Novitatesamerican MUSEUM PUBLISHED by the AMERICAN MUSEUM of NATURAL HISTORY CENTRAL PARK WEST at 79TH STREET, NEW YORK, N.Y

NovitatesAMERICAN MUSEUM PUBLISHED BY THE AMERICAN MUSEUM OF NATURAL HISTORY CENTRAL PARK WEST AT 79TH STREET, NEW YORK, N.Y. 10024 Number 2893, pp. 1-85, figs. 1-19, tables 1-6 October 30, 1987

The Interrelationships of Higher Ruminant Families with Special Emphasis on the Members of the Cervoidea

CHRISTINE M. JANIS' AND KATHLEEN M. SCOTT2

CONTENTS Abstract ...... 2 Introduction ...... 2 Acknowledgments ...... 3 Abbreviations ...... 3 Historical Views of Living Ruminant Relationships ...... 4 Characters Defining Major Groups within the Ruminantia ...... 8 Ruminantia ...... 8 Ruminantia Above Level of Tragulidae ...... 9 Pecora ...... 9 Relevant Characters in Pecoran Phylogeny ...... 10 Cranial Appendages ...... 10 Other Skull Characters ...... 21 General Characters of the Dentition ...... 25 Characters of the Upper Cheek Teeth ...... 27 Characters of the Lower Cheek Teeth ...... 31 Postcranial Characters ...... 38 Features of the Soft Anatomy ...... 45 Defining Characters of Living Pecoran Families ...... 47 Bovidae ...... 47

' Assistant Professor, Division of Biology and Medicine (Program in Ecology and Evolutionary Biology), Brown University, Providence, Rhode Island 02912. 2 Research Associate, Department of Mammalogy, American Museum of Natural History, Associate Professor, Department of Biological Sciences, Rutgers University, Piscataway, New Jersey 08855.

Giraffidae ...... 47 Cervidae ...... 48 Moschidae ...... 49 Antilocapridae. 49 Phylogenetic Relationships of Living Pecoran Families ...... 50 Defining Characters of Fossil Families ...... 51 Gelocidae ...... 51 Palaeomerycidae and Lagomerycidae ...... 52 Dromomerycidae. 56 Hoplitomerycidae. 56 Problematical Ruminant Genera and Families ...... 57 Walangania ...... 57 "Gelocus" whitworthi ...... 59 Propalaeoryx ...... 60 Amphitragulus and Dremotherium ...... 61 Moschidae ...... 64 Palaeomeryx and the Palaeomerycidae ...... 66 Zarafa and Prolibytherium ...... 71 Antilocapridae. 73 Oligocene and Early Miocene "Bovids" ...... 74 Conclusions ...... 75 References Cited ...... 80

ABSTRACT We analyze the interrelationships of the higher ycids. The Cervoidea includes the primitive extinct (Pecoran) ruminants, and suggest possible rela- Eurasian genera Eumeryx and Rutitherium, and tionships between these families and the various (more closely related to the other cervoids) the genera of the polyphyletic assemblage "Geloci- extinct African genus Walangania. The grouping dae." We also review the developmental processes Eucervoidea is proposed for a clade within the of the cranial appendages of the living homed ru- Cervoidea containing the Antilocapridae, the Cer- minant families, and conclude that giraffid ossi- vidae, and the extinct families Palaeomerycidae cones, bovid horns, and cervid antlers cannot be and Hoplitomerycidae (which are deemed as clos- considered to be homologous with each other. The er to the Cervidae than are the Antilocapridae). characters that have been used in the past and in The Palaeomerycidae contains the Old World gen- this paper to distinguish pecoran families are dis- era Palaeomeryx, Amphitragulus, possibly also cussed and evaluated. Prolibytherium, and the North American drom- Within living pecoran families the Giraffidae are omerycids. The Hoplitomerycidae contains the the most primitive, and the Moschidae and An- European genera Hoplitomeryx and Amphimos- tilocapridae are conjoined with the Cervidae in chus. The European genus Triceromeryx remains the superfamily Cervoidea, with antilocaprids as cervoid incertae sedis. A superfamily Giraffo- being closer to cervids than are moschids. The idea is proposed to include the Giraffidae, the ex- Moschidae includes Moschus, the extinct Euro- tinct family Climacoceridae, and possibly also the pean genera Dremotherium, Micromeryx, and His- extinct African genus Propalaeoryx. panomeryx, and the North American blastomer-

INTRODUCTION The Ruminantia is a suborder ofthe order loids, and the more derived pecorans, or Artiodactyla, or even-toed ungulates, and in "higher ruminants," which are characterized Neogene and Recent times has represented by larger body size and possession of cranial the most ecologically and geographically di- appendages (i.e., horns or antlers) in most verse group of living ungulates. Ruminants living species. Traguloids were more diverse are divided into the more primitive tragu- during the early Tertiary (Eocene and Oliogo- 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 3 cene), and today only one family, the Tra- but rather suggest that it should serve as a gulidae (mouse deer or chevrotains), survives hypothetical scheme which can be tested by in the tropical regions of the Old World. In the examination of other characters, either contrast, the Pecora had their main evolu- morphological (e.g., basicranial characters, tionary radiation in the late Neogene, and which we discuss in this paper only in the today comprise between three and five fam- context of the work of Webb and Taylor, ilies (depending on taxonomic opinion), al- 1980) or biochemical (for the interrelation- though a number oforiginal families are now ships of living families). extinct. Today, the families Cervidae (deer) and Bovidae (antelope, sheep, and cattle) rep- ACKNOWLEDGMENTS resent the greatest degree of taxonomic and We would like to thank the following peo- geographical diversity among the Pecora. De- ple and institutions for access to fossil spec- spite the evolutionary success and ecological imens in their collections: Dr. B. Engasser at importance of the Pecora, the subject of the the Naturhistorisches Museum, Basel, Swit- interrelationships of the living families, and zerland; the curators at the Universite Claude of the relationship of fossil taxa to living Bernard and the Musee Gimee, Lyon, France; forms, has remained little studied, and what Dr. V. Eisenmann at the Museum National is known about the subject remains an area d'Histoire Naturelle, Paris, France; Dr. A. of controversy and confusion. Gentry, Dr. J. Hooker, and Dr. J. Clutton- The standard opinion on ruminant evo- Brock at the British Museum ofNatural His- lution for the last few decades has been that tory, London, England; Dr. M. Novacek, Dr. the extinct Gelocidae are the basal group from R. Tedford, and Dr. G. Musser at the Amer- which the families of homed ruminants, i.e., ican Museum ofNatural History, New York, the pecoran ruminants, can be derived U.S.A.; and Dr. M. Rutzmoser and Dr. R. (Simpson, 1945; Viret, 1961; Romer, 1966). Honeycutt at the Museum of Comparative More recently, Webb and Taylor (1980) have Zoology, Harvard University, U.S.A. proposed that the gelocids should be included We appreciate discussion on issues raised in the Pecora, comprising the sister group to in this paper with the following people; Dr. the homed ruminants or "Eupecora," with M. Brunet, Dr. B. Engasser, Dr. J. Leinders, the implication that the presence ofhorns, or Dr. A. Lister, Dr. Y. Jehenne, Dr. N. So- other forms of cranial appendages, is a syn- lounias, and Dr. S. D. Webb. Comments on apomorphy uniting the Eupecoran families. the manuscript were made by Dr. R. Tedford, However, a number of recent papers have Dr. P. Grubb, Dr. S. D. Webb, and Dr. A. shown that horns have evolved more than W. Gentry. once within the Ruminantia (e.g., Janis, 1982; Thanks are also due to Larissa Bilaniuk Leinders, 1983; Scott and Janis, 1987). and Marjorie Thompson for the illustrations, Moreover, recent work by Janis (1987) has to Lyuba Konopasek and Cyril Simsa for bib- shown that the gelocids represent a polyphy- liographic assistance, and to Loren Mitchell letic assemblage rather than a distinct clade. for typing the original manuscript. Therefore the origin ofhomed ruminants and This study was supported in part by a Rut- their interrelationships must be sought among the different genera within the gelocid assem- gers Junior Faculty Fellowship and N.S.F. blage. grant no. BSR-8418349 to KMS, and Brown This paper is a preliminary analysis of the University Biomedical Support Grant characters which can be used to define the RR0785-19 and N.S.F. grant no. BR-8418148 higher ruminant families and to ally them to CMJ. with taxa within the gelocid assemblage, and of the probable relationships among the ru- ABBREVIATIONS minant taxa. It is our intention to use a num- AMNH Department of Vertebrate Paleon- ber of limb, dental, and cranial characters to tology, American Museum of Nat- construct a possible cladogram for the inter- ural History relationships of the Pecora. We do not see AMNH(M) American Museum of Natural His- this proposed cladogram as being immutable, tory, Modern Mammals 4 AMERICAN MUSEUM NOVITATES NO. 2893

BMNH Department of Paleontology, Brit- and poorly defined omasum oftragulids) (see ish Museum of Natural History. Webb and Taylor, 1980; Scott and Janis, F:AM Frick American Mammals, Depart- 1987). ment of Vertebrate Paleontology, The five living families have generally been American Museum of Natural His- characterized by the following distribution of tory MCZ Museum of Comparative Zoology, characters. Harvard University CERVIDAE: Deciduous, branched cranial ap- Ph Naturhistorisches Museum, Basel pendages without a keratin cover (antlers); PQN South African Museum (Cape Town) brachydont (low crowned) to mesodont UCBL Universite Claude Bernard, Lyon (moderately high crowned) cheek teeth; limbs without pronounced elongation of HISTORICAL VIEWS OF LIVING the metapodials, with partial retention of RUMINANT RELATIONSHIPS metapodials II and V (i.e., the "side toes"). MOSCHIDAE: Absence of cranial appendages; Researchers in the past century have rec- three and five living families brachydont to mesodont cheek teeth; mod- ognized between erately elongated metapodials with reten- of pecoran ruminants. Cervids, bovids, and tion of the side toes. giraffids have always been accorded family GIRAFFIDAE: Nondeciduous, unbranched status. Antilocaprids have usually been ac- cranial appendages, covered with skin (os- corded status as the Antilocapridae family sicones); brachydont cheek teeth; elongat- (e.g., Simpson, 1945), although some workers have placed them within the Bovidae (e.g., ed metapodials with complete loss of the O'Gara and Matson, 1975). Moschids (rep- side toes. resented by the living musk deer species of BOVIDAE: Nondeciduous, unbranched cranial appendages, covered by a nondeciduous, the genus Moschus) have traditionally been unbranched keratin sheath meso- regarded as a subfamily Moschinae within (horns); dont to hypsodont (high crowned) cheek the Cervidae (e.g., Simpson, 1945), but a number ofauthors (e.g., Gray, 1821; Brooke, teeth; metapodials may or may not be elon- 1878; Flerov, 1952; Bubenik, 1966; Webb gated, but side toes are invariably absent. ANTILOCAPRIDAE: Nondeciduous, un- and Taylor, 1980; and Groves and Grubb, branched cranial appendages, covered by a 1987) have suggested that Moschus should be deciduous forked keratin sheath (horns); placed in its own family within the superfamily Cervoidea. hypsodont cheek teeth; elongated metapodials with loss of the side toes. The five living families are clearly distin- guishable from each other on the basis of Hydropotes, the only antlerless cervid, can characters of the cranial appendages, limbs, be allied with the Cervidae on the basis of and dentition. They are unified as "higher soft anatomy features, detailed in the next ruminants," or pecorans, distinguishing them section, which are lacking in Moschus. Mos- from tragulids by the following characters: chus also has a less convoluted brain and a possession ofcranial appendages, usually only less complicated omasum than is typical of in the male of the species (although cranial both cervids and bovids (i.e., resembling the appendages are lacking both in Moschus and more primitive pecoran condition seen in gi- the cervid genus Hydropotes); fully seleno- raffids) (Garrod, 1877). Moschus has gener- dont cheek teeth with the absence ofa lingual ally been allied with the Cervidae, if not in- cingulum in the upper molars (as opposed to cluded in the family, on the basis of the bunoselenodont cheek teeth with a lingual brachydont dentition, retention of the side cingulum in tragulids); elongated, fused toes, and possession of a saberlike upper ca- metapodials possessing complete distal artic- nine in males, resembling the condition seen ular keels; a compact, parallel-sided astrag- in Hydropotes, but absent in any other living alus; and a four-chambered, ruminant type pecoran genera. Webb and Taylor (1980) have ofstomach, with a fully formed omasum and shown that Moschus can be grouped with oth- a large rumen (in contrast to the small rumen er fossil pecoran genera into the family Mos- 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 5 chidae on the basis of autapomorphic fea- Leinders (1979) and Leinders and Heintz tures. They consider the Moschidae to be the (1980) have pointed out two morphological sister group to the other living pecorans, which features of Antilocapra which they consider they place in the Eupecora. They consider the link the Antilocapridae more closely with the Eupecora to be united by the presence offron- Cervidae than the Bovidae; these are: the tal appendages, reduced or absent upper ca- presence ofa closed metatarsal gulley (see fig. nines, and loss of the subarcuate fossa. 7) and the possession oftwo lacrimal orifices Antilocaprids have been allied with the situated on the orbital rim (see fig. 4). These Bovidae on the basis of the similarity of the similarities led them to conclude that the horns, despite the difference ofthe deciduous, horns of Antilocapra had a separate evolu- forked horn sheath as opposed to the non- tionary origin from bovid horns, although deciduous, unbranched sheath in the bovids they do not regard them as homologous with (see O'Gara and Matson, 1975, and later dis- cervid antlers, but assume that they represent cussion in this paper, for discussion of peri- an independent evolutionary event within the odic flaking ofthe bovid horn sheath). O'Gara Antilocapridae. They also point out that the and Matson (1975) also ally Antilocapra with genus Moschus, while resembling cervids in the Bovidae based on possession of hypso- the presence of a closed metatarsal gulley, is dont cheek teeth, long limbs with the com- less like cervids than Antilocapra in the re- plete loss of the side toes, and retention of tention ofthe primitive pecoran condition of the gall bladder. a single lacrimal orifice situated within the Giraffids have traditionally been allied with orbit on the ventral anterior orbital wall. Fig- cervids in the superfamily Cervoidea (Simp- ure 1 summarizes the different evolutionary son, 1945; Romer, 1966; Thenius, 1969) be- viewpoints held over the past few decades of cause both groups share possession of the interrelationships of the living pecoran brachydont cheek teeth with many accessory families. styles and ribs, in contrast to the more sim- The cladograms presented in figure 1 il- plified, hypsodont cheek teeth of bovids and lustrate the differential value placed on var- antilocaprids. Hamilton (1978a) has restated ious character states in assessing the inter- the significance ofthe autapomorphic feature relationships between living pecoran families. of the Giraffidae of the possession of a bi- (The numerical characters listed on the lobed lower canine. cladograms are consistent throughout the text More recently, various workers have point- and the figures in this paper, and are refer- ed out that the different types of cranial ap- enced in table 1.) The "traditional" view (fig. pendages in ruminants are not homologous lA) regards the characters ofthe height ofthe in their mode ofdevelopment. Cervid antlers cheek teeth (character 12) and the condition are formed from an outgrowth of the frontal of the metapodials (character 29) as being bone, whereas bovid horns and giraffid os- critical in determining family relationships. sicones are said to be formed by the fusion Cranial appendages are considered a synap- of separate dermal ossification centers with omorphy linking all pecoran families, imply- the skull, although the developmental con- ing that they were secondarily lost in the gen- dition in the Antilocapridae has not been de- era Moschus and Hydropotes. termined (see review in Goss, 1983, and dis- Both present-day views (fig. 1B, 1C) rec- cussion later in this paper). Frechkop (1955) ognize the fact that brachydont cheek teeth and Hamilton (1978a) have regarded this and retention of the side toes are primitive similarity between bovids and giraffids as a characters for the Ruminantia (indeed for all synapomorphy uniting these two families, a tetrapods!), and cannot be used as synapo- view which appears to be generally accepted morphies to unite families within the Pecora. at the present time. However, Todd (1975) They also both acknowledge the fact that cer- presents chromosomal evidence to suggest vid antlers are not homologous with the cra- that the Giraffidae are more primitive than nial appendages -of the other pecoran fami- any other pecoran family (but see discussion lies, and thus that cranial appendages cannot of this paper in Scott and Janis, 1987). be held to be a primitive pecoran character. 6 AMERICAN MUSEUM NOVITATES NO. 2893

TABLE 1 Characters Used in Text and Cladograms 7h. Cranial appendages supposedly secondarily lost 1. "Ruminant" character suite: 7i. Presence of median occipital cranial ap- Fusion of cuboid and navicular in tarsus pendages Upper incisors reduced or lost 8. Presence of lacrimal fossa Lower canine incisiform 9. Double lacrimal orifice on dorsal rim of orbit (Latter two characters also evolved in parallel 10. Laterally enclosed, subcentral tympanohyal va- within the Protoceratidae) gina: 1 a. Suite of characters uniting members of Ruminan- 1Oa. Partially formed tia above the level of the Hypertragulidae: 10b. Fully formed Amastoid condition of skull 11. Upper canine characteristics: Jugular foraman confluent with posterior lac- 1 la. Moderately elongated ("traguloid" type) erate foramen 1 lb. Sabrelike ("moschid" type) Loss of trapezium 1 lc. Secondarily reduced or lost Fusion of magnum and trapezoid 12. Crown height of cheek teeth: Metacarpal I lost 12a. Brachydont Reduction of fibula to malleolar bone 12b. Hypsodont P1 lost 13. Characters of cingulum on upper molars: Mesostyles in upper molars and ectostylids in 13a. Present lower molars 13b. Reduced "Dorcatherium fold" in lower molars 13c. Absent 2a. "Higher ruminant" features (uniting ruminants 14. Condition of protocone on P3: above the level of the Tragulidae): 14a. Posteriorly situated and directed Postorbital bar formed from frontal bone 14b. Centrally situated and lingually directed Axis with odontoid process possessing high dor- 15. Presence ofentostyle formed from anterior wall of sal crest with high anterior articulatory sur- metaconule: faces 15a. Small, incipient entostyle Calcaneum with fibular facet with large proxi- 15b. More prominent entostyle mal concavity and small distal convexity 16. Characteristics of metastyle: 2b. Anterior cingulum present on lower molars 16a. Large 3. Lower premolars with small metaconid, without 16b. Small posterior extension of metaconid forming 17. Characteristics of P4 metacone: posterolingual wall to tooth 17a. Large 4. Characters of p1: 17b. Small 4a. pl large and caniniform 18. Characteristics of M3 metaconule: 4b. pl small and premolariform, separated 18a. Small from p2 by small diastema 18b. Large 4c. p1 lost 19a. Bifurcated posterior wing of protocone on molars 5. Compact, parallel-sided astragalus 19b. Bifurcated posterior wing ofmetaconule on molars 6. Pecoran basicranial character suite: 20. Presence of "Palaeomeryx fold" on molars of Loss of promontorium on petrosal brachydont taxa Enlarged fossa for stapedial muscle on petrosal 21. Metastylids on lower molars Shallow subarcuate fossa on petrosal 22. Condition of postentocristid on molars: Broadened basioccipital with strong flexion stops 22a. Incomplete on condyles 22b. Complete Laterally enclosed postglenoid foramen 23. Bilobed lower canine 7. Characters of cranial appendages: 24. Paraconid on p4 forming lingual wall to tooth 7a. Unbranched ossicones preformed in car- 25. Vertical groove on posterolingual region of p4 tilage (nondeciduous) 26. Double posterior lobe on m3: 7b. Unbranched horns (nondeciduous) with 26a. Closed posteriorly unbranched (?) nondeciduous keratin 26b. Open posteriorly sheath 27. Complete distal metapodial keels 7c. Branched or unbranched nondeciduous 28. Fusion of metapodials: appendages with deciduous branched 28a. Fused with open gully or unbranched keratin sheath 28b. Fused with closed gully 7d. Branched deciduous antlers on top ofnon- 28c. Secondarily open gully (from closed gully) deciduous pedicle 28d. Metapodial fused and elongated over trag- 7e. Branched appendages of uncertain devel- uloid condition. (Metacarpals similar opmental origin (nondeciduous) length to metatarsals) 7f. Unbranched appendages of uncertain de- 29. Condition of side toes: velopmental origin (nondeciduous) 29a. Complete side toes retained 7g. Appendages formed with fusion ofdermal 29b. Side toes partially lost (proximal or distal elements with skull ends) 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 7

TABLE 1-(Continued) sister taxon to the Cervidae, relegating Mos- chus to the position of the sister taxon to the 29c. Side toes completely lost, metapodials fur- Cervidae plus the Antilocapridae. They con- ther elongated 30. Pattern of fusion of remnants of lateral metatar- sider that the derived features that Antilo- sals: capra shares with the Bovidae, such as the 30a. Metatarsal II remnant fused proximally elongation ofthe metapodials, the loss ofthe 30b. Metatarsal V remnant fused proximally side toes, and the hypsodont cheek teeth, were 31. Posterior tuberosity on metatarsus derived in parallel as the result of a similar 32. Condition of cubonavicular facet on proximal metatarsus: evolutionary ecological preference for an open 32a. High and pointed habitat and a fibrous diet. (The development 32b. Very flat and broad of similar characters can be seen in the evo- 32c. Somewhat flattened lution of the Camelidae and the Equidae). 32d. Raised, but not as sharply as in 32a 33. Stomach characters: They also point out that the retention of the 33a. Three-chambered stomach with rumen gall bladder (see O'Gara and Matson, 1975) 33b. Four-chambered stomach with omasum is a primitive character that cannot be used 34. "Cervid" features of soft anatomy, especially ab- to specifically unite Antilocapra with the sence of gall bladder Bovidae, or to exclude the genus from the position of the sister taxon to the Cervidae, and consider that the cranial appendages of By implication, pecoran cranial appendages Antilocapra evolved in parallel with bovid must have evolved in parallel at least once, horns. While Leinders and Heintz consider and the genera Moschus and Hydropotes may that the Bovidae and the Giraffidae may be be assumed to be primarily without cranial united by the synapomorphies ofa metatarsal appendages. with an open gulley, and the development of The view expressed in figure B relies cranial appendages from dermal ossicones, heavily on the assumption that the similarity they point out that this condition ofthe meta- in the mode offormation ofbovid horns and tarsals is probably the primitive condition for giraffid ossicones is a true synapomorphy. the Pecora (as exemplified by the fossil prim- Hamilton (1 978a) did not specifically address itive pecoran genus Gelocus). Further, since the problem of the position of the Antilo- there are obviously only two ways to form a capridae and, as previously mentioned, the cranial appendage (i.e., outgrowth from the mode of formation of their horns is as yet frontal bone, or fusion ofa dermal ossicone), unknown. However, other recent workers similarity in development may not necessar- (e.g., O'Gara and Matson, 1975) have con- ily be indicative of a close relationship. sidered that the similarities between Antilo- Finally, the view presented by Webb and capra and bovids justify the consideration of Taylor (1980) (fig. 1D) provided a basis for Antilocapra as at least the sister taxon to the the definition of the discrete characters that Bovidae, although not all present-day pro- characterize all the ruminants, and also those ponents ofthis view would go so far as O'Gara which characterize the Pecora within the and Matson in placing Antilocapra within the Ruminantia. They also showed that a num- Bovidae. In this scheme, the Antilocapridae ber of fossil taxa of previously dubious taxo- and the Bovidae are usually united in the nomic affinities could be included in the liv- superfamily Bovoidea (which may some- ing family Moschidae on the basis of times be expanded to include the Giraffidae). apomorphic characters of the auditory bulla The view of Leinders and Heintz (1980), (character 10). However, we disagree with represented by figure 1C, takes into account their assertion that the presence of cranial a cervid feature originally noted by Milne- appendages can be used as a synapomorphy Edwards (1864), a double lacrimal orifice on to unite the other pecoran ruminant families the orbital rim (character 9); and a cervid into the "Eupecora," and while we are in feature recognized by Heintz (1963, 1970), a agreement with their inclusion of taxa into closed metatarsal gulley (character 28b). the family Moschidae, do not consider the Leinders and Heintz point out that these fea- Moschidae to represent the sister-group to tures can be used to place Antilocapra as the the other pecoran ruminants. 8 AMERICAN MUSEUM NOVITATES NO. 2893

PECORA PECORA A : CERVOIDEA B --. BOVOIDEA BOVOIDEA CERVOIDEA

PECORA PECORA C BOVOIDEA CERVOIDEA v DUEO

\q)p'op g/294* 128 34

7o12o7A 1

\H2728A,D /

Fig. 1. Historical views on the phylogenetic position of the different living pecoran families. See table 1 for characters. A. "Traditional view," espoused by Stirton (1944) and Romer (1966). B. Present- day view, espoused by Hamilton (1 978a). C. Alternative present-day view ofLeinders and Heintz (1980). D. View of Webb and Taylor (1980).

Our interest in ruminant taxonomy stems this paper to initially describe and evaluate not only from a wish to clarify the interre- the various characters that have been used lationships ofthe living pecoran families, but by different authors in pecoran systematics, also to understand the relationships of fossil and to discuss additional characters stem- genera and families within this scheme, and ming from our own observations which we to ascertain which fossil taxa may represent feel are of classificatory importance. the sister taxa to the living and fossil families. We feel that an evaluation ofthe status ofthe CHARACTERS DEFINING character states held by various authors to be MAJOR GROUPS WITHIN synapomorphies linking various pecoran lin- THE RUMINANTIA eages can only be obtained by an examination RUMINANTIA of the distribution of these states in both living and fossil taxa, and that any scheme which The defining characters which unite the proposes to satisfactorily classify living pec- Ruminantia are the fusion of the cuboid and orans in a cladistic fashion must also be ame- navicular bones in the tarsus, and the pres- nable to the classification of fossil genera ence of an incisiform lower canine (although within the same parameters. We propose in an incisiform lower canine was evolved in 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 9 parallel in the Protoceratidae, the pecoran- molars (character 2b, see fig. 5). Two addi- mimicking tylopod family; Patton and Tay- tional characters can unite Lophiomeryx and lor, 1973). Ruminants are also characterized related genera (Cryptomeryx and Iberome- by the reduction or loss ofthe upper incisors, ryx) with the Pecora: the reduction of the although upper incisors are present in early primitively caniniform pl, widely separated fossil taxa such as Archaeomeryx (Webb and from the other premolars, to a peglike pre- Taylor, 1980) (character suite 1 on clado- molariform tooth (when present), separated grams). The Ruminantia above the level of from p2 by only a small diastema (character the Hypertragulidae (the most primitive ru- 4b); and the advanced ruminant type of p4, minant family) can be distinguished by the in which the metaconid is small, and does following suite of characters (character suite not extend backward to form a posterolingual la on the cladograms): amastoid condition wall to the tooth (character 3) (see Janis, of the skull; jugular foramen confluent with 1987). the posterior lacerate foramen in the basicranium; loss ofthe trapezium and fusion ofthe PECORA magnum and trapezoid in the manus, with loss of metacarpal I; reduction of the fibula The Pecora (sensu Webb and Taylor, 1980) to a malleolar bone; upper P1 lost; presence are characterized postcranially by the loss of ofa mesostyle and (primitively) ofan internal the trapezium in the manus and the posses- cingulum in the upper molars; presence of an sion ofan astragalus with short, parallel sides ectostylid and (primitively) of a "Dorca- (character 5), as opposed to the elongated as- therium" fold in the lower molars (see figs. tragalus with nonparallel sides seen in the 5, 6) (Webb and Taylor, 1980; Janis, 1987). tragulids (Webb and Taylor, 1980). Janis (1987) notes a tragulid type of astragalus in material referred to the problematical trag- RUMINANTIA ABOVE LEVEL OF TRAGULIDAE ulid genera Lophiomeryx and Bachitherium. Webb and Taylor (1980) list a number of A problem with use of astragalus character- characters that unite the extinct Leptomer- istics is that a "pecoran" type of astragalus ycidae with the Pecora, distinguishing them has evidently been evolved in parallel in the from the living family Tragulidae (character supposedly leptomerycid genus Pseudopara- suite 2a on cladograms). Janis (1987) has blastomeryx (Taylor and Webb, 1976), and shown that these features occur in the known in Pseudoceras, a genus of uncertain taxo- material ofthe traguloid genera Bachitherium nomic affinities, possibly a "gelocid" (see and Lophiomeryx. These are: a complete Webb, 1983a). However, Janis (1987) points postorbital bar, composed primarily of the out that the former genus was a survivor in frontal bone (in contrast to the absence of a the late Miocene of North America, where postorbital bar in the Hypertragulidae and there would be strong environmental pres- the presence of a postorbital bar composed sure to develop a more cursorially adapted primarily of the jugal in the Tragulidae); an type of hind limb in parallel with the peco- axis with an odontoid process possessing a rans. The same argument would also apply dorsal crest with high anterior articulatory to Pseudoceras, should this taxon prove to be surfaces (in contrast to the peglike odontoid nonpecoran on other morphological criteria. process in tragulids); and a calcaneum in Characters ofthe petrosal that characterize which the fibular facet has a large proximal the Pecora, seen also in the most primitive convexity and a small distal concavity (as pecoran, Gelocus, are the loss of the prom- opposed to a simple concavity or convexity ontorium, an enlarged fossa for the stapedial in tragulids and hypertragulids; character suite muscle, and a shallow subarcuate fossa. Oth- 2a on cladograms). er basicranial features of the Pecora are a The traguloid genera Lophiomeryx and broadened basioccipital with strong flexion Bachitherium can be united with the Pecora, stops on the condyles, and a laterally enclosed above the level ofthe other traguloids, by the postglenoid foramen (Webb and Taylor, presence ofan anterior cingulum on the lower 1980) (character suite 6). 10 AMERICAN MUSEUM NOVITATES NO. 2893

RELEVANT CHARACTERS IN The bony horn core is frequently described PECORAN PHYLOGENY as developing from a dermal ossicone which fuses with the frontal bone (Frechkop, 1955; CRANIAL APPENDAGES Bubenik, 1966) without further explanation. Webb and Taylor (1980) assume the higher However, this description is somewhat mis- ruminant families to be united into a distinct leading as it implies that at some point in the clade, the Eupecora, characterized by the pos- development of bovid horns a separate os- session ofcranial appendages, or hornlike or- sicone (or os cornu) is present and should gans (character 7), and regard the Moschidae be observable, as is known to be the case in (comprising the living hornless genus Mos- living giraffids. A number of authors have chus, and the fossil genera Blastomeryx, Am- indeed reported the presence of such an os- phitragulus, and Dremotherium) as the sister- sicone in domestic sheep, goats, and cattle group of the Eupecora. Earlier authors such (see review in Dove, 1935). However, Dove as Pilgrim (1941, 1947) also regarded cranial (1935) clearly demonstrates that although appendages as homologous within the Pecora supraperiosteal tissues induce and contribute and considered that the skin-covered ossi- to horn core formation, no separate ossicone cones of giraffids represented the plesiomor- is normally found in bovid horn develop- phous condition for higher ruminants (see ment. also Coope, 1968; Hamilton, 1978a). How- Dove (1935) performed a series of tissue ever, as will be discussed below, there is clear transplants from the horn site of newborn developmental and paleontological evidence calves and kids. He found that development that cranial appendages have evolved at least of a horn core is induced by the connective twice among the ruminants, and possibly in tissue and dermis superficial to the perios- three or more instances. Possession ofcranial teum of the frontal bone in the horn bud appendages therefore cannot be regarded as (which he collectively terms the os cornu) and a synapomorphy which unites higher rumi- that the periosteum and frontal bone in this nants. region are unable to form even a raised boss Roth (1984) has pointed out that the shar- without the inductive action of the os cornu. ing of common developmental pathways, On the other hand, the conective tissue and which presumably are governed by the same dermis will induce horn formation ifthey are batteries ofgenes, is a necessary condition for transplanted elsewhere on the skull; trans- homology. If cranial appendages are to be planted periosteum from the horn bud de- used in assessing relationships within rumi- velops into normal, flat frontal bone. nants, assumptions ofhomology must there- Early in development the os cornu (as yet fore be based on details of ontogeny as well unossified) causes the underlying periosteum as adult structure. One might take into ac- to break down and "draws up" the frontal count the suite of developmental characters bone to form the boss, or the base ofthe horn. which include mode of development, tissues The connective tissue and dermal portions contributing to the formation of the append- ofthe os cornu ossify to form the core proper ages, and tissues actually inducing formation at the same time that the boss forms from of the appendages, as well as shape, place- the frontal, and no demarcation or suture is ment on the skull, and the nature of the cov- seen in normal development. The entire horn ering. We propose here to review both the core becomes surrounded by a membrane formation of cranial appendages to empha- continuous with the periosteum, which Dove size the developmental differences, and to re- (1935) believes forms from the connective view the characteristics of the cranial ap- tissue layer. Dove further showed that where pendages in the living and fossil pecoran the os cornu does ossify separately it is the lineages. result of abnormal development in animals heterozygous for normal and polled horns, a DEVELOPMENT OF HORNS AND ANTLERS conclusion also reached by Fambach (1909). Bovidae Such separate ossification generally results in a "scur," a horn which does not fuse with the The cranial appendages of the Bovidae are skull. We have noted one case of naturally termed horns, and are postorbital in position. occurring bilateral scurs in a wild Cephalo- 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY I1I phus monticolor. In this species females are neath the periosteum is a layer which Diirst polymorphic for the presence of horns, sug- refers to as the "preosseous substance," pro- gesting that this individual was probably het- duced by the fibroelastica. The "preosseous erozygous for the horned/hornless alleles. substance" described by Diirst (see also Although Dove's (1935) study clearly es- Brandt, 1928) is clearly the osteoid tissue or tablishes that it is the supraperiosteal tissues prebone ofmodern histological terminology. which induce horn core formation, and which Osteoid tissue consists of the organic matrix ossify to form the core, he does not discuss of bone which is laid down as the first step further growth of the core itself. We feel that of osteogenesis (Ham and Cormack, 1979). it is important to describe this process, since Thus the location of the preosseous tissue in it is widely stated in the literature that both histological sections is a clear indication of horn cores (e.g., Solounias, 1981) and horn the region where new bone is being formed. sheaths (e.g., Webb, 1973) grow from the base. Deep to the preosseous substance is a layer In fact, neither the core nor the sheath grow ofnewly formed bone, and deep to that older, from the base. The bony core grows both cancellous bone. from the tip and by appositional growth over Since the processes which lead to horn for- the surface (Diurst, 1902a, 1902b, 1926), and mation are a slight modification of the pro- the sheath is continually produced over the cesses of normal growth of the skull bones, entire surface ofthe core (Goss, 1983; Diirst, we will review the mechanism by which the 1 902a). The long-standing but erroneous roofing bones of the skull grow and change contention that horn cores grow from the base shape. The roofing bones are formed by in- may in fact be based on Diirst's rather con- tramembranous ossification: that is, bone is fusing summary, which states that "horn cores deposited within a connective tissue layer, grow from the tip downwards," although he and not in preformed cartilage. Ossification clearly states (and describes) elsewhere that begins from a central point, and bone is added growth is from the tip. Bohlin's (1935) de- to this center only by appositional growth, scription of horn growth in the Miocene bo- which occurs in two places: at the edges, vid Miotragocerus may also have popular- through which the bone increases in size; and ized this belief. However, his statement is on the surface, through which the bone be- based solely on the fact that there is an area comes thicker and changes in shape can be at the base of each horn core which is rich in brought about. Interstitial growth does not foramina, and he believes that this indicates occur in intramembranous bone formation a zone of growth. He indicates the position (Ham and Cormack, 1979). Bone added at of this zone on a photograph of a Miotrago- the surface is produced directly under the cerus skull; it seems clear to us that what he periosteum, and is deposited in horizontal indicates is the boundary between the sheath- layers. These layers of bone may then be re- covered part ofthe core and the skin-covered worked or resorbed from their inner surface. boss or pedicle. Change in the angle ofcurvature ofthe bones Diirst's (1902a) detailed histological stud- of the skull is brought about by differential ies ofa growth series ofdomestic calves forms deposition over the outer surface and differ- the basis for the description presented here. ential resorbtion from the inner surface ofthe The 63 specimens that he described ranged bone (Ham and Cormack, 1979). in age from fetuses of4-5 months gestational The first sign ofthe appearance ofthe horn age to approximately 9 months; he also in- core is the establishment of a connection be- cluded histological comparisons of wild bo- tween the connective tissue layer and the vids of various species and ages. In the ear- preosseous substance, and a thickening ofthe liest fetuses the area destined to give rise to preosseous substance, which was observed in the horn presents a normal appearance. The a week-old calf. At about 4 weeks of age a overlying skin consists of an outer epidermis section of the area shows a mound-shaped and an inner dermis, and apparently contains elevation of the frontal, which is clearly the normally developing anlagen ofhair follicles. beginning of the horn core. Diirst notes at Between the skin and the underlying frontal this stage that the building of the horn core bone is a layer ofconnective tissue. The fron- now differs significantly from that ofthe frontal is bounded by an outer periosteum; be- tal in one important aspect: the surface la- 12 AMERICAN MUSEUM NOVITATES NO. 2893 mellae are now oriented vertically rather grow from the base. The process ofhorn core than horizontally. These lamellae ofdiffering growth appears to be a variation of that of orientation merge at the base of the devel- the other membranous bones ofthe skull, but oping horn core. involving ossification of supraperiosteal tis- Diirst describes the vertical lamellae as ini- sues. tially forming superficial to the periosteum Diirst's descriptions of horn core growth ofthe frontal (see illustrations in Diirst, 1 902a, are entirely consistent with our observations 1 902b, 1926), and gradually elongating. These on growth series of horn cores. If horns grew vertical lamellae then fuse with the horizon- from the base, the tip of the core would not tal lamellae of the frontal. Although Diirst's grow after the first appearance of the core, work preceded that of Dove (1935), we be- but rather would be pushed outward by bone lieve that what Diirst described is the ossi- being added proximal to it (see fig. 32 in So- fication of the dermal/connective tissue os lounias, 1981). One would expect that the tip cornu, and the simultaneous dissolution of of a young horn core would have the same the periosteum described by Dove. Dove's size and taper as that ofan adult core. Instead, (1935) transplantation studies clearly showed young horn cores are thicker at the tip and that the dermal portion ofthe os cornu formed not as finely tapered as adult cores, clearly the tip of the initial core, and the connective indicating that they are not equivalent struc- tissue portion formed the base of the spike, tures. with their inductive effect on the frontal caus- The permanent bony core is covered by a ing it to form the boss. Diirst's histological nondeciduous keratin sheath, which is formed descriptions are thus consistent with Dove's from the basal layers of the epidermis of the experimental results, as far as the two can be horn bud. Dove's (1935) transplantation ex- compared, since Dove did not describe later periments indicate that the epidermis does growth. not induce horn core formation; rather ker- Throughout the remainder of the growth atinization of the epidermis of the horn bud series described by Diirst (1902a) the preos- in induced by the underlying os cornu. How- seous substance is found on the surface and ever, both Diirst (1902a) and Brandt (1928) tip of the horn core, forming a continuous, found that histological changes occur in the cone-shaped layer over the surface of the epidermis of the horn anlagen before birth, growing horn core. New bone is thus added suggesting that the epidermis might have an appositionally in two places: at the tip (that inductive effect prenatally. In any case, this is at the edges of the vertically oriented la- epidermis is evidently "preprogrammed" to mellae), and on the surface of the growing become keratinized, as the os cornu is unable core. Diirst clearly states that the core grows to induce keratinization in epidermis trans- in length through deposition at the tip, and planted from other areas of the head. in thickness through deposition of new bone When the first clearly recognizable keratin- along the sides. As it is laid down the bone ous horn substance appears at 3-4 weeks is reworked internally, forming diploe and postnatally, it is a round disc situated above allowing expansion ofthe frontal sinuses into the tip of the growing horn core (Diirst, the core. 1902a). As the horn core beneath expands, Deposition of bone is intramembranous the original disc gradually grows downward and there is no preformation of the core in to cover the entire core (Diirst, 1902a). The cartilage (Diirst, 1902a; Brandt, 1928). The horn-producing basal epithelium is now a mechanism ofgrowth described by Diirst for cone-shaped layer, covering the entire horn the horn cores is consistent with the growth core and lying parallel to its surface. Succes- mechanisms of the membrane roofing bones sive layers of keratin are produced by this ofthe skull: appositional growth over the sur- epithelium over the entire core. Each new face and at the edges (here at the tip). Neither layer is internal to the last, producing a series the roofing bones of the skull (Ham and Cor- of nested keratin cones, with the oldest cone mack, 1979) nor the growing horn core (Diirst, on the outermost surface, at the tip (see fig. 1902a) exhibit interstitial growth. There is 2). The horn sheath thus does not grow from thus no mechanism by which horn cores can the base, although new growth would only be 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 13 visible at the base, and alterations in rate of growth at the base produce annual growth rings (Diirst, 1926). Although this sheath is never shed completely or on a regular basis, it may be subject to periodic flaking and renewal (exfoliation) in certain bovids, e.g., Bison, Cephalophus, Kobus, Oryx, Bos, Boselaphus, Rupicapra, and Antilope (O'Gara and Matson, 1975). This procedure would seem to involve flaking off, in most cases caused by rubbing, of the outermost layers of the sequentially laid down lamellar keratin. With respect to these phe- nomena, it is important to realize that although the keratin sheath is produced con- tinuously by the germinal layer as described above, two different types ofkeratinous tissue are produced, usually referred to as young Fig. 2. Growth of the keratinous horn sheath horn ("Jugendhorn") and permanent or adult in bovids. (Modified from Goss, 1983) horn ("Dauerhorn") (Diurst, 1902a, 1926; Fambach, 1909; George, 1956). Young horn is softer than permanent horn and the nature chanically difficult for the bovid horn to of the horny fibers of the intertubular horn branch (P. Grubb, in litt.); in order for a differs (George, 1956). It is produced until branched sheath to grow, it would have to be the animal is 4-6 months old (Diirst, 1902a; able to spread outward over the fork, which George, 1956). The layers ofyoung horn are, clearly is not possible for a sheath formed of course, located on the outer surface of the from inflexible keratin. However, it is pos- sheath, near the tip, and it gradually flakes sible for the core to split along all or part of off as the animal becomes older (George, its length, producing two parallel horns. Diirst 1956). This softer keratin tissue would seem (1926) figures the skull of an ibex with two to function as a flexible protective coating for entirely separate horn cores on one side, and the growing horn core, which can easily be the other horn double in the distal third. Diirst split as the core of permanent horn grows up (1926) was able to show experimentally that beneath it. Although the outer layers are re- such split cores could be produced by artifi- moved, exposing shiny new permanent ker- cially dividing the horn anlagen. Depending atin, this process never results in shedding of on the degree to which the parts of the an- the entire keratin sheath, nor is the horn ever lagen are separated, only part ofthe core may left without a functional sheath, as is the case be split, or four separate horns may be pro- in antilocaprids. "Exfoliation" in bovids is duced, as in the relatively common four- clearly growth related and is neither seasonal horned mutants ofdomestic sheep and, in its nor associated with rut-related hormonal most extreme form, in the four-horned an- changes, as it is in Antilocapra. telope Tetracerus quadricornis. In living bovids the sheath generally fol- Although true branching is unknown in any lows the shape of the underlying bony core, living bovid horn core or sheath, either as a but may extend considerably beyond the core, normal or abnormal condition, some authors as for example in the Caprini or musk oxen, have suggested that certain fossil species may Ovibos moschatus (Diirst, 1926). This growth have had a forked horn sheath. It has been beyond the core is largely produced at the tip noted for the fossil genera Miotragocerus (Diurst, 1902a). The permanent horn core and (Bohlin, 1935; Solounias, 1981), and Mesem- its sheath may be straight, curved, or spi- briportax (Gentry, 1974), that the shape of raled, but are never branched or forked. The the horn core is similar to that ofAntilocapra mechanism of growth of the horn core and americana. These authors have suggested that the horn sheath would appear to make it me- the horn sheath may also have been similarly 14 AMERICAN MUSEUM NOVITATES NO. 2893 forked and possibly deciduous. Development infolding ofthe covering skin, but the sheath ofsuch a forked sheath would require a mod- was apparently normal externally. In one case ification of the bovid sheath building mech- a separate "horn core," consisting ofan ovoid anism; in Antilocapra, the upper prong is al- core of poorly defined matrix surrounded by most fully formed before the cornification of concentric layers of horny material, was the anterior prong and proximal part of the formed in the cleft at the tip within the nor- sheath begins (O'Gara and Matson, 1975). mal sheath. It thus seems from Diirst's work Since the sheath does not cornify over its that the horn sheath is not likely to reflect entire surface, formation of a prong does not superficial surface irregularities added late in cause the mechanical problems which would development, and thus that Solounias' (198 1) arise in the formation of a bovid sheath. We suggestion of a multi-tined sheath is unlikely consider the evidence for a forked sheath in to be correct. fossil bovids to be questionable at best. In some specimens of Miotragocerus the anterior surface ofthe horn core has a stepped Cervidae profile, which Solounias (1981) suggests may The cranial appendages of cervids are have had a multi-tined sheath. Bohlin (1935) termed antlers, and are found in all living believed that the core morphology resulted genera with the exception of Hydropotes, from the addition of a series of platelike an- which is assumed to primitively lack cranial terior placations, presumably in conjunction appendages, rather than to have lost the ant- with growth at the base. Solounias (198 1) sug- lers secondarily (see Leinders, 1983; Scott and gests that this method of growth sets Mio- Janis, 1987). Antlers are postorbital in po- tragocerus apart from other bovids. How- sition, forked, branched, or palmate, and are ever, since Bohlin's reconstruction is largely developed as outgrowths of the frontal bone hypothetical, and since bone is normally (Frechkop, 1955; Bubenik, 1966; and nu- added to all surfaces of a bovid horn, there merous other references summarized below). is no reason to believe that the growth mech- Antlers are covered with skin during the anism of this genus was any different from growth phase, but the main portion of the other bovids, past or present, although the antler consists ofnonliving bone when growth final core shape is different. is complete. Antlers are unique among the We do not accept Solounias' (1981) sug- diversity of cranial appendages in living and gestion that the stepped horn core of Mio- fossil ungulates in the fact that the bony por- tragocerus may have been associated with a tion of the appendages is deciduous, and is multi-tined sheath. Since the small points on shed and regrown annually (see Goss, 1983). the core are superficial, they must have been Cervid antlers have long been accepted as deposited late in horn core development, af- outgrowths of the frontal bone, but a more ter the bulk of the core was laid down. This detailed description oftheir development will means that the normal, smooth sheath would point out how different this process is from have been forming over the core during de- that which results in horn formation. Devel- velopment. Even if the sheath-forming epi- opment of cervid antlers is induced by the thelium did follow the shape of the steps on periosteum in the antler bud; the overlying the core, any shape change would be under connective tissue and dermis do not have any the old core, and would not be visible. We role. This has been shown conclusively in a think that it is more likely that the fairly small series of experiments (Goss, 1983). Excision anterior tuberosities on the core are merely of the periosteum from the presumptive ant- slight irregularities in the horn core that were ler site prevents antler formation; transplan- not reflected in the overlying sheath, espe- tation of this periosteum, without the over- cially as they are not present in all specimens. lying connective tissue and skin, to other parts There is some experimental evidence for this. of the body will induce ectopic antler for- Diirst (1926) produced a line of moorland mation (Goss et al., 1964; Hartwig, 1967; sheep by inbreeding in which the tip of the Hartwig, 1968; Hartwig and Schrudde, 1974; horn core was notched or bilobed. He states Goss, 1983). that the notches presumably are produced by Before antler development begins, the site 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 15 where the antler will form is marked by thick- Although skin is necessary for full differ- ening of the frontal bone and its overlying entiation of antlers, its involvement is sec- periosteum. The area which will give rise to ondary (Goss, 1983). Development of the the antler differs from the rest of the skull in overlying dermis and epidermis into velvet being composed of spongy rather than com- is induced by the periosteum; transplanted pact bone (Goss, 1983). New bone is depos- antler bud periosteum can induce formation ited by the periosteum, and the pedicle grows of velvet by integument from other areas of up out of the skull. As the pedicle enlarges, the body (Hartwig and Schrudde, 1974; Goss, chondrogenesis begins at the tip, and the re- 1983). The integument is evidently not "pre- mainder of the antler is preformed in carti- programmed" to produce velvet, as bovid lage before it ossifies (Goss, 1983). Tines form epidermis is to produce a keratinized horn from bifurcations ofthe growing tip, with the sheath. Velvet is a highly specialized type of two parts ofthe bifurcation elongating in dif- integument which is capable of extremely ferent directions. Growth in the tines even- rapid growth. It is unique especially with re- tually decelerates and then ceases, while that gard to the constant regeneration of new hair in the main beam continues at a constant rate follicles at the growing tips ofthe antlers, and (Goss, 1983). in the absence of erector pili muscles (Goss, Cervid antlers consist of two portions: a 1983). In giraffes, the only other living ru- nondeciduous pedicle and a deciduous distal minants with skin-covered cranial append- antler; both are derived from the frontal bone ages, the skin is thickened and highly corni- without contribution from the overlying tis- fied with only a few scattered hairs (Spinage, sue (Goss et al., 1964; Hartwig, 1967, 1968; 1968), but otherwise shows no striking spe- Hartwig and Schrudde, 1974; Goss, 1983). cializations. During initial antler growth, it is evidently not possible to distinguish between the ped- OF AND ANTLERS icle and the antler proper (i.e., the deciduous HOMOLOGIES HORNS part). During initial growth both the pedicle The development ofhorns and antlers thus and the antler proper are covered by skin, differs in a number ofimportant ways. Antler which is later shed from the deciduous por- growth is induced and controlled by the peri- tion. The pedicle remains skin covered osteum, and no supraperiosteal tissue con- throughout life. Commencing with the for- tributes to bone formation, while horn growth mation of the second deciduous antler, the is induced by the overlying connective tissue pedicle is separated from the deciduous por- and dermis, both of which form bony tissue. tion by a bony burr, which is part of the de- Growth in antlers is primarily through en- ciduous portion and is shed with it (Goss, dochondral ossification, while in horns it is 1983). The primitive cervid condition, seen by intramembranous ossification. As Goss in the earliest fossil genus Dicrocerus, and (1983) points out, the developmental pro- retained in the living genera Muntiacus and cesses which result in antler formation on the Elaphodus, is for the antler to possess a long one hand and horn formation on the other pedicle and a short antler proper. Addition- are different in all respects and are clearly not ally, in primitive cervids the entire structure homologous. This developmental evidence, is probably permanent, with later develop- together with the existence of the antlerless ment of the shedding of the distal portion. deer Hydropotes, clearly demonstrates that This was evidently the case in the fossil cer- cervid antlers and bovid horns are not ho- vid genera Lagomeryx and Dicrocerus (Chow mologous structures. and Shih, 1978; Bubenik, 1982), and in the Given the existence of two nonhomolo- living Muntiacus atherodes (Groves and gous types of cranial appendages with differ- Grubb, 1982). Cervids with this type ofantler ent ontogenies, it would be of interest to de- also retain large upper canines. The derived termine the method of development in other cervid condition is for a short pedicle, and ruminants. However, it should be made clear for a long deciduous antler ofincreasing com- that cranial appendages should not be ho- plexity in the number of branches and the mologized on the basis of broad similarities extent of palmation. in mode ofdevelopment. As Leinders (1983) 16 AMERICAN MUSEUM NOVITATES NO. 2893 has pointed out, there are only two ways to nik (1982) states that frontal and supraperi- form a cranial appendage: either as an out- osteal growths can be identified histological- growth of the frontal bone, involving ossifi- ly, we question the reliability ofthis method. cation only of subperiosteal tissues, or from Since no other certainly identified frontal a dermal ossification center, that is, involving outgrowths are available to compare with ossification of supraperiosteal tissue. No cervid antlers, it is difficult to say whether matter how many times bony appendages may their unique histological appearance relates have evolved independently, each evolution- to their mode of development, or to their ary event must have resulted from one of deciduous nature and rapid growth. This these basic developmental pathways. Clearly, methodology requires further study. cranial appendages have evolved at least twice in ruminants and once in protoceratids, in- Giraffidae dicating that this character can arise conver- gently. As Janis (1982) has pointed out, there The cranial appendages of the Giraffidae are good ecological reasons to explain the or- are termed ossicones. They are postorbital in igin of cranial appendages at certain body position, and developed from dermal ossi- sizes. As far as possible, homology of cranial cones which fuse later during development appendages should be based on comparison with the frontal or parietal bones (Lankester, of the details ofthe developmental processes 1907a; Frechkop, 1955; Bubenik, 1966). A and the inducing and contributing tissues; median nasal ossicone is additionally present simply homologizing cranial appendages on in some living subspecies of Giraffa camel- the basis of development by frontal or su- opardalis (Spinage, 1968). Ossicones are skin praperiosteal growth as Bubenik (1982) has covered in living giraffids, and in living and done may be insufficient. Unfortunately, fossil Giraffini and Okapini are straight, un- comparison of developmental processes is branched, and nondeciduous. In the fossil clearly impossible in fossil groups, and it is subfamily Sivatheriinae the ossicones were nearly as impossible in living ruminants, al- branched and palmate, though apparently still though we will summarize the available in- nondeciduous (Hamilton, 1978a). formation below. Among members of the extinct giraffoid One important point should be made re- family Climacoceridae, cranial appendages garding the interpretation of the mode of de- are known in the genera Nyanzameryx (see velopment ofthe cranial appendages offossil Thomas, 1984), Climacoceras, Injanatheri- organisms. In normal development in bovids um (see Heintz et al., 1981), and Zarafa there is never a separate ossicone (os cornu) (Hamilton, 1973). The cranial appendages of as there is in giraffids even though the process Climacoceras were very different in appear- is induced and regulated by the overlying ance from giraffid ossicones, being more connective tissue and integument. If one ex- rounded in cross section, supraorbital rather amines museum specimens of nondomestic than postorbital in position, and possessing bovids, even the earliest stages ofhorn growth multiple branches (but not palmate like siva- (in neonatal and very young animals) will there ossicones) (see Hamilton, 1978b). Su- appear as an outgrowth of the frontal bones. perficially they bear greater resemblance to We have done this for a series of Antilope cervid antlers or merycodontine cranial ap- cervicapra skulls, which includes both new- pendages than to giraffid ossicones, although born and very young animals, in the AMNH; there is no evidence that they were deciduous in this series it would not be possible to tell (see Goss, 1983), and Bubenik (1982) sug- that the developmental stimulus for horn core gests histological similarities with frontal formation was an os cornu rather than a true outgrowths. In fact, Climacoceras was orig- frontal outgrowth as is seen in cervids. Unless inally described by Maclnnes (1936) as a cer- a separate ossicone is normally formed dur- vid. There is no apparent line of fusion with ing development, as it is in living giraffids, the skull at the base of these appendages, as both frontal outgrowths and appendages that is evident in giraffid ossicones. The cranial are induced by supraperiosteal ossification appendages of Nyanzameryx appear to be centers will appear identical. Although Bube- similar to those of Climacoceras (Thomas, 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 17

1984). In the genus Zarafa, the ossicones are gests that boss formation is induced by the short, supraorbital, and dorsolaterally pro- overlying ossicone. However, some of the jecting (Hamilton, 1978a). A skull of a me- median bosses which develop in Okapia are dium-size giraffoid with cranial appendages not associated with an ossicone (Lankester, similar to those seen in Zarafa has been de- 1 907a), suggesting that some control may ad- scribed from the late Miocene ofIraq (Injan- ditionally reside in the periosteum of the atherium hazami, Heintz et al., 1981). This frontal bone. This condition may be second- animal may well be closely related to Zarafa! ary, as median ossicones are present in some Canthumeryx. The situation in Canthume- male Giraffa (Spinage, 1968). ryx is not clear. No cranial appendages can Giraffid ossicones thus differ develop- clearly be ascribed to Canthumeryx, unless mentally from those of bovids in the initial Hamilton (1 978b) is correct in synonymizing formation in cartilage, in the later fusion with Zarafa with Canthumeryx, which we ques- the skull, and probably also in the extent to tion (see below). which the periosteum is involved. The pro- Giraffid ossicones are widely considered cess is clearly not homologous with the for- homologous with bovid horns (e.g., Hamil- mation of antlers in cervids. Despite the ton, 1978a), although the supposed homol- greater apparent similarity with horn for- ogy is based largely on the erroneous as- mation in bovids, we do not consider giraffid sumption that a separate ossified os cornu is ossicone formation to be homologous to it, routinely formed in bovids. The develop- since this would require a shift from an in- ment of cranial appendages in giraffids is tramembranous to an endochondral form of not as well understood as in bovids and ossification. cervids, since no experimental evidence ex- This conclusion is further supported by ists. Nevertheless, it is clear that in living evidence suggesting that ossicones must have giraffids a separate ossicone is formed in the evolved within the Giraffidae. The epiphy- integument overlying the frontal bone. The seal nature ofgiraffid ossicones with the zone ossicone is evidently preformed in cartilage, of growth at the base may preclude the pos- with subsequent endochondral ossification sibility ofbranching. Although the epiphyseal (Lankester, 1907a; Spinage, 1968), in con- ossicone might also grow at the tip, it is clear, trast to the bovid condition where ossifica- at least in Okapia where the tip is bare, that tion is intramembranous (Diirst, 1902a; this is not the case. Excepting sivatheres, no Brandt, 1928). Ossification proceeds from the giraffid ossicones branch, although male tip ofthe ossicone ventrally toward the skull, specimens of Giraffa may grow knobs on the with growth occurring in the zone ofcartilage side of the horns (Spinage, 1968). Even in between the ossicone and skull in the nature sivatheres, the "ossicones" are palmate, with ofa true epiphysis. In neonatal animals there small knobs rather than distinct tines along is no corresponding boss on the frontal be- the edges, and might be produced by the flat- neath the ossicone (see illustrations in Lan- tening and elongating of the ossicone. kester, 1907a). We could find no histological It thus seems improbable that the type of studies ofthe bone, so it is not known wheth- ossicone possessed by living giraffids could er there is any thickening of the periosteum develop into a branched structure in the ofthe frontal, although a section in Lankester course of evolution. Yet, as previously de- (1907a) indicates that the bone in this region tailed, branched cranial appendages are pres- is somewhat thicker. ent in the giraffoid genus Climacoceras and The formation ofa boss on the frontal bone in the giraffid subfamily Sivatherinae. The ob- is, however, apparent later during develop- vious conclusion is that the "ossicones" of ment, and the enlarged ossicone eventually these fossil taxa (or at least of Climacoceras) comes to articulate with and finally to fuse are not homologous with the ossicones seen with this boss. It is not known whether the in other giraffids. This view was also ex- development of this boss is induced by the pressed by Geraads (1986), although Soloun- overlying ossicone, or whether it develops ias (personal commun.) considers that the independently under the influence ofthe peri- cranial appendages ofsivatheres are not truly osteum. The sequence of development sug- "branched" and may be homologous with 18 AMERICAN MUSEUM NOVITATES NO. 2893 those of living giraffids. Climacoceras and the other giraffids, which all appear to share sivatheres are clearly giraffoids, as evidenced the type of ossicone seen in the living taxa. by the possession of a bilobed lower canine Additional evidence for this suggestion comes (Hamilton, 1978b). Sivatheres are included from the observation that an early mem- in the Giraffidae because they possess the gi- ber of the Sivatherinae (Helladotherium) raffid synapomorphous features of a central (Churcher, 1978) apparently lacked cranial lingual cuspid on p4 and a vertical groove on appendages. Cranial appendages may also be the posterolingual region of p4 (Hamilton, absent in the climacocerid Canthumeryx, al- 1978a). However, Janis and Lister (1985) though of course this type of negative evi- have shown that the first character is not dence cannot be taken as conclusive. Al- unique to the Giraffidae, and we argue later though we would speculate that is is possible in this paper that the second character cannot that the palmated form ofcranial appendages be regarded as a unique giraffid autapomor- in sivatheres could be derived from a giraffid phy either. type of ossicone, we cannot see any possible Traditional phylogenies of the Giraffidae homology between giraffid ossicones and the (e.g., Hamilton, 1978a) place the Okapinae form of cranial appendages in the Climaco- as the most primitive subfamily, with the ceridae. Sivatheriinae more closely related to the Gi- The obvious conclusion from this discus- raffinae, although Hamilton (1973) originally sion is that if the type of cranial appendage considered sivatheres to form their own dis- present in living giraffids was developed tinct family Sivatheriidae, with an early (late within the Giraffidae, rather than represent- Oligocene) divergence from the Giraffidae. ing a character shared by all members of the This phylogeny is not compatible with the Giraffoidea, it clearly cannot share an evo- suggestion that, while the cranial appendage lutionary origin with the bovid type of horn of the living giraffid genera Okapia and Gi- core. raffa are clearly homologous, their ossicones may not be homologous with the cranial appendages of sivatheres. While it is not our Antilocapridae intention to revise the phylogeny of the Gi- Living antilocaprine antilocaprids (genus raffidae in this paper, we would point out that Antilocapra) possess an unbranched, non- the characters used by Hamilton (1978a) to deciduous, supraoccipital bony horn core, in unite the Sivatheriinae with giraffids above combination with a forked, deciduous kera- the level of Okapia are all characters that can tinous horn sheath (see O'Gara and Matson, be functionally related to habitat choice (Scott, 1975). The cranial appendages are usually unpublished data) or to feeding style (Janis, termed "horns." Fossil antilocaprines var- unpublished data). These characters are listed iously possessed horn cores that were by Hamilton (1978a) as: further lengthening branched (e.g., Texoceras) or spiraled (e.g., of limbs; skull flexed; orbits positioned pos- Ilingoceros) (see Frick, 1937). The nature of teriorly; back of skull shortened; and ossi- their covering is uncertain, but at least some cones large. (See also Radinsky, 1984, for dis- genera (Hexameryx, Hexobelameryx) appar- cussion of the acquisition of some of these ently had keratinous sheaths (Webb, 1973), cranial characters in the evolution of the although it is not clear whether these were Equidae.) In addition, sivathere "ossicones" deciduous. are formed of cancellous bone, in contrast to It is apparently not known whether antilo- the compact bone seen in the ossicones of caprine horn cores are formed from an out- living giraffids (Spinage, 1968). growth of the frontal bone, or from a dermal We suggest that cranial appendages were ossification center. If the skulls of a series of evolved independently three times within the young Antilocapra are examined, the horn Giraffoidea: in the Climacoceridae, in the core first appears as a triangular knob over Sivatherinae, and in the other Giraffidae. This the posterior of the orbit. There is no trace would imply, accepting Hamilton's place- of a separate ossicone but, as previously ment of the Sivatherinae within the Giraffi- pointed out, this does not mean that the horn dae, that sivatheres form the sister-group to core is a frontal outgrowth, although it has 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 19

TABLE 2 TABLE 2-(Continued) Distribution of Cranial Appendages in Ruminant Genera Dromomerycids Supraorbital (and median occipital in some), un- TRAGULINA branched, nondeciduous, Hypertragulus Absent ?skin-covered Tragulus Absent HOPLITOMERYCIDAE Leptomeryx Absent Bachitherium Absent Hoplitomeryx Postorbital and nasal, un- Absent branched, nondeciduous, Lophiomeryx covered with keratin "GELOCIDAE" sheath Gelocus Absent Amphimoschus ?Absent Notomeryx/Gobiomeryx Absent CERVIDAE Prodremotherium Absent Eumeryx Absent Hydropotes Absent Rutitherium ?Absent Dicrocerus Nondeciduous, skin-covered pedicle. Deciduous, GIRAFFOIDEA branched naked antler Propalaeoryx ?Absent on top of pedicle Climacoceras Supraorbital, branched, Other cervids Pedicle often shorter, de- nondeciduous, ?skin- ciduous antler more covered elaborate Canthumeryx/Zarafa Supraorbital, dorsolaterally projecting, unbranched, ?skin-covered been interpreted as such (Noback, 1932; So- Sivatheres Postorbital, nondeciduous, lounias, in press). As the core increases in may be branched or pal- size the area anterior to the developing horn mate, ?skin-covered core becomes elevated and a sinus forms in Giraffines Postorbital, nondeciduous, unbranched, skin-cov- the bone over the orbit. The area ofbone and ered its sinus are incorporated into the horn core, BOVIDAE and form its anterior expansion, resulting in Eotragus Postorbital, unbranched, the broad horn core spanning the entire su- nondeciduous, keratin praorbital area in adult animals. The position sheath covering and shape of the immature horn core are in- Other bovids Similar, may be completely teresting in that the core is similar in position curved, twisted, or spi- and shape to that of immature merycodon- raled tines. Since no developmental evidence exists MOSCHINA to suggest that antilocaprine horn cores are Walangania ?Absent or in- Dremotherium Absent either frontal outgrowths dermally Blastomeryx Absent duced ossifications, they cannot be homolo- Parablastomeryx Absent gized with other living ungulates with any Micromeryx Absent degree ofconfidence. Bubenik (1982) consid- Moschus Absent ers them to be similar in their histology to ANTILOCAPRIDAE frontal outgrowths, but is uncertain of their Paracosoryx Supraorbital, branched, true mode of development. nondeciduous, ?naked The fossil subfamily Merycodontinae pos- Other merycodontines As Paracosoryx Antilocaprines Supraorbital, unbranched, sessed branched, supraorbital cranial ap- branched, or spiraled. pendages, which were rounded in cross sec- Deciduous keratin cover tion (as opposed to oval cross section of in some antilocaprine horns). Many merycodontines PALAEOMERYCIDAE have a burr or series of burrs near the base Prolibytherium Supraorbital, horizontally of the cranial appendage, and this was ini- projecting, flattened and as an indication that the lobate, ?skin-covered tially interpreted Amphitragulus Absent distal portion of the appendage was shed Palaeomeryx Supraorbital and occipital, (Cope, 1874, 1877; Matthew, 1904, 1918). unbranched, nondecid- However, a number of authors have con- uous, ?skin-covered vincingly demonstrated that these append- 20 AMERICAN MUSEUM NOVITATES NO. 2893 ages were not deciduous (see for example tine appendages show wear on the tines (Frick, Furlong, 1927; Voorhies, 1969). The burr has 1937; Voorhies, 1969). On this basis Goss been interpreted by various authors as the (1983) has termed these appendages "non- line of demarcation between the covering of deciduous antlers." Although there is no de- the appendage proper and that of the skull velopmental information for merycodon- (see for example Furlong, 1927; Frick, 1937). tines, Bubenik (1982) has suggested that they Pilgrim (1941) suggested that the burrs rep- are frontal outgrowths based on histological resented the point from which the keratinous characters and branching patterns. horn sheaths were shed annually. However, We consider that there is extremely strong as discussed earlier, it is not possible to grow evidence that merycodontine cranial ap- a bovid-type sheath over a branched struc- pendages were initially skin-covered during ture, since the sheath grows internally over the period of growth, and progressively be- the entire surface of the core. Neither would came naked as the animal aged. Certainly the it be possible for such a core to be shed entire evidence precludes the possibility that mer- as is that of Antilocapra, or to increase in ycodontines had a keratinous horn sheath. size. It seems much more likely that the ap- This has a number of important taxonomic pendage was covered by skin, as is suggested implications. If we accept, for the moment, by the smooth surface marked by vascular the widely accepted belief that merycodon- grooves similar to those of cervids. The burr tines are ancestral to antilocaprines, then it would then mark the boundary between head is clear that the keratinous sheath of antilo- and cranial appendage integuments (Furlong, caprines (or Antilocapra) has evolved inde- 1927). Frick (1937) suggested that multiple pendently and cannot be used as a character burrs might have been formed by periodic to link bovids with antilocaprines. This is retreat and regrowth ofthe skin covering, but further supported by differences in formation as Goss (1983) points out, in no living system of the sheath. Conversely, if one accepts the does skin grow over dead bone. Voorhies homology ofthe keratinous sheath in bovids (1969) noted that each of the pseudoburrs and Antilocapra, then merycodontines can- was associated with a different outer layer of not be ancestral to antilocaprines. bone. As Voorhies (1969) and Goss (1983) We would also like to stress the other dif- discuss, this is consistent with the hypothesis ferences between merycodontine and antilo- that these appendages were covered with an caprine cranial appendages. The cranial ap- integument which laid down successive lay- pendages of merycodontines are rounded ers of bone each year, and which retreated rather than oval in cross section and may proximally each year, with the burr repre- have branches with acute angles between shaft senting the level to which the covering had and tines. It should be noted that even Frick regressed. An analogous situation exists in (1937) remarks on the differences between the living Okapia, where the skin retreats from the appendages of merycodontines and an- the tip, leaving a distinct groove in the os- tilocaprines, and he bases his assignment of sicone at the level to which it has moved these two groups to the same family on other (Lankester, 1907b). The appendage would characters. There seems little reason to as- have been covered with skin as it grew, but sume the homology ofmerycodontine antlers when it reached adult size the skin covering with antilocaprine horns, although it is not would die back, gradually leaving a dead, na- impossible that such a homology exists. The ked structure (Voorhies, 1969; Goss, 1983). only morphological similarity between them This is supported by the fact that the cranial rests in their supraorbital position, which is appendages do not increase in size in older shared with the condition in the dromomer- individuals (Voorhies, 1969), as they do in ycids and in the giraffoid Climacoceras. bovids or cervids. Goss (1983) has pointed out that the bony structure ofmerycodontine Dromomerycidae appendages, with an inner layer of spongy bone and an outer layer of lamellar bone, is Dromomerycids possessed supraorbital or consistent with this hypothesis. This is also postsupraorbital cranial appendages, which supported by the fact that some merycodon- were variably round or oval in cross section, 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 21 and were straight or curved, or flattened and pendage. Leinders offered no evidence for extended laterally (as in Dromomeryx) but their mode of formation, but considers the never forked or branched. Members of the animal to represent a parallel evolution of at Cranioceratini also possessed a single me- least a superficial bovidlike form of cranial dian occipital cranial appendage, and Sin- appendage within the Cervoidea. clairomeryx possessed paired nasal bosses in addition to the frontal appendages. The tex- "Palaeomerycidae" ture of the appendages was usually smooth, suggesting a skin cover. In some genera, such Ginsburg and Heintz (1966) ascribe an iso- as Dromomeryx, the appendage surface was lated cranial appendage to Palaeomeryx kau- more rugose and pitted in places, suggesting pi. The appendage appears to be similar in a keratinous thickening of the skin cover, al- texture and shape to a giraffid ossicone, sug- though there is no evidence for a bovid or gesting a skin cover and a postorbital posi- antilocaprine type of horn sheath. The con- tion, and its isolated occurrence and the ap- cept of a skin covering (as opposed to the pearance of the lower surface suggest that it type of naked appendage seen in cervid ant- is a dermal ossicone which would later fuse lers) is strengthened from the examination of with the skull. Qiu et al. (1985) recently de- certain specimens ofSinclairomeryx (includ- scribed a complete skull of Palaeomeryx tri- ing the type of S. sinclairi, AMNH 33791) cornis, which possessed supraorbital giraffid- where the appendage has clearly suffered pre- like cranial appendages, and had in addition mortem breakage and subsequent healing a single median occipital appendage, resem- over the tip region. The tips of dromomer- bling the condition in the cranioceratine ycid cranial appendages have a more spongy dromomerycids. texture, reminiscent ofthe surface ofthe horn The genus Triceromeryx (Crusafont-Pairo, cores ofAntilocapra, and may have been rein- 1952), which has been linked with the genus forced with keratin at this point. However, Palaeomeryx by some authors (e.g., Hamil- as pointed out by Webb (1983b), there is no ton, 1973; Leinders, 1983), but with the Gi- evidence for the type of forked, deciduous raffidae by other authors (e.g., Crusafont-Pai- keratinous tips restored to these appendages ro, 1952; Churcher, 1970; Hamilton, 1978a), by Frick (1937). These appendages have been also possessed a pair of postsupraorbital ap- referred to as "ossicones" (e.g., Janis, 1982; pendages and a single median occipital ap- Leinders, 1983; Webb, 1983b), because ofthe pendage. The texture of the appendages is superficial similarity to giraffid ossicones. smooth, suggesting a skin covering, but there However, there is no direct morphological is no evidence as to their mode of formation. evidence to suggest that they were formed Figure 3 shows in diagrammatic form the from giraffidlike dermal ossicones, as there is different kinds ofcranial appendages in living no apparent line of fusion between the ap- and fossil artiodactyl families. pendage and the skull in young animals. Bubenik (1982) suggests that they were fron- OTHER SKULL CHARACTERS tal outgrowths on histological grounds. Antorbital Vacuity and Lacrimal Fossa Both an antorbital vacuity, sometimes re- Hoplitomerycidae ferred to as the ethmoidal vacuity (e.g., Sigo- Leinders (1983) described the ruminant gneau, 1968), and a lacrimal fossa are present Hoplitomeryx matthewi, which appears to be in living cervids and were included by Brooke a cervoid on other morphological criteria, yet (1878) among his defining characters of the possesses unbranched cranial appendages re- Cervidae (see fig. 4). These characters have sembling bovid horn cores. The surface tex- been used by several authors as defining cer- ture of these appendages is heavily ridged, vid characters. For example, Sigogneau (1968) suggestive of a bovidlike nondeciduous ker- uses them in her discussion of the cervoid atin sheath. Unlike true bovids, this animal affinities ofDremotherium, and Ginsburg and possessed two pairs of suprapostorbital fron- Heintz (1966) state that the apparent absence tal appendages and a single median nasal ap- of an antorbital vacuity in Palaeomeryx im- 22 AMERICAN MUSEUM NOVITATES NO. 2893

(Protoceratidae) (Palaeomerycidae)

Cervidae (Hoplitomerycidae)

6,~~ Giraffidae (Climacoceridae)

Antiloccaprinae n fLOPODA Tragulidae Bovidae RUMINANTIA KEY NN- bone LII deciduous bone NEOSELENODONTIA keratin deciduous keratin skin ( ) = extinct families Fig. 3. Types ofcranial appendages found in living and fossil families of the Neoselenodontia (sensu Webb and Taylor, 1980). plies giraffoid rather than cervoid affinities. era Giraffokeryx and Bohlinia (Colbert, 1933), However, an examination ofthe distribution and may also be present in Zarafa (Hamilton, of these character states in fossil and living 1973). A small vacuity is also present in the artiodactyls suggests that they cannot be tak- camelid genus Llama and has been described en as cervid (or cervoid) apomorphies. in Prodremotherium (Jehenne, 1977). The antorbital vacuity forms at the junc- The antorbital vacuity has obviously aris- tion of the lacrimal, frontal, nasal, and max- en in parallel a number of times, making its illa in cervids. In defining the Cervidae, taxonomic value questionable. The character Brooke (1878) notes that the antorbital va- would still be of value if it were possible to cuity in cervids is of such dimensions as to distinguish a cervid-type of vacuity from a exclude the lacrimal from articulation with bovid-type, but this does not appear to be the nasals. Brooke also notes that an antor- the case. The vacuity always seems to form bital vacuity ofthis type is not unique for the at the junction between the aforementioned Cervidae, as it is also present in the bovid bones; failure ofthese bones to meet is prob- genera Gazella and Oryx. In addition, an ably the simplest way, developmentally, to antorbital vacuity is also present in a number form a vacuity in this area. The position of ofother bovid genera, although in some cases the vacuity in some bovids may differ be- the vacuity is ofsmaller dimensions and there cause ofthe differing relationships ofthe four may be nasal/lacrimal contact. These genera bones. For example, in Neotragus batesi, include Raphicerus, Gazella, Tragelaphus, which lacks an antorbital vacuity, the lacri- Neotragus, Oreotragus, Kobus, and Ourebia. mal has an anterior extension which blocks In the Giraffidae, a small fossa is present in the maxilla from contact with the nasal. In the living genus Okapia and in the fossil gen- the other two species of this genus (N. mos- 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 23

Fig. 4. Skulls of living ruminant taxa, illustrating certain cranial characters (not to scale). A. Lateral view ofMuntiacus reevesi, MCZ 25858, showing cervoid skull features and primitive condition ofcervid antlers with long nondeciduous pedicle. B-E. Anterior views, showing direction of flaring canines: B. Tragulus memmina, MCZ 30251. C. Moschus sifanicus, MCZ 13272. D. Hydropotes inermis, MCZ 11526. E. Muntiacus reevesi, MCZ 25885. AOV, antorbital vacuity; DA, deciduous portion of antler; BUR, antler burr; DLO, double lacrimal orifice; LV, lacrimal vacuity; PED, nondeciduous antler pedicle. chatus and N. pygmaeus) which have an which contains the suborbital gland. In species antorbital vacuity, it forms at the junction of in which the gland is large, the depression the nasal, frontal and lacrimal. The vacuity may extend onto the maxilla and may also in this genus could thus be distinguished from involve the jugal. A lacrimal fossa is present that ofother bovids as well as cervids, but in in Hydropotes and all living antlered cervids, general there appears to be no way to deter- but is absent in Moschus and most fossil cer- mine when the vacuity is homologous, and voids. This character also appears to have it should therefore not be used as a taxonomic arisen in parallel within the Pecora, as it is character. Although an antorbital vacuity is present in the dromomerycids Barbouromer- characteristic of most extinct cervoid lin- yx, Sinclairomeryx, and Dromomeryx (Frick, eages, such as dromomerycids, blastomer- 1937), in Dremotherium (Sigogneau, 1968), ycines, antilocaprines, and merycodontines, in Zarafa and Prolibytherium (Hamilton, it has obviously arisen in parallel among oth- 1973), in certain bovids, including Cephalo- er artiodactyl lineages, and cannot be consid- phus, Madoqua, Neotragus, Oreotragus, and ered diagnostic of cervoids. Ourebia (personal observ.), in Dacrytherium, The lacrimal fossa (character 8) is a depres- and in many oreodonts. The lacrimal fossa sion in the facial plate of the lacrimal bone may well be a symplesiomorphic character 24 AMERICAN MUSEUM NOVITATES NO. 2893 for bovids, lost in more derived taxa (P. that only a single orifice was present. (The Grubb, personal commun.). The lacrimal condition of the lacrimal orifice in these two fossa varies in depth and extent among the genera will be discussed further in a later sec- species of cervids and bovids in which it oc- tion.) The derived double orifice has evi- curs. In both bovids and cervids in which the dently evolved at least twice among the ru- fossa is large the walls of the fossa are very minants, since it is present in the bovid tribes thin, and may become fenestrated. The size Bovini and Tragelaphini (Leinders and and depth of the fossa are clearly related to Heintz, 1980), and occasionally in the Al- the degree of development of the suborbital celaphini, Hippotragini, and Caprini (per- gland, which is in turn related to aspects of sonal observ.), as well as in the Cervidae. The social behavior. There is therefore no con- double orifice has clearly evolved indepen- sistent or taxonomically useful pattern in its dently in these bovid tribes, since the orifice distribution, nor are there any characteristics is single in the more primitive Boselaphini of the fossa which might be used to establish (Leinders and Heintz, 1980); however, the homology. The character should therefore be condition in Eotragus (the earliest known bo- dropped from taxonomic considerations in vid, considered to be a boselaphine) is not distinguishing cervoids from other pecorans, known. The placement of the double orifice but we consider that it may have some value in Tragelaphini and Bovini is highly variable, at the lower taxonomic level of linking Hy- and in some individuals there may be only a dropotes with antlered cervid genera. single orifice, also variably placed (Leinders and Heintz, 1980). Lacrimal Orifice (character 9) The cervoid condition is for the dorsal orifice to be situated on the orbital rim, or some- Flower (1875) and Brooke (1878) first what inside the rim, and for the ventral orifice pointed out the significance of the number to be situated on the external side ofthe rim. and placement ofthe lacrimal orifices, noting Usually, a small spur of the lacrimal bone that the presence of two lacrimal orifices on juts out laterally from a position within the or near the rim of the orbit was one of the two orifices (see fig. 4). The condition in An- features which defined the family Cervidae tilocapra, and in those fossil antilocaprine taxa (see fig. 4). Leinders and Heintz (1980) used in which we have been able to observe this this character more broadly, to define the su- feature, is similar to the cervoid condition (in perfamily Cervoidea within the higher ru- those instances where a double orifice is pres- minants. However, as we pointed out else- ent). The condition in bovids is more vari- where (Scott and Janis, 1987), there are able. Individuals of the genus Tragelaphus several problems associated with use of the usually have both orifices situated slightly number of orifices as a taxonomic character, within the orbit, and there is no spur of the namely that the number of orifices may be lacrimal bone. However, we have observed individually variable within individuals in a individuals ofthe Bovini where the situation species, and that the character has clearly is identical to the condition in cervids, de- evolved more than once within the mam- spite variability within each species. Thus, mals. For example, it is seen in certain suid the simple presence ofthe two orifices cannot and kangaroo species. be assumed to be homologous with the cer- The primitive condition for the Ruminan- void condition, although the presence oftwo tia is the presence of a single orifice located orifices with the cervoid configuration may within the rim of the orbit. This is the con- be useful as part of a suite of defining char- dition seen in all of the Tragulina, the Mos- acters. chidae, and in Prodremotherium (Jehenne, Use of the double lacrimal orifice as a de- 1977). The lacrimal orifice in giraffids is either rived character is also problematical since the single, reduced, or absent. Although a double number of orifices can be variable within a lacrimal orifice, situated within the orbit, was species. This was noted by Leinders and described for the genera Zarafa and Pro- Heintz (1980) for Tragelaphus scriptus and libytherium by Hamilton (1973), Leinders by Scott and Janis (1987) for Antilocapra (1983) decided on reexamining this material and Moschus. The character should therefore 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 25 be used with caution, especially in fossil the tips directed downward. Viewed ante- groups where only one or a few specimens riorly, the lateral surface appears concave are known. proximally and convex distally. This type of canine is seen in Moschus, blastomerycids, Hydropotes, Hoplitomeryx, Dremotherium, Characters of the Auditory Bulla Amphitragulus, the dromomerycid Bar- The primitive condition of the auditory bouromeryx (see Frick, 1937), and has re- bulla in the Pecora is an unexpanded bulla cently been described in Palaeomeryx (Qiu with a posteriorly positioned, exposed at- et al., 1985) and Lagomeryx (Chow and Shih, tachment for the tympanohyal. Although ex- 1978). In these genera, the canine is much pansion of the bulla and changes in the po- larger relative to body size than is the en- sition of the tympanohyal are common with larged canine of the tragulid type. The same the Pecora, personal observation has shown type of canine is present in Micromeryx, but that, with the exception ofthe Moschidae, no it is smaller relative to body size. Reduction specializations of the bulla can be found to of this type of canine is seen in the dromo- characterize any one Pecoran family. How- merycid genera Aletomeryx and Sinclairo- ever, Webb and Taylor (1980) showed that meryx. The tragulid type of canine also oc- living and fossil moschid taxa can be char- curs in living Muntiacus and Elaphodus and acterized by a subcentral position ofthe tym- related fossil genera; its occurrence here is panohyal, with a laterally enclosed tympa- presumably secondary (see fig. 4). nohyal vagina (character 10). Although the differences in canine morphology are not strictly allometric, that is, the size and shape ofthe canine do not change GENERAL CHARACTERS OF regularly with body size, the difference in THE DENTITION morphology is probably related to the abso- lute size of the tooth relative to the skull. In Upper Canines (character 11) those genera with the tragulid type of canine, The presence of large upper canines, pre- the tooth is short relative to body size and sumably used in intraspecific combat, has the tip would have little clearance below the been considered a primitive characteristic jaws if it were relatively straight as in the among ruminants (e.g., Webb and Taylor, moschid type. The outward flare allows the 1980). The presence of large upper canines tip to be used more effectively in combat. In has been noted in a number of traguloid and the relatively larger moschid type canine, the pecoran genera and families. However, no tip is clear without the lateral flare. In previous attempt has been made to distin- muntjaks, the reduction ofthe canine in con- guish among the types oflarge upper canines junction with the evolution of antlers pre- seen among ruminants. While an enlarged sumably led to the redevelopment of the lat- upper canine appears to be a primitive char- erally flared canine from the reduction of the acter for the Ruminantia in general (see Webb moschid type of canine. and Taylor, 1980), the primitive type of en- Large upper canines are absent in all known larged upper canine, seen in modem tragu- giraffids and bovids, although small upper lids, which curves laterally, differs markedly canines may be present as an individual vari- from the loose-socketed ventrally projecting ation in some Miocene bovids, e.g., Protrago- canine seen in Moschus or Hydropotes. In ceras (Gentry, 1970: 249). They are also pres- modem and fossil tragulids the canines curve ent in a variety of living cervid species, e.g., laterally and posteriorly, so that the pointed Cervus elephus. We have also observed a small tips are widely divergent. This type ofcanine upper canine alveolus in the merycodontine also is found in the fossil traguloid genera genus Paracosoryx (AMNH 31191, Paraco- Hypertragulus, Archaeomeryx (Webb and soryx wilsoni). The anterior part of the skull Taylor, 1980), and Bachitherium (Bouvrain is unknown in all Eurasian gelocids, although and Geraads, 1985). small, apparently tragulid type canines have In contrast, the moschid type of canine is been ascribed to Prodremotherium elongatus curved slightly posteriorly and laterally, with (AMNH 10336) and Eumeryx culminus TABLE 3 Distribution of Other Cranial Characters in Ruminant Genera Lacri- mal Antorbital Lacrimal Upper orifices vacuity fossa Auditory bulla canine TRAGULINA Hypertragulus ?1 Absent Absent Unexpanded Large Tragulus 1 Absent Absent Expanded, Large cancellous bone Leptomeryx 1 Absent Absent Unexpanded Large Bachitherium ?I ? ?7 Large Lophiomeryx ?I ? Large "GELOCIDAE" Gelocus ?1 ? Unexpanded Notomeryx/Gobiomeryx ?1 ? Prodremotherium ?I Present Large Eumeryx ?I ? Absent Large Rutitherium ?I ? I? GIRAFFOIDEA Propalaeoryx ?I ? I? '?Large Climacoceras ?l ? ? I? Absent Canthumeryx/Zarafa 1 Present Present Moderately expanded Absent Giraffids 1 Small or absent Absent Expanded in some Absent BOVIDAE Eotragus ?I Absent Absent Unexpanded ? Other bovids 1 or 2 Present in some Present in some Expanded in some Small/absent MOSCHINA Walangania ?I ? I? 9.Small Dremotherium 1 Present Present Less pronounced than Sabrelike other moschids Blastomeryx 1 Present Absent Laterally enclosed Sabrelike Parablastomeryx 1 Present Absent Subcentral Sabrelike Moschus 1 Present Absent Tympanohyal vagina Sabrelike Micromeryx ? Present ?7 Sabrelike ANTILOCAPRIDAE Paracosoryx 2 Present Absent ?Somewhat like Small/absent moschids Other merycodontines 2 Present Absent As Paracosoryx Absent Antilocaprines 2 Present Absent Moderately expanded Absent PALAEOMERYCIDAE Prolibytherium ?2 ? Present ?Moderately expanded Amphitragulus 2 ? Absent Unexpanded Sabrelike Palaeomeryx ?I Present Present ?Unexpanded Sabrelike Barbouromeryx 2 Present Present Unexpanded Sabrelike Aletomeryx 2 Present Absent Unexpanded Moderate Other dromomerycids 2 Present Present in some Unexpanded Mod./absent HOPLITOMERYCIDAE Hoplitomeryx 2 ? Absent Expanded in unique Sabrelike fashion Amphimoschus ? ? Similar to ? Hoplitomeryx CERVIDAE Hydropotes 2 Present Present Unexpanded Sabrelike Dicrocerus 2 Present Present Unexpanded Large Other cervids 2 Present Present Unexpanded/slightly Large/absent expanded 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 27

(AMNH 19147) in the AMNH collections. A similar type ofcanine is associated with Pro- A palaeoryx austroafricanus from Namibia (BMNH 36963). The North American Pseu- doceras, which has been referred to the Gel- ocidae by Webb (1983a), has short upper ca- MC nines of the tragulid type. This suggests that the tragulid type canine is primitive for the psmic ruminants, that the absence of a canine in bovids and giraffids is a derived condition for these groups, and that the moschid type canine is a derived condition uniting the blastomerycines and various other cervoid genera with the early true cervids. B ESd CO Height of Cheek Teeth (character 12) PrCd The primitive ruminant condition is for AntC \ low-crowned or brachydont cheek teeth. Prprcd \\ High-crowned, or hypsodont, cheek teeth are phcd generally defined as deeper than they are long psprcd from root to crown (Janis, 1984; Fortelius, prmcd 1985), although many hypsodont mammals have cheek teeth that have exceedingly high MCd psecd crowns in the unworn condition, and the term psmcd ECd mesodont is often applied to teeth that are MtSd precd moderately hypsodont. Despite past asser- Fig. 5. Guide to dental morphology terms used as in the section, that tions, detailed previous in this paper: Upper and lower molars of Dremo- bovids and antilocaprids can be united by the therium (Montaigu, Allier). A. M2, Ph 52933. B. possession of hypsodont cheek teeth, the m2, Ph 32488. Abbreviations: AntC, anterior cin- character is obviously ofno taxonomic value, gulum; CO, cristid obliqua; ECd, entoconid; ES, as it can be shown to have evolved in parallel entostyle; ESd, ectostylid; HCd, hypoconid; MC, many times within the Mammalia. Among metaconid; MCL, metaconule; MsS, mesostyle; living ungulate families alone, this has oc- MtS, metastyle; MtSd, metastylid; PaC, paracone; curred in the Proboscidea, Hyracoidea, PF, Palaeomeryx fold; phcd, posthypcristid; PrC, Rhinocerotidae, Equidae, Suidae, and Ca- protocone; PrCd, protoconid; precd, preentocris- melidae. tid; prmc, premetacrista; prmcd, premetacristid; prmlc, premetaconule crista; prpac, preparacrista; prprc, preprotocrista; prprcd, preprotocristid; psecd, postentocristid; psmc, postmetacrista; CHARACTERS OF UPPER CHEEK TEETH psmcd, postmetacristid; psmlc, postmetaconule Figure 5 depicts the tooth nomenclature crista; pspac, postparacrista; psprc, postprotocris- used in this paper. Figure 6 shows diagram- ta; psprcd, postprotocristid. matic dentitions illustrating "primitive" and "derived" conditions in pecorans. in the tragulid genera Lophiomeryx and Bachitherium, and is retained in primitive pecoran genera such as Gelocus, Eumeryx, Internal Cingulum on Upper Molars and Prodremotherium, although it is absent An internal cingulum on the upper molars from all living pecoran genera (see Janis, (see fig. 6) is often considered a diagnostic 1987). Traces of this internal cingulum are characteristic of the Tragulidae (see, e.g., retained in the extinct higher pecoran genera Sudre, 1984), and is a common feature of Climacoceras, Canthumeryx, Zarafa, Pro- traguloid taxa above the level of the Hyper- palaeoryx, Walangania, Dremotherium, Pa- tragulidae. Although absent (presumably sec- laeomeryx, Barbouromeryx, and Dicrocerus ondarily lost) in the Leptomerycidae, it is seen (see table 4). 28 AMERICAN MUSEUM NOVITATES NO. 2893

Posteriorly Situated Protocone on P3 presence in Hypertragulus and Lophiomeryx, (character 14) it is apparently not a character which has Webb and Taylor (1980) unite the family evolved numerous times, as evidenced by its Gelocidae on the basis of a narrow waisted, absence in most nonruminant selenodonts. posteriorly situated, and posteriorly inclined Among living pecorans, it is present in cer- protocone on P3. While this character is cervids, with the exception of the genera Hy- tainly present in all of the gelocids for which dropotes, Moschus, Capreolus, and Pudu, and the upper dentition is known (see Prodremo- in bovids in the tribes Bovini, Cephalophini, therium, fig. 11), it is not unique to the gel- Hippotragini, Reduncini, and Tragelaphini, ocids as it is also present in the tragulid genera but generally absent among the Giraffidae. Bachitherium and Lophiomeryx (Janis, 1987). Among fossil pecorans it is present in the However, all other pecorans have P3 with an earliest bovids and cervids, as well as in other unconstricted, lingually situated and lingual- cervoid genera (see table 4), and is also seen ly inclined protocone, with the possible ex- in Prolibytherium. It is generally absent in ception of Walangania (e.g., BMNH 21364) antilocaprids, with the exception of the early and Propalaeoryx (e.g., BMNH 36962), which genus Paracosoryx, and is absent in Propa- show some resemblance to the gelocid con- laeoryx. The entostyle is absent in all non- dition (see fig. 12). The functional signifi- giraffoid giraffids, and in giraffids with the cance of this character is not known. The exception of its occasional occurrence in cer- development of a "normal" pecoran type of tain individuals of the species Giraffa ca- P3 might be considered to be a character link- melopardalis and in Palaeotragus coelophysis ing the higher ruminant families, but as cer- (specimens in the collections ofthe AMNH), tain derived cervoid characters are developed where it is derived from the posterior face of within the gelocids, as will be discussed in a the protocone, as in Prosynthetoceras, and so later section, we consider it more likely that may have evolved in parallel with other pec- the lingually directed protocone was evolved orans. A small entostyle derived from the independently in the higher pecoran families. anterior face of the metaconule is evident on the unerupted molars of specimens ascribed to Zarafa (Hamilton, 1973). Despite Ham- ilton's later synonymizing of this genus with Entostyle (character 15) Canthumeryx (1978a), we have never ob- The entostyle in the upper molars (see figs. served this feature in other specimens as- 5, 6) is a dental character which we consider signed to Canthumeryx, and regard the taxo- to be of phylogenetic significance. The func- nomic assignation ofthese Gebel Zelten teeth tional significance of this character appears to be problematical. They are clearly too big to be related to the puncture-crushing offood to belong to the same species represented by items with rounded contours, such as seeds the type of Zarafa (and, indeed, are consid- and hollow grass stems (Janis, 1979), but erably larger than the fragmentary remains its presence is not uniformly distributed of molars on this skull, which clearly show throughout the selenodont ruminants. It is the absence of an entostyle). Moreover, the absent from the tylopod artiodactyls with the larger teeth ascribed to Zarafa have a very exception of the protoceratids Pseudoproto- different type of preservation from those on ceras and Prosynthetoceras (where it is de- the skull, possessing enamel that is brownish rived from the posterior face of the proto- black as opposed to the medium tan color of cone, rather than from the anterior face of the enamel on the skull of Zarafa and also the metaconule, or from the lingual cingu- on the associated dental material ascribed to lum, as in the Ruminantia) and is seen as an Prolibytherium. occasional variant in the living genus Ca- An entostyle is generally absent among melus. It is also absent from the traguloid primitive Pecora, although an incipient ento- ruminants, with the exception of the Hyper- style, formed as a protuberance ofthe lingual tragulidae and Lophiomeryx. While it is clear cingulum, is present in the genera Prodremo- that this feature may evolve in parallel among therium and Eumeryx (see fig. 11), and a selenodont artiodactyls, as evidenced by its similar condition exists in some specimens 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 29

All 6l 76 1 BiA-

9 16A 11 12 1416

Fig. 6. Diagram of composite ruminant teeth, illustrating dental character: Ai.FEComposite primitive pecoran condition: 1, posteriorly situated and posteriorly directed protocone on P3; 2, large metacone on P4; 3, large metastyle; 4, internal cingulum; 5, small M3 metaconule; 6, p1 present; 7, long premolar row, unmolarized premolars; 8, incomplete premetacristid and postentocristid; 9, ectostylid; 10, simple posterior loop to m3. Aii. p4 and ml in more primitive traguloid condition: 11, p4 metaconid forming posterolingual wall to tooth; 12, "Dorcatherium fold." Bi. Derived pecoran condition: 1, lingually directed and centrally situated protocone on P3; 2, attenuated P4 metacone; 3, bifurcated protocone; 4, bifurcated metaconule; 5, reduced metastyle; 6, entostyle; 7, cingulum lost; 8, large M3 metaconule; 9, pl lost; 10, shorter, more molarized premolar row; 11, "cervoid" type of molarization, with expanded metaconid; 12, "giraffoid" type of molarization, with suppression of the metaconid and posterior extension of the paraconid; 13, vertical groove on posterolingual region of p4; 14, metastylid; 15, "Palaeomeryx fold"; 16, double posterior lobe on m3: A, closed. B, open. Also shows condition in more hypsodont teeth, where metastylids, ectostylids, and "Palaeomeryx fold" tend to be reduced or lost. of Walangania (e.g., BMNH 35251), al- represent a precursor of the more derived though it is absent in other specimens. The condition. condition in Rutitherium is not known. We A problem using the entostyle as a diag- assume this incipient condition to be ho- nostic character is that it appears to be a trait mologous with the more developed entostyle that is lost in more hypsodont members of a apparent in the bovids and cervoids, and to lineage, presumably correlated with the re- 30 AMERICAN MUSEUM NOVITATES NO. 2893 duction of small, crushable items in the diet in most dromomerycid genera, Prolibytheri- when hypsodonty is developed for better um, and Parablastomeryx (see table 4), the mastication offibrous material (Janis, 1979). exterior rib is reduced, and there is little labial For example, an entostyle is present in some projection of the metacone in occlusal view specimens of Paracosoryx minor but not in (see figs. 13, 15, and 16). This condition seems the other merycodontines. It tends to be lost to accompany the attenuation of the meta- in the more hypsodont living cervids and bo- style on the molars, a feature also seen in vids, though it is characteristically present as these genera, and may be taken as a derived a pillarlike style in the Bovini and the Hip- pecoran condition, possibly indicative of in- potragini. Within the dromomerycids, it is terrelationships between the genera in which greatly reduced in the more hypsodont genera it occurs. Aletomeryx and Pediomeryx (see Frick, 1937). Thus the absence of an entostyle in a lineage that is initially fairly hypsodont may not Small M3 Metaconule (character 18) be of phylogenetic significance, although we The presence ofa small metaconule on M3 feel that its absence in an initially brachydont appears to be a primitive ruminant character, lineage, such as the Giraffoidea probably is seen in traguloids such as Dorcatherium and of phylogenetic significance. Despite the Lophiomeryx and is also characteristic ofgel- variable presence of an entostyle in rumi- ocids (e.g., Prodremotherium, see fig. 1 1). The nants, its distribution among the range ofliv- metaconule is the same size as the protocone ing and fossil pecorans suggests a closer as- in the earliest members of the Bovidae, Gi- sociation of the Bovidae with the Cervoidea raffidae, and Cervidae, but the primitive state than with the Giraffoidea. is variously retained in other fossil pecorans such as Paracosoryx (but not in other antilo- Reduced Metastyle (character 16) caprids), some specimens of Palaeomeryx, The presence of a metastyle in the upper Amphimoschus, and Dremotherium. The molars appears to characterize the pecoran metaconule is larger in the dromomerycids, genera, as it is present in all gelocids, although blastomerycids, and in Amphitragulus, and it has also been evolved in parallel in the is also large in the traguloid Leptomeryx (see traguloid genera Leptomeryx and Lophio- table 4). It is apparent that the larger size meryx (Janis, 1987). However, certain pec- metaconule has arisen in parallel among the oran genera, namely Parablastomeryx, Am- Pecora several times, as also evidenced by phitragulus, Palaeomeryx, Prolibytherium, the parallel evolution ofthe condition in Lep- and all of the dromomerycids with the ex- tomeryx. However, the character is ofinterest ception of Aletomeryx and Rakomeryx, ap- because it may be presumed that an animal pear to have secondarily reduced the meta- with a large metaconule represents a derived style (figs. 6, 13, 15, and 16, and see table 4). condition. The functional significance of this character state is not clear. However, we believe that Bifurcated Protocone (character 1 9a) this is a derived condition that may be of Ginsburg and Heintz (1966) state that the phylogenetic significance. bifurcated posterior wing of the protocone is a derived cervid character, which forms part Attenuated Metacone on P4 (character 17) oftheir argument for the conclusion that Pa- Sigogneau (1968) mentions the presence of laeomeryx cannot be a cervoid. A bifurcated an attenuated P4 metacone in the upper den- protocone is present in the earliest antlered tition ofAmphitragulus and Palaeomeryx. In cervids such as Dicrocerus and Euprox, but the primitive pecoran condition, retained in is not present in any of the other noncervid most pecoran genera, the P4 metacone is cervoids except for Amphimoschus (Lein- prominent, forming a posterior ridge on the ders, 1983) (see fig. 6). We consider this to labial wall of the tooth, and projecting labi- be an advanced cervoid character, possibly ally from the ectoloph in occlusal view (see linking Amphimoschus, and by association fig. 6). In the aforementioned genera, and also Hoplitomeryx (see Leinders, 1983), with the 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 31

true cervids. To the best of our knowledge, ilton, 1973), but the teeth in question from this dental character does not appear as a Gebel Zelten cannot be assumed to belong to variant in any other pecoran genera, although the skull of Zarafa, as previously discussed. it can be seen in the possibly tragulid genus In addition, we have never seen a Palaeo- Dorcabune and in oromerycid tylopods. meryx fold in other specimens referred to However, this character is not universally Canthumeryx. Further, we have not noticed present among living cervids, being found a Palaeomeryx fold in any giraffid genus (an only in the genera Odocoileus, Blastocerus, observation confirmed by Solounias, person- Alces, Mazama, Pudu, and Capreolus. al commun.), and are dubious as to the homology of such a fold in these giraffids with Bifurcated Metaconule (character 19b) the typical cervoid condition. Janis (1987) has shown that the Palaeomeryx fold quoted A bifurcated posterior wing of the meta- by Matthew (1908) as being present in Lep- conule is a feature of early giraffids such as tomeryx is actually a remnant of the buccal Palaeotragus, and is occasionally seen in portion of the traguloid Dorcatherium fold, specimens of Canthumeryx [e.g., Canthu- which involves a bifurcation ofboth the post- meryx sirtensis from Moruorot, in the col- paracristid and the postmetacristid (see fig. lections of the University of California 6). A similar situation may exist in these gi- (Berkeley), V 4898/41981]. However, among raffid genera. Thomas (1984) presents evi- other African genera, it is present in Propa- dence of a "Palaeomeryx" fold in the laeoryx, but absent in Walangania. The char- premolars of Palaeotragus. However, a Pa- acter may be of limited phylogenetic signif- laeomeryx fold is not seen in cervoid pre- icance, as it has arisen in parallel in other molars, and we are dubious about the ho- selenodonts; for example, it is seen in the mology ofthis structure in Palaeotragus with living cervid Capreolus, in Palaeomeryx, and the condition in cervoid molars. We consider in many dromomerycid genera [e.g., Barbou- the Palaeomeryx fold to be a derived cervoid romeryx, Bouromeryx, Drepanomeryx, condition. Among the earliest representa- Dromomeryx, and Rakomeryx (see Frick, tives ofthe living pecoran families, it is pres- 1937), although this character is not invari- ent in early cervids but absent in early mem- ably present in all specimens ofthese genera]. bers of the other families (see table 5). However, we believe that it may be of im- The problem with this character is that it portance in establishing the sister-groups of is often seen as a primitive feature in certain the Giraffoidea and the Dromomerycinae, as lineages and is lost in later forms. For ex- will be discussed later. ample, it is present as a diagnostic character in Miocene cervids, but has been lost in all CHARACTERS OF LOWER CHEEK TEETH Recent genera. Among dromomerycids, its presence is apparently linked with tooth crown Palaeomeryx Fold (character 20) height. It is present in the brachydont genera The Palaeomeryx fold in the lower molars Barbouromeryx, Bouromeryx, Procraniocer- (see figs. 5, 6) has been generally considered as, and Dromomeryx, variably present in the to be a primitive ruminant character (e.g., mesodont genera Sinclairomeryx, Rakome- Hamilton, 1973). This assumption probably ryx, Drepanomeryx, and Cranioceras, and stems from the phylogeny of Stirton (1944), absent in the more hypsodont genera Aleto- where he assumed that the Palaeomerycidae meryx and Pediomeryx, representing parallel (including Palaeomeryx and the North loss with increasing hypsodonty in three American blastomerycids and dromomer- lineages. It is generally absent in the blas- ycids) were the basal family for the origin of tomerycids, but present in the most brachy- both cervids and giraffids (see also Matthew, dont early genus Problastomeryx and in the 1934). Hamilton (1973) repeated Stirton's as- brachydont Parablastomeryx. The Palaeo- sertion, and claimed that a Palaeomeryx fold meryx fold is absent in all known antilocap- is irregularly present in the giraffid genera rids, bovids, and giraffids. Among other Palaeotragus and Honatherium. He also de- primitive pecoran genera, it is absent in Gel- scribed a Palaeomeryx fold in Zarafa (Ham- ocus, Notomeryx, Gobiomeryx, and Pro- 0 0

0 00 00 0 0 00 0 0 0~ 0 0

0 0 0 00 000 0 o 0 0 0 zzZz z z-zZZZ. zzZ

Q -

.CCoW0o Q

Ev ~~ ct ~ ~ 0 at 0. Po 0. C- ao a4 ao A4 a4 a4

4 S 0 0 0 00 0 0 0 0 0 0 0 0 0 0

C) z z t , z0z zz zz .0 0 c ; ; ;

o~~;ZZZZ ZZZZ. .v. ZZZZZn .u-O4) - S SH 0 e

00 a0 0o .o 00P00 0000 00

S E q~

| 7 E S g t E < U ts t t t rm'0< 0

32 a) 0 0 0 0 0 0-C 3 co3Si-0 0 a 3 to00l

0 100 a) z Qo zo

000 0 0 0i~ a) zo;zo zo S- o

Ut 0 0 v 0

a) cn 000e o o o C1 W iC GM C = CuCu a)a) C-IS I- 0 0 a) a )I 'o o o CuCu a)a) a)a)a) C.) a) ) ) a Cd cis u C.Y')4 Cu CuCu Cu a)a)a) JD)JD '0 a )a)a C' o eowe ._ ~~ ~ 4 U, U, U, 14 o~~~~~~~~~~*,0 0 0 a)a) a) a) a) a)~ a) cuccuc Cu (uu cn( cU(A zz'0 bObO a)a)a) zzoo 0'-'- 0 ct CZ c cd CQC.) C. Cu C.

'00

oma 00 0

- a) 04 ow oCo 02 00 CZ co Cd co Cld d cis '0Oa 04 '0'0 0 0*--*-* .;., .> C6) ) .Cu ~ ~ .

Q ~ ~ t '0m a)0D

33 34 AMERICAN MUSEUM NOVITATES NO. 2893 dremotherium, but present in some speci- evidenced by its absence in the recent Tra- mens ofEumeryx and Rutitherium (see Janis, gulidae, but the retention of the character in 1987). It can also be seen in some specimens certain genera is of interest in that it suggests of the poorly known African taxon Gelocus that they cannot be derived directly from an- whitworthi (e.g., in the holotype and the para- imals which have already lost pl. pl gener- type from Songhor BMNH. K.Sgr. 365-1949 ally is present in primitive pecorans and in and K.Sgr. 581-1949). Among more derived the genera Propalaeoryx and Amphitragulus, extinct pecorans, it is absent in the mesodont although it is admittedly difficult to distin- taxa Walangania gracilis, Amphimoschus, guish with certainty a p 1 from a retained dp 1 and Hoplitomeryx, but present in the more in fossil taxa. It is generally absent in Dremo- brachydont taxa Walangania africanus, Am- therium, although it is present in the earliest phitragulus, Dremotherium, Micromeryx, members of the genus from La Milloque in Palaeomeryx, and Triceromeryx. We would France (M. Brunet, personal commun.). pl conclude, on the basis of observational evi- is generally absent in Palaeomeryx, although dence, that the presence or absence of the it is present in early species such as P. kaupi, Palaeomeryx fold may be considered a di- as shown by the presence of an alveolus for agnostic character for brachydont taxa, but p1 in specimens ofthis species from Artenay for lineages where no early brachydont genera (L. Miocene, France) (Basel, 502342). pl is are known (as is the case for the Antilo- generally absent in blastomerycids, although capridae), its apparent absence in the known it is present in Problastomeryx, and in drom- hypsodont species cannot be taken as diag- omerycids, though a deciduous p1 is present nostic. in Dromomeryx, Rakomeryx, and Drepan- omeryx (Frick, 1937). However, p1 is com- Metastylids (character 21) pletely absent in the earliest antilocaprids Metastylids in the lower molars (see fig. 5) (such as Paracosoryx), bovids (e.g., Eotra- are present in all pecoran taxa, with the ex- gus), cervids (e.g., Dicrocerus), and giraffoids ception of Gelocus (although small metasty- (e.g., Climacoceras). It is also absent in Ho- lids are present in the African "Gelocus" plitomeryx. The condition in Canthumeryx, whitworthi), and are not seen in any traguloid Walangania, and Micromeryx is unknown genera (Janis, 1987). Metastylids are fre- (see table 5). quently reduced or lost in more hypsodont taxa, and this reduction has occurred in par- Postentocristid (character 22) allel in the Bovidae, Antilocapridae, Hopli- The postentocristid (see figs. 5, 6) forms tomerycidae, Cervidae, and Dromomeryci- the completion ofthe posterior lingual selene nae. In the living genus Antilocapra, small in the lower molars, making the tooth fully metastylids are apparent on newly erupted selenodont. The formation of a completely cheek teeth, but are not apparent on teeth selenodont lower molar with complete lin- that have started to wear. gual selenes has evidently occurred in parallel among the ruminants many times (Janis, Condition of p1 (character 4) 1987). For example, complete lingual selenes P1 is absent in all the Ruminantia with the are present in Hypertragulus, but not in the exception ofthe Hypertragulidae (Webb and tragulids; among the leptomerycids they are Taylor, 1980), but pl is retained, and the present in the Oligocene genus Leptomeryx, primitive condition for the ruminants is ap- but not in the Eocene Archaeomeryx. Taking parently for pl to be large and caniniform Gelocus as the sister taxon to the other pec- (Janis, 1987). p1 is small and premolariform orans (Janis, 1987), the lower molars are very in Lophiomeryx and in primitive Pecora with primitive, lacking both the postentocristid the exception ofNotomeryx where it is absent and the premetacristid. The premetacristid is (Savage et al., MS). The presence or absence present in the other pecoran genera, but the of p1 is a character of similar interest to the postentocristid is not fully formed in a num- condition ofthe metaconule: p1 has obvious- ber ofearly higher pecorans. As with the pres- ly been lost in parallel numerous times, as ence of a p1, this character is of interest as 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 35 the retention of the primitive condition im- lingual "waist," was also held by Hamilton plies that an animal with a fully formed post- (1978a) to be diagnostic of the Giraffidae, entocristid cannot be directly ancestral to one although he did note in an earlier paper in which the postentocristid is not fully (Hamilton, 1973) that this dental feature was formed, although it does not exclude the pos- also present in "some palaeomerycids." We sibility of their being sister taxa. have not performed a thorough investigation A complete postentocristid is absent from of the distribution of this trait in all pecoran all gelocids with the exception of Prodremo- taxa, but would point out that a similar p4 therium. The postentocristid is also incom- morphology is present in Amphitragulus (see plete in the early giraffoids Canthumeryx and fig. 13), Eotragus, and Prolibytherium, and it Climacoceras, as well as the other early Mio- appears to the usual form of premolar mor- cene African genera Propalaeoryx and Wal- phology among the dromomerycids (see fig. angania. It is incomplete in Amphitragulus, 16) and blastomerycids (see Frick, 1937). early blastomerycids such as Problastomeryx, Thus this character cannot be considered di- and early dromomerycids such as Barbou- agnostic for the Giraffidae. romeryx. In contrast, it is complete in early bovids (Eotragus), cervids (Dicrocerus), gi- Double Loph on Posterior Lobe of raffids (Palaeotragus), antilocaprids (Paraco- m3 (character 26) soryx), and also in other cervoid genera (see table 3). This character has been included in Signogneau (1968) mentions the presence our cladistic analysis, but we do not consider of a double loph on the posterior loph of m3 it to have a high degree of phylogenetic sig- in Amphitragulus and Palaeomeryx, which nificance. she suggests may be indicative of a relationship between these two genera. In the usual Bilobed Lower Canine (character 23) pecoran condition, the m3 hypoconulid aris- es close to the base of the entocristid, fre- Hamilton (1978a) unites the Giraffoidea quently forming a closed loop by the traverse (families Giraffidae and Climacoceridae) on ofa cristid around the posterior border ofthe the basis of a bilobed lower canine. We have tooth, returning towards the entostyle on the not observed any instances ofthe parallel ac- lingual side ofthe tooth (see condition in Dre- quisition of this character in any other ru- motherium, fig. 13). In the case ofthe double minant taxon, and consider it to be a valid lophed m3, the hypoconulid has a double apomorphy of the Giraffidae. origin, with a distinct cusp arising on the posterolingual margin of the tooth, somewhat Condition of Paraconid on p4 (character 24) posterior to the entostyle, in addition to the Hamilton (1978a) asserted that the pres- more labially positioned "true" hypoconulid. ence of a posteriorly directed paraconid on Posteriorly directed cristids extend from both p4, forming a lingual wall to the tooth, in labial and lingual hypoconulids, approaching combination with the suppression of the each other at the posterior border ofthe tooth metaconid, was a unique feature of the Gi- to form a closed loop. This morphology of raffidae (see fig. 6). Janis and Lister (1985) the lower dentition is also seen in most drom- have shown that, while this feature is not omerycids, in Parablastomeryx, and possibly variable among living giraffids, it cannot be also in Prolibytherium and the teeth from Ge- taken as diagnostic of the Giraffidae, as it is bel Zelten ascribed to Zarafa (see table 5). seen as a variant of premolar morphology in This type of m3 morphology is sometimes cervids, bovids, and dromomerycids. also evident in the living giraffid species, though it is not typical of all giraffids. It is a derived condition though evidently not a Vertical Groove on Posterolingual unique one. A parallel condition, seen in Am- Region of p4 (character 25) phimoschus and Hoplitomeryx, and used by The presence ofa posterolingual groove on Leinders (1983) to link these two genera, is p4, dividing the tooth into a posterior third for a similarly double condition of the m3 separated from the anterior two-thirds by a hypoconulid in which the posteriorly extend- 0 00 0 00 0 0 0 00 00 0 0

0~~~~~~~~~~~~~

0 z z z z z z zzz zzz zz z z

0 V 0 ut c_ 'P.4E;0 zz0 0zgo0 zz0 zz0 0z0 z00z z00z00z zz0 0 zz00z0

U,, E z zE zb z C/z / < / P -~~~~ -* 4-~~~w 4-50 00- 0Z _t_n

"' w 0 &A0 (A0 0 0 A0- CO00(

0 Cd ~~~~~~~~~~~~00 3 ~.0 00 0 00 0 ,940000

03 0 0 -

0 ~~~~~~~~~~C1Cd CIScq CZ CZUCIS C,3tvC'S0 0 0

UUU C-~~~~~~U,

<~ ~ ~ ~ ~ OQ Z R~ ,, t -

36 G)00 00 000 Q 0 a >-zz0 z zz >.> > ~> zz zzz

0o C)

600 Q QQQ QQ U U Q QQ QU QQ Q o o0o o o o o o o o o o o u u u u V u u u u u u u u uu

00 0 000 0 0000 U) (A 0 00 z zz zzzz z zzz z z 0 4) zz z

(A 0 0 0 0 C) C) C) C,) 00 000 zzzO O >vQ zO z z zz z

4- 4-- 0 0 0 u rA 0 u I-- 0.4

;0 un ° Ci) Cd Ci) C C) Ci (L) 4ADrA V V 0 w LLIU] 3 88 m 0 4-. I- a) C,),CA,) C,i ) 000 0 0 +- 0 0.zzz zz.40. C, Ci) C,

CA a)a 00; 0 o u: *-j *ce C- o 000~~~4 0A .D .C . C,) C,) E4o ;°iC o ut CA cn,) u( r 4 0 Q 1-i 0 0 0 0 001 C 04 -4 -4 -.

Ch a) -o ._ 004 > 00d-o a)o IA II L-

Eg;.6E5cZi @ } j 4 $, , t02E > i |

37 38 AMERICAN MUSEUM NOVITATES NO. 2893 ing cristids do not meet at the posterior mar- 1970), who noted that the state of the gully gin of the tooth, and an open valley extends could be used to distinguish bovids from cer- between the two parallel cristids from the vids in all deposits where they co-occur. Lein- posterior end of the tooth to the twinned hy- ders and Heintz (1980) discussed the useful- poconulids (see fig. 6). ness of the metatarsal gully as a taxonomic character which distinguished cervoid (as well POSTCRANIAL CHARACTERS as cervid) artiodactyls from the bovoids. They point out that the character is not known to Complete Distal Metapodial be variable in any living bovid, cervid, or Keels (character 27) giraffid, and this invariability makes it appear In higher ruminants (including the mem- to be a useful taxonomic character. Although bers ofthe homed ruminant families, the fos- Leinders and Heintz (1980) do not specifi- sil homless genera which are not included in cally address the point, it is implicit in their the Gelocidae, and the genus Pseudoceras) the discussion that the closed gully is a derived distal articular surfaces for the first phalanges condition for the Cervoidea and that an open have a strongly marked keel which extends gully is the primitive condition for the Pec- completely around the articular surface (see ora; other authors have also assumed this fig. 7 and table 6). In more primitive rumi- (e.g., Groves and Grubb, 1987). nants, including the traguloids and the genera The invariability of the character in living assigned to the Gelocidae, the keels are in- ruminants makes it appear to be particularly complete, being present only on the posterior strong. No variability has ever been reported part of the articular surface. Completion of in living bovids and cervids, and none has the distal keels presumably stabilizes the been observed in the over 1000 skeletons of metapodial-phalangeal joint, preventing dis- living bovids and cervids we have examined. location of the phalanges. The development Neither have we observed variation in meta- ofthis character might therefore be expected tarsals of Eotragus or early cervids such as to arise in conjunction with a more cursorial Euprox, Heteroprox, and Dicrocerus, an ob- habit. Although the development ofthe char- servation also noted by Heintz (1963). How- acter itself does not present any direct evi- ever, the character does show variability in dence that it may have evolved more than two groups, the Moschidae (sensu Webb and once, the strong functional basis of the char- Taylor, 1980) and the merycodontine anti- acter suggests that it might have done so. locaprids. An open gully was observed in one Complete distal keels appear independently specimen of Moschus moschiferous in the in equids, suggesting that the character may Rijksmuseum von Naturhistorische, Leiden readily be evolved in parallel. Although the (osteology catalog d, no number) and we have character does have taxonomic significance, observed both open and closed gullys in the we would suggest that when it appears in con- North American genus Blastomeryx from the flict with other characters it may be assumed Olcott and Trinity River Formations. Vari- to have evolved in parallel. ability is also seen in Amphitragulus and Dremotherium from San Gerard-le-Puy (Al- lier, France). Among merycodontines the Fusion of Metapodials and Condition of character is variable in Paracosoryx, Mery- Metatarsal Gully (character 28) codus, Ramoceros, and Cosoryx. In ruminants in which the third and fourth The significance of the variability can be metatarsals have fused to form a single can- understood only in the context of the ontog- non bone, the line of fusion on the anterior eny of the character. In living cervids the surface is marked by a groove, the metatarsal immature metatarsus consists ofa single shaft, gully. This gully may be open along its entire or diaphysis, with an open gully on the an- length, as in living bovids and giraffids, or it terior surface, and two distal epiphyses (see may be closed at its distal end by a thin bridge fig. 8). The edges ofthe distal end of the gully of bone, as in living cervids and moschids project toward the median axis of the shaft. (see fig. 7). The taxonomic significance ofthis These ridges begin just distal to the point at gully was first pointed out by Heintz (1963, which a nutrient foramen passes through the 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 39

A CN D G

T 11 T G .71 II

CN CN CN

B H

C F I L MtG

MK MK Fig. 7. Morphology of pecoran metatarsals (not to scale). A, D, G, J, proximal views (posterior to top of page). B, E, H, K, posterior views of proximal metatarsus. C, H, I, L, anterior views of distal metatarsus. A-C, giraffid (Okapia johnstoni, MCZ 38015). D-H, cervid (Axis porcinus, MCZ 1703). G-I, bovid (Tragelaphus scriptus, MCZ 14222). J-L, primitive cervoid (Eumeryx culminis, type, AMNH 19147). Abbreviations: CN, cubonavicular facet; G II, groove for metatarsal II; G V, grove for metatarsal V; MK, metapodial keels; MtG, metatarsal gully; T II, tubercle for metatarsal II; T V, tubercle for metatarsal V. bone. As the bone grows, the edges of the the two articular condyles. As growth contin- gully continue to grow inward until they meet. ues at the distal epiphyseal plate the length Initially there is a visible suture where the of the bridged portion increases. The epiph- two sides of the bridge meet, but it is rapidly yses also contribute to the bridge as they fuse obliterated. The nutrient formation is inter- to the diaphysis. The shorter bridge seen in rupted as growth continues, and is no longer Moschus may result from some minor dif- continuous through the bone. A small fora- ferences in timing of the developmental pro- men usually remains on the posterior surface. cess, but the process can be seen to be similar Anteriorly the foramen is bridged over, and in young moschids we have observed. We now ends at the distal end of the shaft. The will refer to the type of gully formed in this end of this foramen is formed by the epiph- way as the cervid-type gully. yses, and in the adult it terminates between In the Amphitragulus and Dremotherium 40 AMERICAN MUSEUM NOVITATES NO. 2893

A B C D E

I' .1-. I.',

F G H I

,1. Ii

I, I

; II

K L M

"I ,I, "I 'i

Fig. 8. Developmental stages of the fusion of the metatarsus and formation of the metatarsal gully in bovids, cervids, and tragulids (not to scale). Youngest individuals in each series are on the left. A-E, cervid developmental stages (all of Odocoileus virginianus). F-J, tragulid developmental stages (all of Hyemoschus aquaticus). K-O, bovid developmental stages (K and N are ofAepyceros melampus, L, M, and 0 are of Antilope cervicapra). A, MCZ 50991. B, MCZ 50987. C, MCZ 50985. D, MCZ 51000. E, MCZ 50986. F, AMNH(M) 53601. G, AMNH(M) 53640. H. AMNH(M) 53611. 1, AMNH(M) 53617. J, AMNH(M) 53604. K, AMNH(M) 33317. L, AMNH(M) 35218. M, AMNH(M) 54481. N, AMNH(M) 82045. 0, AMNH(M) 35957. specimens from St. Gerard Le Puy with a there is clearly a partial bridge: two thin ridges closed gully (in the collections of the Musee ofbone extend medially from the raised areas Gimee, Lyon, France) the bridge is relatively on either side of the gully. The bridge is very low as it is in Moschus. In some, there is a thin, and some ofthe "open" specimens may visible suture in the ventral midline of the in fact have been closed in life and been bro- bridge. In specimens with an "open" gully ken or worn after preservation. However, not 1987 JANIS AND SCOTT: RUMINANT PHYLOGENY 41 all ofthe open metatarsals show signs ofwear, become permanently open in Merycodus fur- and it is clear that some of the variation at ther supports our contention that a closed least is real. It is possible that in life this gap gully can give rise to an open one. was bridged by connective tissue or cartilage, In living Antilocapra the gully is closed, and that the extent ofossification is variable. and as in cervids the bridge occupies ap- Metatarsals ofAmphitragulus and Dremo- proximately the lower one-third ofthe meta- therium were examined from a number of tarsal. In fossil antilocaprines the gully is sim- other localities. In only one, La Milloque, a ilar to that of Antilocapra. In the earliest French Chattian locality under study by Y. material we have been able to examine (Pli- Jehenne ofPoitiers, was any variability seen; oceros from the Ash Hollow Formation, the nature and extent ofthe variability is sim- Cherry County, Nebraska) the gully is always ilar to that at St. Gerard-le-Puy. It is signif- closed, although it tends to be closed more icant that in younger deposits the gully is distally than in living cervids, as is more typ- invariably closed, indicating that this char- ical ofearly cervids and the living genus Mos- acter became more stable over time, although chus. There is no evidence of variability in we would still interpret the closed condition any antilocaprine, as one might expect ifthey of the gully as primitive. It is also clear that were derived from a group like the meryco- continuation ofthe trend toward incomplete dontines in which variability in ossification ossification of the edges ofthe gully in meta- seems to be typical. However, it is clearly tarsal development would result in an open possible that a variable gully may be fixed in gully, probably one with raised edges at its the closed state as well as in the open one. distal end. This is precisely the condition seen We have assumed throughout this discus- in some bovids (i.e., Gazella). It thus seems sion that there is no functional significance clear that a closed gully in which the degree to the closed versus open gullies. Leinders of ossification of the bridge is variable can (1979), in his analysis of functional differ- give rise to an open gully which would be ences in the foot ofcervids and bovids, notes interpreted as a bovid type gully. that the metatarsal gully contains the dorsal In the merycodontines the character is digital artery in both groups; the artery is variable in the earliest genus Paracosoryx, accompanied by the dorsal common digital and in Merycodus, Ramoceros, and Cosoryx. nerve (Ghoshal, 1975). Leinders (1979) finds The nature of the variability is very similar, no connection between morphology of the and, judging from the immature specimens gully and the functional differences he de- we have observed, the metatarsus develops scribes between bovids and cervids. It is in- in the same way. The bridge appears to form deed difficult to postulate any functional sig- in the same way as it does in living cervids, nificance for the various states of the gully. with thin shelves of bone extending toward A deeper gully, with or without a bridge, the midline from either side of the gully on would provide protection for the artery and the diaphysis before fusion begins between nerve and would reduce pressure from the epiphysis and diaphysis. The extent to which extensor tendons on these structures. The this bridge is ossified evidently remains vari- depth of the gully is greatly reduced in the able in all genera of merycodontines except Bovini and the Caprini, which would seem Meryceros, in which it is consistently open. to suggest that this is not an important func- In Meryceros the gully is similar to that of tion of the gully. However, in these groups some smaller bovids, with a relatively deep the metatarsals are strongly compressed in gully bordered at the distal ends by raised the anteroposterior plane, presumably to ridges which extend slightly toward the mid- counter lateral stresses during locomotion line. Although the gully is variable in the ear- (Scott, 1985). The gully may become reduced liest genus of this subfamily, we hypothesize as a secondary result ofthis flattening. Selec- that the primitive condition for the mer- tive pressures for increased transverse di- ycodontines is to have a closed gully on the ameter of these bones may outweigh any ad- basis of the morphological and develop- vantage of a deeper gully. mental similarities with the bridge in the Al- Even if the gully did function to reduce lier specimens. The fact that the gully has pressure on the artery and nerve, it is difficult 42 AMERICAN MUSEUM NOVITATES NO. 2893 to see why the groove should be more pro- most fossil genera of Tragulidae the meta- tected in the Cervidae which are less spe- tarsals are unfused, and the fused metatarsals cialized for rapid locomotion. Nor does the of living tragulids are clearly an indepen- loss ofthe bridge appear to be correlated with dently derived condition. Of the other trag- more rapid locomotion; if this were the case uloid families, the metatarsals are unfused the gully would be expected to have the same also in the Hypertragulidae. In leptomerycids form in the gazelles and in Antilocapra, in the metatarsals are fused, although the post- which the gully is in fact closed. Cervids gen- crania are primitive in other respects. In Lep- erally do emphasize the use of the hindlimb tomeryx, the gully forms a shallow, open in locomotion more than do bovids (Scott, groove, and the suture between metatarsals 1987) and this may be the reason why the III and IV is not entirely obliterated. How- gully is more protected in cervids. However, ever, in metatarsals referred to the lepto- none of these suggestions has a strong func- merycid Pseudoparablastomeryx by Taylor tional basis and we would agree with Leinders and Webb (1976), the gully is closed. The that the gully is essentially a neutral character association ofthe metatarsal with the cranial with little functional significance. and dental material is not certain, but the Caution must be used in drawing conclu- metatarsal is primitive in other respects and sions based on the presence of a closed gully, there do not seem to be any other taxa with since the closed condition has evolved in- which it might more reasonably be associ- dependently in at least one other instance. In ated. Although there is no developmental se- the living tragulid Hyemoschus the gully is ries available for this genus, the metatarsal is closed in older individuals but both the mor- nearly identical to that ofHyemoschus. Ifthis phology of the distal end of the bone and the specimen is a leptomerycid, then the closed sequence of development ofthe bridge differ gully of Pseudoparablastomeryx is probably from those in cervids. In young individuals an independent closure from the condition the metatarsals are unfused. Proximally they in the cervoids, as is that of Hyemoschus. are closely joined by connective tissue but Among primitive pecorans, postcranial re- distally they are separated by a distinct gap. mains are known in the genera Gelocus, Pro- Fusion appears to begin proximally and an- dremotherium, andEumeryx. Janis (1987) has teriorly and proceed distally and posteriorly. suggested that Gelocus is the sister taxon to As the metatarsals fuse anteriorly the edges all other Pecora, and this genus is therefore ofthe groove grow toward each other and the of considerable interest. The specimen de- gap is filled by connective tissue. The epiph- scribed by Kowalevsky (1876) includes a yses fuse completely before fusion of the metatarsal, and this appears to have a shal- metatarsal shafts is completed. When fusion low, open gully, although it is difficult to be has been completed along the anterior surface certain of the condition as the distal end is ofthe metatarsals the connective tissue filling broken. Other metatarsals from Ronzon in the gap begins to ossify, forming a bridge. the collections at Le Puy (PAR 27-60), Paris The metatarsals do not fuse completely at the (RZN 42), and Basel (Ro 41) all clearly show distal end and there is always a line offusion the open gully type of morphology, so that it between the third and fourth metatarsals (see seems reasonable to assume that the meta- fig. 8). As in Tragulus, there is a large foramen tarsals are fused and that the gully is open in passing posteriorly through the bone. A closed Gelocus. This would suggest that the primi- gully has evidently also occurred indepen- tive condition for ruminants is an open gully. dently in the genus Pseudoceras. A cervoid However, the possibility should also be con- type of closed gully is present in this genus, sidered that the metatarsals have fused in- which also has a pecoran type of compact, dependently several times in higher rumi- parallel-sided astragalus and complete distal nants and Gelocus, as has occurred in metapodial keels. Webb (1 983a) refers to tragulids, and among other artiodactyl groups Pseudoceras as a gelocid, but the relation- in the camelids, amphimerycids, xiphodon- ships of this genus are under further study tids, tayassuids, and entelodontids. Pro- (Webb, personal commun.). dremotherium (Viret, 1961; Bouvrain and The primitive condition for ruminants is Geraads, 1985) and Eumeryx (Matthew and for metatarsals III and IV to be unfused. In Granger, 1924) both possess a cervid type of 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 43 gully with a short bridge, in contrast to Gel- morphologies of the metatarsus in fossil and ocus. The morphology of the preceding two living giraffid species, where the metatarsals genera may be regarded as a derived condi- are somewhat laterally flattened, and the gul- tion, linking them with cervids (see Janis, ly is shallow with low ridges. However, the 1987, and further discussion later in this pa- giraffoid Canthumeryx has a deeper meta- per). tarsal gully, more reminiscent oftragelephine Interpretation of the gully would be facil- bovids, as seen in an unnumbered Nairobi itated if it were known whether metatarsals Museum specimen from Moruorot. had fused independently in bovids and cer- Based on the evidence available at present, vids. Ifso, both open and closed gullys would we can only state that an open gully appears be derived conditions, with an unfused meta- to be the primitive condition for ruminants, tarsus representing the primitive condition. since this is the condition in Gelocus. If this If bovids and cervids were derived from a is true, then an open gully must have given common ancestor with a fused metatarsus, it rise to the closed cervid type. However, we would imply that a closed gully was derived know of no metatarsals among primitive ru- from an open one (or vice versa) in evolution. minants which have an open gully, but have Detailed examination of the ontogeny of fu- ridges which may give rise to a closed one. sion in the two groups might help to clarify Neither is there fossil evidence of any inter- the situation. Unfortunately, complete series mediate type. No ridges form along the groove of young individuals of bovid species with a in Gelocus, and it is difficult to see how a pronounced gully were not available for us Gelocus-like metatarsal would give rise to a to examine. However, by combining speci- cervidlike one. Further work will have to be mens of three species, Alcelaphus buselaphus done on the postcrania of primitive rumi- (the kongoni), Antilope cervicapra (the black- nants before this issue can be resolved. How- buck), and Aepyceros melampus (the impala), ever, it is clear that a closed gully can give in which the morphology of the adult gully rise to an open one within a population, in- is similar, we have been able to describe the dicating that a mechanism does exist for this process at least in part. In the youngest spec- shift, and that this possibility must be con- imens available to us the distal ends of the sidered in assessing relationships. diaphyses are just fusing distal to the nutrient Presence of a closed cervid type gully ap- foramen. There are no ridges along the sides pears to be a good derived character, al- of the groove on either the diaphysis or the though the possibility of independent acqui- epiphyses (which are not yet fused at this sitions cannot be entirely ruled out. This type stage). In older specimens there is a distinct ofgully is seen in all living and fossil cervids, ridge on either side of the groove, and these Palaeomeryx, Amphitragulus, Dremotheri- ridges are continued on the epiphyses, even um, Moschus, Walangania, Eumeryx, Pro- before they fuse. However, the edges ofthese dremotherium, dromomerycids, antiloca- raised areas do not project towards the mid- prines, and Hoplitomeryx, and is present but line as they do in cervids, and they appear variable in merycodontines and blastomer- later, after fusion of the distal ends of the ycines (see table 6). We will continue to re- diaphysis around the nutrient foramen (see gard the presence of an open gully as a re- fig. 8). This suggests that in bovids the raised tained primitive condition at this time, edges along the gully are not remnants of the although the open gully may prove to be sec- cervid type of shelf, but that they form in- ondary in some groups. An open gully is pres- dependently, perhaps as attachment for a ent in all bovids and giraffoids, Gelocus, Lep- connective tissue covering for the blood ves- tomeryx, Bachitherium, and Propalaeoryx sels and nerve passing along the groove. (see table 6). However, it is difficult to judge how significant the differences in the relationship of the artery and nerve to the fusing metapodials Condition of Digits I, II, and V are, and we do not consider it to be conclusive The primitive condition for ruminants is proofofeither homology or nonhomology of the presence of five complete digits in the fusion. The situation in bovids differs from forelimb and four complete digits (II-V) in the admittedly more limited diversity of the hind limb, a condition known only in 44 AMERICAN MUSEUM NOVITATES NO. 2893

Hypertragulus and perhaps also in Archaeo- The characteristics of digits II and V can meryx (Webb and Taylor, 1980). In primitive provide some information oftaxonomic val- ruminants metatarsal I is already reduced and ue, but it must be recognized that the fusions present as a proximal remnant which artic- can, and probably have, occurred indepen- ulates with the posterior diarthrodial facet. dently. However, since a fused remnant pre- The presence of this facet thus indicates that sumably does not secondarily become free, a metatarsal I rudiment is present. In more it can be assumed that genera without fusion advanced ruminants, digits III and IV have of metapodials II and V represent the prim- fused to form a cannon bone and digits II and itive condition. The presence or absence of V are reduced so that only proximal remnants a distinct tubercle has been used by Heintz ofthe metatarsals and distal remnants of the (1963, 1970) to distinguish Eotragus from phalanges remain. Dicrocerus. Although a distinct tubercle for This loss of the lateral digits is commonly II characterizes Eotragus, the character is not referred to as the "loss ofthe side toes" (char- consistent through the Bovidae; neither is the acter 29c). Among living pecorans, the side presence ofa tubercle for V consistent for the toes are completely lost (with the exceptions Cervidae. The presence of a tubercle may be of proximal remnants that fuse with the can- suggestive, but it is clearly not a reliable char- non bone) in the Bovidae, Antilocapridae, acter. (See table 6.) and Giraffidae, while proximal or distal remnants of the side toes are variously present among taxa in the Cervidae and Moschidae. Posterior Tuberosity (character 31) In certain fossil cervids complete, though re- Another character used by Heintz (1963, duced, lateral digits were retained. As noted 1970) to distinguish cervids from bovids is in the historical review section, this feature the presence ofa posterior tuberosity, a raised has been accorded taxonomic significance in eminence on the lateral side of the posterior ruminant phylogeny, but the reduction or loss surface of the metatarsal, visible when the ofthe side toes has occurred in parallel many bone is viewed from the proximal end, as well times within the Artiodactyla (e.g., in the Ca- as when viewed in posterior aspect (fig. 7). melidae, Tayassuidae, Entelodontidae, and This character is uniformly present in the cer- Xiphodontidae). The common evolutionary vids we have examined and is absent in bo- pattern is for the hind limb to be more pro- vids (except Neotragus batesi, Bate's ante- gressive than the forelimb in the develop- lope). It is uniformly absent in all giraffoids ment of this feature. For example, in drom- and antilocaprids, and absent in Walangania omerycids and in the early merycodontines, and Propalaeoryx. The character is some- such as Paracosoryx, the side toes have been what problematical for two reasons. First, it lost in the hind limb, but complete reduced is sometimes difficult to distinguish between side toes or distal remnants are retained in no tuberosity and a slight one ofthe type that the forelimb (Frick, 1937). would be expected in the early evolution of The configuration of the proximal rem- this character. Additionally, in genera in nants of the metatarsals can provide some which we would characterize the posterior information of taxonomic importance (char- tuberosity as "none" or "slight," there is also acter 30). The remnants are initially present usually some individual variation. This ap- but not fused to the cannon bone, and they pears to be the case in Amphitragulus, Pa- fit into distinct notches on the proximal sur- laeomeryx, and some dromomerycids. Sec- face ofthe metatarsal. The presence ofa notch ond, the character appears to have arisen on either the medial or lateral side of the independently in dromomerycids, although cannon bone indicates the presence ofan un- interpretation is difficult because the char- fused proximal remnant II or V, respectively. acter is both variable and only slightly de- These remnants fuse with the cannon bone veloped in some genera. However, as the independently, so that only one or the other character is slight or absent in the early drom- may be present. After fusion, the remnant omerycids Barbouromeryx and Sinclairo- may form a distinct tubercle, as was pointed meryx, seems to be invariably absent in Al- out by Heintz (1963, 1970) (see fig. 7). etomeryx, but is most marked in the later 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 45

Tertiary cranioceratines, we conclude that it the cubonavicular facet is more flat, but might have been developed independently in among living bovids this condition is some- this group. In addition to living cervids, a what variable, and it is more cervidlike in well-developed posterior tuberosity is found certain genera (e.g., Tragelaphus). The Giraf- in Micromeryx, Moschus, and Parablasto- foidea have a truly derived condition with meryx, but not in other blastomerycines. (See respect to this character. In living giraffids table 6.) Although the presence of the tu- the facet is extremely flat and elongated; we berosity in Neotragus indicates that it can have also observed this character in every evolve independently, its development in this fossil giraffoid we have examined, including genus may be associated with secondary size the primitive Canthumeryx (BMNH 30179, reduction. In this genus both lateral and me- from Moruorot). A raised lip ofthe posterior dial edges of the posterior surface are ex- cubonavicular facet is thus suggestive of cer- panded, and this condition may have devel- void status, but cannot be treated as a diag- oped secondarily to provide additional area nostic character since it represents a modified for insertion of the flexor muscles. retention of the primitive ruminant condition. (See table 6.) Posterior Cubonavicular Face (character 32) Heintz (1963, 1970) described differences FEATURES OF THE SoFr ANATOMY in the shape of the posterior cubonavicular All living ruminants are characterized by facet which he found to be useful in distin- an enlarged, subdivided stomach, with a ru- guishing Dicrocerus from Eotragus. In Dicro- men forming the initial chamber (character cerus this facet is rectangular, long, and nar- 33). The primary enlargement of a rumen row, while in Eotragus it is diamond-shaped, area appears to be shared with living camel- short, and broad (see fig. 7). Although these ids, but the subsequent development of the differences are sufficient to distinguish be- enlarged stomach appears to be independent tween Eotragus and Dicrocerus, as Heintz in- (Langer, 1974). Living tragulids have a small dicated, they are not consistent throughout rumen, a reticulum, and an abomasum (the the Bovidae and Cervidae. However, the equivalent ofthe true stomach in other mam- morphology of the anterior edge of the facet mals, serving as the site of protein degrada- does provide a taxonomically useful char- tion). Living pecorans have an enlarged acter. In cervids, the anterior edge of the rumen, and an additional chamber, the facet forms a distinct lip where it meets the omasum, inserted between the reticulum and small cuneiform facet, while in bovids no the abomasum. We follow Webb and Taylor distinct lip is formed (see fig. 7). In addition (1980) in assuming a three-chambered stom- to cervids, this "raised lip" is found in drom- ach with a small rumen to be typical of the omerycids, moschids, Amphitragulus, Tragulina and a four-chambered stomach Dremotherium, Walangania, Palaeomeryx, with a larger rumen and an omasum to be Hoplitomeryx, and antilocaprids. However, typical of the Pecora. the raised lip appears to be a primitive pec- The giraffe (Giraffa camelopardalis) has oran character, since a more prominent type some features ofstomach anatomy which dif- ofraised lip is seen in the gelocids Prodremo- fer from the condition in bovids and cervids therium and Eumeryx (see fig. 7), and pos- (the okapi has not been investigated), notably sibly also in Gelocus. It is also present in the cranially positioned entrance of the material assigned to the advanced traguloid esophagus to the rumen (Hofmann, 1973). Bachitherium from Marseilles (St. Andre), Judging from the diagrams in Langer (1974), France (UBCL 8900), as well as in limb ma- this appears to represent the primitive con- terial assignable to Lophiomeryx from the dition for the Artiodactyla. Further work is Phosphorites du Quercy (UCBL 8900). Al- needed on this character, especially a more though the condition in cervoids is distinc- detailed examination of the relative position tive among higher ruminants, it clearly rep- of the esophagus in the okapi and the Tra- resents a modified retention of a primitive gulidae. The condition in the giraffe may rep- condition. Bovids are more derived in that resent a secondarily derived state in associ- TABLE 6 Distribution of Postcranial Characters in Ruminant Genera Com- plete distal meta- Posterior Fusion of podial Fusion of Posterior cubonavicular lateral keels metapodials tuberosity facet metapodials TRAGULINA Hypertragulus No Unfused Absent Raised, pointed No Tragulus No Unfushed/fused w. Absent Raised, pointed No closed gully Leptomeryx No Fused w. open gully Absent Raised, pointed Both Bachitherium No Fused w. open gully Absent Raised, pointed II Lophiomeryx No Unfused Absent Raised, pointed ?Both "GELOCIDAE" Gelocus No Fused w. open gully Absent Raised, pointed No Notomeryx/Gobiomeryx ? ? A? e p Prodremotherium No Fused w. closed gully Absent Raised, pointed No Eumeryx No Fused w. closed gully Absent Raised, pointed No Rutitherium ? ? I? ? ? GIRAFFOIDEA Propalaeoryx Yes Fused w. open gully Absent Raised w. lip II Climacoceras Yes Fused w. open gully Absent Elongated, flat II Canthumeryx/Zarafa Yes Fused w. open gully Absent Elongated, flat II Giraffids Yes Fused w. open gully Absent Elongated, flat Both BOVIDAE Eotragus Yes Fused w. open gully Absent Fairly flat, Both Other bovids Yes Fused w. open gully Absent no distinct lip Both MOSCHINA Walangania Yes Fused w. closed gully Absent Raised with lip II Dremotherium Yes Fused w. closed gully Absent Raised with lip Both Blastomeryx Yes Fused, closed or open Absent Raised with lip Both gully Parablastomeryx Yes Fused w. closed gully Present Raised with lip Both Moschus Yes Fused w. closed gully Present Raised with lip Both Micromeryx Yes Fused w. closed gully Present Raised with lip Both ANTILOCAPRIDAE Paracosoryx Yes Fused, open or closed Absent Raised with lip II gully Other merycodontines Yes Fused, open or closed Absent Raised with lip Both gully Antilocaprines Yes Fused w. closed gully Absent Raised with lip Both PALAEOMERYCIDAE Prolibytherium Yes ?Fused w. open gully R w l Amphitragulus Yes Fused w. closed gully Absent or small Raised with lip II Palaeomeryx Yes Fused w. closed gully Variably present Raised with lip Both Barbouromeryx Yes Fused w. closed gully Absent or small Raised with lip II Aletomeryx Yes Fused w. closed gully Absent Raised with lip II Other dromomerycids Yes Fused w. closed gully Variably present Raised with lip II or both HOPLITOMERYCIDAE Hoplitomeryx Yes Fused w. closed gully Present Raised with lip Both Amphimoschus ? ? CERVIDAE Hydropotes Yes Fused w. closed gully Present Raised with lip Both Dicrocerus Yes Fused w. closed gully Present Raised with lip Both Other cervids Yes Fused w. closed gully Present Raised with lip Both 1987 JANIS AND SCOTT: RUMINANT PHYLOGENY 47 ation with the high, narrow thorax and the gus from Artenay (Burdigalian of France) on elongated neck. However, the possibility re- the basis of dental remains, but assign it to mains that a more ventral displacement of the Bovidae primarily on the basis ofthe ab- the esophageal entrance to the rumen is a sence of cervoid characters (e.g., lack of a derived feature within the Ruminantia, link- Palaeomeryx fold) (character 20) and on gen- ing together the Bovidae and the Cervidae eral appearance (molars somewhat higher- (the condition in Antilocapra is not known). crowned and longer relative to the width than Cervids have been noted to have a number seen in the contemporary cervoid genus of of features of soft anatomy which are spe- similar size, Amphitragulus), rather than on cialized over the general pecoran condition. any uniquely defined "bovid" autapomor- These are: the absence ofa gall bladder (Flow- phy. er, 1875); the possession of a placenta with Bovid limbs, too, show little in the way of few cotyledons (as opposed to many cotyle- uniquely characterizing specializations. dons as is typical ofother pecorans) (Brooke, Heintz (1963, 1970) describes a number of 1878); the absence of an ileocecal gland; and features that may be used to distinguish bo- 21/2 (as opposed to 31/2) colic coils (Garrod, vid from cervid limbs in a Pleistocene assem- 1877) (character suite 34). These specialized blage, but our examination ofthese character features are shared by Hydropotes (Forbes, states in a range of living and fossil rumi- 1882; Garrod, 1877), clearly indicating that nants shows that, while many of the defining this genus belongs in the Cervidae. cervid characters represent an autapomorphic condition for the family, the bovid char- DEFINING CHARACTERS OF acters are mainly plesiomorphic characters of LIVING PECORAN FAMILIES the Pecora. A possible derived feature of the As previously discussed, cranial append- bovids is the presence of a fused metatarsal ages have obviously evolved independently with an open gully (character 28a). However, within the higher ruminant families. It then as previously discussed, this condition may follows that the homed ruminants cannot be represent the plesiomorphic condition for the assumed to have a single common ancestor, Eupecora, and is also shared with the Giraf- and other characters should be found to iden- foidea. tify the sister taxa of the homed families The upper molars ofrelatively brachydont among the homless ruminants. In this sec- bovids (such as the first bovid genus Eotra- tion, we will review the characters that have gus) can be distinguished from the giraffids been used to identify living and fossil pecoran by the presence of an entostyle, but this fea- families. ture also characterizes brachydont cervoids. The auditory bulla of many bovid genera is distinctive, with the attachment ofthe styloid BOVIDAE process toward the anterior edge, in contrast Bovids are undoubtedly the most success- to the more posterior positioning in the cer- ful ruminant family today, in terms of mor- vids. However, many living bovid species phological and ecological diversity, and many (e.g., many species ofthe genus Cephalophus) genera possess the derived characters ofhyp- lack any expansion ofthe auditory bulla, and sodont cheek teeth (character 12b) and elon- so this character cannot be taken as diagnos- gated limbs with loss of the lateral digits tic of the family. (character 29d). Consequently, it tends to be In summary, it appears that no apomorph- assumed that bovids are "advanced" peco- ic characters, apart from the presence ofhorn rans but, as previously discussed, such "ad- cores, can be used to characterize the Bovi- vanced" features characterize many open dae, and on the basis of our presently avail- habitat ungulates. Bovid genera can be united able evidence, it is impossible to recognize a by their type of cranial appendages, or horns hornless bovid ancestor with any degree of (character 7b). However, as far as we can de- certainty. termine, no other morphological feature uniquely characterizes the Bovidae. Eotra- gus, the earliest known bovid, was identified GIRAFFIDAE by the presence of postorbital horn cores. The Giraffoidea was defined by Hamilton Ginsburg and Heintz (1968) identify Eotra- (1978a) to include all ruminants possessing 48 AMERICAN MUSEUM NOVITATES NO. 2893 a bilobed lower canine (character 23). This outside the orbital rim (character 9); lacrimal comprises the Miocene East African genera fossae (character 8); antorbital or ethmoidal Climacoceras, Nyanzameryx, and Canthu- vacuity, cutting off the lacrimal from artic- meryx, probably also the genus Injanatheri- ulation with the nasals; first molar brachy- um from Iraq (Heintz et al., 1981), assigned dont; the parietosquamosal suture nearer the to the family Climacoceridae, and the true upper than lower border ofthe temporal fos- giraffid genera, of which Palaeotragus is the sa; placenta with few cotyledons (character earliest known and most primitive. Hamilton 34). Flower (1875) noted the absence ofa gall also included the Miocene European Tricero- bladder (character 34), and Heintz (1963) meryx in the Giraffoidea, but not in the noted the presence ofa closed metatarsal gul- Giraffidae, on the basis of p4 morphology ley (character 28b). Living cervids, with the (characters 24 and 25) alone, as the anterior exception ofHydropotes, are characterized by dentition is unknown. However the use of the presence of antlers (character 7d). Living premolar morphology to assign ruminants cervids, including Hydropotes, also possess a without question to the Giraffoidea has been lacrimal depression or fossa and an antorbital refuted (Janis and Lister, 1985). vacuity, although, as previously discussed, As previously discussed, living and fossil these characters are not unique to cervids or giraffids possess distinctive cranial append- cervoids. ages or ossicones (character 7a) but it seems While these features, excluding the posses- unlikely that these are homologous with the sion of antlers, typify living cervids (defined cranial appendages of the giraffoids Cli- to include Hydropotes but to exclude Mos- macoceras and Nyanzameryx. Other giraf- chus), none ofthem is a unique defining char- foids, such as Zarafa and Injanatherium, acter of the group that would clearly distin- genera which may both be synonymous with guish them from fossil "cervoids." The the genus Canthumeryx, known only from dromomerycids (Scott and Janis, 1987) and dental remains (Hamilton, 1978a; Heintz et the antilocaprines (Leinders and Heintz, al., 1981), have cranial appendages which are 1980) share the characteristics of the "cer- peculiar in their dorsolateral orientation, but void" type double lacrimal orifices and the which present no evidence as to their mode antorbital vacuity. The earliest antlered cer- of formation. Within the Giraffidae, the cra- vids, such as Dicrocerus and Procervulus, pos- nial appendages of the Sivatherinae may not sess a protocone with a posterior bifurcation be homologous with the ossicones ofthe oth- (character 19a) but this is also seen in Am- er giraffids. In addition, all giraffoids possess phimoschus (Leinders, 1983). Living cervids the derived condition of a very flat and elon- possess a posterior tuberosity on the meta- gated posterior cubonavicular facet on the tarsal (character 31) and a distinct raised lip proximal metatarsus (character 32b). All gi- on the posterior cubonavicular facet (char- raffoids possess an open type of metatarsal acter 32d), and lack a posterior diarthrodial gully (character 28a). facet (the facet for the articulation ofthe rem- In summary, the Giraffoidea can be char- nant of metatarsal I, character 30c; see fig. acterized by the presence of a bilobed lower 7). However, the first two features are also canine and the form ofthe proximal and dis- seen in a number of fossil cervoids, such as tal metatarsus; the Giraffidae can be identi- Palaeomeryx and the dromomerycids. The fied within this group by two features oftheir latter feature, along with a number of the p4 morphology, although these are not unique cervid limb characters described by Heintz features within the Pecora. The type of os- (1963, 1970), may be characteristic ofRecent sicone found in living giraffids is probably an and Pleistocene antlered cervids, but have autapomorphic character for the Giraffidae, obviously evolved within the family, and with the possible exception of the Sivatheri- these features do not necessarily characterize nae, but cranial appendages evidently arose the limbs of Tertiary antlered cervids. more than once within the superfamily. Most of the cervid characters can be used to unite a number offossil genera with living cervids in a superfamily Cervoidea (see Scott CERVIDAE and Janis, 1987). However, the presence The Cervidae were diagnosed by Brooke ofdeciduous antlers is the only feature which (1878) as follows: two lacrimal orifices, on or can define a true cervid in the fossil record, 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 49 leaving the relative position ofHydropotes in ofMoschus that they consider make it on the some confusion with regard to fossil cervoids one hand more advanced than Gelocus (such (see Leinders, 1983). At this level the pres- as a large supraorbital fissure and the loss of ence of the lacrimal fossa may serve to unite the promontory artery, part ofcharacter suite it with antlered cervids, despite the parallel 6), and on the other hand more primitive evolution ofthe lacrimal fossa within the oth- than the homed ruminants (such as the re- er pecoran genera. As previously mentioned, tention of a subarcuate fossa on the endo- Hydropotes is clearly allied with the Cervidae cranial side of the petrosal and the retention by features of its soft anatomy, but such fea- ofa median branch ofthe carotid artery; that tures may also have characterized fossil cer- is, lacking the rest ofcharacter suite 6). How- voids above the level of Moschus or Antilo- ever, as they base their conclusions on a com- capra. parison of Moschus with Bos and Ovis, further work needs to be done to elucidate MOSCHIDAE whether those features are common to all of the Eupecora (sensu Webb and Taylor, 1980) This family is represented today by a single or are merely autapomorphic ofthe Bovidae. genus, Moschus (musk deer). Moschus pos- We follow Leinders and Heintz (1980) in con- sesses a closed metatarsal gully (character sidering the Moschidae to belong to the su- 28b), a posterior tuberosity on the metatarsal perfamily Cervoidea. (character 31), and a raised lip on the posterior cubonavicular facet (character 32d), but has only a single lacrimal orifice. It lacks any ANTILOCAPRIDAE form of cranial appendage, possesses large The family Antilocapridae consists of two sabrelike canines (character 11 b), and retains subfamilies, the Miocene Merycodontinae the gall bladder. Leinders (1979) and Lein- and the Miocene to Recent Antilocaprinae. ders and Heintz (1980) consider that the pres- As previously discussed, both possess su- ence of a closed metatarsal gully indicates praorbital cranial appendages, but these ap- that Moschus should be placed in the Cer- pendages differ considerably in morphology, voidea, but that the single lacrimal orifice and may not be homologous. Antilocaprids excludes it from the Cervidae. traditionally have been allied with the Bovi- Webb and Taylor (1980) expanded the dae, especially by authors familiar with living Moschidae to include other small homless species only, on the basis of the horn cores genera with sabrelike canines. These are the and the keratinous coverings (e.g., O'Gara North American blastomerycines (Para- and Matson, 1975), and the hypsodont cheek blastomeryx, Blastomeryx, Problastomeryx, teeth and the loss of the side toes (e.g., Mat- Pseudoblastomeryx, Longirostromeryx, and thew, 1904; Pilgrim, 1941). But, as previ- Machaeromeryx) and the European genera ously discussed, none of these features can Amphitragulus and Dremotherium. They cite serve as synapomorphies to unite the Antilo- the laterally enclosed, subcentral tympano- capridae with the Bovidae. O'Gara and Mat- hyal on the auditory bulla as the autapo- son (1975) also link antilocaprids with bovids morphous condition which unites the family based on the presence of a gall bladder and (character 10). However, while this feature the supposedly single lacrimal orifice. How- appears to be shared by Moschus and the ever, these characters are plesiomorphic for blastomerycines and probably to a lesser ex- ruminants, and the lacrimal orifice (character tent by Dremotherium, the auditory bulla of 9) is in fact double in most specimens ofAn- Amphitragulus is more like a primitive cer- tilocapra. Leinders and Heintz (1980) use this void, with a posteriorly positioned tympa- character and the closed metatarsal gully nohyal vagina with little enclosure (Webb, (character 28b) typical of all antilocaprines personal commun.). The sabrelike canines to classify antilocaprids in the superfamily cannot be used to unite these genera into a Cervoidea. We agree with this placement, but single clade, for, as already discussed, this is the phylogenetic position of the Antilocapri- a primitive cervoid feature, although it may dae within the Cervoidea is still problemat- be used to identify cervoids of moschid or ical. No unique features can be found to de- higher status. Webb and Taylor (1980) men- fine the Antilocapridae (other than the tion a number offeatures ofthe basicranium biogeographic feature of only occurring in 50 AMERICAN MUSEUM NOVITATES NO. 2893

PECORA - - ..d- CERVOIDEA r >~~

,3134

a / 11B 20 28B 32D

3156

x3 521 27 28A,B 29B630A 336

1A 11A 32A 33A

Fig. 9. Cladogram of living pecoran families. See table 1 for key to characters.

North America!). Similarly, no unique fea- dontines in the complete loss of the side toes tures can be found to link the merycodontines in all genera, and in the larger body size of with the antilocaprines. most genera, but these features can hardly be Fossil antilocaprids appear to have had a used as antilocaprine autapomorphies. We double lacrimal orifice, although this detail will return to a discussion ofthe phylogenetic is often difficult to ascertain in fossil speci- position of the Antilocapridae in the light of mens, as the orbital rim is frequently broken, other cervoid characters in a later section. and the orbit itself usually filled with matrix. However, in the collections of the American Museum of Natural History we have ob- PHYLOGENETIC RELATIONSHIPS served probable double lacrimal orifices in OF LIVING PECORAN FAMILIES the following specimens: With regard to the distribution ofcharacter states among the living pecoran families and Antilocaprines: "Plioceros" dehlini (AMNH: following our review of the differences in de- AINS 793: Pros. 500.40), "Plioceros" tex- velopment of the cranial appendages, taking anus (F:AM 52089), and Stockoceras the position that these appendages have been onusrosagris (F:AM 42250). evolved independently within each ofthe liv- Merycodontines: Paracosoryx wilsoni (F:AM ing families, we would construct a cladogram 31191), Meryceros joraki (F:AM 31163), of the living pecoran families as follows (see and Cosoryx furcatus (F:AM 32902 and fig. 9). 51025). The Pecora can be united within the Rumi- The condition appears to be variable in mer- nantia by the following features: a compact, ycodontines, as in the living genus Antilo- parallel-sided astragalus (character 5); elon- capra, as only a single orifice is apparent in gated, fused metapodials with an open meta- a specimen of Paracosoryx wilsoni (F:AM tarsal gulley (character 28a) and complete 32899-A). Both subfamilies have a closed distal metapodial keels (character 27); a four- metatarsal gully as the primitive condition, chambered stomach with an omasum (char- but this character state is variable in most acter 33b); and a suite ofderived molar char- merycodontine genera. Antilocaprines are acters such as a lingually situated protocone somewhat more "advanced" than meryco- on P3 (character 14b), a large M3 metaconule 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 51

(character 16a), the reduction or absence of the bifurcation of the posterior wing of the an internal cingulum on the upper molars protocone in primitive forms (character 19a), (characters 13b, c), metastylids on the lower and the absence of the gall bladder (along molars (character 21), and the postentero- with other features ofsoft anatomy discussed cristid usually complete (character 22b). previously, character 34). Cervids above the The Giraffidae can be characterized by the level of Hydropotes are characterized by the presence ofa bilobed lower canine (character possession of antlers (character 7d). 23), a flat cubonavicular facet on the proxi- Figure 9 summarizes this information in a mal metatarsus (character 32b), and ossi- cladogram of the living pecoran families. cones which preform in cartilage (character A consideration ofthe distribution ofchar- 7a). Other pecoran families are united by the acter states among fossil pecorans does not presence of an entostyle in the upper molars alter this scheme of the interrelationships of (in brachydont forms only), derived from the the living families, but does result in the ad- anterior face ofthe metaconule (character 15), dition of information to the cladogram, sup- and may also be united by the more ventral porting the use of these particular characters entrance of the esophagus to the rumen (see at the nodes, in some cases, and refuting the discussion in Characters section). nodal position ofcertain characters in others. The Bovidae are characterized by the pos- In particular, the suggestion that the posterior session ofhorns with a nondeciduous keratin tuberosity of the metatarsal evolved in par- sheath (character 7b). No other morpholog- allel within the Cervoidea is confirmed. The ical features can be found to uniquely define position of the Giraffidae is also in accor- this family. Other pecoran families can be dance the conclusion reached by Todd (1975) united into the Cervoidea by the possession on the basis of chromosomal studies of the of the following characters: a closed meta- Ruminantia, who claimed that the Giraffidae tarsal gulley (character 28b); a sabrelike up- represented the most primitive pecoran fam- per canine in forms lacking cranial append- ily in the absence of an X-autosome trans- ages (character 1 lb); a raised lip to the location fusion (seen also in the Tragulidae, cubonavicular facet on the proximal meta- but not in other pecorans). However, despite tarsus (character 32d); and the presence of a the concurrence of Todd's conclusion with Palaeomeryx fold in brachydont members of our own, based on morphological characters, the lineages (character 20). (This latter feature we reserve doubts about the validity of this is assumed in the case ofthe Antilocapridae, chromosomal character as a true synapo- as all known fossil taxa are highly hypso- morphy ofall living pecorans above the level dont.) of the Giraffidae (see discussion in Scott and The Moschidae are distinguished by the Janis, 1987). presence of a laterally enclosed, subcentral tympanohyal vagina (character 10), and have DEFINING CHARACTERS OF developed a posterior tuberosity on the meta- FOSSIL FAMILIES tarsus (character 31) in parallel with the Cer- vidae. The Cervidae and Antilocapridae can GELOCIDAE be united by the possession of a double lac- Pecorans ofthe gelocid grade comprise the rimal orifice on the orbital rim (character 9). following Oligocene genera: Gelocus [Europe Antilocaprids can be characterized only by and possibly Africa (see Hamilton, 1973)]; the unique form of their cranial appendages, Notomeryx [Asia (see Savage et al., MS)]; Go- with a deciduous keratin sheath covering a biomeryx [Asia (see Sudre, 1984)]; Prodremo- bony horn core in the living species, Antilo- therium (Europe); Eumeryx (Asia); and Ru- capra americana. This bony horn core, while titherium [Europe (see Sudre, 1984)]. unbranched in Antilocapra (although the de- Many authors consider the Gelocidae to be ciduous keratin sheath is forked in this the basal group ofhigher ruminants, although species), had a propensity for branching in of traguloid status (e.g., Simpson, 1945; Vi- fossil taxa (character 7c). The Cervidae are ret, 1961; Romer, 1966). Webb and Taylor characterized by the possession ofa posterior (1980) transferred this group to the Pecora tuberosity on the metatarsus (character 31), on the basis of a number of advanced basi- 52 AMERICAN MUSEUM NOVITATES NO. 2893 cranial features (character 6) and the posses- polyphyletic group composed of taxa of var- sion of a compact, parallel-sided astragalus ious phylogenetic affinities within the Pecora. (character 5). They consider that the gelocids form the sister group to the other pecorans, and that they are united into a distinct clade PALAEOMERYCIDAE AND LAGOMERYCIDAE by the presence of a narrow-waisted, backwardly situated protocone on P3 (character The Palaeomerycidae were originally de- 14a). Janis (1987) has argued that this char- fined by Lydekker (1883) in his description acter cannot be used to unite the family, as ofPropalaeomeryx sivalensis, taking the name it is also present in the traguloid genera Lo- of the family from Palaeomeryx Von Meyer, phiomeryx and Bachitherium, and suggests 1834. Roger (1904) established the genus that the gelocids are better considered a prim- Lagomeryx to include small species of "Pa- itive grade of pecorans. laeomeryx" with branched cranial append- Gelocids are further characterized by the ages. Zittel (1925) and Colbert (1936) origi- following features: the apparent absence of nally considered Lagomeryx to be a cervoid cranial appendages; moderately large, but not belonging to the subfamily Cervulinae sabrelike canines (character 11 a); brachydont (=Muntiacinae). Roman and Viret (1934) cheek teeth (character 12a); the presence of later suggested that the two genera were syn- a small premolariform p1 separated from p2 onymous. Genera added to this family later by a small diastema (character 4b); lower pre- in the 1900s included Palaeomeryx, Blasto- molars in which the metaconid is small, and meryx, Dremotherium, Dicrocerus, and Mi- does not extend backwards to form a postero- cromeryx. Teilhard de Chardin (1939) sug- lingual wall to the tooth (character 3; see fig. gested the formation of a separate family to 6); lower molars which possess an anterior include Lagomeryx and Procervulus, and Pil- cingulum (character 2b) and an ectostylid (part grim (1941) proposed the name Lagomeryc- ofcharacter suite la, see fig. 6), and in which idae as a family to include Lagomeryx, Pro- the postentocristid is usually incomplete cervulus, and Climacoceras. (The latter genus (character 22b); metastyles on the upper mo- was previously assigned to the Cervidae by lars (character 16a) and the absence of a dis- Maclnness, 1936.) Both Teilhard de Chardin tinct entostyle; a small metaconule on M3 and Pilgrim considered the Lagomerycidae (character 18a); the presence of a lingual cin- and the Palaeomerycidae to have giraffoid gulum on the upper molars (character 13a); affinities. However, Stirton (1944) included and fused, somewhat elongated metatarsals the North American dromomerycids in the (character 28d), with the absence ofcomplete family Palaeomerycidae as the subfamily distal keels. However, all of these features Dromomerycinae and added the Old World represent characters which are either primi- genera Lagomeryx, Procervulus, Climacocer- tive for the Pecora or primitive retentions of as, Amphitragulus, and Dremotherium to the conditions seen in various traguloid genera, subfamily Palaeomerycinae. He considered and none can be considered to uniquely char- the palaeomerycids to form the basal pecoran acterize the "Gelocidae" (see Janis, 1987). "cervoid" group from which the moschids, Gelocus is apparently the most primitive ge- cervids, and giraffids were derived. On the nus, as it lacks metastylids in the lower mo- other hand, Simpson (1945) placed the Pa- lars that are characteristic of all other Pecora laeomerycinae as a subfamily of the Cervi- (character 21). Prodremotherium and Eu- dae, including Palaeomeryx, Amphitragulus, meryx possess the derived cervoid feature of Dremotherium, Eumeryx, and the North a closed metatarsal gully (character 28b) and American blastomerycine genera; he consid- also possess an incipient entostyle (character ered the genera Lagomeryx, Procervulus, and 1 Sa). Eumeryx and Rutitherium possess the Climacoceras to belong to the family Lago- derived feature of a Palaeomeryx fold (char- merycidae within the Giraffoidea (see Ham- acter 20). The condition of the metatarsals ilton, 1978b; Scott and Janis, 1987). in Rutitherium is unknown. In our opinion, Whitworth (1958) supported Stirton's con- the Gelocidae as presently considered is a tention that Palaeomeryx and Lagomeryx 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 53 were synonymous, and considered the family sider the position of the Bovidae in this Palaeomerycidae to be closely related to the scheme). Cervidae. Ginsburg and Heintz (1966) con- Hamilton (1978b) later pointed out that sidered Palaeomeryx to be a separate genus the Palaeomerycidae as defined previously from Lagomeryx, and restated the giraffoid had no unifying characters. Most "paleo- affinities of Palaeomeryx on the basis of a merycids" have a Palaeomeryx fold, as the giraffidlike ossicone found in association with nomenclature suggests but, as previously dis- other material of the genus. Crusafont-Pairo cussed, this has long been considered a prim- (1952) placed the North American Blasto- itive pecoran character. He removed Cli- merycinae in the Palaeomerycidae, and the macoceras to the Giraffoidea on the basis of Dromomerycidae in the Giraffoidea. Gins- the bilobed lower canine (character 23), and burg and Heintz (1966) removed the genera pointed out that the large species of Palaeo- Walangania and Heterocemas from the ge- meryx found in Africa was in fact synony- nus Palaeomeryx, and suggested that those mous with the giraffoid Canthumeryx. The Oligocene genera which lacked cranial ap- other "palaeomerycid" described from Af- pendages (Dremotherium, Amphitragulus, rica by Whitworth (1958), Palaeomeryx afri- and the blastomerycids) should be placed in canus, was synonymized by Hamilton (1973) the family Dremotheriidae. Hamilton (1973) with Walangania gracilis, which he tenta- proposed that a superfamily Dremotherio- tively assigned to the Bovidae. Hamilton sug- idea should be erected to include the New gested that all other genera ascribed to the World Blastomerycidae and the Old World Palaeomerycidae be treated as Pecora incer- Dremotheriidae, refuting the affinity of the tae sedis. Subsequently, Webb and Taylor Mongolian genus Eumeryx with the blasto- (1980) assigned the blastomerycines, plus merycids that was suggested by Simpson Amphitragulus and Dremotherium, to the (1945). Viret (1961) assigned Eumeryx to the Moschidae. Neither Hamilton nor Webb and Gelocidae, and the lagomerycid genera Pro- Taylor attempted to deal with the North cervulus and Heterocemas to the Cervidae. American dromomerycids in these consid- Hamilton (1973) assigned the Palaeomer- erations. In our review of cervid interrela- ycidae to the Giraffoidea. He agreed with two tionships (Scott and Janis, 1987) we sug- ofWhitworth's (1958) assertions that the fea- gested that the following genera should remain tures uniting Palaeomeryx with giraffids incertae sedis until further information is ("unequal development of anterior and pos- available: Palaeomeryx, Lagomeryx, Procer- terior external ribs on the upper molars" and vulus, Prolibytherium, Propalaeoryx, and "characteristically corrugated enamel of the Triceromeryx. teeth") could well be primitive pecoran fea- Since the 1978 paper by Hamilton the fam- tures seen in early giraffoids and cervoids ilies Lagomerycidae and Palaeomerycidae alike. However, Hamilton did not support have been more or less abandoned by current Whitworth's third statement that "the pos- workers in the field of ruminant taxonomy session of non-deciduous, velvet-covered (e.g., Leinders, 1983). However, new fossil antlers ... was as likely to be the primitive finds have thrown further light on the identity cervoid condition as the giraffoid" (Whit- ofthe problematical genera Palaeomeryx and worth, 1958: 19). Hamilton instead support- Lagomeryx. A complete skull of Lagomeryx ed the assertion of Ginsburg and Heintz from the middle Miocene of China (Chow (1966) that the "ossicones" of Palaeomeryx and Shih, 1978) shows a number of features were a true synapomorphy shared with the suggestive of true cervid affinities. The genus Giraffidae. Hamilton (1973) saw the Oligo- possessed nondeciduous antlers, consisting of cene Dremotheriidae giving rise to three dif- a long, supraorbital pedicle supporting a whorl ferent lineages: the Old World Giraffoidea of palmated tines at the tip; a lacrimal fossa; (including the Palaeomerycidae and presum- large, muntjaklike canines; and lower molars ably also the Lagomerycidae), Cervoidea, and possessing a Palaeomeryx fold. They consid- a New World lineage of the Blastomerycidae er the skull to be very similar overall to the and the Dromomerycidae (he did not con- living cervids Elaphodus (the tufted deer) and 54 AMERICAN MUSEUM NOVITATES NO. 2893

GIRAFFOIDEA A

GIRAFFOIDEA CERVIDAE c _~~ ~ ~ ~ ~~~1 D

GIRAFFOIDEA E

Fig. 10. History ofideas on phylogenetic position offossil Eupecoran genera (sensu Webb and Taylor, 1980). (Position of Bovidae is excluded.) A. Composite diagram incorporating views of Teilhard de Chardin (1939) and Pilgrim (1941). "Lagomerycidae" includes the genera Lagomeryx, Procervulus, and Climacoceras. "Palaeomerycidae" includes the genera Palaeomeryx, Blastomeryx, Dremotherium, Am- phitragulus, Dicrocerus (placed in Cervidae in all later schemes), and Micromeryx. B. View of Stirton (1944). "Palaeomerycidae" includes the divisions "Dromomerycini" (considered to be somewhat closer to the giraffids) and "Palaeomerycini" (considered to be somewhat closer to the cervids). "Dromomer- ycini" includes the dromomerycids and the blastomerycids. "Palaeomerycini" includes the genera Pa- laeomeryx, Procervulus, Climacoceras, Dremotherium, and Amphitragulus. C. Composite diagram incorporating the views of Simpson (1945) and Whitworth (1958). "Lagomerycidae" includes the genera Lagomeryx, Procervulus, and Climacoceras. "Palaeomerycinae" includes the taxa Palaeomeryx, Am- phitragulus, Dremotherium, Eumeryx (considered to be a gelocid by other authors), Walangania (=Pa- laeomeryx africanus), and the blastomerycids. Propalaeoryx included in the Bovidae. D. Composite diagram incorporating the views ofGinsburg and Heintz (1966) and Crusafont (1952): "Dremotheriidae" includes the genera Dremotherium and Amphitragulus. "Palaeomerycinae" includes the blastomerycids. Triceromeryx is included in the Giraffidae. E. Composite diagram incorporating the views ofViret (196 1) and Crusafont (1961): "Lagomerycinae" includes the genera Lagomeryx, Procervulus, and Climacoceras. "Palaeomerycinae" includes the Blastomerycini and the Palaeomerycini, the latter group containing the genera Palaeomeryx, Amphitragulus, and Dremotherium. The "other cervid subfamilies" include the 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 55

F djt - 9<, - V7-ov-

G / G

GIRAFFOIDEA CERVOIDEA

H 0~~~~~~0D GIRAFFOIDEA CERVOIDEA

Muntiacinae, containing the extinct genera Dicrocerus, Euprox, Heteroprox, and Stephanocemas. F. View of Hamilton (1973). "Dremotheriidae" includes the genera Dremotherium and Amphitragulus. "Pa- laeomerycidae" includes the genera Palaeomeryx, Zarafa, Climacoceras, and Propalaeoryx. Tricerom- eryx and Prolibytherium are included in the Giraffidae. Walangania (=Palaeomeryx africanus) is included in the Bovidae. G. Composite diagram combining the views of Hamilton (1978b) and Webb and Taylor (1980). "Moschidae" includes the blastomerycids, the living genus Moschus, and the fossil genera Am- phitragulus and Dremotherium. "Giraffidae" includes the fossil genera Triceromeryx, Climacoceras, and Canthumeryx (=Zarafa). Walangania is included in the Bovidae. Genera considered as incertae sedis are: Palaeomeryx, Lagomeryx, Procervulus, Prolibytherium, and Propalaeoryx. H. Current viewpoint, incorporating views ofLeinders (1983), Qiu et al. (1985), and Chow and Shih (1978). "Palaeomerycidae" includes the genera Palaeomeryx and Triceromeryx. "Moschidae" is used sensu Webb and Taylor (1980). "Hoplitomerycidae" includes the genus Hoplitomeryx (Leinders, 1983). Lagomeryx and Procervulus are included in the Cervidae (subfamily Muntiacinae). Walangania is included in the Bovidae. Proliby- therium and Propalaeoryx remain as genera Pecora incertae sedis. I. View proposed in this paper. "Climacoceridae" includes the genera Climacoceras, Nyanzameryx, Injanatherium, and Canthumeryx (=Zarafa). Propalaeoryx is considered as the sister taxon to the Giraffoidea. "Moschina" is a paraphyletic assemblage, containing the monophyletic Moschidae (including the blastomerycids, Moschus, and the fossil genera Dremotherium and Micromeryx) and Walangania as the sister taxon to the Moschina plus Eupecora. "Palaeomerycidae" includes the dromomerycids and the fossil genera Palaeomeryx, Amphi- tragulus, and possibly also Prolibytherium. ("Palaeomeryx" also includes the isolated teeth from Gebel Zelten ascribed to Zarafa; Hamilton, 1973.) "Hoplitomerycidae" includes the genera Hoplitomeryx and Amphimoschus. Lagomeryx and Procervulus are included in the Cervidae. Triceromeryx remains as Cervoid incertae sedis. 56 AMERICAN MUSEUM NOVITATES NO. 2893

Muntiacus (the muntjak), and consider Lago- of the double lacrimal orifice can be unques- meryx to belong in the Muntiacinae (=Cer- tionably observed in specimens of all three vulinae) within the Cervidae. dromomerycid subdivisions (see Frick, 1937): A complete skull and several complete in the Aletomerycini (e.g., Aletomeryx sp., skeletons of a small species of Palaeomeryx F:AM 42883), the Dromomerycini [e.g., Sub- (Palaeomeryx tricornis) have recently been dromomeryx scotti (type), F:AM 33758], and described from the Miocene ofChina (Qiu et in the Cranioceratini (e.g., Cranioceras granti, al., 1985), confirming the assertion by Gins- F:AM 31270). The giraffidlike features oftheir burg and Heintz (1966) that this genus had cranial appendages evidently represent an ex- giraffidlike ossicones. They agree with Gins- ample ofparallel evolution. Dromomerycids burg and Heintz in the placement ofPalaeo- also possess cervoid features of the limbs in- meryx within the Giraffoidea, although not- cluding, in addition to the closed metatarsal ing that it lacked the giraffid autapomorphy gulley, a somewhat raised lip to the cubon- of a bilobed lower canine. We will discuss avicular facet (character 32d), and a posterior our hypotheses of the phylogenetic affinities tuberosity to the metatarsus (character 31) in Palaeomeryx and other "palaeomerycid" most genera (see later discussion). They pos- genera in a later section and attempt to argue sess the dental character ofa sabrelike canine for the resurrection of the Palaeomerycidae, (character 1 b) in the early genera (Barbou- although on the basis of different characters romeryx and Aletomeryx), plus a Palaeo- than those in previous use. meryx fold (character 20) in the more brachydont genera. They are excluded from the Cer- DROMOMERYCIDAE vidae by the lack of a naked, deciduous portion of the cranial appendages and the more Dromomerycids were an endemic North ventral position of the parietosquamosal su- American Miocene ruminant assemblage, ture (Frick, 1937). characterized by supraorbital nondeciduous We would define dromomerycids as ru- unbranched cranial appendages (character 7f), minants possessing general cervoid postcra- with the presence of a median occipital cranial and dental features, including a double nial appendage in the subfamily Cranioce- lacrimal orifice, an elongated occipital region, ratinae (Webb, 1983b). They were considered and supraorbital unbranched, nondeciduous cervoids during the early part ofthis century cranial appendages. Unfortunately, as with (e.g., Matthew, 1926; Frick, 1937) but were the Antilocapridae, no unique autapomor- classified by Stirton (1944) as a subfamily of phies are known that unequivocally unite the Palaeomerycidae, which he allied with these genera into a single clade, but they are the Giraffidae on the basis of the similarities unified by their biogeographic occurrence in of the cranial appendages. Since that time the Miocene of North America. Their phy- most authors have allied them with the gi- logenetic position in relation to the cervid raffids (e.g., Crusafont-Pairo, 1952, 1961; Vi- lineages will be discussed further in a later ret, 1961; Hamilton, 1978a), although Ro- section. mer (1966) classified them with the cervoids. Since the concept of the family Palaeo- merycidae was abandoned, following Ham- HOPLITOMERYCIDAE ilton (1973, 1978b), dromomerycids have Leinders (1983) has defined this family tended to be referred to as a discrete family, based on the finding ofa new ruminant genus, the Dromomerycidae (e.g., Hamilton, 1978b; Hoplitomeryx matthewi from the Late Mio- Webb and Taylor, 1980; Janis, 1982; Webb, cene island fauna of Monte Gargano (Italy). 1983b; Leinders, 1983), but their status as a The animal appears to have the cervoid fea- family has never been formally defined. tures of a closed metatarsal gully (character Dromomerycids have the derived cervid 28b), a cervidlike double lacrimal orifice on features of a closed metatarsal gully (char- the orbital rim (character 9), and a sabrelike acter 28b), a double lacrimal orifice (char- upper canine (character 1 b), but possesses acter 9), and can clearly be considered cer- five cranial appendages (two pairs of post- voids (see Janis, 1982; Leinders, 1983; Scott supraorbital appendages and a median nasal and Janis, 1987). A cervidlike condition appendage) which resemble the nondecid- 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 57 uous, keratin-covered horn core otherwise try (in Hendey, 1978) has expressed the view found in the Bovidae. (The deep grooving of that while there may be no reason to debar the horn cores of Hoplitomeryx is reminis- Walangania from bovid ancestry (as it is a cent ofthe condition of bovid horn cores with suitably hyposodont pecoran, predating the a permanent keratin sheath, rather than the appearance of the first horned bovid in Af- spongy texture ofthe antilocaprid horn core.) rica), there is also no reason to assume it Leinders considers the genus to be closely might not be a small moschid, and considers related to the Cervidae on the basis of these that it may be synonymous with Amphitra- cervoidlike cranial and postcranial features, gulus or Dremotherium. but to be excluded from the family by pos- Palaeomeryx africanus (Whitworth, 1958) session of apparent horn cores rather than was synonymized with Walangania gracilis antlers, and places it in a family of its own. (Whitworth, 1958) by Hamilton (1973). De- He also considers that, on basicranial and spite Hamilton's classification, we prefer to dental grounds, the pecoran genus which most regard Walangania africanus and Walan- closely resembles Hoplitomeryx is the poorly gania gracilis as separate species for the time known European genus Amphimoschus. Both being, as they do possess different morpho- possess a large, bicuspid third lobe on m3 in logical characteristics (for example, the pres- which the median valley between both cusps ence of a Palaeomeryx fold and an incipient is open posteriorly (character 26b), which ap- entostyle in Walangania africanus.) Whit- pears to be a unique feature among the Pecora worth originally assigned the larger species to (see fig. 6), and share a large, inflated, smooth- the genus Palaeomeryx because of the pres- surfaced auditory bulla with a number ofsim- ence of a Palaeomeryx fold and the cervoid ilarities in structural detail. Both genera lack nature of the supposedly associated meta- a Palaeomeryx fold, but both have fairly high- tarsus (which possesses a closed metatarsal crowned molars, so this probably represents gully), whereas Walangania gracilis lacks a a secondary loss from the primitive cervoid Palaeomeryx fold. Hamilton (1973) synon- condition, and both have lost p1 and possess ymized the two genera and placed them ten- unmolarized premolars (in contrast to the tatively in the Bovidae, on the supposition condition observed in more advanced cer- that the Palaeomeryx fold is a primitive pec- vids and dromomerycids). Amphimoschus, oran character. While we consider the Pa- unlike Hoplitomeryx, has a bifurcated pro- laeomeryx fold to be a derived cervoid char- tocone (character 19a), the retention of p2, acter, we do not consider this character to and a strongly developed entostyle and ec- unite Walangania with Palaeomeryx, al- tostylid (Leinders, 1983). though we do consider this character to ally Figure 10 summarizes the historical de- Walangania with the Cervoidea. Its absence velopment of ideas about the systematic po- in Walangania gracilis is probably related to sition of fossil Eupecoran taxa. the greater degree ofhypsodonty in the smaller species. PROBLEMATICAL RUMINANT The available material of Walangania pre- GENERA AND FAMILIES sents an interesting assemblage ofcharacters. Material of Walangania africanus from Ru- WALANGANIA singa includes an upper dentition in which Walangania is a small ruminant known M3 has a small metaconule and the upper from partial remains from the lower Miocene molars generally lack an entostyle, although of East Africa (Whitworth, 1958) and also a small entostyle is apparent in BMNH 35251. possibly from South Africa (Hendey, 1978). The metastyle is not pronounced, resembling The genus contains two species: Walangania the condition in Eumeryx. Another specimen (=Palaeomeryx) africanus and the smaller assigned to this species (BMNH 21364) has Walangania gracilis. Walangania apparently a P3 with a backwardly directed gelocidlike lacked cranial appendages, although a com- protcone resembling the condition in Pro- plete cranium is not known, and has generally dremotherium (see fig. 1 1). The low molars been considered to be a bovid (e.g., Hamil- have a moderate metastylid and small ecto- ton, 1973) largely on the basis of its some- stylids and the postentocristid is incomplete what hypsodont cheek teeth. However, Gen- (see fig. 12). 58 AMERICAN MUSEUM NOVITATES NO. 2893

Fig. 11. Prodremotherium elongatum, right P3-M3, BMNH 1809, Phosphorites du Quercy (early Tertiary), France. Bar equals 10 mm.

Material of Walangania gracilis from more advanced than those ofthe East African Mfwangano includes upper molars (BMNH Walangania. The Arrisdrift specimens have 21379) similar to those of Walangania af- a smaller metastylid, weaker anterior ribs, ricanus, but which have a slight lingual cin- and a complete postentocristid. However, gulum, as seen in gelocids. An immature low- there is also a small distal metapodial frag- er jaw shows a possible alveolus for the ment in the collection at the South African deciduous p 1. A proximal metatarsal as- Museum (PQ-AD-503) that has complete signed to Walangania shows a raised lip on distal keels and appears to have a closed the cubonavicular facet, but lacks a cervid metapodial gully, but we were unable to as- type of posterior tuberosity. Metatarsal II is certain with certainty whether this was a fused, but V is unfused, and a facet is present metacarpal or metatarsal. for the first metatarsal remnant. A distal On the basis of the dental evidence, Wa- metapodial fragment from Mfwangano in the langania appears to be similar to primitive Kenya National Museum, Nairobi, which is cervoid genera such as Eumeryx despite the assigned to Walangania gracilis, shows the relative hypsodonty ofthe teeth. Similarities presence of complete distal keels. include the retention of a slight lingual cin- Walangania africanus is also known from gulum (character 13b) in Walangania gra- Songhor. An unassigned small distal meta- cilis, the backwardly directed protocone on tarsal (BNNH Sgr. 232-1949) from this lo- P3 (character 1 4a), the presence of a Palaeo- cality shows a closed metatarsal gully and meryx fold (character 20), and an incipient complete distal keels. Whitworth (1958) as- entostyle (character 1 5a) in some specimens signed this metatarsal to Walangania afri- of Walangania africanus, and the incomplete canus. However, it could belong to "Gelocus" postentocristid (character 22a). However, the whitworthi, a similar size animal ofuncertain limbs suggest more advanced cervoid affin- affinities in the same fauna. (Gelocus whit- ities; in addition to the closed metatarsal gul- worthi is not known from dental remains at ly (character 28b) (seen also in Prodremo- Mfwangano, so the small postcrania from that therium and Eumeryx), it also apparently had locality are more certainly assignable to Wa- complete distal keels on the metapodials langania.) Walangania may also be present (character 27). We would assign Walangania in the Arrisdrift fauna from the early Mio- to the Cervoidea on the basis ofthe presence cene of South West Africa. Hendey (1978) ofa Palaeomeryx fold (character 20), a closed describes this fauna, and comments on the metatarsal gully (character 28), and the raised similarities ofthe molars to Walangania, but lip on the posterior cubonavicular facet follows Gentry (quoted as personal commun. (character 32d). The presence of complete in Hendey, 1978) in assigning this animal to distal keels (character 27) make it a more the Bovidae because the lower molars are derived cervoid than Eumeryx, but the re- 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 59

I ,omo

I mn

Fig. 12. A. Walangania africanus, left DP3-DP4 (reversed), BMNH 21362, right ?M2, BMNH 21363 and left p3-m3, BMNH 30137 (from cast), Songor (early Miocene), East Africa. B. Propalaeoryx austroafricanus, right P3-Ml, BMNH 369 (cast of PQN 57) and right p2-m3, BMNH 369.59 (cast of PQN 51), Lagental (middle Miocene), Namibia. Bar = 10 mm. "GELocus" WHITWORTHI tention ofthe aforementioned primitive pecoran dental characters relegates it to a more This taxon was described by Hamilton primitive position than other cervoids. There (1973), the type consisting ofa mandible frag- appears to be no evidence for the presence of ment containing a lightly worn m2 and m3. a moschidlike upper canine in this genus. The type specimen and most of the other 60 AMERICAN MUSEUM NOVITATES NO. 2893 material of this taxon are from Songhor, but fold (character 20). The similarities of the teeth assigned to "Gelocus" whitworthi are dentition of "Gelocus" whitworthi to Gelocus also known from Maboko and Rusinga. The communis probably represent primitive pec- taxon is only known from remains ofthe low- oran features. However, we have not made a er molars. thorough study ofthis taxon, and present this "Gelocus" whitworthi differs from the trag- taxonomic assignation as a possible hypoth- ulid Dorcatherium (also present at Songhor) esis, awaiting the discovery of further main the absence of a Dorcatherium fold. The terial. teeth are ofa similar size to Walangania, but differ from his taxon in certain dental features. The metastylid in "Gelocus" whit- PROPALAEORYX worthi is closely conjoined to the metaconid, Propalaeoryx austroafricanus from the Na- and is situated lingual to the posterior end of mib Desert was described by Stromer (1926), the metaconid, resulting in a median valley and he originally assigned it to the Bovidae. on the lower molars that is open lingually. In However, Arambourg (1933) considered the contrast, in Walangania the metastylid is sit- dental characters to be more cervidlike. uated lingual to the anterior end of the ento- Whitworth (1958) described Propalaeoryx conid, resulting in a closed median valley. nyanzae from Rusinga (lower Miocene), al- "Gelocus" whitworthi resembles Gelocus lied this animal with Stromer's Propalaeoryx communis from Europe in the roundness of austroafricanus, and also concluded that the the metaconid, and in the forked posterior genus was more cervoid in nature, due to the end of the entoconid, with the labial branch strong metastylid and entostylid on the lower meeting the posterolingual end of the hypo- molars. He described a metatarsal ascribed conid and the lingual branch being produced to Propalaeoryx as possessing complete distal posterolingually. In the m3 of "Gelocus" keels and as being bovidlike in the housing whitworthi there is a strong entostylid that for the extensor tendon, with a cervidlike runs posterolaterally to fuse with the hypo- grooving on the posterior shaft, but having a conulid, giving a unique appearance among bovidlike anterior shaft with an open meta- Pecora to the posterior part ofthis tooth, and tarsal gully. Hamilton (1973) described Pro- an accessory stylid is present in the labial end palaeoryx from Rusinga as lacking a Palaeo- of the posterior valley. meryx fold and possessing a metastyle and These dental details, and others described entostyle in the upper molars and a small p1 by Hamilton (1973), warrant the distinction separated from p2 by a small diastema. He of "Gelocus" whitworthi from the small con- also concluded that the dentition bears a temporaneous pecoran taxa, such as Wa- stronger resemblance to Palaeomeryx than to langania and Propalaeoryx. However, we bovids, and placed it in the Palaeomerycidae. would question the inclusion of this taxon However, an entostyle is only apparent on with the genus Gelocus. As previously dis- one specimen of Propalaeoryx from Moruo- cussed, the type of Gelocus communis, and rot, consisting of a worn and badly damaged other specimens of this species from Le Puy specimen (BMNH MT 67'5 1). This tooth was (France), lack metastylids, and metastylids originally ascribed to "Palaeomeryx" appear to be a good synapomorphy of pec- (= Walangania) africanus by Whitworth orans above the level ofGelocus as evidenced (1958), and may have been misidentified by by the European fossil material (see Janis, Hamilton. We have never observed an ento- 1987). A few specimens of "Gelocus" whit- style on any other specimens of Propalaeo- worthi have a faint, but definite, Palaeo- ryx, in the range ofmaterial available in Brit- meryx fold. This is visible in the holotype ish and African museum collections. (BMNH K.Sgr. 365-1949) and in the para- Additional characteristics ofPropalaeoryx type (BMNH K.Sgr. 581-1949) from Son- that we have noted include the presence of a ghor. We consider this taxon to have had bifurcated posterior crista of the metaconule affinities with the primitive pecoran taxa Eu- (character 19b) (also depicted in Hamilton, meryx and Rutitherium on the basis of the 1973), a small metaconule on M3 (character derived cervoid character of a Palaeomeryx 1 8a), a P3 with a posterior situated protocone 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 61

(character 14a) seen in the specimen of P. on to the Giraffoidea). The proximal surface austroafricanus (BMNH 36928) from Lan- ofthe metatarsus (character 32) is cervoidlike gental, Namibia (see fig. 12), and the presence rather than giraffoidlike, but this merely rep- of a slightly raised lip on the cubonavicular resents the primitive pecoran condition, and facet (character 32d) (BMNH 36928, also again would not exclude Propalaeoryx from from Namibia). A fragment oftusk like upper affinities with the Giraffoidea. canine from Namibia (BMNH 36963) may In the absence of any other available evi- also be assignable to Propalaeoryx. The bi- dence, we would follow Gentry's suggestion furcated posterior wing of the metaconule is for the time being in placing Propalaeoryx as a character shared with the early giraffoids the sister taxon to the Giraffoidea, but note Palaeotragus and Canthumeryx, although this that the synapomorphic character (19b) of character, as previously mentioned, also ap- the bifurcation of the posterior wing of the pears in certain other pecoran genera as well metaconule is a rather weak one. as in early giraffoids. Despite the assertion ofthe cervoidlike nature ofthe dentition in Propalaeoryx, includ- AMPHITRAGULUS AND DREMOTHERIUM ing Gentry's (1978) comments that the Amphitragulus (Pomel, 1853) and Dremo- shallow ramus of the mandible is more rem- therium (Geoffroy Saint-Hilaire, 1833) were iniscent of the cervid condition than of the small, apparently homless ruminants that bovid one, we see no reason to group Pro- were common in the European middle Oli- palaeoryx with cervoids, as these characters gocene to early Miocene. They are known represent the primitive pecoran condition from dental remains and postcrania that have seen, for example, in Prodremotherium. Pro- been assigned primarily on the basis of the palaeoryx lacks any derived cervoid charac- size sorting. Sigogneau (1968) discusses the ters, such as a Palaeomeryx fold, and the history ofthe genus Dremotherium and notes presence of an open metatarsal gully and the that while the original type first described by lack ofan entostyle make it a more primitive Pomel (1845-1846) consisted of a complete pecoran than Prodremotherium. However, the articulated skeleton, most ofthe material has presence of complete distal metapodial keels been lost, and only the skull remains. Ap- (character 27) elevates this genus above the parently the complete type ofAmphitragulus "gelocid" grade. Gentry (in Hendey, 1978) has also been lost (M. Brunet, personal com- has suggested a possible affinity of Propa- mun.). This had led to much confusion as to laeoryx with early giraffoids, based on the the identification of specimens of these gen- general similarity of the teeth with those of era; for example, several faunal sites in the the larger genus Climacoceras, and also be- BMNH have all the small ruminant limbs cause its early presence in Africa makes its labeled "Dremotherium" and all the dental role as a giraffoid ancestor biogeographically material labeled "Amphitragulus"! The two plausible. The only possible synapomorphic genera have usually been considered as a pair, feature that we can find to link Propalaeoryx and treated as sister taxa in various classifi- with the Giraffoidea is the bifurcation of the cations. We consider the two genera to have posterior wing of the metaconule (character some significant differences, resulting in a 19b). This is a pronounced feature ofthe early distinct separation between them in our pro- giraffid Palaeotragus but, as previously men- posed cervoid phylogeny. tioned, is seen only occasionally in specimens With the exception ofPomel (1845-1846), of Canthumeryx, although it is present in the who considered these genera to be tragulids, problematical teeth assigned to Zarafa from and Gervais (1859), who classified Dremo- Gebel Zelten. Unfortunately, the anterior therium with the antelopes, Dremotherium portion of the dentition of Propalaeoryx is and Amphitragulus have been considered to not known, so it cannot as yet be determined be cervoids of some sort. Geoffroy Saint-Hi- whether this genus had a giraffoid type of laire (1833) and Richard (1946) considered bilobed lower canine (character 23) (though Dremotherium to be a true cervoid, but closer ofcourse the absence ofthis character would to the tragulids than Moschus. Milne-Ed- not exclude it from the position of sister tax- wards (1864) related Dremotherium to Mos- 62 AMERICAN MUSEUM NOVITATES NO. 2893 chus, while Riitimeyer (1881, 1883) placed it ledge is not only more strongly defined in in the Cervidae in the subfamily Cervulinae. Dremotherium, but projects laterally to form More recently, these genera have been lumped a convex supraorbital roof, bearing closer re- with a number of other early homless ru- semblance to the condition in true cervids. minants in the Palaeomerycidae (e.g., Simp- In addition, the supraorbital depressions on son, 1945; Stirton, 1944; Viret, 1961; Romer, the skull roof are more narrow and covered 1966). by bone for a greater length of their traverse Sigogneau (1968) considered Dremotheri- in Amphitragulus than in Dremotherium. um to be a true cervoid, resembling Moschus Sigogneau provides measurements to show and Hydropotes (though retaining more that the occipital region is considerably higher primitive characters), and possibly closely rein Dremotherium than in Amphitragulus. She lated to the North American blastomerycids. questions the supposed lack ofthe antorbital She considered Amphitragulus to be possibly vacuity in Amphitragulus, as she points out related to another cervoid lineage, and hints that no cranium is well enough preserved to at certain similarities between Amphitragulus be absolutely certain of the condition. She and Palaeomeryx. Webb and Taylor (1980) also discusses some primitive, more tragu- discuss a number of similarities between the lidlike features of Amphitragulus, including basicranial regions of Dremotherium, Mos- the relatively smaller orbits with the anterior chus, and Blastomeryx, and, linking Amphi- border placed relatively further forward (i.e., tragulus with Dremotherium primarily on the at the level of the junction of the first and basis of the sabrelike canines, group these second molars, as opposed to at the level of genera together in the Moschidae (which in the border between the second and third mo- their classification is considered as the sister lars in Dremotherium), a Hyemoschus-like group to the Eupecora, or homed ruminants). orbital extension of the palatine, and a con- There is little doubt that both Amphitra- tribution from the maxilla to the floor of the gulus and Dremotherium are true cervoids. orbit. In contrast to Dremotherium, the or- They both possess elongated, fused meta- bitosphenoid does not attenuate toward the podials (character 28d) with complete distal sphenopalatine orifice, and the infraorbital keels (character 27) and a closed metatarsal portion of the lacrimal extends less postero- gully (character 28b); a raised lip on the cu- ventrally. Sigogneau also considers Dremo- bonavicular facet (character 32d); sabrelike therium to have a relatively larger basioccip- upper canines (character 1 lb); an entostyle ital region than Amphitragulus. However, in on the upper molars (character 15b); and a her discussion ofthe amount ofvariation seen "Palaeomeryx" fold in the lower molars in the limited amount of cranial material (character 20). Viret (196 1) describes the dif- known of the two genera, she considers that ferences between the two genera as follows: none ofthese differences could be considered Amphitragulus has a short facial region, with- as diagnostic for the separation of the two out lacrimal fossae or antorbital vacuities; genera. She also points out that Amphitra- short, thickset premolars and a relatively short gulus has two lacrimal orifices (character 9), diastema with p1 usually present; little ele- whereas there is only a single orifice in vation ofthe occipital region, and only slight Dremotherium, but she does not comment projection ofthe supraorbital ledge; and nor- on the potential significance ofthis character. mal length of cervical vertebrae. In contrast, With regard to dental differences, Sigo- Dremotherium has a longer facial area, with gneau adds to Viret's list by pointing out that both lacrimal fossae and antorbital vacuities the ectostylid and metastylid are less pro- present; generally only three premolars, with nounced in Amphitragulus than in Dremo- a large diastema and more elongated pre- therium, the Palaeomeryx fold is less well- molars (see tables 3, 5); braincase more ex- developed in Amphitragulus, and the lower panded, with a more elevated occipital area premolars are less molarized, especially with and a more strongly projecting supraorbital regard to the development of the metaconid. ledge; and elongated cervical vertebrae. In the upper molars, the entostyle is less well- Sigogneau (1968) discusses these differ- developed in Amphitragulus, and the molars ences and also notes that the supraorbital are more bunoselenodont in the development 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 63

a~~~~~~~~~~~~~~~I

6 I Immm ,

Fig. 13. A. Amphitragulus sp., left P3-M3, Ph 3107 and left p3-m3, Ph 4141. B. Dremotherium feignouxi, right P3-M3, Ph 52933 and left p2-m3, Ph 32488, Montaigu, Allier (early Miocene), France. Bar = 10 mm. of the crescents. The paracone on P4 is less acter 1 7b), and a double loph on the posterior well developed in Dremotherium. Amphitra- lobe of m3 (character 26a) (see fig. 13). gulus differs from Dremotherium in having a To this list of dental differences, we would reduced metastyle, especially on M3 (char- add the following: the molars of Dremo- acter 1 6b), a reduced metacone on P4 (char- therium are more fully selenodont than those 64 AMERICAN MUSEUM NOVITATES NO. 2893

facial region and diastema. The elongated cervical vertebrae, similar to those of the modem bovid genera Litocranius and Am- modorcas, suggest that Dremotherium may have had a high browsing type of diet, as do these gazellines, which also resemble Dremo- therium in the possession of a long basioccipital and a high occipital region (Janis, unpubl. data). The cranial and molar morphology and dental wear of Amphitragulus suggest a more tragulid or duikerlike mode Fig. 14. Amphitragulus sp., Ph 3107. Bar = 5 oflow-level browsing on more succulent ma- cm. terial (Janis, 1979; unpubl. data). Thus the skull appears to be superficially more tragu- of Amphitragulus, particularly with respect lidlike. However, it is clear that Amphitra- to the development of the postentocristid in gulus has two definitive characters that make the lower molars. The molars of Amphitra- it a more derived type of cervoid than gulus have a "chubby" appearance, and can Dremotherium. These are possession of the be shown to be relatively broader than those posterior tuberosity on the metatarsal (char- ofDremotherium. In addition, the upper mo- acter 31), and in particular the presence of lars of Dremotherium retain the primitive two lacrimal orifices situated in the cervid pecoran condition of the reduced M3 meta- position on the orbital rim (character 9) (see conule (character 18a) and traces of an in- fig. 14). This specimen ofAmphitragulus was ternal cingulum (character 13b), whereas in figured in dorsal view by Sigogneau (1968), Amphitragulus the metaconule is equal in size but she did not comment on the significance to the protocone (character 1 8b) and there is of the double lacrimal orifice. However, the little evidence ofthe internal cingulum (char- significance of this feature in cervoid phy- acter 1 3c) (see fig. 13). (These differences were logeny was not widely recognized until the also noted by Sigogneau.) work ofLeinders and Heintz (1980). We have Although there is no firm association of also observed a double lacrimal orifice on a limb material with either ofthese two genera, partial cranium ofAmphitragulus lemanensis we have examined many different faunal col- from St. Gerard-le-Puy, Allier, France in the lections in museums where different size fos- collection of the Musee Gimet, Lyon (St. G. sil species of the two genera are found to- 600). We consider these differences between gether, and have found some invariant the two genera to ally Amphitragulus with the characters of the two genera on the basis of more advanced true cervids, whereas Dremo- the size sorting. The metatarsals of Dremo- therium should be considered to be more therium have the tubercle for metatarsals II closely allied to the moschids (see Webb and and V fused (characters 30a, b) and lack a Taylor, 1980). The exact systematic position posterior tuberosity. In contrast, the meta- of these genera within the phylogeny of the tarsals of Amphitragulus have only V fused Cervoidea will be discussed in more detail in (character 30b), and have a variably present a later section. small cervidlike posterior tuberosity (character 31). MOSCHIDAE Amphitragulus is generally considered to be more primitive and more "tragulid-like" Webb and Taylor (1980) linked the living in its cranial and dental features than Dremo- genus Moschus with the fossil genera Am- therium (e.g., Sigogneau, 1968). However, we phitragulus, Dremotherium, Blastomeryx, and consider the "advanced" characters of the other blastomerycine genera in the family Dremotherium to be associated with a more Moschidae on the basis of sabrelike upper folivorous dietary habit, as suggested by the canines (character 1 b) and a laterally en- more selenodont cheek teeth, the greater mo- closed subcentral tympanohyal on the audi- larization of the premolars, and the longer tory bulla (character 10). We have argued in 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 65 this paper and elsewhere (Scott and Janis, arisen in parallel within the dromomerycids. 1987) for the inclusion ofthe Moschidae with A posterior tuberosity is present in Moschus the Cervidae in opposition to Webb and Tay- and Parablastomeryx, but absent in Dremo- lor (1980), who place the Moschidae as the therium and Blastomeryx. Parablastomeryx sister group to the homed ruminants. How- also differs from Blastomeryx in the following ever, the question of whether moschids rep- characters: a large premolar row, lower mo- resent a clade or a grade within the cervoid lars with a pronounced Palaeomeryx fold phylogeny remains to be answered, and will (absent in Blastomeryx but present in Pro- be discussed here. blastomeryx), more prominent entostyle in As previously stated, Amphitragulus pos- the upper molars, entostylid and metastylid sesses some derived cervoid characters ab- in the lower molars, a reduced metastyle in sent in Dremotherium, and should not be the upper molars (character 16b), an atten- grouped with the other moschid genera. In uated protocone on p4 (character 1 7b) (fea- their discussion of basicranial anatomy, tures 16b and 17b resemble the normal pe- Webb and Taylor (1980) discussed only the coran condition in Blastomeryx), posterior condition in Dremotherium, and included lobe of m3 with a double loph that is closed Amphitragulus because ofthe presence ofthe posteriorly (character 26a) (the molar is com- moschidlike upper canines. However, as pre- pressed with a single loph in Blastomeryx), viously discussed, such canines appear to be and a P3 with a somewhat posteriorly di- a derived character ofcertain higher cervoids, rected protocone (character 14a) (lingually but their presence may well serve to unite a directed in Blastomeryx). clade of cervoids that are more derived than Some of these differences between Para- Eumeryx and Walangania. blastomeryx and Blastomeryx were noted by We would define moschids as cervoids with Frick (1937), who also commented on the a closed metatarsal gully (character 28b), a shorter diastema ofParablastomeryx and the Palaeomeryx fold (character 20) primitively more lightly proportioned limbs. Frick present in brachydont members of the lin- grouped the genera Longirostromeryx, Blas- eage, a raised lip on the cubonavicular facet, tomeryx, and Machaeromeryx in the subfam- upper dentition with an entostyle, a pro- ily Longirostromerycinae, and the genera nounced metastyle, and a P3 with a lingually Parablastomeryx, Pseudoblastomeryx, Pro- directed protocone, but with a single lacrimal blastomeryx, and Pseudoparablastomeryx in orifice and lacking any form of cranial ap- the Parablastomerycinae. Pseudoparablasto- pendages. The fossil genera retain the gelocid meryx was referred to the Leptomerycidae by condition ofa small metaconule on M3 (char- Taylor and Webb (1976). We believe that a acter 18a). However, all moschids are de- thorough review of the interrelationships of rived in the fusion ofmetatarsal V (character the "blastomerycids" and of the distribution 30b). of basicranial characters throughout the Webb and Taylor's (1980) comparison of "moschids" and the other higher Pecora is the ear regions of Moschus and the blasto- essential before the phylogenetic reality ofthe merycids presents a convincing case for strong "Moschidae" can be evaluated. However, at similarities, although the situation in Dremo- the present time we consider the scheme of therium is not as pronounced. Examination Webb and Taylor (1980), based on ear region ofthe type ofDremotherium in the Paris mu- characters (character 10), to be a working hy- seum shows that the tympanohyal is placed pothesis (but excluding the genus Amphitra- more posteriorly on the bulla, with a lesser gulus), and consider the posterior tuberosity degree of lateral enclosure than is seen in in Moschus and Parablastomeryx to have Moschus or Blastomeryx (character 1 Oa). evolved in parallel with the condition in the However, the distribution of the posterior other cervoids (though it should be probably tuberosity (character 3 1) among the moschid considered a synapomorphy linking these two genera is problematical. As previously dis- genera within the Moschidae). cussed, this appears to be a fixed trait in true Further genera which may be included cervids, present in other cervoids such as Pa- within the Moschidae are the European Mi- laeomeryx and Hoplitomeryx, but it may have cromeryx and Hispanomeryx. The interre- 66 AMERICAN MUSEUM NOVITATES NO. 2893 lationships of Micromeryx with the blasto- berosity of the metatarsus) than to Dremo- merycids was first suggested by Frick (1937), therium or the other blastomerycids. and reiterated by Leinders (1983). Micro- However, since we did not see the original meryx possesses a closed metatarsal gully and material, we would not presume to comment a Palaeomeryx fold, so it is undoubtedly a further on the affinities of this taxon within cervoid. It lacks cranial appendages, pos- the Moschidae. sesses a sabrelike upper canine, and has a distinct posterior tuberosity on the metatarsus. Metatarsal V is fused, as in other mos- PALAEOMERYX AND THE chid genera. To the best of our knowledge, PALAEOMERYCIDAE the condition of the lacrimal orifice is un- Palaeomeryx has traditionally been placed known, and the ear region has not been stud- within its own family, the Palaeomerycidae, ied. We tentatively place it in the Moschidae and the family has been variously considered on the assumption ofa single lacrimal orifice, cervoid (Viret, 1961; Simpson, 1945) or gi- and would group it with Moschus and Para- raffoid (e.g., Stirton, 1944). Although the blastomeryx on the basis of the posterior tu- composition ofthe Palaeomerycidae has var- berosity. ied, the family has traditionally included a Hispanomeryx duriensis was recently dis- number ofproblematical genera, as discussed covered in the Vallesian ofSpain (Morales et in a previous section. al., 1981). Morales et al. place this taxon in A recent definitive statement on the sys- the Moschidae (sensu Webb and Taylor, tematic position ofPalaeomeryx was that by 1980), and comment that it may also bear a Ginsburg and Heintz (1966). They found what relationship to Walangania, which they sug- appears to be a dermal ossicone, similar to a gest should also be classified with the Mos- giraffid dermal ossicone, associated with a chidae. We have not had the opportunity to specimen ofPalaeomeryx kaupi, and consid- examine the original material of Hispano- er this to be strong evidence of giraffoid as- meryx. However, to judge from the illustra- sociation. They also point out a number of tions in Morales et al. (1981), the lower molars other characteristics of Palaeomeryx that resemble the condition in Parablastomeryx seem to be more giraffid than cervid, such as in the possession of an m3 with a double the lack of bifurcation of the protocone, the posterior lobe closed posteriorly (character absence of an antorbital vacuity, the posses- 26b) and a vertical groove on the posterolin- sion of a ridge on the proximal metacarpus, gual region ofp4 (character 25), but resemble the absence of a diarthrodial facet on the the situation in the more hypsodont Moschus proximal metacarpals, and the equal length in the absence of a Palaeomeryx fold and the ofthe metacarpals and metatarsals. Leinders reduction in the size of the metastyles. The (1983) points out that these characters are all upper molars resemble Moschus in the ab- primitive pecoran characters, also seen in sence of an entostyle and in the presence of most early cervids (as defined by the presence a large metastylid. Morales et al. (1981) de- of antlers). Other giraffid characters pos- scribe the metatarsus as possessing a closed sessed by Palaeomeryx from Sansan (Helve- metatarsal gully and a posterior tuberosity, tian), such as the morphology ofp4, the form and having metatarsal II (but not metatarsal of the distal articulation of the humerus, the V) fused with the proximal surface of the grooved surface on the anterior side of the metatarsus. Their illustration ofthe proximal distal part of the radius, and the groove on metatarsal surface appears to resemble the the posterior surface of the metapodials, are generalized cervoid condition, with a raised said by Leinders (1983) to be absent from the lip for the posterior cubonavicular facet. earlier specimen of Palaeomeryx from Ar- We would agree with Morales et al. (1981) tenay, and so must have been derived within in assigning Hispanomeryx to the Moschi- the evolution of this genus, in parallel with dae, despite the fact that no upper canine is the condition in giraffids. known from this taxon, and consider the While the form ofthe associated ossicones genus to be closer to Moschus and Para- is similar to the condition in giraffids (char- blastomeryx (on the basis ofthe posterior tu- acter 7a), this need not be indicative ofa close 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 67 relationship, as previously discussed. Lein- The skull had both a lacrimal fossa (char- ders (1983) discusses the cervoid condition acter 8) and an antorbital vacuity. There was of the closed metatarsal gully in Palaeome- apparently only one lacrimal orifice, but this ryx, which he included in the Cervoidea. He was only discernible on a single specimen in points out that even if a complete cranium which the anterior orbital rim was damaged. of the animal were known, the absence of a The lower canine, while not fully preserved, double lacrimal orifice would not exclude the apparently did not exhibit the giraffoid type genus from the Cervoidea, as it could rep- of bilobed condition (character 23). P4 had resent a moschid level of cervoid evolution. a weak metacone (character 17b), and the Leinders suggests that Palaeomeryx and Tri- molars exhibited weak metastyles (character ceromeryx may be related to the North Amer- 16b) and a bifurcated posterior wing of the ican dromomerycids, and may have reinvad- metaconule (character 1 9b). Weak entostyles ed Eurasia from North America in the early (character 15a) were variably present. Both Miocene along with the equid Anchitherium. P4 and the upper molars retained traces of Whitworth (1958) commented on a feature an internal cingulum (character 1 3b). p1 was in the upper molars of Palaeomeryx that he present in some specimens, separated from considered to be cervidlike, rather than gi- p2 by a small diastema (character 4b). The raffid. That is, the position of the metacone, lower molars show the presence of a strong which in Palaeomeryx resembles the cervoid Palaeomeryx fold (character 20), and the ec- condition, being placed parallel to the labial tostylids and metastylids were also strong. edge of the fold (in contrast to the giraffid The posterior lobe of m3 was double and condition, where the metacone is placed more closed posteriorly (character 26a), described obliquely). However, this feature is probably by the authors as "horseshoe shaped." The a derived giraffid feature, as the position of lower premolars were little molarized. p4 had the metaconid in Canthumeryx resembles that a small metaconid and a posterolingual groove in Palaeomeryx and the other cervoids (and (character 25). The neck of the animal was indeed is the condition in primitive peco- not elongated. The metapodials were com- rans), so the character is probably not sig- pletely fused, with slender, complete meta- nificant for assessing the phylogenetic rela- carpals II and V. The metatarsal gully was tionships of Palaeomeryx. closed, with the hind limb longer than the Recent discoveries from the Miocene of forelimb. China of a complete skull and several com- Qiu et al. (1985) remain convinced that plete skeletons ofPalaeomeryx tricornis (Qiu Palaeomeryx was a giraffoid, or at least rep- et al., 1985) have thrown new light on the resents the sister group to the other giraffoids, anatomy of this genus. They describe the as it apparently lacked the giraffoid synapo- specimens as having sexually dimorphic cra- morphy of a bilobed lower canine, although nial appendages, consisting in the (presumed) they note that it shared many plesiomorphic males of a pair of supraorbital giraffidlike features with cervids. They also point out that "ossicones," which were triangular at the base Ginsburg and Heintz (1966) were incorrect and sloped posteriorly, with a bulbous or in stating that there was no antorbital vacuity pointed tip, a rough and cancellous surface in this genus (although they also note that an (suggestive ofa giraffidlike skin covering), and antorbital vacuity is indeed found within the a suture demarking the base of the append- Giraffidae, and is probably a plesiomorphic ages from the skull in one specimen. In ad- condition for the Pecora, or one evolved a dition, these animals possessed a dromo- number oftimes in parallel, as we concluded merycid-like single, unbranched occipital in an earlier section). appendage with a smooth surface, the distal Qiu et al. (1985) consider that the mor- part of the appendage forming a laterally phology of the p4 of Triceromeryx justifies compressed, bulblike structure. Sexual di- its inclusion in the Giraffidae (after Hamil- morphism was also shown in the possession ton, 1978a), and note that the supraorbital ofupper canines, which were large and sabre- cranial appendage in this genus is comparable like in the males, and small and peglike in to the situation in Palaeomeryx tricornis. the females. However, it has been demonstrated recently 68 AMERICAN MUSEUM NOVITATES NO. 2893 that this type of p4 morphology cannot be the Moschidae. The lacrimal fossa is a less used as a diagnostic giraffid autapomorphy useful character, but we note that it is also (Janis and Lister, 1985). They also comment characteristic of primitive dromomerycids. that the dolicocephaly ofthe skull, the central Qiu et al. are very impressed by the similar- position of the orbit, the relatively small de- ities of the cheek teeth of Palaeomeryx tri- gree offlexion ofthe face on the basicranium, cornis with those of Zarafa, as described by and the long basicranial region all resemble Hamilton (1973). We agree with them, but the condition seen in the giraffoids Zarafa as noted in our previous discussions in this and Giraffokeryx. However, this type ofskull paper, we feel that there is no reason to as- morphology is also apparent in many drom- sociate these isolated molars with the type of omerycid genera. They note that the appar- Zarafa. The feature of the bifurcated poste- ently single lacrimal orifice is a plesiomorph- rior wing of the metaconule is seen in many ic character for the Pecora, not necessarily dromomerycids as well as in early giraffids. suggestive of giraffoid affinities, although it This seems to us a convincing case for as- would exclude Palaeomeryx from the Cer- signing the Gebel Zelten "Zarafa" teeth to voidea above the level ofthe Moschidae (see an unknown African species ofPalaeomeryx. Leinders, 1983). However, we would ques- Qiu et al. (1985) also comment on the cer- tion their assertion that this is the true con- voidlike nature ofthe closed metatarsal gully dition in Palaeomeryx. As previously dis- in Palaeomeryx, but note that this condition cussed, the condition of the lacrimal orifice could be independently derived with the is variable among the antilocaprids (see also Ruminantia. They cite as evidence (from Scott and Janis, 1987), even though most Frick, 1937) the presence of a closed meta- specimens do possess a double lacrimal ori- tarsal gully in the Antilocapridae, but the fice, and so a single specimen of a taxon that presence of an open gully in some meryco- exhibits the plesiomorphic condition cannot dontines. However, as previously discussed, be taken as representative of the taxon as a our studies ofthe Antilocapridae clearly show whole. (In contrast, ifa single specimen shows that the polarity of this character within the a derived character condition, this is a much Merycodontinae is from a closed gully to a more convincing situation, although ofcourse secondarily open one (although we concede the possibility ofparallel evolution cannot be that this type of closed metatarsal gully may ruled out, as shown by the occurrence of a have evolved more than once within the double lacrimal orifice in some of the bovid Ruminantia, as evidenced by the condition tribes). We have discovered, in our exami- in Pseudoceras). Finally, they are obviously nation of fossil material in the Frick collec- impressed by the fact that it can at last be tions, AMNH, where many complete skulls conclusively shown that Palaeomeryx did in- exist for dromomerycid and antilocaprid deed possess giraffidlike ossicones, as origi- species, that a single specimen with a dam- nally hypothesized by Ginsburg and Heintz aged orbital rim is a poor basis for the de- (1966) on the basis of an isolated "ossicone" termination of this character in a taxon or in found at Artenay. However, previous dis- a pecoran lineage. The condition of the lac- cussions in this paper have clearly shown that rimal orifices is only clearly discernible in a the cranial appendages in the giraffoids Cli- few individuals in the Frick collections. In- macoceras and Nyanzameryx cannot be ho- terestingly enough, their drawings ofPalaeo- mologous with true giraffid ossicones. If cra- meryx tricornis (Qiu et al., p. 175) suggests a nial appendages have evolved independently double lacrimal orifice. within the Giraffoidea, as defined by Ham- Qiu et al. (1985) note that the lacrimal fos- ilton (1973) on the basis ofthe possession of sa and sabrelike upper canine present in Pa- a bilobed lower canine, then the cranial ap- laeomeryx are generally considered to be cer- pendages of Palaeomeryx cannot be homol- void characters, but that the polarity ofthese ogous with giraffid ossicones, despite their characters is unknown to them. However, superficially similar appearance, since the lack previous discussions in this paper have of a bilobed lower canine excludes Palaeo- pointed out that a sabrelike upper canine is meryx from the Giraffoidea. a derived feature of cervoids at the level of Our analysis ofPalaeomeryx is as follows: 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 69

Fig. 15. Palaeomeryx bojani, left P3-M3, BMNH 29623 and right p3-m3, BMNH 21868, Sansan (middle Miocene), France. Bar = 10 mm.

the genus possesses a number of characters acter suggests a level of cervoid evolution which suggest that it is a more derived pec- above the level of the Antilocapridae. oran than the level of the Giraffoidea. The Sigogneau (1968) hinted at a relationship presence of an entostyle (character 15) raises between Palaeomeryx and Amphitragulus in the genus to the level of the cervoid/bovid describing a number of dental synapomor- dichotomy, and the character of a closed phies; these were: a weak metastyle in the metatarsal gully (character 28b) is a synap- upper molars, especially on M3 (character omorphy linking Prodremotherium with the 16b), an attenuated metacone on P4 (char- cervoids. The raised lip ofthe cubonavicular acter 1 7b), and a posterior lobe of m3 which facet (character 32d) is a plesiomorphic fea- had a double loph with a closed posterior ture for the Pecora, and while the morphol- valley (character 26a). We would add to this ogy of the proximal metatarsus of Palaeo- list the posterolingual groove on p4 (char- meryx is generally cervoid in appearance, this acter 25), a feature which is also characteristic feature cannot be used to exclude the genus ofgiraffoids (Hamilton, 1 978a). These dental from the position of sister taxon to the Gi- characteristics are also seen in the blasto- raffoidea. Palaeomeryx also possesses a num- merycid Parablastomeryx, but this genus has ber of characters which are undoubtedly of a a moschid type of auditory bulla (character derived cervoid condition. These include: a 10) (Webb and Taylor, 1980), and an un- Palaeomeryx fold (character 20); a sabrelike doubtedly single lacrimal orifice. However, upper canine (character 1 b, also noted in a what is interesting about this suite of cranial specimen attributed to Palaeomeryx mag- and dental characters is that-along with the num in the collections of the Musee Gimet derived cervoid character of a closed meta- in Lyon, France, from the la Grive St. Alban tarsal gully (character 28b), a variably present collections ofthe Middle Miocene ofFrance, posterior tuberosity (character 3 1), sabrelike LaG. 15280); and a posterior tuberosity on upper canines in the males (character 1 b), the metatarsus (character 31). This last char- and the presence of a double lacrimal orifice 70 AMERICAN MUSEUM NOVITATES NO. 2893

Fig. 16. Subdromomeryx wilsoni, left P4-M3, F:)AM 52933 and right p2-m3, F:AM 32488, Olcott formation (middle Miocene), North America. Bar = 10 mm.

(character 9) (at least in Amphitragulus, the progressive characters may merely be linked situation in Palaeomeryx being as yet un- with the level of hypsodonty, as their teeth determined in our opinion)-they are all resemble the more simplified teeth of hyp- shared with the North American dromomer- sodont bovids and antilocaprids. The com- ycids. (See figs. 15 and 16 for dentition of plete lack of a posterior tuberosity is also Palaeomeryx and Subdromomeryx.) Fea- problematical, but as limb proportions oftheir tures shared by Palaeomeryx and early dro- genus suggest a preference for a more open momerycid genera include the bifurcation of habitat than that seen in any other dromo- the posterior wing of the metaconule (char- merycid, or in most living cervid genera acter 1 9b) and the presence ofa lacrimal fossa (Scott, unpubl. data; see also Janis, 1982), the (character 8) although, as previously dis- absence of a posterior tuberosity could be cussed, neither of these features is unique to correlated with ecological considerations. these genera (although their occurrence in (Among the Bovidae, a slight posterior em- combination may well be unique). The limbs inence is seen in the woodland-dwelling tribes of Palaeomeryx are somewhat more derived Tragelaphini and Cephalophini, but is absent than in Amphitragulus and the dromomer- in the more open-habitat lineages.) In light ycids, as the tubercle for metatarsal V is fused of the absence of good synapomorphic char- with the proximal surface of the metatarsal acters uniting either the Antilocaprinae or the (character 30b) (although we do not know the Dromomerycidae, and their similarity in the condition in the Chinese species of Palaeo- position of cranial appendages, we have con- meryx.) sidered the possibility that Aletomeryx may Aletomeryx is the only dromomerycid that represent an early offshoot of the Antilo- does not possess this derived suite of dental caprinae. However, pursuit ofthis idea awaits characters, and additionally lacks a posterior more detailed studies ofruminant basicranial tuberosity on the metatarsus. It is a proble- regions. The genus Rakomeryx is also prob- matical genus in many respects. It is dentally lematical because, despite the variable pres- advanced for a dromomerycid in its posses- ence ofa posterior tuberosity and the derived sion of small entostyles, its lack of a Palaeo- features ofthe lower dentition, it has a prom- meryx fold, and its lack ofmetastylids (except inent metastyle and a large metacone on P4. for a small metastylid on m3). However, these These problematical genera notwithstand- 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 71 ing, we suggest that this suite of characters cluding the development of an occipital ap- constitutes a basis for reestablishment of the pendage) may have been shared by this com- family Palaeomerycidae, to include Palaeo- mon ancestor. We are impressed by the meryx, Amphitragulus, and the dromomer- similarity of Palaeomeryx tricornis with the ycids. Further material of Palaeomeryx and earliest known dromomerycid Barbourome- study ofbasicranial similarities between Am- ryx trigonocorneus, both superficially in gen- phitragulus and the dromomerycids would be eral morphology (especially ofthe cranial ap- needed to confirm this hypothesis, as would pendages), but also in the details of the skull confirmation of our hypothesis that the con- and dentition discussed in this section. Fi- dition ofthe lacrimal orifices in Palaeomeryx nally, the limbs of Palaeomeryx are some- may actually have been double. Prolibytheri- what more derived than in Amphitragulus um may also be included in this ruminant and the dromomerycids, as the tubercle for clade (see discussion in following section). metatarsal V (character 30b) is fused with the The dental material of Triceromeryx depict- proximal surface of the metapodial. ed by Crusafont-Pairo (1952) does not dem- onstrate any of these derived dental characters, with the exception ofthe posteroventral ZARAFA AND PROLIBYTHERIUM groove on p4 used by Hamilton (1978a) to Zarafa zeltini and Prolibytherium mag- unite this genus with the Giraffidae, although neiri were pecoran taxa known only from the the presence of a Palaeomeryx fold implies Gebel Zelten, in the lower Miocene of Libya. cervoid affinities. We would leave the genus Zarafa was about the size of a fallow deer as cervoid incertae sedis, rather than include (Dama dama) and Prolibytherium was about it with the Palaeomerycidae. As previously the size of a roe deer (Capreolus capreolus) mentioned, the isolated molars from Gebel (Hamilton, 1973). Both possessed lacrimal Zelten ascribed to Zarafa by Hamilton (1973) fossae (character 8), which Hamilton (1973) bear dental similarities to Palaeomeryx and considered to be a primitive pecoran feature, may possibly represent an African palaeo- and Zarafa also appears to have possessed merycid. an antorbital vacuity. Both were character- Certain members of the Palaeomerycidae ized by unusual forms ofcranial appendages: are also characterized by the evolution of su- Zarafa had supraorbital appendages with a praorbital, unbranched, nondeciduous cra- dorsolateral orientation (the basal part of the nial appendages. These may well have been ossicone alone is preserved), and Proliby- homologous between Palaeomeryx and the therium had flattened, horizontal winglike dromomerycids. The isolated appendage de- cranial appendages divided into anterior and scribed by Ginsburg and Heintz (1966), and posterior lobes in a butterflylike pattern. The the line ofsuture between the appendage and grooves on the surface ofthe cranial append- the skull described by Qiu et al. (1985) in one ages ofProlibytherium suggest a vascularized, ofthe Chinese specimens ofPalaeomeryx tri- skin-covered structure (see Hamilton, 1973; cornis, suggest that the appendages were Churcher, 1978). Zarafa has been variously formed initially from a dermal structure fus- assigned to the "Palaeomerycidae" (Hamil- ing with skull, as in giraffids. However, in ton, 1973), or to the Giraffoidea (Hamilton, other respects (especially in the possession of 1978a, 1978b; Churcher, 1978). Churcher an occipital appendage) the appendages are described Zarafa as a "palaeotragine," while at least superficially similar to those pos- Hamilton (1978a) synonymized the genus sessed by the North American dromomer- with Canthumeryx. Heintz et al. (1981) de- ycids, although it has been suggested that these scribed the skull of an upper Miocene giraf- appendages represent frontal outgrowths foid from Iraq (Injanatherium hazimi) which (Bubenik, 1982). The additional dental syn- appears to be very similar to Canthumeryx, apomorphies shared by Palaeomeryx and the and may represent part ofa radiation ofMio- dromomerycids suggest that this genus shares cene giraffoids with these peculiar types of a closer common ancestor with the North cranial appendages. Hamilton (quoted per- American forms than does Amphitragulus, sonal commun. in Patterson, 1981), later sug- and that the type of cranial appendages (in- gested that Zarafa/Canthumeryx might also 72 AMERICAN MUSEUM NOVITATES NO. 2893 be synonymous with the genus Progiraffa, de- ical "palaeomerycid" features ofan entostyle scribed from India by Pilgrim (1911). Pro- (character 15), a small metastyle (character libytherium was originally referred to the Gi- 16b), and an attenuated metacone on P4 raffidae, subfamily Sivatheriinae, on the basis (character 17b) (seen in BMNH 21901). The of its palmated cranial appendages (Hamil- lower molars show a condition that could ton, 1973; Churcher, 1978), but was later possibly represent an incipient double loph reassigned to "Pecora insertae sedis" (Ham- on m3 (character 26a) (BMNH 21899), noted ilton, 1978a, 1978b). The anterior dentition by Hamilton (1973) as a "weak ectostylid in is unknown in both taxa. addition to the hypoconulid." The teeth of As previously noted, there is a problem Prolibytherium also show the primitive fea- with the dental material from Gebel Zelten tures of a small M3 metaconule (character referred to Zarafa by Hamilton (1973). These 18a), and a somewhat posteriorly situated teeth are too large to belong to the same taxon protocone on P2 (P3 is missing from the den- as the type specimen of Zarafa (a skull with tition) (character 14a). p4 shows the primi- fragmentary dental remains), and their pres- tive pecoran condition of a small metaconid, ervation appears to be of a different type. but has a vertical groove on the posterolin- These teeth resemble those of Canthumeryx gual region (character 25). As previously not- in the possession of a bifurcated metaconule ed, this could be suggestive ofeither giraffoid (character 19b), but this dental feature is also or palaeomerycid affinities. seen in Palaeomeryx (Qiu et al., 1985). In Hamilton (1973) described both Zarafa and addition, these teeth possess an entostyle de- Prolibytherium as having a double lacrimal rived from the anterior surface of the meta- orifice (character 9), although this assertion conule (character 15) which, as previously was refuted by Leinders (1983). Reexami- noted, is a character seen primarily in Pecora nation of the material shows that the type of ofa higher level than the Giraffoidea. We have Zarafa (BMNH 26670) appears on first in- never observed an entostyle in any specimens spection to have had a double orifice, but the of Canthumeryx, and an entostyle is not ob- dorsal "orifice" is clearly a postmortem break servable in the type of Zarafa, where the lin- in the anterodorsal orbital wall, and in any gual portion of M3 is preserved. Hamilton case is situated too far dorsally above the (1973) also assigned postcranial material from more ventral orifice to resemble the cervoid Gebel Zelten to Zarafa on the basis of size condition. (Additional breaks of similar ap- sorting and general "palaeotragine-like" sim- pearance are also apparent in the upper dorsal ilarities, in which the metatarsal gully was orbital rim.) The condition in Prolibytherium open. We concede that Hamilton may be cor- is more convincing. In one preserved skull rect in synonymizing Zarafa with Canthu- fragment (BMNH 26679) the orbit is broken meryx, but see no compelling reason for it. in a critical position midway along the an- We regard the disassociated teeth ascribed to terior rim. However, there is a definite dorsal Zarafa as problematical, and possibly be- orifice visible on the preserved portion ofthe longing to a palaeomerycid taxon (sensu our rim, and just ventral to this is a bony spur definition in this paper). highly reminiscent of the spur seen between In contrast, Prolibytherium presents a more the two lacrimal orifices in cervids. However, complex problem. The teeth associated with the orbital rim is completely broken offjust the type in the BMNH are all heavily worn, ventral to this spur, so it is impossible to and it is impossible to tell if a Palaeomeryx ascertain ifa more ventral orifice was actually fold (character 20) was present in the unworn present. (None of the other skull fragments condition. Later material assigned to this ge- preserve any anterior portion of the orbital nus in l'Institut de Palaeontologie, Paris, in- rim.) However, what is preserved ofthe mor- cludes a lightly worn lower dentition which phology of the orbital rim does appear very lacks a Palaeomeryx fold. However, as noted cervoidlike, and we are prepared to believe by Hamilton (1973), Prolibytherium is a fair- that a cervoidlike double lacrimal orifice was ly hypsodont taxon (about as hypsodont as indeed present. the dromomerycid Cranioceras, in which the Hamilton (1973) described the brain of Palaeomeryx fold has already been lost). The Prolibytherium, and decided that it was closely upper molars ofProlibytherium show the typ- comparable to the brain of Dremotherium, 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 73 but also stated that these similarities may is never seen in known specimens of antilo- simply reflect primitive pecoran features. He caprids. While such canines are secondarily also noted that the petrosal was more similar lost in antlered cervids, they tend to be re- to that of the living cervid Capreolus than to tained as a primitive feature in early antlered present-day giraffids, but the polarity ofthese cervids (such as Dicrocerus, persisting today characters has not as yet been ascertained. in Muntiacus and Elaphodus) and in drom- Hamilton (1973) also assigned certain disas- omerycids (i.e., Barbouromeryx, Aletomeryx, sociated postcranial elements, which were and Sinclairomeryx), although they are similar to those of Capreolus in certain re- somewhat reduced in these genera. Antilo- spects, to Prolibytherium because he believed caprines also lack any evidence of a Palaeo- certain features of its skull were cervidlike. meryx fold, although as the earliest known These postcranial elements included an as- antilocaprids are highly hypsodont, the ab- tragalus and calcaneum that bore similarities sence ofthis character may not be significant. to those of Palaeomeryx, and a metatarsal The entostyle and ectostylids are generally with an open gully, which is not typical of lacking in antilocaprids, but a small entostyle either Palaeomeryx or Capreolus (or any oth- and ectostylid are present in some species of er cervid). However, he describes the gully the earliest merycodontine genus Paracoso- in this metatarsal as being very restricted in ryx (e.g., Paracosoryx minor AMNH ESP the region just proximal to the condyles. It 622-4745). Antilocaprids lack metastylids in may thus represent an immature specimen in the lower molars, but the reduction and loss which the distal bridge over the gully was not of the metastylids is another progressive fea- yet fully formed. We have not had the op- ture associated with increasing hypsodonty portunity to examine this specimen and thus in ruminants, e.g., within the Bovidae and in cannot comment further on its probable af- Aletomeryx within the dromomerycids. finities. In any event, there seems to be little (However, small metastylids may be seen in reason to firmly associate this postcranial the unerupted molars of Antilocapra ameri- material with Prolibytherium. cana.) Both merycodontines and antilo- We would suggest, on the basis ofthe den- caprines possess an antorbital vacuity. tal morphology and the probable cervoidlike In their limb characters, all antilocaprids condition of a double lacrimal orifice, with share the characters of the raised lip of the the additional assumption that the metatarsal cubonavicular facet (character 32d), but uni- with an open gully was misinterpreted, or formly lack the posterior tuberosity on the wrongly assigned to this genus, that Proliby- metapodials that characterizes most higher therium may have been a palaeomerycid of cervoid lineages, that is completely absent some sort. Obviously, additional material only in the genera Blastomeryx, Dremo- would be necessary before we could be con- therium, and Aletomeryx. The form of the fident of this assignation. cubonavicular facet is not a uniquely cervoid feature, but the antilocaprid form bears a ANTILOCAPRIDAE greater resemblance to the condition retained in the cervoids than that seen in the earliest Both merycodontines and antilocaprines bovids. share the derived characters ofa closed meta- Antilocaprids are problematical because tarsal gully (character 28b) (variably open in they possess no unique defining characters as merycodontines) and a double lacrimal ori- a group; no unique characters can typify either fice (character 9) (variably single in both the merycodontines or the antilocaprines. subfamilies). Both subfamilies share the fea- However, both subfamilies share a similar tures of nondeciduous supraorbital cranial mixture of primitive and derived cervoid appendages that appear to be formed from characters. A major problem is the question outgrowths of the frontal bone, rather than of whether the double lacrimal orifice in an- by fusion of a separate dermal ossicone. tilocaprines is homologous with that ofother However, as previously discussed, there is no cervoids. The double orifice in antilocaprines firm evidence for the homology of these cra- is not completely cervidlike in nature (in con- nial appendages between the two subfamilies. trast with the condition in dromomerycids, The cervoid feature of large upper canines Amphitragulus, and Hoplitomeryx), as the 74 AMERICAN MUSEUM NOVITATES NO. 2893 dorsal orifice may be placed inside the orbital asiaticus as the earliest known bovid, from rim, and the presence of a double orifice is the middle Oligocene ofMongolia. The spec- variable (see Scott and Janis, 1987). How- imen consists of the two lower molars (m2 ever, the double orifice of Antilocapra bears and m3), which are moderately hypsodont, a greater resemblance to the cervid condition and are about the size ofthe teeth ofthe living than to the condition of the double lacrimal tragulid genus Hyemoschus. The teeth clearly orifice occasionally found in bovine or tra- belong to a pecoran ruminant, as they possess gelaphine bovids (see Leinders and Heintz, metastylids (character 21), but are simplified 1980, and discussion in Characters section). in their morphology, lacking an anterior cin- Leinders and Heintz (1980) consider Antil- gulum and ectostylids, and show no trace of ocapra to be definitely a cervoid and consider a Palaeomeryx fold. Trofimov (1958) con- this trait to be a synapomorphy linking an- siders these teeth to belong to a bovid because tilocaprids and cervids; on the basis of cur- of the crown height, and the absence of a rently available evidence we would agree with Palaeomeryx fold and other accessory styles their conclusions. that usually characterize cervoid molars. Assuming the double lacrimal orifice to be However, as previously discussed, such fea- a true synapomorphy linking antilocaprids tures are also lost in the teeth of mesodont with cervids, we would consider the antilo- or hypsodont cervoids (for example, the pro- caprids to have branched off the cervoid lin- gressive loss of these features is documented eage above the level of the moschines, but in the three tribes of dromomerycids). The below the level of the "Palaeomerycidae," absence of such dental features is certainly because ofthe lack ofeven a variably present not diagnostic ofbovids, as such dental mor- posterior tuberosity. (In addition, antilo- phology is also typical of antilocaprids (to caprines lack any of the derived dental char- give but one example). Indeed, some Amer- acters of the Palaeomerycidae.) This would ican palaeontologists have claimed Palaeo- assume the original possession and subse- hypsodontus as the earliest merycodontine quent loss of sabrelike canines in the antilo- antilocaprid (Earl Manning, personal com- caprines (which, of course, has occurred nu- mun.). Other problematic early "bovids" in- merous times within other cervoid lineages, clude Gobiocerus mongolicus from the lower linked with the evolution of cranial append- Miocene of Mongolia (Sokolov, 1952), Hyp- ages), and would also assume the parallel evo- sodontus miocenicus (Sokolov, 1949), and lution of the posterior tuberosity in Moschus Kubanotragus sokolovi (Gabunia, 1973), both and Parablastomeryx with the condition in known from Belometcheskaia (middle Mio- higher cervoids. Regardless of the phyloge- cene ofthe north Caucasus). Hypsodontus may netic position of the Antilocapridae, if the also be present in the Miocene ofYugoslavia "Moschidae" (sensu Webb and Taylor, 1980) (Pavlovic and Thenius, 1959). Thomas (1984) are to be considered as a monophyletic group, discusses these taxa, concluding that Gobi- this hypothesis would also entail the parallel ocerus and Hypsodontus may be synony- evolution of the posterior tuberosity within mous, and that the differences in size between this family independent of the cervid con- the fragmentary portions of cranial append- dition. We consider the phylogenetic position ages at Belometcheskaia, attributed to Hyp- and the cladistic reality ofthe Antilocapridae sodontus (small) and Kubanotragus (large), to remain unresolved, and hope that the sug- may be due to sexual dimorphism rather than gestions in this paper will stimulate other to the presence of two different taxa. workers to further research on this topic. All these dental remains do undoubtedly resemble those of present-day neotragine ("dwarf") antelope in size and morphology, OLIGOCENE AND EARLY but there is nothing specific that allies them MIOCENE "BOVIDS" with bovids, and they may equally well rep- A variety of small, hypsodont taxa known resent small cervoids of some sort (mos- primarily or exclusively from dental remains choids or antilocaprids). Neither are the frag- have been ascribed to the Bovidae. Trofimov mentary remains of cranial appendages at (1958) claims the taxon Palaeohypsodontus Belometcheskaia unequivocally bovid in na- 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 75

4 11A 14B 17A 18A 29A 32AO

Fig. 17. Position of the Pecora within the Ruminantia. See table 1 for key to characters. O = primitive characters for the Ruminantia (not necessarily unique). * = unique characters for clade. * = character derived in parallel with condition in pecorans.

ture. It is our opinion that these "bovid" taxa tocone on P3 (character 14b). Evidence for should be classified as Pecora incertae sedis the parallel evolution ofthese character states until more material is known. To classify them can be gained by noting the conflict of these as bovids is misleading in terms of drawing characters with those held synapomorphous conclusions about the time and place of the of higher pecoran lineages. Both primitive origin of the Bovidae. dental features are retained in the genera Walangania and Propalaeoryx, animals which have complete distal metapodial keels. CONCLUSIONS A posteriorly directed protocone on P3 is re- The presence of derived cervoid charac- tained to a lesser extent in both Dremotheri- ters, such as the closed metatarsal gully (char- um and Eotragus, and many antlerless cer- acter 28b) and the Palaeomeryx fold (char- voids retain an M3 with a small metaconule. acter 20) within the primitive Pecora assigned Given the assumptions that the aforemen- to the "Gelocidae," leads us to the conclusion tioned characters have been acquired in par- that the higher Pecora did not have a single allel by higher pecoran lineages, we have ar- common ancestry from within the "Geloci- rived at the following conclusions about the dae," but must have been derived indepen- origins of the pecoran superfamilies. Figure dently from within the diversity of primitive 17 depicts the position of the Pecora within pecoran taxa. If this is indeed the case, then the Ruminantia, and figure 18 presents a certain conclusions must be drawn about the cladogram of the families within the Pecora. probable parallel evolution of certain char- The origin of the Cervoidea (including the acter states that are shared by more advanced Cervidae, Moschidae, Antilocapridae, and ruminants. We previously discussed the fact "Palaeomerycidae," as we have redefined it) that cranial appendages (character 7) and is the most readily apparent. The genera Eu- complete distal metapodial keels (character meryx, Rutitherium, and "Gelocus" whit- 27) are likely to have evolved in parallel worthi possess the derived cervoid character among the higher pecorans. Further charac- ofa Palaeomeryx fold (character 20), and Eu- ters that apparently evolved in parallel are meryx (at least) possesses the derived char- the presence of a larger metaconule on M3 acters ofan incipient entostyle (character 1 5a) (character 1 8b) and a lingually directed pro- and a closed metatarsal gully (character 28b). 76 AMERICAN MUSEUM NOVITATES NO. 2893

GIRAFFOIDEA CERVOIDEA r GIRAFFIDAE CLIMACOCERIDAE

'40

30A 32 D

Fig. 18. Interrelationship of pecoran families. See table 1 for key to characters.

We would thus include our definition of the we have clearly shown in this paper that an Cervoidea (in the cladistic sense) to include open metatarsal gully can be secondarily de- such genera. rived from a closed one, there remains a pos- The origins of the Giraffoidea and the sibility that bovids may share a more recent Bovidae are less clear. If we consider that the common ancestor than Prodremotherium presence of an entostyle derived from the an- with cervoids. terior face of the metaconule (character 15) The Giraffoidea are defined by two distinct to be a true synapomorphy linking the Bovi- apomorphies: the presence ofa bilobed lower dae with the Cervoidea, it follows that the canine (character 23) (Hamilton, 1978a) and Giraffoidea remain as the most primitive of the possession of a posterior cubonavicular the pecoran superfamilies. Bovids and cer- facet on the proximal metatarsus that is elon- voids may also be united by the soft anatomy gated and flat (character 32b). An entostyle feature of a more ventrally positioned en- derived from the anterior face of the meta- trance of the esophagus to the rumen (see conule is rarely a feature of giraffoids, which discussion in Characters section). With the suggests a branching offfrom the pecoran lin- exception ofthe postorbital horn cores (char- eage below the level of the Bovidae or Pro- acter 7b), no apomorphic feature can be found dremotherium. However, giraffoids possess to unite the Bovidae. The absence of a closed metastylids in the lower molars (character 21), metatarsal gully (character 28b) suggests that which suggests a more derived pecoran an- their divergence from the post-giraffoid pec- cestry than Gelocus. The only known taxa oran lineage was below the level of Pro- which fall in this appropriate pecoran grade dremotherium. However, given the fact that for giraffoid ancestry are the poorly known 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 77

Asian genera Notomeryx and Gobiomeryx. fig. 19 for summary). Eumeryx (and Ruti- The condition ofthe metatarsal gully in these therium) are clearly the most primitive cer- genera is not known, although the type of voids, possessing a closed metatarsal gully open gully seen in giraffoids (character 28a) (character 28b) and a Palaeomeryx fold might, as previously mentioned, represent the (character 20); but lacking complete distal plesiomorphic condition for the Pecora. The metapodial keels and possessing a primitive African genus Propalaeoryx may possibly pecoran type of dentition. Walangania is represent the sister taxon to the Giraffoidea, clearly a cervoid based on the presence of a as originally suggested by Gentry (in Hendey, Palaeomeryx fold in Walangania africanus, 1978). Like the giraffoids, Propalaeoryx lacks and is more derived than Eumeryx in the an entostyle and possesses an open metatarsal possession ofcomplete distal keels (character gully, but these are both plesiomorphic fea- 27) (Whitworth, 1958). However, the prim- tures. Propalaeoryx does not possess the gi- itive nature ofthe dentition, with the variable raffoid autapomorphy of the flat and elon- presence of a lingual cingulum in the upper gated cubonavicular facet on the posterior molars (character 1 3b), the small metaconule metatarsus. To the best ofour knowledge, the on M3 (character 18a), the somewhat pos- anterior dentition is not known, so it is im- teriorly situated and directed protocone on possible to ascertain the condition ofthe low- P3 (character 14a), and the apparent lack of er canine. a sabrelike upper canine, relegates Walang- The only possible synapomorphy between ania to the most primitive branching off Propalaeoryx and the Giraffoidea is the pres- point in the cervoid phylogeny after Eume- ence of a bifurcated posterior wing of the ryx. Like Eumeryx, Walangania lacks the fu- metaconule (character 1 9b). However, this is sion of metatarsal V. not a truly unique feature, as it is also seen We consider the Moschidae to be the next in many palaeomerycids, and in a few cer- group to branch offthe cervoid lineage. They vids. Neither can a strong biogeographical are more advanced than Walangania in the case be made for uniting Propalaeoryx with following characters: a lingually situated and the Giraffoidea. If Hamilton (quoted in Pat- directed protocone (character 14b), a prom- terson, 198 1) is correct in synonymizing Pro- inent entostyle in the more brachydont gen- giraffa with Canthumeryx, then the Giraf- era (character 15b) (such as Dremotherium foidea may well have had an Asian origin, and Parablastomeryx), and the possession of rather than an African one, as commonly a large, sabrelike upper canine (character 1 1b). supposed. Clearly, more fossil material of The genus Dremotherium and the blasto- both Asian and African "gelocids" would merycines retain the primitive feature, also prove invaluable in determining giraffoid re- seen in Walangania, of a small metaconule lationships. For the moment, we would con- on M3 (character 1 8a), but are more dentally sider that a common ancestor for the Bovidae advanced than more primitive cervoids in and the Cervoidea would lie among the prim- possessing a complete postentocristid in the itive pecorans at just below the level of Pro- lower molars (character 22b) (in all but some dremotherium, while the ancestry of the Gi- individuals of the early genus Problastome- raffoidea is more distant, among more ryx). Moschids are more primitive than higher primitive pecorans which possess metasty- cervoids in the possession ofa single lacrimal lids but which lack any evidence of an ento- orifice. style. We consider that Propalaeoryx may well We follow Webb and Taylor (1980) in unit- represent the sister taxon to the Giraffoidea, ing the moschid genera by the presence of a but on the basis of the presently available laterally enclosed, subcentral tympanohyal evidence it is impossible to make a strong vagina on the auditory bulla (character 10), case for this hypothesis. but note that the problem with this classifi- Within the Cervoidea, which we would de- cation scheme is that it necessitates the par- fine as pecorans possessing (primitively at allel evolution of the posterior tuberosity of least) a Palaeomeryx fold (character 20) as the metatarsus (character 31) in Moschus, the defining autapomorphic character, we Parablastomeryx, Micromeryx, and His- perceive the interrelationships as follows (see panomeryx with the condition in higher cer- 78 AMERICAN MUSEUM NOVITATES NO. 2893

"MOSCHINA" EUCERVOIDEA A______k_ ____ MOSCHIDAE ANTILO- PALAEO- HOPLITO- CERVIDAE CAPRIDAE MERYCIDAE MERYCIDAE

_lo

13B 14A 15A 16A 28B 29B 32AO

Fig. 19. Interrelationships within the superfamily Cervoidea. See table 1 for key to characters. voids. All but the latter genus share the de- era Palaeomeryx and Amphitragulus, possi- rived feature of the fusion of metatarsal V bly also the genus Prolibytherium, and the (character 30b). Dromomerycidae (which we would now rel- The position of the Antilocapridae, as the egate to subfamily status as the Dromomer- next group to branch off the cervoid lineage, ycinae). As previously discussed, we are here reflects our assumptions that the double lac- making the assumption that more complete rimal orifice (character 9) is a synapomorphy cranial material of Palaeomeryx will reveal linking them with higher cervoids, and also the presence of a double lacrimal orifice that the posterior tuberosity (character 31) (character 9). Palaeomerycids can otherwise has evolved more than once within the Cer- be united by: the variable presence of a pos- voidea. However, we note that no unique fea- terior tuberosity (character 31); the absence ture can be found to unite the Antilocapridae of fusion of metatarsal V in primitive mem- and, with the exception of the form of the bers ofeach lineage; the retention ofp1 (char- cranial appendages, no unique feature can acter 4b) in the primitive members of each distinguish the Antilocaprinae from the Mer- lineage; a sabrelike upper canine (character ycodontinae. It is not clear whether the cra- 11 b) in at least the early members of each nial appendages of the two subfamilies are lineage; and a dentition in which the meta- homologous although both contain genera style is reduced (character 1 6b), the metacone with branched appendages and they may share is attenuated in P4 (character 17b), there is a common developmental origin as an out- a posterolingual groove on p4 (character 25), growth of the frontal bone. and the posterior lobe on m3 possesses a dou- The next family to branch off the cervoid ble loph that is closed posteriorly (character lineage consists of our tentatively redefined 26a). We note that none of these features is family Palaeomerycidae, including the gen- unique in defining the family, but that this 1 987 JANIS AND SCOTT: RUMINANT PHYLOGENY 79 suite of characters does characterize mem- lematical genus, Prolibytherium, which like bers of the three lineages (with the exception Triceromeryx has also been linked to the gi- of the genus Aletomeryx, as mentioned pre- raffoids in past classificatory schemes, may viously, for which we make the assumption actually be a better candidate for palaeomer- that these have been secondarily lost). It is ycid affinities on the grounds of dental mor- not clear to us, on the basis ofcurrently avail- phology, although the very peculiar cranial able evidence, whether the posterior tuber- appendages can probably not be homologized osity evolved independently within the with those of Palaeomeryx and the dromo- Dromomerycinae, or was a primitive char- merycids. We also note that the isolated mo- acter, lost in some early taxa that were more lars at Gebel Zelten ascribed to Zarafa by open-habitat in their ecology (Scott, unpubl. Hamilton (1973) cannot belong to the skull data). However, on the assumption that some of Zarafa from this locality, and may belong form ofposterior tuberosity characterizes the to some unknown African species of Palaeo- primitive members of all lineages within the meryx. Palaeomerycidae, and the presence of a dou- Cervoids higher than the palaeomerycids ble lacrimal orifice in all taxa in which the can be characterized by the invariable pres- orbital region is clearly observable, this char- ence ofthe posterior tuberosity (character 3 1), acter combination leads to the more derived combined with the fusion of metatarsal V position ofthe Paleomerycidae in the cervoid (character 30b); by the complete absence of phylogeny than the Antilocapridae. (Despite p1 (character 4c); and possibly also by the the fact that the posterior tuberosity has ob- primitive presence of a bifurcated protocone viously arisen more than once within the Cer- (character 1 9a). The bifurcated protocone is voidea, as evidenced by the presence of this present in Dicrocerus and some living cervids character in the Moschidae, we consider this (although not in Hydropotes), and while it is placement of the Palaeomerycidae to be the absent in Hoplitomeryx, it is present in Am- most parsimonious, as it only entails the par- phimoschus, which on other criteria appears allel evolution of this character state twice.) to be the closest related genus to Hoplito- Within the Palaeomerycidae, we note that meryx (see Leinders, 1983). It may have been the genus Palaeomeryx can be united with primitively present in Hoplitomeryx, and lost the dromomerycids (or at least with some of with the other cervoid features of the denti- the earlier genera) by the presence of a bi- tion that are usually lost in cervoids with the furcation of the posterior wing of the meta- evolution of increasing hypsodonty, such as conule (character 1 9b), although we note that the entostyle, the ectostylids, and the Pa- this character is also seen in Propalaeoryx laeomeryx fold. (The first two of these fea- and in some early giraffoids (Canthumeryx tures are present in the more brachydont and Palaeotragus). If Palaeomeryx and the Amphimoschus, but have been lost in Hopli- dromomerycids really do share a more recent tomeryx.) We follow Leinders (1983) in as- common ancestry than Amphitragulus, then signing Hoplitomeryx to its own family, the it might be reasonable to assume that the Hoplitomerycidae, and tentatively place Am- form ofcranial appendages in both (supraor- phimoschus in the same family, despite the bital appendages that appear to have been apparent lack of cranial appendages, on the formed in a giraffidlike fashion from fusion basis of similarities of the cranium and the of a dermal ossicone with the skull, and the possession of a double loph on the posterior additional presence ofa single median occip- lobe ofm3 that is open posteriorly (character ital ossicone) was inherited from a common 26b) (see Leinders, 1983). ancestor. However, we would note, while en- Finally, we would group Hydropotes, the dorsing this hypothesis, that a similar mor- living antlerless cervid, with antlered cervids phology of the cranial appendages is seen in on the basis ofpossession ofthe lacrimal fos- the genus Triceromeryx, which lacks many sa (character 8), as well as the common fea- of the palaeomerycid derived dental char- tures of soft anatomy (character suite 34). acters. We would prefer to leave Tricero- The lacrimal fossa is found throughout ant- meryx as "cervoid insertae sedis" for the time lered cervids, and while it does occur in par- being. However, we note that another prob- allel among other isolated pecoran genera (as 80 AMERICAN MUSEUM NOVITATES NO. 2893 previously discussed), at this level ofcervoid character 20). Our final conclusions in figures evolution it may be considered a valid char- 18 and 19 represent our tentative summary acter state, especially as a lacrimal fossa is of the interrelationships within the pecoran absent in Hoplitomeryx (Leinders, 1983). (The genera that we consider to be the most par- condition of the lacrimal fossa in Amphi- simonious, based on the presently available moschus is unknown.) We would agree with evidence from living animals and the existing Chow and Shih (1978) in assigning Lago- fossil record. It is our hope that these pro- meryx and other related "lagomerycid" gen- posed phylogenies, with or without the dis- era to the subfamily Muntiacinae within the covery of additional fossil evidence, will Cervidae. stimulate other workers in ruminant taxon- On the assumption that the double lacri- omy to support or refute our hypotheses, as mal orifice (character 9) has evolved only once we consider the evolution of the higher ru- within the Cervoidea, we would group the minant grade, with the capacity of its mem- Antilocapridae, Palaeomerycidae, Hoplito- bers for enormous physiological and ecolog- merycidae, and Cervidae in the higher cer- ical flexibility in the evolution of cursorially void assemblage Eucervoidea. The term adapted animals capable of dealing with "Moschina," originally used by Webb and vegetation ofhigh cellulose content, to be one Taylor (1980), could then be used as a para- of the more interesting areas of mammalian phyletic grouping of other cervoids (as de- evolution. fined by the presence of a Palaeomeryx fold,

REFERENCES CITED Arambourg, C. Chow, Ben-shun, and Mo-chuang Shih 1933. Mammiferes Miocenes du Turkana 1978. A skull of Lagomeryx from the middle (Afrique orientale). Ann. Palaeontol. 22: Miocene of Linchu, Shantung. Vertebr. 121-148, pls. 12, 13. Palasiat., 16(2): 111-122, fig. 1, pl. 1. Bohlin, Birger [in Chinese and English] 1935. Kritische Bemerkungen uiber die Gat- Churcher, C. S. tung Tragocenus. Nova Acta Regiae Sac. 1970. Two new upper Miocene giraffids from Ups., ser. 4, 9: 1-18, pls. 1-5. Fort Ternan, Kenya, East Africa. In L. Bouvrain, Genevieve, and Denis Geraads S. B. Leakey and R. J. G. Savage (eds.), 1985 Un squelette complet de Bachitherium Fossil vertebrates of Africa, vol. 2, pp. (Artiodactyla, Mammalia), de l'Oligo- 1-106. London: Academic Press. cene de Cerestes (Alpes de Haute-Prov- 1978. Giraffidae. In V. I. Maglio and H. B. S. ince). Remarks sur la systematique des Cooke (eds.), Evolution of African Ruminants primitives. C. R. Acad. Sci. mammals, pp. 509-535, figs. 1-10. Paris., ser. 2, 300(2): 75-78, fig. 1. Cambridge, Mass.: Harvard University Brandt, Kurt Press. 1928. Die Entwicklung des Homes beim Rinde Colbert, Edwin H. bis zum Beginn der Pneumatisation des 1933. A skull and mandible of Giraffokeryx Hornzapfens. Morphol. Jahrb., 60: 428- punjabensis. Ann. Mus. Novitates, 632: 468, figs. 1-9. 1-14, figs. 1-6. Brooke, Sir Victor 1936. Tertiary deer discovered by the Amer- 1878. On the classification of the Cervidae, ican Museum Asiatic expeditions. Am. with a synopsis of the existing species. Mus. Novitates, 854: 1-15. Proc. Zool. Soc. London: 833-928, figs. Coope, G. R. 1-19, pl. LV. 1968. The evolutionary origin ofantlers. Deer, Bubenik, Dr. Anton B. 1: 215-217. 1966. Das Geweih. Hamburg and Berlin: P. Cope, E D. Parey Verlag, 214 pp., 84 figs. 1874. Report on the stratigraphy and Pliocene 1982. Taxonomy ofPecora in relation to mor- vertebrate palaeontology of Northern phophysiology of their cranial append- Colorado. Bull. U.S Geol. Geogr. Surv. ages. In Robert D. Brown (ed.), Antler Ten., 1(1): 9-28. development in Cervidae, pp. 163-186, 1877. The extinct Vertebrata obtained in New figs. 1-17. Kingsville, Texas: Caesar Mexico by parties of the expedition of Kleberg Wildlife Research Institute. 1874. Report ofthe U.S. Geol. Surv. W. 1987 JANIS AND SCOTT: RUMINANT PHYLOGENY 81

100th Meridian (Wheeler), 4(2): xii + tions. Acta Zool. Fenn., 180: 1-76, figs. 370 pp., 83 pls., 1 woodcut. 1-43. Crusafont-Pair6, Miguel Frechkop, S. 1952. Los Jirafidos f6siles de Espania. Me- 1955. Sous-ordre des Ruminants ou Sel&no- morias y Communicaciones del Insti- donts. In P. Grassse (ed.), Traite de tuto Geol6gico (Barcelona), 8: 1-69, figs. zoologie, vol. 17(1), pp. 568-693. Paris: 1-4, pls. 17-35. Masson et Cie. 1961. Giraffoidea. In J. Piveteau (ed.), Traite Frick, Childs de Pal'aeontologie, tome 6, pp. 1022- 1937. Horned ruminants of North America. 1037, figs. 120-135. Paris: Masson et Bull. Am. Mus. Nat. Hist., 69: xxvii + Cie. 669 pp., 103 figs. Dove, W. Franklin Furlong, E. L. 1935. The physiology of horn growth: a study 1927. The occurrence and phylogenetic status ofthe morphogenesis, interaction oftis- of Merycodus from the Mohave Desert sues, and the evolutionary processes of Tertiary. Univ. Calif. Berkeley, Publ. a Mendelian recessive character by Geol. Sci., 17: 145-186, figs. 1-28, pls. means of transplantation of tissues. J. 24-28. Exp. Zool., 69: 347-405, figs. 1-36, pls. Gabunia, L. K. 1-5. 1973. The fossil vertebrate fauna from Belo- Diirst, Dr. J. Ulrich metcheskaia. Tbilisi: Metsniereba, 136 1902a. Versuch einer Entwicklungsgeschichte pp., 35 figs. [in Russian] der H6rner der Cavicornia nach Unter- Garrod, A. H. suchungen am Hausrinde. Sonderab- 1877. Notes on the anatomy of the Chinese druck aus Forshungen auf dem Gebiete water deer (Hydropotes inermis). Proc. der Landwirtschaft. (Festschrift zur Feier Zool. Soc. London, 1877: 789-793, figs. des 70. Geburstages von Prof. Dr. Ad. 1-25. Kramer.) Zurich, 1902, pp. 1-47, figs. Gentry, A. W. 1-14. 1970. The Bovidae (Mammalia) of the Fort 1902b. Sur le developpement des cornes chez Ternan fossil fauna. In L. S. B. Leaky les Cavicornes. Bull. Mus. Hist. Nat. and R. J. G. Savage (eds.), Fossil ver- (Paris), 8: 197-203, figs. 1-5. tebrates of Africa, vol. 2, pp. 243-323, 1926. Das Horn der Cavicornia. Denksschrift. figs. 1-22. London: Academic Press. d. Schweiz. Naturforsch. Gesell., 68: 1- 1974. A new genus and species of Pliocene 180, figs. 1-80. boselaphine (Bovidae, Mammalia) from Fambach, Dr. South Africa. Ann. South Afr. Mus., 1909. Gewieh und Gehorn. Ein kritisches Re- 65(5): 145-188, figs. 1-28. ferat. Z. Naturwiss., 81: 225-264, figs. 1978. Bovidae. In V. I. Maglio and H. B. S. 1-18. Cooke (eds.), Evolution of African Flerov, C. C. mammals, pp. 540-572, figs. 1-9. Cam- 1952. Musk deer and deer. Fauna of the bridge: Harvard University Press. U.S.S.R., Mammals, 1(2). Moscow and Geoffroy Saint-Hilaire, E. Leningrad: The Academy ofSciences of 1833. Considerations sur des ossements fos- the U.S.S.R., 257 pp., 79 figs. [translat- siles, la plupart inconnus, trouves et ob- ed through Israel program for scientific serves dans les bassins de l'Auvergne. translation, 1960] Revue Encyclopedique, 59: 76-95. Flower, William Henry George, A. N. 1875. On the structure and affinities of the 1956. The post-natal development ofthe horn musk deer (Moschus moschiferous tubules and fibres (intertubular horn) in Linn.). Proc. Zool. Soc. London, 1875: the horns of sheep. Br. Vet. J., 112(1): 159-190, figs. 1-14. 30-34, pls. 1-8. Forbes, W. A. Geraads, Denis 1882. Supplementary notes on the anatomy of 1986. Remarques sur la systematique et la the Chinese water deer (Hydropotes phylog6nie des Giraffidae (Artiodactyla, inermis). Proc. Zool. Soc. London, 1882: Mammalia). Geobios, 19(4): 465-477, 636-638, figs. 1, 2. figs. 1, 2. Fortelius, Mikael Gervais, P. 1985. Ungulate cheek teeth: developmental, 1859. Zoologie et Palaeontologie Fran,aise, functional, and evolutionary implica- 2nd ed. Paris. 82 AMERICAN MUSEUM NOVITATES NO. 2893

Ghoshal, N. G. Entwicklungsphysiologie der Stangen- 1975. Ruminant heart and arteries, pp. 960- bildung beim Reh (Capreolus c. capreo- 1023, and Ruminant spinal nerves, pp. lus L. 1758). Wilhem Roux Arch. Ent- 1124-1151. In R. Getty (ed.), Sisson and wicklungs Mech. Org., 158: 358-384. Grossman's The anatomy of the do- 1968. Verhinderung der Rosenstoch und mestic animals, 5th ed. Philadelphia: W. Stangenbildung beim Reh, Capreolus B. Saunders. capreolus, durch Periostauschaltung. Ginsburg, L., and E. Heintz Zool. Garten (Leipzig), n. ser., 35: 252- 1966. Sur les affinites du genre Palaeomeryx 255. (ruminent du Miocene Europeen). C. R. Hartwig, H., and J. Schrudde Acad. Sci. Paris, D, 262: 979-982, 1974. Experimentelle Untersuchungen zur fig. 1. Bildung der Primaren Stirnauswiuchse 1968. La plus ancienne antilope d'Europe, beim Reh (Capreolus capreolus L.). Z. Eotragus artenensis du Burdigalien Jagdwiss., 20: 1-13. d'Artenay. Bull. Mus. Nat. d'Hist. Nat. Heintz, E. Paris, ser. 2, 40(4): 837-842, figs. 1-4. 1963. Les characteres distinctifs entre meta- Goss, Richard J. tarses de Cervidae et Bovidae actuels et 1983. Deer antlers: regeneration, function and fossils. Mammalia, 27: 200-209, figs. 1- evolution. New York: Academic Press, 6. 316 pp., 153 figs. 1970. Les Cervides Villafranchiens de France Goss, Richard J., C. W. Severinghaus, and S. Free et d'Espagne. MWm. Mus. Nat. d'Hist. 1964. Tissue relationships in the development Nat. Paris, n. ser., ser. C, Sciences de la of the pedicles and antlers in the Vir- Terre. XXII, vols. 1 and 2, 303 pp., figs. ginia deer. J. Mamm., 45: 61-68, pl. 1. 1-43. Gray, J. E. Heintz, Emile, Michel Brunet, and Sevket Sen 1821. On the natural arrangement of verte- 1981. Un nouveau Giraffid6 du Miocene su- brate mammals. London Med. Repos- perieur d'Irak, Injanatherium hazimi, itory, 15(1): 296-310. n.g., n.sp. C. R. Acad. Sci. Paris, 292(11): Groves, Colin P., and Peter Grubb 423-426, fig. 1. 1982. The species of muntjac (genus Muntia- Hendey, Q. B. cus) in Borneo: unrecognised sympatry 1978. Preliminary report on the Miocene ver- in tropical deer. Zool. Med., 17: 203- tebrates from Arrisdrift, South West Af- 216, figs. 1-4. rica. Ann. South African Mus., 76: 1- 1987. Relationships of living deer. In C. 41, figs. 1-13. Wemmer (ed.), The biology and man- Hoffman, R. R. agement ofthe Cervidae. National Zoo- 1973. The ruminant stomach. In T. R. logical Society Symposium Volume. Odhiambo (ed.), East African mono- Washington, D.C.: Smithsonian Insti- graphs in biology (2). East African Lit- tution Press. erature Bureau, 349 pp. Ham, Arthur W., and David H. Cormack Janis, Christine M. 1979. Histology, 8th ed. Philadelphia: J. B. 1979. Aspects ofthe evolution ofherbivory in Lippincott Co., 966 pp. ungulate mammals. Unpubl. Ph.D. the- Hamilton, W. R. sis, Harvard University, Cambridge, 484 1973. On the Lower Miocene ruminants of pp., 84 figs. Gebel Zelten, Libya. Bull. Br. Mus. (Nat. 1982. Evolution of horns in ungulates; ecol- Hist.) Geol., 21: 75-150, figs. 1-10, pls. ogy and palaeoecology. Biol. Rev., 57: 1-5. 261-318, figs. 1-6. 1978a. Fossil giraffes from the Miocene of Af- 1984. The use of fossil ungulate communities rica, and a revision of the phylogeny of as indicators of climate and environ- the Giraffoidea. Philos. Trans. Roy. Soc. ment. In P. J. Brenchley (ed.), Fossils (London), ser. B., 282: 165-229, figs. 1- and climate, pp. 85-104, figs. 1-3. 13, pis. 1-5. Chichester: Wiley. 1978b. Cervidae and Palaeomerycidae. In V. I. 1987. Grades and clades in hornless ruminant Maglio and H. B. S. Cooke (eds.), Evo- evolution: the reality of the Gelocidae lution of African mammals, pp. 496- and the systematic position of Lo- 508. Cambridge, Mass.: Harvard Uni- phiomeryx and Bachitherium. J. Vert. versity Press. Paleontol., 7(2): 200-216, figs. 1-10. Hartwig, H. Janis, Christine M., and Adrian Lister 1967. Experimentelle Untersuchungen zur 1985. The morphology of tJle lower fourth 1987 JANIS AND SCOTT: RUMINANT PHYLOGENY 83

premolar as a taxonomic character in Mus. Nat. Hist., 24: 535-562, figs. 1- the Ruminantia (Mammalia, Artio- 15. dactyla) and the systematic position of 1918. Contributions to the Snake Creek Fau- Triceromeryx. J. Paleontol., 59(2): 405- na, with notes upon the Pleistocene of 410, fig. 1. Western Nebraska. Bull. Am. Mus. Nat. Jehenne, Y. Hist., 38(7): 183-229, figs. 1, 2, pls. 4- 1977. Description du premier crane du genre 10. Prodremotherium Filhol, 1877. (Ru- 1926. On a new primitive deer and two trag- minant primitif de l'Oligocene Eurasi- uloid genera from the Lower Miocene atique). Geobios, Mem. Spec., 1: 233- ofNebraska. Am. Mus. Novitates, 215: 237, p1. 1. 1-8, figs. 1-3. Kowalevsky, Vladimir 1934. A phylogenetic chart of the Artiodac- 1876. Osteologie des Gelocus aymardi. Pal'ae- tyla. J. Mamm., 15: 207-209, fig. 1. ontographica, n. ser., 4: 145-162, taf. Matthew, W. D., and Walter Granger 21, 22. 1924. New insectivores and ruminant faunas Lankester, E. Ray from the Tertiary of Mongolia, with re- 1907a. The origin of the lateral horns of the marks on the correlation. Am. Mus. giraffe in foetal life on the area of the Novitates, 105: 1-7, figs. 1-3. parietal bones. Proc. Zool. Soc. London, Meyer, H. von 1907: 100-115, figs. 24-36. 1834. Die fossilen Ziihne und Knochen und 1907b. On the existence ofrudimentary antlers ihre Ablagerung in der Gegend von in the okapi. Proc. Zool. Soc. London, Georgensmiind in Bayern. Abh. Sen- 1907: 126-135, figs. 49-55, pls. 6, 7. ckenb. Naturf. Ges., Frankfurt a.m., Langer, P. l(suppl.), viii + 126 pp., 14 pls. 1974. Stomach evolution in the Artiodactyla. Milne-Edwards, Alphonse Mammalia, 38(2): 295-314, figs. 1-8. 1864. Recherches anatomique, zoologique et Lavocat, Rene pal'aeontologique sur la famille dt s 1955. Revision de la faune des mammiferes Chevrotains. Ann. Sci. Nat., ser. 5, oligocenes d'Auvergne et du Velay. Sci- Zoologique, 2: 49-167, pls. 2-11. ences et Avenir (Paris), 153 pp., 26 pls. Morales, J., S. Moya-Sola, and D. Soria Leinders, J. J. M. 1981. Presencia de la familia Moschidae (Ar- 1979. On the osteology and function of the tiodactyla, Mammalia) en el Vallesiense digits in some ruminants and their bear- de Espafia; Hispanomeryx duriensis ing on taxonomy. Z. Saugetier., 44(5): novo gen nova sp. Estudios Geol., 37: 305-318, figs. 1-5. 467-475, figs. 1-4, pl. 1. 1983. Hoplitomerycidae fam. nov. (Rumi- Nobach, Charles V. nantia, Mammalia), from Neogene fis- 1932. The deciduous horns of the pronghorn sure fillings in Gargano (Italy). Scripta antelope, Antilocapra americana. Bull. Geol., 70: 1-51, figs. 1-4, pls. 1-10. New York Zool. Soc., 35(6): 195-207, Leinders, J. J. M., and E. Heintz figs. 1-13. 1980. The configuration ofthe lacrimal orifice O'Gara, Bart W., and Gary Matson in pecorans and tragulids (Artiodactyla; 1975. Growth and casting of horns by prong- Mammalia) and its significance for the horns and exfoliation of horns by bo- distinction between Bovidae and Cer- vids. J. Mamm., 56: 829-846, figs. vidae. Beaufortia, 30: 155-160, pls. 1-8. 1-4. Patterson, C. Lydekker, R. 1981. Methods of palaeobiogeography. In G. 1883. Siwalik Suina. Palaeontol. Indica, 10: Nelson and D. E. Rosen (eds.), Vicari- 142-177. ance biogeography. A critique, pp. 446- Maclnnis, D. G. 500. New York: Columbia University 1936. A new genus offossil deer from the Mio- Press. cene of Africa. Zool. J. Linn. Soc. Lon- Patton, Thomas Hudson, and Beryl E. Taylor don, 39: 521-530, figs. 1-5. 1973. The Protoceratinae (Mammalia, Tylo- Matthew, W. D. poda, Protoceratidae) and the system- 1904. A complete skeleton ofMerycodus. Bull. atics of the Protoceratidae. Bull. Am. Am. Mus. Nat. Hist., 20: 101-129, figs. Mus. Nat. Hist., 150(4): 347-414, figs. 1-21, p1. 1. 1-15. 1908. Osteology of Blastomeryx and phylog- Pavlovic, M., and E. Thenius eny ofthe American Cervidae. Bull. Am. 1959. Gobiocyon macrognathous (Canidae, 84 AMERICAN MUSEUM NOVITATES NO. 2893

Mammalia) aus dem Mioziin Jugosla- schichte der Hirsche. Abhandl. Schweiz. wiens. Anz. Ost. Akad. Wiss., 11(3): 1- Paliieontol. Gesell., 10: 1-122, pls. 1-6. 8, figs. 1-5. Savage, D. E., Zhai Renjie, Tong Yong Sheng, and Pilgrim, Guy E. R. L. Ciochon 1911. The fossil Giraffidae of India. Mem. Ms. Concerning the small artiodactyls from Geol. Surv. India., 4(1): 1-29, pls. 1-5. the late Eocene and early Oligocene of 1941. The dispersal of the Artiodactyla. Biol. Asia and China. Rev., 16: 155-158. Scott, Kathleen M. 1947. The evolution of the buffaloes, oxen, 1985. Allometric trends and locomotor ad- sheep and goats. J. Linn. Soc. London, aptations in the Bovidae. Bull. Am. Mus. 41: 272-286, figs. 1-6. Nat. Hist., 179(2): 197-288, figs. 1-16. Pomel, A. 1987. Allometry and habitat-related adap- 1845-1846. Memoire pour servir a la geologie tations in the postcranial skeleton of palaeontologique des terrains tertiares the Cervidae. In C. Wemmer (ed.), Bi- du departement de l'Allier. Bull. Societe ology and management of the Cervi- Geologique de France, ser. 2, vol. 3, 353 dae. National Zoological Society Sym- pp- posium Volume. Washington, D.C.: 1853. Catalogue methodique et descriptif des Smithsonian Institution Press. Vertebres fossils de la Loire. Paris: Bail- Scott, Kathleen M., and Christine Janis liere, 193 pp. 1987. The phylogenetic position of the Cer- Qiu, Zhauxiang, Defa Yan, Hang Jia, and Bo Sun vidae, and the case for a superfamily 1985. Preliminary observations on the newly Cervoidea. In C. Wemmer (ed.), The found skeleton of a Palaeomeryx from biology and management of the Cervi- Shanwag, Shandong. Vert. Palasiat., dae. National Zoological Society Sym- 23(3): 173-195, figs. 1-3, pls. 1-5. [in posium Volume. Washington, D.C.: Chinese and English] Smithsonian Institution Press. Radinsky, L. B. Sigogneau, Denise 1984. Ontogeny and phylogeny in horse skull 1968. Le genre Dremotherium (Cervoidea). evolution. Evolution, 38: 1-15, figs. Anatomie du crane, denture et moulage 1-6. endocranien. Ann. Palaeontol. (Ver- Richard, M. t'ebre's), 54(l): 3-100. 1946. Contribution a l'etude du bassin Simpson, George Gaylord d'Aquitaine. Les gisments de Mammi- 1945. The principles of classification, and a feres tertiares. Mem. Soc. Geol. France, classification of mammals. Bull. Am. 24(52): 380 pp. Mus. Nat. Hist., 85: xvi + 350 pp. Roger, Dr. Otto Sokolov, I. I. 1904. Wirbeltierreste aus dem Obermiociin der 1949. Sur des restes de Cavicornes (Bovidae, bayerish-schwabischen Hochebene. Ber. Mammalia) trouves dans le Miocene Naturw. Ver. Schwaben, Ausberg, 36: moyen du Caucase septentrional. Dokl. 1-22, pls. 1-4. Akad. Nauk. SSSR, 67(4): 1101-1104, Roman, F., and J. Viret figs. 1-3. [in Russian] 1934. La faune de Mammiferes du Burdiga- 1952. Remains of Bovidae, Mammalia, from lien de la Romieu (Gers). Mem. Soc. lower Miocene deposits ofwestern Gobi. Geol. France Paris, n. ser., 9(21): 1-67, Tr. Palaeontol. Inst. Akad. Nauk. SSSR, pls. 1-12. 41: 155-158, fig. 1. [in Russian] Romer, Alfred Sherwood Solounias, Nikos 1966. Vertebrate palaeontology, 3rd ed. Chi- 1981. The Turolian fauna from the island of cago: University of Chicago Press, viii Samos, Greece. In M. K. Hecht and F. + 468 pp., 443 figs. S. Szalay (eds.), Contributions to ver- Roth, V. Louise tebrate evolution, vol. 6, pp. vi-xv, 1- 1984. On homology. Biol. J. Linn. Soc., 22: 232, figs. 1-67, map 1. Basel: S. Karger. 13-29. In press. Evidence from horn morphology on Riitimeyer, L. the phylogenetic relationships of the 1881. Beitriige zu einer natiirlichen Ge- pronghorn (Antilocapra americana). J. schichte der Hirsche. Abhandl. Schweiz. Mamm. Palaeontol. Gesell., 8: 1-93, pls. 1, 2. Spinage, C. A. 1883. Beitrage zu einer natuirlichen Ge- 1968. Horns and other bony structures of the 1987 JANIS AND SCOTT: RUMINANT PHYLOGENY 85

skull of the giraffe, and their functional lution ofartiodactyls. Paleobiology, 1(2): significance. East Afr. Wildl. J., 6: 53- 175-188, figs. 1-3. 61, figs. 1-5, pls. 1, 2. Trofimov, B. A. Stirton, R. A. 1958. New Bovidae from the Oligocene of 1944. Comments on the relationships of the Central Asia. Vert. Palasiat., 2(4): 243- Palaeomerycidae. Am. J. Sci., 242:633- 247, fig. 1. 655, figs. 1, 2. Viret, Jean Stromer, E. 1961. Artiodactyla. In J. Piveteau (ed.), Traite 1926. Reste Land- und Siisswasser-Bewohn- de Palaeontologie, vol. 6, pp. 887-1021 ender Wirbeltiere aus den Diamant- and 1038-1084. Paris: Masson et Cie. feldern Deutsch Siidwestafrikas. In E. Voorhies, Michael R. Kaiser (ed.), Die Diamantwuste Sud- 1969. Taphonomy and population dynamics west-africas, Berlin, vol. 2, pp. 107-153, ofearly Pliocene vertebrate fauna, Knox figs. 1-6. County, Nebraska. Contrib. Geol., spec. Sudre, Jean pap. no. 1. Laramie: University of Wy- 1984. Cryptomeryx (Schlosser, 1886), tragu- oming Press, 69 pp., 29 figs. lide de l'Oligocene d'Europe: relations Webb, S. David du genre et considerations sur l'ori- 1973. Pliocene pronghorns of Florida. J. gine des ruminants. Palaeovertebrata, Mamm., 54: 203-22 1, figs. 1-7. Montpellier, 14(1): 1-31, figs. 1-25. 1983a. The rise and fall of the late Miocene Taylor, Beryl, and S. David Webb ungulate fauna in North America. In M. 1976. Miocene Leptomerycidae (Artiodacty- H. Nitecki (ed.), Coevolution, pp. 267- la, Ruminantia) and their relationships. 306. Chicago: University of Chicago Am. Mus. Novitates, 2596: 1-22, figs. Press. 1-7. 1983b. A new species of Pediomeryx from the Teilhard de Chardin, P. late Miocene of Florida, and its rela- 1939. The Miocene cervids from Shantung. tionships within the subfamily Cranio- Bull. Geol. Soc. China, 19(3): 269-278, ceratinae (Ruminantia, Dromomer- figs. 1-7. ycidae). J. Mamm., 64(2): 261-276, figs. Thenius, Erich 1-7. 1969. Phylogeny de Mammalia. Berlin: Wal- Webb, S. David, and Beryl E. Taylor ter Greyter. 1980. The phylogeny of hornless ruminants Thomas, Herbert and a description of the cranium ofAr- 1984. Les Giraffoidea et les Bovidae Miocenes chaeomeryx. Bull. Am. Mus. Nat. Hist., de la Formation Nyakach (Rift Nyanza, 167(3): 117-158, figs. 1-11. Kenya). Palaeontographica, abt. A, 183: Whitworth, T. 64-89, figs. 1-7, pls. 1-3. 1958. Miocene ruminants ofEast Africa. Brit. 1985. Les Bovidae (Artiodactyla: Mammalia) Mus. Nat. Hist., Foss. Mammals Afr., du Miocene du sous-continent Indien, 15: 1-50, figs. 1-18. de la peninsule Arabique et de l'Afrique: Zittel, K. A. von biostratigraphie, biogeographie et 1925. In A. S. Woodward (ed.), Text-book of ecologie. Palaeogeogr., Palaeoclimatol., palaeontology, 3. Mammalia. London: Palaeoecol., 45: 251-299, figs. 1-5. Macmillan, viii + 316 pp., 374 figs. Todd, Neil B. 1975. Chromosomal mechanisms in the evo-

Recent issues of the Novitates may be purchased from the Museum. Lists of back issues of the Novitates, Bulletin, and Anthropological Papers published during the last five years are available free of charge. Address orders to: American Museum of Natural History, Department D, Central Park West at 79th St., New York, N.Y. 10024.