus. Nov. 3047: 1-67.

Sanderson, M. J. 1993. Reversibility in evolution - a maximum-likelihood approach to character gain loss bias in phylogenies. Evolution 47: 236-252.

Savage, D.E. and D.E. Russell. 1983. Mammalian paleofaunas of the world. Addison-Wesley Pub. Co. Reading, MS.

Schluter, D. 2000. The ecology. of . Oxford Univ. Press, Oxford.

Schluter, D., E. A. Clifford, M. Nemethy, and J. S. McKinnon. 2004. Parallel evolution and inheritance of quantitative traits. Am. Nat. 163: 809-822.

Schluter, D., T. Price, A. O. Mooers, and D. Ludwig. 1997. Likelihood of ancestor states in adaptive radiation. Evolution 51: 1699-1711.

Schmidt-Kittler, N. 1987. The Carnivora (Fissipedia) from the Lower Miocene of East Africa. Palaeontographica 197(3): 85-126.

Schwarz, M.P., Fuller, S., Tierney, S.M., and S.J.B. Cooper. 2006. of the exoneurine allodapine bees reveal an ancient and puzzling dispersal from Africa to Australia. Syst. Biol. 55(1): 31-45.

Schwenk, K., and G. P. Wagner. 2001. Function and the evolution of phenotypic stability: connecting pattern to process. Am. Zool. 41: 552-563.

———. 2004. The relativism of constraints on phenotypic evolution. In: M. Pigliucci and K. Preston, (eds.). Phenotypic Integration. Oxford Univ. Press, Oxford. Pp. 390-408

Sheldon, P. R. 1996. Plus ça change - a model for stasis and evolution in different environments. Paleogeography, Palaeoclimatology, PalaeoEcol. 127: 209-227.

Siddall, M. E., D. R. Brooks, and S. S. Desser. 1993. Phylogeny and the reversibility of parasitism. Evolution 47: 308-313.

Simon, C. 1983. A new coding procedure for morphometric data with an example from periodical cicada wing veins. in J. Felsenstein, (ed.). Numerical . Springer, Berlin. Pp. 378-382.

Simpson, G. G. 1945. The principles of classification and a classification of mammals. Bull. Am. Mus. Nat. Hist. 85: 1-350.

Slowinski, J.B. 2001. Molecular polytomies. Mol. Phylogenet. Evol. 19: 114-120.

135

Slowinski, J.B., and C. Guyer. 1993. Testing whether certain traits have caused amplified diversification – am improved method based on a model of random speciation and extinction. Am. Nat. 142: 1019-1024.

Smith, J.F. 2001. High species diversity in fleshy-fruited tropical understory plants. Am. Nat. 157: 646-653.

Smith, J. M., R. Burian, S. Kauffman, P. Alberch, J. Campbell, B. Goodwin, R. Lande, D. Raup, and L. Wolpert. 1985. Developmental constraints and evolution. Quarterly Rev. Bio. 60: 265-287.

Sokal, R. R., and F. J. Rohlf. 1998. Biometry. W.H. Freeman and Company, New York.

Sotnikova, M. and P. Nikolskiy. 2006. Systematic position of the cave lion (Goldfuss) based on cranial and dental characters. Quat. Int. 143-143: 218-228.

Steppan, S. J. 1997a. Phylogenetic analysis of phenotypic covariance structure 1. Contrasting results from matrix correlation and common principal component analyses. Evolution 51: 571-586.

———. 1997b. Phylogenetic analysis of phenotypic covariance structure 2. Reconstructing matrix evolution. Evolution 51: 587-594.

Steppan, S.J., R.M. Adkins, and J. Anderson. 2004a. Phylogeny and divergence-date estimates of rapid radiations in muroid rodents based on multiple nuclear genes. Syst. Biol. 53: 533- 553.

Steppan, S.J., Storz, B.L., and R.S. Hoffmann. 2004b. Nuclear DNA phylogeny of the squirrels (Mammalia, Rodentia) and the evolution of arboreality from c-myc and RAG-1. Mol. Phylogenet. Evol. 30: 703-719.

Sullivan, J. and D.L. Swofford. 1997. Are guinea pigs rodents? The utility of models in molecular phylogenetics. J. Mamm. Evol. 4: 77-86.

Swofford, D.L. 2004. PAUP*. Phylogenetic Analysis Using Parsimony (*and other methods). Version 4.0. Sinauer Associates. Sunderland, MA.

Swofford, D. L., G. J. Olsen, P. J. Waddell, and D. M. Hillis. 1996. Phylogenetic inference. In: D. M. Hillis, C. Moritz, and B. Mable (eds.). Molecular Systematics (2nd ed.). Sinauer Associates, Sunderland, Massachusetts. Pp. 407-514

Tedford et al. 1987. Faunal succession and biochronology of the Arikareean through Hemphillian (late Oligocene through earliest Pliocene epochs) in North America. In M.O. Woodburne (ed.), Cenozoic Mammals of North America, Geochronology and Biostratigraphy. Univ. Cal. Press. Berkeley. Pp. 153-210.

Teeling, E.C., Madsen, O., Van Den Bussche, R.A., de Jong, W.W., Stanhope, M.J., and M.L. Springer. 2002. Microbat paraphyly and the convergent evolution of a key innovation in Old World rhinolophoid microbats. Proc. R. Soc. Lond. 99: 1431-1436.

136

Teilhard de Chardin, P. 1915. Les carnassiers des Phosphorites du Quercy. Ann. Paleontologie 9(3,4): 103-192.

Turner, A. and M. Anton. 1997. The big cats and their fossil relatives. Columbia Univ. Press, NY.

Van Valkenburgh, B. 1988. Trophic diversity in past and present guilds of large predatory mammals. Paleobiology 14: 155-173.

———. 1989. Carnivore dental adaptations and diet: a study of trophic diversity within guilds. In: Gittleman, J.L., (ed). Carnivore behavior, ecol. and evolution. Cornell Univ. Press, Ithaca. Pp. 410-436.

———. 1990. Skeletal and dental predictors of body mass in carnivores. In: J. Damuth and B.J. MacFadden, (eds.). Body size in mammalian paleobiology: estimation and biological implications. Cambridge Univ. Press, Cambridge. Pp. 181-205.

———. 1991. Iterative evolution of hypercarnivory in canids (Mammalia, Carnivora) - evolutionary interactions among sympatric predators. Paleobiology 17: 340-362.

———. 1999. Major patterns in the history of carnivorous mammals. Ann. Rev. Earth Plan. Sci. 27: 463:493.

Van Valkenburgh, B. Grady, F. and B. Kurten. 1990. The Plio-Pleistocene cheetah-like cat Miracinonyx inexpectatus of North America. J. Vert. Paleo. 10: 434-454.

Van Valkenburgh, B. and C.B. Ruff. 1987. Canine tooth strength and killing behavior in large carnivores. J. Zool. 212: 379-397.

Van Valkenburgh, B., X. M. Wang, and J. Damuth. 2004. Cope's rule, hypercarnivory, and extinction in North American canids. Science 306: 101-104.

Veron, G. 1995. La position systematique de Cryptoprocta ferox (Carnivora). Analyse cladistique de carateres morphologiques de carnivores Aeluroidea actuels et fossils. Mammalia. 59: 551-582.

Veron, G. and F.M. Catzeflis. 1993. Phylogenetic relationships of the endemic Malagasy carnivore Cryptoprocta ferox (Aeluroidea): DNA/DNA hybridization experiments. J. Mamm. Evol. 1: 169-185.

Veron, G. and S. Heard. 1999. Molecular systematics of the Asiatic Viverridae (Carnivora) inferred from mitochondrial cytochrome b sequence analysis. J. Zool. Syst. and Evol. Res. 38: 209-217

Veron, G., Colyn, M., Dunham, A.E., Taylor, P., and P. Gaubert. 2004. Molecular systematics and origin of in mongooses (Herpestidae, Carnivora). Mol. Phylogenet. Evol. 30: 582-598.

Viret, J. 1929. Les faunes des mammiferes de l’Oligocene superieur de la Limagne Bourbonnaise. Ann. L’Univ. Lyon Ser. 47: 1-328.

137

Wagner, G. P., and G. B. Muller. 2002. Evolutionary innovations overcome ancestral constraints: a re-examination of character evolution in male sepsid flies (Diptera: Sepsidae). Evol. and Dev. 4: 1-6.

Wagner, G. P., and K. Schwenk. 2000. Evolutionarily stable configurations: functional integration and the evolution of phenotypic stability. Evol. Biol. (vol. 31.) Pp. 155-217

Wang, X. 1994. Phylogenetic systematics of the Hesperocyoninae (Carnivora: Canidae). Bull. Am. Mus. Nat. Hist. 221: 1-207.

Wang, X., Tedford, R.H., and B.E. Taylor. 1999. Phylogenetic systematics of the Borophaginae (Carnivora: Canidae). Bull. Am. Mus. Nat. Hist. 243: 1-391.

Warheit, K. I., J. D. Forman, J. B. Losos, and D. B. Miles. 1999. Morphological diversification and adaptive radiation: a comparison of two diverse lizard clades. Evolution 53: 1226- 1234.

Weisrock, D.W., Shaffer, H.B., Storz, B.L., Storz, S.R., and S.R. Voss. 2006. Multiple nuclear gene sequences identify phylogenetic species boundaries in the rapidly radiating clade of Mexican ambystomatid salamanders. Mol. Ecol. 15: 2489-2503.

Werdelin, L. 1983. Morphological patterns in the skulls of cats. Biol. J. Linn. Soc. 19: 375-391.

———. 1985. Small Pleistocene felines of North America. J. Vert. Paleo. 5(3): 194-210.

———. 1987. Supernumerary teeth in Lynx lynx and the irreversibility of evolution. J. Zool. 211: 259-266.

———. 1989. Constraint and adaptation in the bone-cracking canid Osteoborus (Mammalia: Canidae). Paleobiology 15: 387-401.

———. 1996. Carnivoran Ecomorphology: a phylogenetic perspective In: J. L. Gittleman, (ed). Carnivore behavior, ecology and evolution, Vol II. Cornell Univ. Press, Ithaca.

Werdelin, L., and N. Solounias. 1991. The Hyaenidae: taxonomy, systematics and evolution. Fossils and Strata 30: 1-104.

Werdelin, L. and N. Solounias. 1996a. The evolutionary history of hyaenas in Europe and Western Asia during the Miocene. In: Bernor, R.L., Rietschel, S. and W. Mittmann, (eds.). The evolution of the Western Eurasian Miocene mammal faunas. Columbia Univ. Press, New York. Pp. 290-306.

———. 1996b. Carnivores, exclusive of Hyaenidae, from the later Miocene of Europe and Western Asia. In: Bernor, R.L., Rietschel, S. and W. Mittmann, (eds.). The evolution of the Western Eurasian Miocene mammal faunas. Columbia Univ. Press, New York. Pp. 271-289.

138

Wesley-Hunt, G.D. and J.J. Flynn. 2005. Phylogeny of the Carnivora: Basal relationships among the carnivoramorphans, and assessment of the position of “Miacoidea” relative to Carnivora. J. Syst. Palae. 3(1): 1-28.

Wiegman, B.M., Mitter, C. and B. Farrell. 1993. Diversification of carnivorous parasitic insects – extraordinary radiation or specialized dead end. Am. Nat. 142: 737-754.

Wiens, J. J. 1999. Phylogenetic evidence for multiple losses of a sexually selected character in phrynosomatid lizards. Proc. R. Soc. Lond. B 266: 1529-1535.

Wiens, J.J. and T.W. Reeder. 1995. Combining datasets with different numbers of taxa for phylogenetic analysis. Syst. Biol. 44: 548-558.

Wilgenbusch J.C., Warren D.L., Swofford D.L. 2004. AWTY: A system for graphical exploration of MCMC convergence in Bayesian phylogenetic inference. http://ceb.csit.fsu.edu/awty

Wills, M. A., D. E. G. Briggs, and R. A. Fortey. 1994. Disparity as an evolutionary index - a comparison of and Recent arthropods. Paleobiology 20: 93-130.

Wozencraft, W.C. 1984. A phylogenetic reappraisal of the Viverridae and its relationship to other Carnivora. Ph.D. dissertation. Univ. Kansas, Lawrence.

———. 1989. The phylogeny of the recent Carnivora. In: Gittleman, J.L., (ed.). Carnivore behavior, Ecol. and evolution. Cornell Univ. Press, Ithaca. Pp. 495-535.

———. 2005. In Wilson, D. E., and D. M. Reeder (eds). Mammal Species of the World. Johns Hopkins University Press. Pp. 532-572.

Wyss, A.R., and J.J. Flynn. 1993. A phylogenetic analysis and definition of the Carnivora. In: Szalay, M., Novacek, M., McKenna, M., (eds.). Mammal Phylogeny, Placentals (Vol. 2). Springer-Verlag, New York. Pp. 32-52.

Yang, Z. 1998. On the best evolutionary rate for phylogenetic analysis. Syst. Biol. 47(1): 125- 133.

Yu, L., and Y. Zhang. 2005. Phylogenetic studies of pantherine cats (Felidae) based on multiple genes, with novel application of nuclear b-fibrinogen intron 7 to carnivores. Mol. Phylogenet. Evol. 35: 483-495.

Zwickl, D. and D.M. Hillis. 2002. Increased taxon sampling greatly reduces phylogenetic error. Syst. Biol. 51(4): 588-598.

139

BIOGRAPHICAL SKETCH

Jill Holliday was born in Orlando, Florida, on November 26, 1968. She grew up in Lakeland, Florida, where she received her undergraduate degree in Biology from Florida Southern College in 1998. Jill entered graduate school at Florida State University in Tallahassee, Florida in the fall of 1999, where she pursued a Ph.D. in Biology with a focus on Ecology and Evolution. Jill has two children, Alexandra and Benjamin.

140