Comparative osteology of early Tertiary tapiromorphs (Mammalia, Perissodactyla)

Transcription

1 -l_-_l_l_ - Zoological Journal of the Linnean Society (ZOOl), 132: With 21 figures doi: Jzj1s , available online at hstp;(hww.idealibrary.com on I E )r-l --_ Comparative osteology of early Tertiary tapiromorphs (Mammalia, Perissodactyla) LUKE T. HOLBROOK Department of Biological Sciences, Rowan University, Glassborn, NJ , USA Received September 2999; accepted for publication May 2000 Characters of cranial and postcranial osteology provide important data for examining the interordinal relationships of mammals. Understanding variation in the cranial and postcranial skeleton is necessary for adequately representing ancestral character states for each mammalian order. This paper provides a comprehensive description and discussion of cranial and post-rsnial osteology in one of the major perissodactyl lineages, the Tapiromorpha, the lineage including the extant rliinoceroses and tapirs and their fossil relatives. Tapiromorph skeletal morphology is described, and tapiromorphs axe compared to each other, to other perissodactyls, and to a number of other eutherians. Characters of potential phylogenetic significance are emphasized. A phylogenetic analysis of tapiromorphs using cranial and postcranial osteological characters provides a number of interesting results. The data do not support slacing Homogalar and Chalicotherioidea within Tapiromorpha. The data support several novel hypotheses regarding the phylogeny of rhinocerotoids, including a close relationship between rhinocerotids and amynodontids, relative to hyracodontids. Finally, examples of the relevance of tapiromorph osteology to understanding perissodactyl origins are discussed, particularly new interpretations of certain primitive character states for perissodactyls The Linnean Society of London ADDITIONAL KEYWORDS: Tapiromorpha - perissodactyls - osteology - skull - postcranial. INTRODUCTION Cranial and postcranial morphology has figured prominently in phylogenetic studies of eutherian mammals, particularly studies of interrelationships among orders (e.g. Court, 1992; Novacek, 1986; Novacek & Wyss, 1986; Novacek, Wyss & McKenna, 1988; Thewissen & Domning, 1992; Prothero, Manning & Fischer, 1988). It is easy to score characters of the skull and postcranial skeleton for orders with low diversity, but characterizing the morphology of a diverse order is difficult, because a broader range of taxa must be considered, and their phylogenetic relationships can affect the interpretation of the primitive morphology for the order. Thus, many interordinal studies are compelled to use a less rigorous approximation of the primitive osteology of a diverse order. For instance, only the extant members of the order might be examined, or some most primitive fossil or fossils might be used to represent the whole order. For diverse orders, comprehensive studies of osteological variation within the order are necessary for overcoming this problem. holbrook@rowan.edu /01/ $35.00/0 1 The order Perissodactyla includes the extant horses, rhinoceroses, and tapirs, as well as an excellent fossil record dating from the earliest Eocene. Despite a wealth of material and a large body of previous work on this group, the intraordinal and interordinal phylogenetic relationships of perissodactyls are not well understood. One reason is that most studies of perissodactyl phylogeny have focused on tooth morphology. While teeth do provide interesting data for phylogenetic analysis, they are only one source of potentially useful information. Furthermore, few dental characters are utilized in higher-level studies, and the focus on teeth has come at the expense of comparative studies of perissodactyl skulls and postcrania (e.g. compare characters from Hooker, 1989, and Novacek, 1992b). A rigorous comparative study of perissodactyl osteology is long overdue and important to our understanding of how perissodactyls are related to other mammals. This paper discusses the comparative osteology of a major lineage of the Perissodactyla, the Tapiromorpha. The purpose of this study is two-fold. One aim is to describe variation in skeletal morphology among tapiromorph genera and use this information to better understand ancestral character states for Tapiromorpha and Perissodactyla. The second aim is to use The Linnean Society of London

2 2 L. T. HOLBROOK character information from the skull and postcranial skeleton to elucidate tapiromorph phylogeny. The first part of this paper will describe and compare the morphology of each bone among a comparative set of taxa including tapiromorph perissodactyls, non-tapiromorph perissodactyls, and other selected mammals. The second part will utilize selected characters of the skeleton for a phylogenetic analysis of tapiromorph perissodactyls. Finally, selected characters will be used to illustrate and discuss the relevance of tapiromorph osteology to perissodactyl origins. INSTITUTIONAL ABBREVIATIONS Throughout this paper, references are made to specimens housed in many museums. I examined specimens from the following institutions (listed with their abbreviated acronyms): the AmericanMuseum of Natural History, New York (AMNH); the Carnegie Museum of Natural History, Pittsburgh (CM); the Denver Museum of Natural History, the Los Angeles County Museum, Los Angeles (LACM and LACM(CIT)); the Museum of Comparative Zoology, Harvard University, Cambridge (MCZ); the University of Colorado, Boulder (UC); the University of California Museum of Paleontology, Berkeley (VCMP); the University of Michigan Museum of Paleontology, Ann Arbor (UM); the United States Geological Survey (USGS; specimens stored at Johns Hopkins University, Baltimore); the Zoology Museum of the University of Massachusetts, Amherst (UMA); the University of Oregon Museum of Natural History, Eugene (UOMNH); the National Museum of Natural History, Smithsonian Institution, Washington (USNM); the University of Wyoming, Laramie 0; the Peabody Museum of Natural History, Yale University OCPM); the Princeton University collection, housed in the Peabody Museum (YF M(PU)). TAF IROMOFWHA Perissodactyla has traditionally been divided into two suborders: the Hippomorpha (including the superfamilies Equoidea, Chalicotherioidea, and Brontotherioidea) and the Ceratomorpha (including the superfamilies Tapiroidea and Rhinocerotoidea) (Wood, 1937). Scott (1941) and Radinsky (1964) both recognized a third suborder, Ancylopoda (originally established as a separate order by Cope [1889]), including only the Chalicotherioidea. The classifications of Scott and Radinsky differed only in that Scott separated Ancylopoda from his Chelopoda (Hippomorpha + Ceratomorpha), whereas Radinsky considered Ancylopoda to be of equal rank to Hippomorpha and Ceratomorpha. Hooker (1984, 1989) established the Tapiromorpha to recognize a close phylogenetic relationship between the traditional Ceratomorpha and Chalicotherioidea. In Hooker s (1989) scheme, chalicotherioids are allied with a group of Old World tapiroids, the Lophiodontidae, in Ancylopoda. He considered Ancylopoda and a more exclusive Ceratomorpha to be sister-taxa, and they in turn are the sister-group to the Isedolophidae, a primitive group historically placed in the Ceratomorpha. Subsequent phylogenies (e.g. Colbert & Schoch, 1998) followed Hooker s classification, although Schoch (1989) proposed the term Moropomorpha as a substitute for Tapiromorpha. In this paper, Hooker s term Tapiromorpha will be used. Table 1 lists the putative tapiromorph genera examined in this study. Hooker s (1989) classification is the basis for the scope of Tapiromorpha. For those families whose monophyly is reasonably well established, at least two genera that were thought to demonstrate the primitive condition for the family were analysed. For poorly defined families (i.e. those that are not likely to be monophyletic), all members represented by cranial or postcranial material were included. A more detailed discussion of tapiromorph families is given below. ISECTOLOPHIDAE This family is considered by Hooker (1984, 1989) to be the sister-group of all other tapiromorphs. The monophyly of this taxon, considered to include Homogalax, Isectolophus, and Cardiolophus, is uncertain. Schoch (1989) considered the Isectolophidae to be paraphyletic. Schoch proposed that Isectolophus is more closely related to other ceratomorphs than to Homogalax, although he himself described his only synapomorphy for this grouping as variable. Hooker (1989) and Colbert & Schoch (1998) have presented synapomorphies uniting Homogalax and Isectolophus. Hooker (1989, fig. 6.6) indicated three unique dental synapomorphies for Homogalax and Isectolophus. Interestingly, Hooker s results indicated that Homogalax may be paraphyletic. Colbert & Schoch (1998) also included Cardiolophus in a monophyletic Isectolophidae, although they admitted that the evidence is weak. Indeed, isectolophids may turn out to include some of the most primitive perissodactyls, and they may not be more closely related to tapiromorphs than to non-tapiromorph perissodactyls. Because of the uncertainty surrounding isectolophid monophyly, all three genera are considered in this study, although little of the postcranial skeleton is known for any of these genera. The skulls of Homogalax and Isectolophus have never been described, while the skull of Cardiolophus was described by Gingerich (1991). Most of the postcrania of Homogalax were unknown, but Rose (1996) described a recently discovered skeleton of this genus, which provides our

3 TAPIROMORPH OSTEOLOGY 3 Table 1. List of principal putative tapiromorph genera studied. Traditional familial assignments, relevant specimens, agdocality data, and references are given for each genus. Abbreviations: E. =early; M. =middle; L. =late; Eo =Eocene; 01 = Oligocene; Mi = Miocene; Rec = Recent; Eur = Europe; NA =North America; SA = South America; num. = numerous. Institutional abbreviations are listed in Chapter 1 Family Genus Age and locality Specimens Fkferences Isectolophidae Homogalax E. Eo; NA, Asia UM 87027,92584; USGS 25032, 21843, Eomoropidae Chalicotheriidae Helaletidae Cardiolophus Zsectolophus Eomoropus Litolophus Moropus E. Eo; NA E.-M. Eo; NA M. Eo; NA M. Eo; Asia L. 01 -E. Mi; NA, Eur UM AMNH AMNH 5096 AMNH AMNH,CM num. Heptodon Helaletes Plesiocolopirus Colodon E. Eo; NA, Asia E.-M. Eo; NA, Asia L. Eo; NA M. Eo-E. Mi; NA, MCZ 17670; AMNH 294 USNM: several UOMNH 20377, 28313,27929 AMNH 658, Asia E.-L. Eo; Asia Rose (1999); Gingerich (1991) Gingerich (1991) Osborn (1913) Colbert (1934) Holland & Peterson (1914); Coombs (1978a) Radinsky (1965a) Schoch (1984) Schoch (1989) Scott (1941) Lophialetidae Lophialetes AMNH: num. Radinsky (1965b); Fkshetov (1979) Deperetellidae Tapiridae Schlosseria Deperetella Protapirus 'llzpirus E.-L. Eo; Asia M.-L. Eo; Asia L. EO -L. 01; NA, EU AMNH: num. AMNH: num. AMNH 662 AMNH, USNM: num. Radinsky (196515) Radinsky (1965b) Scott (1941) Radinsky (1965a) Hyrachyidae Amynodontidae Hyracodontidae Indricotheriidae Rhinocerotidae Zncertae sedis Hyrachyus Amynodon Rostriamynodon Metamynodon Sha ram ynodon Triplopus Hyracodon Forstercooperia E. Mi-Fkc; NA, SA, Eur, Asia E. Eo; NA, Asia, Eur M. Eo; NA L. Eo.; Asia L. Eo-E. 01; NA L. Eo; Asia M. Eo; NA, Asia L, Eo-L. 01; NA M.-L. Eo: Asia AMNH, USNM: num. AMNH 533 AMNH AMNH: several AMNH AMNH 5095 AMNH: num. AMNH Juxia M. Eo-E. 01; Asia AMNH 20287, 20288, Paracera theriu m AMNH: several Trigonias Su bhyracodon Uintaceras M. 01-E. Mi; Asia, Eur L. Eo; NA L. Eo; NA M.-L. Eo; NA best knowledge of isectolophid postcrania. Additional information comes from some partial skeletons in the University of Michigan collection, reported by Gingerich (1991). CHALICOTHERIOIDEA This group has been divided into two families: the Eomoropidae and the Chalicotheriidae (Coombs, 1989). AMNH, CM, DMNH: num. AMNH: num. UCMP 69722; CM Cope (1884) Wall & Manning (1986) Scott (1941) Cope (1884) Scott (1941) Wood (1938); Lucas et al. (1981); Lucas & Sobus (1989) Granger & Gregory (1936) Scott (1941) Scott (1941) Holbrook & Lucas (1997) Eomoropidae, including Eomompus, Litolophus, and Grangeria (see Lucas & Schoch, 1989 for review), appears to be a paraphyletic group of primitive chalicotheres (Coombs, 1989). Eomompus and Lito- Zophus, which were described by Osborn (1913) and Colbert (1934), respectively, are included in this study. Material of Grangeria was unavailable for this study. Some chalicotheriids, including the schizotheriines

4 4 L. T. HOLBROOK Mompus and Borissiakia, and the chalicotheriines Chalicotherium and Macmtherium (sensu de Bonis et al., 1995), are also discussed. HELALETIDAE This is another tapiroid family of dubious monophyly. Radinsky (196%) considered the Helaletidae to include the ancestors of the Tapiridae and Rhinocerotoidea. Radinsky (1967b) included Hyrachyus, considered by many workers to be the most primitive rhinocerotoid, in this family, but for this study Hyrachyus will be considered as its own family, the Hyrachyidae. Regardless of the position of Hyrachyus, no worker has proposed a monophyletic Helaletidae. Besides Hyrachyus, Heptodon is the helaletid genus most often placed outside of any monophyletic grouping of tapiroids. Schoch (1989) placed Heptodon in a polytomy with rhinocerotoids and a monophyletic group of tapiroids; Hooker (1989) placed Heptodon as the sistertaxon of Tapiroidea and Rhinocerotoidea, although Dashzeveg & Hooker (1997) have found support for including Heptodon in a monophyletic Tapiroidea. Given the uncertainty of the monophyly of the Helaletidae, all helaletid genera represented by skulls and/or postcrania, namely Heptodon, Helaletes, Colodon, and Plesiocolopirus, are included in this analysis. Scott (1941) described Colodon, and Radinsky (1965a) described an excellent skeleton of Heptodon. Excellent undescribed material of Helaletes is present in the collections of the USNM, and material of Pleswcolopirus is present at UOMNH and UCMP. Schoch placed Plesiocolopirus in the Helaletidae, whereas Colbert & Schoch (1998) placed it in the Tapiridae. Hanson (1996) included Plesiocolopirus in the tapirid genus Pmtupirus, but cranial differences, as discussed later, appear to warrant generic separation. LOPHIALETIDAE AND DEPERETELLIDAE Tapiroids from the early Tertiary of Asia, most recently reviewed by Schoch (1989), Reshetov (1979), Dasheveg & Hooker (1997), and Radinsky (1965b), include the families Lophialetidae and Deperetellidae. The lophialetids Lophialetes and Schlosseria and the deperetellid Deperetella are included in this study. TAPIRIDAE This is an apparently monophyletic group that includes the extant genus Tupirus and several extinct genera. The earliest recognized member of this family is Pmtupirus, known from the Oligocene of Europe and North America. Pmtapirus is represented by incomplete postcranial material but is still the best known extinct tapirid genus in terms of postcrania. Although Tapirus does not appear before the Miocene, it is used here, along with Pmtapirus, to provide a more complete picture of tapirid postcranial morphology. HYRACHYIDAJ?. Wood (1934) included four genera in this family, but Radinsky (1967b) reduced the family to a single genus (Hyrachyus) in a subfamily of the Helaletidae. Hyrachyus has historically been placed near the ancestry of rhinocerotoids, a position interpreted more recently as within Rhinocerotoidea and as the sister-taxon to other rhinocerotoids (Prothero, Manning & Hanson, 1986; Schoch, 1982, 1984). While the evidence for the rhinocerotoid affinities of Hyrachyus is debatable (Radinsky, 1983; Emry, 1989), it is generallyrecognized that Hyrachyus should not be included in the Helaletidae and is best considered as forming a monotypic family. The skeleton of Hyrachyus is well-known from numerous specimens, mostly collected from the middle Eocene (Bridgerian) of the Bridger Basin of Wyoming. AMYNODONTIDAJ?. Wall (1981, 1982, 1989; also Wall & Manning, 1986) most recently revised this family. Generally considered to be a monophyletic group of rhinocerotoids, amynodonts are one of the earliest perissodactyl groups to diverge from the primitive subcursorial morphotype. Amynodonts were generally large in body size and some lineages apparently became semi-aquatic. Amynodonts were Holarctic in distribution and are known from the middle Eocene to the early Miocene. The genera Rostriamynodon, Amynodon, Sharamynodon, Amynodontopsis, and Metamynodon are all considered in this study. HYRACODONTIDAE This is the least well-defined of the three major rhinocerotoid families. It is possibly not monophyletic, especially when it is considered sensu Radinsky (1966a, 1967a), who first included the giant rhinoceroses of the Eocene and Oligocene of Asia in a subfamily of Hyracodontidae, the Indricotheriinae. Radinsky, however, appeared deliberately to treat the Hyracodontidae as a waste-basket taxon, since his criterion for inclusion in this family was the absence of the derived characters of amynodontids and rhinocerotids. Other workers (Lucas, Schoch & Manning, 1981; Lucas & Sobus, 1989; F rothero et al., 1986) have attempted to diagnose a monophyletic Hyracodontidae matching Radinsky s membership for this family. Lucas & Sobus (1989) most recently reviewed the evidence for hyracodontid monophyly and proposed two characters as valid synapomorphies for the family: elongate metapodials, and a tridactyl manus. These characters were thought to support the notion that the

5 lack of graviportal adaptations in indricotheres (which attained sizes greater than those of modern elephants [Fortelius & Kappelman, 19931) could be explained by the retention of a morphology derived from a cursorial ancestor. The smaller hyracodontines, such as Hyracodon, possess a cursorial morphology with elongated distal limb elements. Cursorial adaptations, however, are also seen in many other primitive perissodactyls and may be primitive traits for tapiromorphs. Because the evidence for an indricothere/hyracodont clade is limited, indricotheres will be considered as a separate family (Indricotheriidae). The genera Hyracodon and Triplopus are used to represent the Hyracodontidae, but this usage is not meant to suggest that non-indricothere hyracodonts definitely form a monophyletic group; rather, it is because other reasonably well-known hyracodontines are from Europe and Asia, and are found in collections not available for this study. INDRICOTHERIIDAE The monophyly of the indricotheres is based on the presence of a preorbital fossa in the maxillary (Lucas & Sobus, 1989). Wall (1981, 1989) also used this character as a synapomorphy of the Amynodontidae, so some doubt is cast on the validity of this character as an indricothere synapomorphy. Lucas & Sobus (1989) recognized four valid genera of indricotheres, all of which are restricted to the middle Eocene to early Miocene of Eurasia (see also Holbrook & Lucas, 1997). Skulls of Forstercooperia, Juxia, and Paraceratherium are known. Of the Eocene genera Forstemooperia and Juxia, only Juxia is represented by postcranial material, but much of the skeleton is represented for this genus. Of the later genera Urtinotherium and Paraceratherium, Urtinotherium is known from only a few postcranial elements, whereas a significant portion of the skeleton of Paraceratherium is known (Cooper, 1923; Granger & Gregory, 1936). Paraceratherium was one of the largest land mammals known (Fortelius & Kappelman, 1993). Most of the information on indricotheres used here is derived from study of specimens of Forstercooperia, Juxia and Paraceratherium. RHINOCEROTIDAE The family Rhinocerotidae includes the extant genera Ceratotherium, Dicems, Dicemrhinus, and Rhinocems, as well as a large number of fossil taxa. Radinsky (1966a) provided the diagnosis which established the monophyly of this family, namely the derived chisel/ tusk relationship and morphology of I1 and i2. The earliest rhinocerotids are known from the late Eocene (Duchesnean) of North America, as represented by the primitive genus Teletaceras. Teletaceras is known from TAPIROMORPH OSTEOLOGY 5 excellent skulls, but only a relatively small amount of postcranial material. Other primitive rhinocerotid genera, such as Trigonias and Subhyracodon, are known from complete skeletons. RHINOCEROTOIDEA ZNCERTAE SEDIS Finally, the genus Uintaceras, described by Holbrook & Lucas (1997), may be the sister-taxon to the Rhinocerotidae, but it is otherwise classified as incertae sedis. The type specimen, CM12004 from the Uintan of Wyoming, includes most of the postcranial skeleton, and the skull of this genus is also known, mainly from UCMP COMPARATIVE TAXONOMIC SET In order to make inferences about the evolution of skeletal features in tapiromorphs, particularly the polarity of characters, comparisons were made with a variety of non-tapiromorph eutherian mammals. Table 2 lists the non-tapiromorphs examined for this study, which are discussed below. NON-TAPIROMOFPH PERISSODACTYLS Non-tapiromorph perissodactyls are the outgroups that provide the closest comparisons for this study. These include taxa traditionally assigned to the families Equidae, Palaeotheriidae, and Brontotheriidae. However, the precise membership and interrelationships of these families are not well understood (see Hooker, 1989,1994; Mader, 1989; Froehlich, 1999), so discussion of the relevant genera is necessary here. Equidae and Palaeotheriidae Equids and palaeotheriids are often grouped together in the superfamily Equoidea. Since there is little agreement as to which genera belong in which families, I discuss both families together. The skull of the genus Palaeotherium, from the Eocene and Oligocene of Europe, has recently been described in detail by Remy (1992). Hyracotherium, of the early Eocene of North America and Europe, is generally considered to be the most primitive equid, and its skull has been described by several workers, including Cope (1884), Simpson (1952), and Kitts (1956) (see also MacFadden, 1976). Postcrania of Hyracotherium have been described (Cope, 1884; Kitts, 1956) and are reasonably well-known. Hooker (1989, 1994) has disputed the monophyly of Hyracotherium, and it may be the case that some species in this genus are closer to the Tapiromorpha than others. Hooker has revised many of the European species of Hyracotherium, and Froehlich (1999) has studied American

6 6 L. T. HOLBROOK Table 2. Comparative taxonomic set. Traditional familial assignments, principal specimens, agdocality data, and references are given. Abbreviations: E. = early; Pa = Paleocene; Eo =Eocene; 01 =Oligocene; P1= Pliocene; Rec = Recent; Eur = Europe; NA = North America; SA = South America Order (0.) and Genus Age and locality Specimen@) References Familv (F.) 0. Condylarthra F. Phenacodontidae Phenacodus Ectocion Pa-E. Eo; NA, Eur Pa-E. Eo; NA F. Meniscotheriidae Meniscotheriurn Pa-E. Eo; NA 0. Hyracoidea F. Procaviidae F. Pliohyracidae 0. Proboscidea F. Moeritheriidae F. Numidotheriidae 0. Sirenia F. Prorastomidae 0. Embrithopoda F. Arsinoitheriidae 0. Desmostylia F. Desmostylidae 0. Perissodactyla F. Equidae Pmavia Megalohyrax Moeritheriurn Nurnidotheriurn P1-Rec; Africa M. Eo-M. Mi; Africa Eo-01; Africa Eo; Africa YPM(PU) UM Thewissen (1990) Gazin (1956); Thewissen (1990) Williamson & Lucas (1992) UM 4174,4176 Fischer (1986) Matsumoto (1926) Tassy (1981) Court (1995) Pmrastornus Eo; Jamaica Savage et al. (1994) Arsinoitherirn 01; Africa Court (1992) Desrnosty lus Mi; NA USNM 8191 Vanderhoof (1937) Hyracotheriurn E. Eo; NA, Eur, Asia AMNH 4831, ,55269 YPM(PU) Mesohippus Pachynolophus L. Eo-Ol; NA Eo; Eur Cope (1884); Kitts (1956) Scott (1 94 1) Savage et al. (1965); Remy (1972) Equus Pl-Fk; NA, SA, Ew, Asia, Africa UMA 2182 F. Palaeotheriidae F. Incertae sedis Palaeotheriurn Hallensia Eo-E. 01; EU Eo; Eur Remy (1992) Franzen (1990) F. Brontotheriidae Eotitanops E. Eo; NA UCMP Osborn (1929) Palaeosyops E.-Eo; NA AMNH 1544 Osborn (1929) Lam bdotheriurn E. Eo; NA CM Osborn (1929) 0. Incertae sedis F. Phenacolophidae Radinskya Pa; Asia McKenna et al. (1989) Hyracotherium and argued for its paraphyly. Regardless of its status, Hyracotherium is generally not considered to be a tapiromorph. Two other equoid taxa whose affinities are not as clear are discussed in this study. The skull of Pachynolophus has been described by Savage, Russell & Louis (1965) and Remy (1972), but the former assigned it to the Equidae, whereas the latter considered it to be a palaeotheriid. Hooker (1994) has recently proposed that Pachynolophus is the sister-group of the Tapiromorpha and, therefore, not an equoid. The skeleton of Hallensia, from the Eocene of Messel (Germany), was described by Franzen (1990), who considered it to be the sister-taxon of the Equidae. Hooker (1994) considered Hallensia to be the sister-taxon of the Pachynolophus-tapiromorph clade. Brontotheriidae Mader (1989, 1991) has most recently reviewed the early members of the Brontotheriidae. Osborn (1929) considered Lambdotherium and Eotitanops to be the most primitive members of this family, but Mader excluded the enigmatic genus Lam bdotherium, judging that, of the two genera, only Eotitanops possesses the diagnostic synapomorphies that ally it to other brontotheres. The skull of neither genus was particularly well known, but a specimen in the University of California Museum of Paleontology (UCMP ) includes the most complete skull of Eotitanops yet discovered. Both Lambdotherium and Eotitanops are included in this analysis, keeping in mind that the two forms may or may not be closely related. Neither Lambdotherium nor Eotitanops is particularly well-

7 known from postcrania, so observations on other brontotheriids, such as Palaeosyops and Bmntops, are also used here. Hooker (1989) included both Lambdotherium and Eotitanops in the Titanotheriomorpha, which he suggested might be the sister-taxon of all other perissodactyls. It should be noted here that perissodactyl interrelationships are still not well understood. The phylogenies of Hooker (1989,1994) are noteworthy for their broad taxonomic scope and cladistic rigour, but many of the clades recognized by Hooker are based on equivocal synapomorphies. Indeed, one aim of this paper is to test the monophyly of Hooker s Tapiromorpha (Moropomorpha of Schoch [ 19891). Froehlich (1999) has recently reanalysed the interrelationships of basal perissodactyls, but his analysis focused more on relationships among non-tapiromorphs and included relatively few non-dental characters. Historically, arguments have been made for almost every possible arrangement of the major perissodactyl lineages. NON-PERISSODACTYLS The sister group of Perissodactyla is somewhat controversial. Radinsky (1969) considered phenacodontids, a group of archaic ungulates, to include the ancestors of perissodactyls. Fischer (1986, 1989; see also Prothero et al., 1988; Fischer & Tassy, 1993) argued, on morphological grounds, that hyracoids are the sister-taxon to perissodactyls, an idea also proposed by McKenna (1975; but see Novacek et al., 1988) and McKenna & Manning (1977). In fact, Fischer has included Hyracoidea in the Perissodactyla and resurrected the term Mesaxonia (Marsh, 1884) for the traditional Perissodactyla, reflecting Owen s (1848) original concept of Perissodactyla. Some workers (Novacek, 1986, 1992a,b; Novacek & Wyss, 1986; Novacek et al., 1988; Tassy & Shoshani, 1988; Shoshani, 1993) have argued that hyracoids are the sister-group to tethytheres (Proboscidea, Sirenia, Desmostylia, and Embrithopoda). Recent morphological studies on ungulate phylogeny (Thewissen & Domning, 1992; Fischer & Tassy, 1993; Shoshani, 1993) suggest that the relationships of the Recent ungulate orders are still unclear or controversial, but it appears that Phenacodonta (an extinct taxon including phenacodontids and meniscotheriids) is the sister-group to Pantomesaxonia (perissodactyls, hyracoids, proboscideans, sirenians, and desmostylians). The following non-perissodactyls (listed with their traditional familial or ordinal assignments) have been considered: Phenacodus, Ectocion (Phenacodontidae); Meniscotherium (Meniscotheriidae or Phenacodontidae); Moeritherium, Numidotherium, and other proboscideans; Pmrastomus and other sirenians; the TAF IROMORPH OSTEOLOGY 7 Figure 1. Skull of Heptodon (after Radinsky, 1965a). Abbreviations: de: dentary; fr: frontal; ju: jugal; la: lacrimal; mx: maxilla; na: nasal; pa: parietal; pmx: premaxilla; sq: squamosal. order Desmostylia; the order Embrithopoda (represented by Arsinoitherium); and fossil and Recent Hyracoidea. Thewissen (1990) most recently reviewed the Phenacodontidae, from which he excluded Meniscotherium, a taxon historically placed in its own family. Williamson & Lucas (1992) most recently reviewed Meniscotherium and placed this genus in the Phenacodontidae. Unless otherwise stated, the term phenacodont as used in this paper refers to Meniscotheriurn as well as members of Thewissen s Phenacodontidae. One other interesting non-perissodactyl outgroup is Radinskya yupingae (?Family Phenacolophidae) from the late Paleocene of China. McKenna et al. (1989) described this taxon as being closely related to perissodactyls. Fischer & Tassy (1993) included Radinskya in their phylogenetic analysis of various ungulates, and concluded that this genus falls outside of the clade including hyracoids, perissodactyls, and tethytheres, but is otherwise closely related to these taxa. Additionally, comparisons are also made between tapiromorphs and selected non-ungulate mammals, such as the marsupial Didelphis and the fossil insectivore Leptictis. CRANIAL OSTEOLOGY Description and discussion of skull morphology in tapiromorphs and the comparative taxonomic set is given below, bone by bone. The primitive proportions and general arrangement of the skull are illustrated by the skull of Heptodon (Fig. 1). Nasal Description. In tapiromorphs with unmodified rostra, the posterior part of the nasals is expanded laterally

8

9 The morphology of the nasals seen in early tapiromorphs like Heptodon is primitive for perissodactyls; it is seen in Hyracotherium, some species of Palaeotherium, and most brontotheres (and possibly Pachynolophus, although the nasals are not preserved in the skull described by Savage et al. [1965]). This morphology is uncommon in other eutherians. Recent hyraxes are similar, except that contact with the lacrimals is absent or variably present. Fossil hyraxes (Matsumoto, 1926), Phenacodus, and Meniscotherium lack posterolateral expansion and lacrimal contact. Instead, the nasals in these taxa are more uniform in width, and the frontal suture wraps around laterally, so that the frontal extends between the nasal and lacrimal to contact the maxilla. Osborn (1929) figured a similar condition occurring in at least one brontothere ( Telmatherium ; Metatelmatherium of Mader [1991]), but Mader (1991) claimed that Osborn s reconstruction was based on a specimen with obscured sutures. Mader reconstructed the skull of Metatelmatherium with a nasolacrimal contact. Other brontotheres, e.g. Manteoceras and Pmtitanotherium, show remarkable lateral extensions of the nasals (also seen in some species of Palaeotherium [Remy, 19921) that maintain lacrimal contact even when the nasals have been modified for horns (Osborn, 1929). These nasal extensions further attest to the primitive nature of the nasolacrimal contact within perissodactyls. Phenacodus possesses lateral apophyses on the free portion of the nasals, but this trait is absent in Meniscotherium, Ectocion, equids, palaeotheres, and brontotheres, indicating that it is probably not primitive for perissodactyls. Premuxilla Description. The premaxilla is one of the least wellpreserved bones of the tapiromorph skull. In Homogalax, Cardiolophus, eomoropids, Lophiodon, Heptodon, Hyrachyus, and a number of primitive rhinocerotoids, the premaxilla is restricted to the anterior of the rostrum, anteroposteriorly short, but with a relatively tall ascending process that contacts the nasals (and thus forms part of the border of the narial incision) (Fig. 3A). Radinsky (1963a), in a figure summarizing the evolution of the tapiroid proboscis, illustrated a skull of Isectolophus showing the primitive condition of the premaxilla. He made no allusion, however, to any specimen showing this morphology; AMNH and 12222, the only skulls oflsectolophus known, do not adequately preserve this portion. Living tapirs and rhinos, the only extant tapiromorphs, have lost the contact between premaxillae and nasals (Fig. 3B,C), but this is certainly an independently evolved trait. The premaxillae of Tbpirus are stout and fused medially. According to Wall (1980), A B C TAPIROMORPH OSTEOLOGY 9 \-- Figure 3. Skulls of Heptodon (A) and Tapirus (B), and anterior skull of Trgonias (C) in lateral view with nasals and premaxillae shaded. (A and B modified from Radinsky, 1965a; C redrawn from Scott, 1941.) this morphology is typical of other taxa that are thought to have possessed a proboscis (e.g. the amynodontid Caduntodon). The retraction of the narial incision leaves the premaxilla unsupported by other bones; in order to support the incisors, which are unreduced in size or number from the primitive condition, the premaxilla must become robust. The loss of premaxilla-nasal contact in Tapirus is related to the reduction of the nasals combined with the retraction of the narial incision. A similar condition occurs in

10 10 L. T. HOLBROOK Helaletes, Cokmbn, Pmtapirus, PlesiocOlopirus, and Lophiaktes, as well as cadurcodontine amynodonts. The premaxillae of all unequivocal rhinocerotids have reduced or lost the ascending process, so that the remaining horizontal ramus juts out from the anterior of the skull bearing the chisel-like I1 and any other incisors. More often than not, the premaxilla has broken off of rhinocerotid skulls and is not preserved in fossils. Even Tektacerus, which does not have a particularly deep narial incision, has a reduced ascending ramus so that the maxilla forms most of the lateral border of the narial incision. There is a general trend in rhinocerotids to lose the posterior incisors, and the premaxillae of the living genera Cemtotherium and Dicems are small, edentulous, and often loosely attached to the maxillae. Some chalicotheres (e.g. the chalicotheriine Macmtherium brevimstris; see Colbert ) also lack a premaxilla-nasal contact, but this again is probably an independent acquisition. M. brevimstris, for instance, has a short facial region, a derived feature within chalicotheriines that almost certainly is responsible for the loss of premaxillqlnasal contact. Although it is difficult to identify sutures on the crushed skulls of specimens of Litolophus, the narial incision is not retracted, indicating that the premaxilla was probably not reduced. The premaxilla of Eomompus is not known. Coombs (1978b) described an isolated premaxilla from Mompus. Although the exact relationship between the premaxilla and nasal cannot be ascertained, the ascending process of the premaxilla is relatively unreduced. There is a general trend within chalicotheres to reduce the anterior upper dentition, to the extent that the premaxilla of Mompus is edentulous. The bone was probably still important in feeding, functioning in a manner similar to that of the premaxillae of cervids and bovids (Coombs, 1978b), and therefore has remained relatively unreduced. Hyrachyus and a number of primitive rhinocerotoids show the primitive condition seen in Homogalax. This condition is present in Uintacerus and Hyramcbn, but the extent of the premaxilla is not known for Triplopus. Among indricotheres, Forstemperia andjuxia retain the primitive condition. The nasal incision of Paruceratherium is greatly retracted, although the nasals are unreduced, and the premaxilla does not contact the nasals. The morphology of the premaxilla is unusual, however, in that the maxillary suture runs obliquely, so that it might be said that the ascending process is not reduced but horizontally displaced. The premaxilla still forms a large part of the border of the narial incision, only now it has been displaced to the ventral edge. The morphology seen in Paraceratherium is similar to that of tapiroids like Tapirus and Helaletes. Rostriamynodon and Amynodon retain the primitive condition of the premda, but some other amynodontids reduce or lose the nasal contact (Wall, 1989). The incisive foramina are a pair of foramina or a single foramen that pierce the palatal portion of the premaxilla and are bordered posteriorly by the maxilla. In Cardwlophus, Mompus, Lophiodon, Heptodon, Lophiuletes, Helaletes, Hyrachyus, and Hyracodon, paired foramina are present. In Pksiocolopirus, Pmtapirus, Zzpirus, Paraceratherium, and rhinocerotids, there is a single, large, median incisive foramen. Discussion. The primitive condition is seen in hyracoids and phenacodontids, where there is a prominent ascending process that contacts the nasals and forms the lateral border of the narial incision. In hyracoids, such as Pmviu, the premaxilla is posteriorly expanded, whereas in phenacodontids (and Meniscotherium), the premaxilla is slender rostrocaudally. The latter condition is seen in primitive perissodactyls such as Hyracotherium and early brontotheres. In summary; the loss of premaxilla-nasal contact seems to have been acquired independently in a number of tapiromorph lineages. Different morphological changes have led to this condition in different lineages, so it is best to consider this character in a more detailed context for phylogenetic purposes, i.e. the differences in the shape of the premaxilla are more informative than the simple absence of contact with the nasal. Maxilla Description. The maxilla dominates the lateral aspect of the rostrum in tapiromorphs, contacting the premaxilla, nasal, jugal, and lacrimal on the face. Posteriorly, the maxilla extends into the floor of the orbit as the maxillary tubercle. In the orbit, the maxilla contacts the jugal, lacrimal, palatine, and alisphenoid in all tapiromorphs that preserve this region. Contact with the frontal in the orbit appears to be precluded by a palatine-lacrimal contact. In some specimens, however, it is difficult to tell whether a suture is really a crack in the bone, and thus, it is impossible to tell whether a small piece of bone at the juncture of the four bones in question is separated from any one of them by a suture or by a crack. It is also possible that this small piece of bone actually represents an exposure of the ethmoid. In ventral view, the maxilla contacts the premaxilla anteriorly, forming the posterior border of the incisive foramina or foramen, and contacts the palatine posteriorly, from the level of Ml-M2 posteriorly. This latter contact is such that the palatine is laterally flanked by the maxilla for the entire length of the palatine s exposure on the palate. An elaboration of the facial exposure of the maxilla is the presence of a maxillary or preorbital fossa in

11 B C Figure 4. Preorbital fossae. Skulls of Heptodon (A), Zbpirus (B), and Shurumynodon (C) in lateral view. (A and B modified from Radinsky, 1965a; C modified from Osborn, 1936.) Abbreviations: mxe maxillary (preorbital) fossa. some tapiromorphs. This fossa is present to at least a slight degree in all tapiromorph taxa (Fig. 4A), but it is most pronounced in some tapiroids, amynodontids, and some indricotheres. Wall (1981, 1989) has used the presence of a prominent preorbital fossa as a synapomorphy for Amynodontidae (Fig. 4C). Likewise, Lucas & Sobus (1989) claimed that presence of a maxillary fossa united Forstemooperia, Juxia, and Paraceratherium in the subfamily Indricotheriinae of Hyracodontidae. (The skull of Urtinotherium is unknown.) My own observations indicate that a fairly TAPIROMORPH OSTEOLOGY 11 prominent fossa is present in Forstercooperia and Paraceratherium, but injuxia it appears to be shallower and more similar to that of other tapiromorphs. The preorbital fossae of Forstercooperia and amynodontids are somewhat similar, in that they are depressions in the maxillary wall roofed over by the nasals, but the amynodont fossa is in all cases deeper and forms a distinct pocket anterior (or in some cases medial) to the orbit. The preorbital fossa of Puruceratherium is more difficult to compare, because the narial incision is greatly expanded posteriorly and the nasals, being mostly free, cannot be said to form a roof over the fossa. A preorbital fossa is also found in certain tapiroids, although it is somewhat different from that of rhinocerotoids and appears to have a separate origin. In lbpirus, the fossa is a vertical groove just posterior to the narial incision and extending dorsally to the posterior aspect of the nasals (Fig. 4B). The posterolateral corners of the nasals of lbpirus are usually notched or scrolled due to the association with this fossa. Gregory (1920b) and Radinsky (1963a, 1965a) stated that the preorbital fossa of Tapirus accommodates the nasal diverticulum, which is expanded and displaced from the nasal passage. Witmer, Sampson & Solounias (1999) have provided strong anatomical evidence this structure (which they term the meatal diverticulum) is not homologous with the nasal diverticulum of extant horses. Radinsky (1963a, 1965a) hypothesized that the development of the fossa was related to the retraction of the narial incision. This view is consistent with the distribution of this trait; a vertical preorbital fossa is found in Tapirus, Protapirus, Colodon, Plesiocolopirus, and Helaletes. Radinsky (1963a, 1965a) considered the long nasals of Helaletes to be evidence of absence of a mobile proboscis, and thus he considered the preorbital fossa to be not functionally related to proboscis development. The infraorbital foramen, which transmits the infraorbital artery and nerve, opens above the vicinity of F 2-P3 in Homogalax, Cardiolophus, Litolophus, Heptodon, Hyrachyus, Hyracodon, Forstercooperia, Juxia, rhinocerotids, and Uintaceras. In a number of other taxa, the foramen lies more posteriorly, a position possibly correlated with certain aspects of snout morphology. In Helaletes, Colodon, Protapirus, Tapirus, Plesiocolopirus, Lophialetes, and Paraceratherium (all taxa with greatly retracted narial incisions), the infraorbital foramen lies between P3 and P4. Moropus, which does not show any remarkable rostra1 modifications, has the foramen lying above the border between P4 and M1. The most striking displacement of the infraorbital foramen is seen in amynodontids, where the foramen may lie as far back as M2 and often faces anteriorly from within the preorbital fossa. As mentioned above, a number of tapiromorphs have

12 12 L. T. HOLBROOK greatly retracted the narial incision, including a number of tapiroids and Paracemtherium. This retraction typically involves reduction of much of the dorsal portion of the lateral maxillary wall. In a few taxa, namely Helaletes, Colodon, and Plesiocolopirus, the posterior portion of the narial incision is deeply rounded, especially ventrally, giving a stepped appearance to the narial incision. Discussion. The extensive facial exposure of the maxilla is a primitive trait, as seen in phenacodonta and primitive perissodactyls (including Hallensia, Pachynolophus, Hyracotherium, and Eotitanops). The large premaxilla of hyracoids comes at the expense of the maxilla, but the latter element is still anteroposteriorly longer, especially in long-snouted forms like Megalohymx (Gregory, 1920a). In tethytheres, the premaxilla dominates the rostrum. The maxilla of most perissodactyls is expanded posteriorly into the maxillary tubercle. In many juvenile mammals, the maxillary tubercle is the site of the formation of the molars, but it is retained in most adult perissodactyls, hyraxes, and tethytheres even after all the molars have erupted. Fischer (1986,1989) considered the retention of the maxillary tubercle in adults to be a synapomorphy uniting perissodactyls and hyracoids. He argued that the morphology seen in Recent Sirenia and Proboscidea was independently derived in each of these taxa and that this condition was a product of their horizontal tooth succession. This idea is supported by the absence of a maxillary tubercle in early proboscideans like Moeritherium, but the presence of this structure in the early sirenian Prorastomus (Savage, Domning & Thewissen, 1994), which does not appear to have undergone horizontal tooth replacement, weakens Fischer s hypothesis. The various preorbital fossae of tapiromorphs have an interesting phylogenetic distribution. The vertical groove for accommodating the expanded nasal diverticulum, seen in a number of tapiroids, may be a synapomorphy for those taxa. The fossae of indricotheres and amynodontids were certainly a separate acquisition relative to the tapiroid fossa, but it could be argued that the preorbital fossa is a synapomorphy uniting these two rhinocerotoid groups. Indeed, Wall & Manning (1986) noted the gross similarity of the skulls of Rostriamynodon to Forstemperia. There are, however, some important differences in the morphology of the fossa between these two groups (as well as between Forstemperia and Paracemtherium) that obscure the homology or non-homology of these structures, namely differences in the extent and shape of the fossa and differences in the position of the infraorbital foramen. The differences in the extent of the fossa have been mentioned already, but the differences in the relative position of the infraorbital foramen are also interesting. This foramen transmits the infraorbital nerve and artery, branches of the maxillary nerve (cranial nerve V2) and the maxillary artery, respectively. The infraorbital nerve is a sensory nerve, and does not contribute to the motor innervation of the facial muscles. Thus, repositioning of the infraorbital foramen is not indicative of the presence of elaborated snout musculature in the preorbital fossa. Gregory (1920b) noted that the shallow preorbital fossae of living rhinoceroses may bear expansions of the nasal diverticula, and it is possible that this was the function of the preorbital fossa in other tapiromorphs. Expansion of the narial incision is a prerequisite for developing a mobile proboscis (Wall, 1980), although Radinsky (1963a) argued that the long nasals of Helaletes precluded the musculature for such a proboscis. The posteriorly rounded incision seen in Helaletes, Colodon, and Plesiocolopirus may be a synapomorphy for a monophyletic Helaletidae. The fact that Protapirus does not possess this feature warrants separation of Plesiocolopirus from Protapirus, contra Hanson (1996). Palatine Description. In ventral view, this bone forms the posterior portion of the hard palate; posterior extensions of the maxilla wrap around the palatines laterally. On either side, the anterior palatine foramina are found on the suture of the palatine and maxilla. The posterior end of the palatine bones is emarginated by the internal nares, forming a broad notch that extends anteriorly to the anterior edge of the second molar. In the orbit, the palatine forms the anteroventral portion of the orbital wall. In the extant tapirid Tupirus and the extant rhinocerotid Dicemrhinus, the palatine contacts the maxilla, frontal, orbitosphenoid, alisphenoid, and lacrimal. In most fossils, the region of the anterior orbit is not preserved well enough to determine whether or not a lacrimal contact is present. In the extant rhinocerotid Diceros, contact between the frontal and maxilla in the orbit separates the palatine and lacrimal. Two foramina are associated with the palatine in the orbit. The sphenopalatine foramen usually takes the form of a large oval situated near the palatine s dorsal contact with the frontal. The posterior palatine foramen is usually posteroventral to the sphenopalatine foramen. Discussion. The notch for the internal nares at the posterior end of the palate is deeper in perissodactyls and extends more anteriorly than that of phenacodonts and Pmavia. The contact between palatine and lacrimal in the

13 orbit is likely a primitive feature for tapiromorphs, as this contact is seen in Hyracotherium (Simpson, 1952), Pachynolophus (Savage et al., 1965), and phenacodonts (Thewissen, 1990; Williamson & Lucas, 1992). Contact between the frontal and maxilla, however, is seen in Equus and Pmcavia. In Pmcauia, the palatine and lacrimal are well separated, but, in Equus, they are separated by only a narrow strip of bone. It is possible that the condition seen in Equus exists in some fossil perissodactyls. Simpson (1952), in his discussion of the orbit of Hyracotherium, described a large foramen in the anterior part of the palatine that he identified as the maxillary foramen. As Simpson noted, the maxillary foramen is normally found in the maxilla. I suggest that the foramen Simpson described as the maxillary foramen is actually the sphenopalatine foramen; the sphenopalatine foramen is always found in the palatine bone in perissodactyls. If this interpretation is correct, then the sphenopalatine foramen of Hyracotherium is situated more anteriorly than that of tapiromorphs. An anteriorly positioned sphenopalatine foramen is also seen in Equus and may be an equid synapomorphy. The tapiromorph condition is also seen in Pachynolophus. In phenacodonts, the sphenopalatine and posterior palatine foramina share a common recess, but the situation is more similar to tapiromorphs than to equids in the sense that this recess is not anteriorly placed. Pterygoid Description. The extent of the pterygoid is not evident in most fossil tapiromorphs, but a distinct hamulus is present and extends a short distance ventrally. The tip of the hamulus is usually expanded or even doubled. Discussion. A doubled pterygoid hamulus is seen in some non-perissodactyl outgroups, particularly extant hyracoids and Radinskya (McKenna et al., 1989). This morphology is found in a variety of eutherians and is probably primitive for perissodactyls. Lacri ma 1 Description. Nasolacrimal contact (Fig. 5A) is present in isectolophids, Mompus, primitive amynodontids, hyracodontids, indricotheres, and most rhinocerotids. The skulls of primitive chalicotheres are not sufficiently well-known to determine the relation of the nasal to the lacrimal, but, based on its taxonomic distribution, a nasolacrimal contact appears to be primitive for the Chalicotherioidea. Among rhinocerotids, only the Dicerotini, including the Recent two-horned genera Ceratotherium and Dicems, lose the contact, TAF IROMORPH OSTEOLOGY 13 Figure 5. Nasolacrimal contact in perissodactyls. Skulls of Hyracotherium (A) and Heptodon (B) in lateral view with nasals and lacrimals shaded. (A modified from Simpson, 1952; B modified from Radinsky, 1965a.) possibly as a consequence of the nasal horn (Gregory, 1920a). (Not all tandem-horned taxa, however, lose this contact.) The only extinct tapiromorphs that do not possess nasolacrimal contact (Fig. 5B) are certain tapiroids and metamynodontine amynodontids. Wall (1980, 1981, 1989) has described how the primitive condition is retained in most amynodontids, including cadurcodontines, which maintain the contact despite the tapir-like modifications for a proboscis. Metamynodontines, whose rostra are otherwise rather primitive, do not possess nasolacrimal contact. It is interesting to note that the absence of nasolacrimal contact is not related to modifications of the skull that accommodate a proboscis. Wall (1981) mentioned that this character could be useful in tapiroid phylogeny. In Heptodon, Helaletes, Colodon, Plesiocolopirus, Protapirus, and lbpirus, the lacrimal is a small bone restricted to the anterior rim of the orbit. The narial incision is retracted in most of these taxa, but not so in Heptodon, demonstrating that possession of small lacrimals (and absence of nasolacrimal contact) is a character independent of the development of a proboscis.

14 14 L. T. HOLBROOK Discussion. Novacek (1986) considered a narrow (or lack of) maxillofrontal facial contact (character 6a in his table 3) to be primitive for eutherians, whereas he considered a broad maxillofrontal contact on the face to be derived. He therefore considered presence of nasolacrimal contact to be primitive, and perissodactyls were one of the groups that Novacek considered to possess the primitive condition. The polarity of this character appears to have been based on the presence of a nasolacrimal contact in didelphid marsupials. This contact is often not present in Didelphis (pers. observations of skulls of Didelphis in the University of Massachusetts collection). Furthermore, nasolacrimal contact is consistently absent from Australian marsupials, including dasyuroids, peramelids, macropodids, phalangerids, and vombatids. It is likewise absent in most eutherians, including lipotyphlans, carnivorans, macroscelideans, and rodents. Of xenarthrans, Bradypus may possess a nasolacrimal contact, but Dasypus and Tamandua clearly lack it. Whereas many of these groups do have a narrow maxillofrontal contact on the face (usually due to their narrow skulls), the nasolacrimal contact of perissodactyls is clearly a different and derived condition. Cretaceous eutherians, such as Asioryctes and Kennalestes, reportedly possess a nasolacrimal contact (e.g. Kielan-Jaworowska, 1969). Kielan-Jaworowska (1981) has recently changed her view, at least regarding the palaeoryctoid Asioryctes. Apparently, nasolacrimal contact may be absent in this genus. While the state of this character at the origin of eutherians is by no means completely clear, I would argue that the presence of nasolacrimal contact in perissodactyls is indeed a derived condition and synapomorphic. Perissodactyls primitively possess nasolacrimal contact, which appears to be derived and has been used as a perissodactyl synapomorphy (Fischer & Tassy, 1993). Presence of this contact occurs in equids, some species of Paheotherium, and most brontotheres, and it appears to be primitive for tapiromorphs. Spurious contacts between these two bones occur in some individuals of Phenacodus and some Recent hyracoids, but the absence of this contact appears to characterize all of the non-perissodactyl outpups employed here. Within Perissodactyla, the absence of nasolacrimal contact due to reduction of the lacrimal is derived and may be a synapomorphy uniting Tapirus, Pmtapirus, Plesiocolopirus, Colodon, Helaletes, and Heptodon. Novacek (1986) has also argued that a facial exposure of the lacrimal is a primitive eutherian character. Gregory (1920a) argued that the alternate condition, restriction of the lacrimal to the orbit and orbital rim, is primitive. Facial exposure of the lacrimal is small or absent in most extant eutherians, with the exceptions of xenarthrans, macroscelideans, perissodactyls, and artiodactyls. The prominent facial ex- posure of the lacrimal seen in artiodactyls, perissodactyls, various condylarths, and fossil (but not all Recent) hyracoids is a condition of uncertain polarity. Tethytheres possess small (if any) lacrimals, restricted to the orbital rim. It is not certain whether the absence of a large facial exposure of the lacrimal in tethytheres is primitive (as Gregory [ 1920al suggested) or derived. For comparisons made here, it seems clear that a large facial exposure of the lacrimal is primitive for perissodactyls, as it is seen not only in primitive perissodactyls, but in artiodactyls, phenacodonts, and fossil hyracoids (Gregory, 1920a) as well. Recent hyracoids have reduced the facial exposure, but this is probably related to the overall shortening of the face in these taxa. Jugal Description. The morphology of the jugal is quite conservative among tapiromorphs. The jugal forms the anteroventral border of the orbit, where it broadly contacts the maxilla and the lacrimal. The jugal extends posteriorly as the anterior portion of the zygomatic arch, curving dorsally to meet the zygomatic process of the squamosal, with which it forms a horizontal suture. Anarrow process of the jugal continues posteriorly, running along the ventral edge of the squamosal process, and extends to a point just anterior to the glenoid fossa. Discussion. The jugal morphology of perissodactyls, including tapiromorphs, is very similar to that of several other eutherian taxa and appears to be primitive. In hyracoids, sirenians, and proboscideans, the maxilla forms the anterior part of the zygomatic arch, and the jugal is reduced anteriorly, so that it no longer contacts the lacrimal. This condition is true of adult hyraxes, but ontogenetically the jugal starts out in contact with the lacrimal (pers. obs., UMA 4174 and 4176). Based on the distribution of this character, the adult hyrax condition (seen in proboscideans and sirenians) is probably derived. (A similar condition is also seen in some lipotyphlans and rodents.) Jugal/lacrimal contact, the primitive condition, is observed in other putative tethytheres, including Desmostylus (Vanderhoof, 1937) and Arsinoitherium (Court, 1992). Hyracoids also show another unusual feature of the jugal. In hyracoids, the jugal extends posteriorly to the glenoid fossa and actually contributes to the articular surface for the mandibular condyle. This feature is seen in marsupials and some eutherians, including rodents. Many more eutherians, however, including carnivorans and most artiodactyls, show the same condition seen in perissodactyls. The perissodactyl condition is usually interpreted as primitive for eutherians. Novacek & Wyss (1986) considered the

15 Figure 6. Skull of Eomompus (AMNH 5096) in dorsal view. (Modified from Osborn, 1913.) Abbreviations: sot supraorbital foramen. hyracoid condition to be derived and similar to the posteriorly extended jugal of tethytheres (which does not contribute to the jaw articulation). They considered the posteriorly extended jugal to be a synapomorphy for tethytheres and hyracoids, but it might also be argued that the jugal contribution to the glenoid in hyracoids is simply the primitive condition for therians. Frontal Description. The frontal forms the dorsal roof of the skull in the vicinity of the orbits. The nasal and parietal bones broadly contact the frontal anteriorly and posteriorly, respectively, in roughly transverse sutures. In some cases (discussed above in the section on the lacrimal), the maxilla also contacts the frontal anteriorly between the nasal and lacrimal. A postorbital process extends from the frontal in all tapiromorphs. In Eomompus and Mompus, this process is pierced by a supraorbital foramen (Fig. 6), which presumably transmitted the supraorbital artery and nerve, as in extant horses (Sisson & Grossman, 1938). This foramen is present in some other chalicotheres, including the schizotheriine Ancylotherium (Garevski, 1974) and the chalicotheriine Macmtherium (de Bonis et al., 1995), but is not visible in any of the crushed skulls of Litolophus. The supraorbital foramen of Macmtherium macedonicum is incomplete (de Bonis et al., 1995), i.e. the lateral edge is open, so that it is more properly called a supraorbital notch. sof TAPIROMORPH OSTEOLOGY 15 Discussion. The significance of the facial contact of the frontal and maxilla has been discussed in the section on the lacrimal in terms of the presence or absence of nasolacrimal contact. The supraorbital foramen develops first as a notch and later in life becomes closed laterally. In humans, a notch or foramen may be present in adults. The presence of the supraorbital foramen in Eornompus and Mompus was noted by Osborn (1913) as evidence of the chalicotherioid affinities of the former genus. A supraorbital foramen is also seen in Equus and Palaeotherium (Remy, 1992), but its absence is clearly primitive for tapiromorphs, as this foramen is absent in early equids and brontotheres, as well as nonperissodactyl outgroups. This foramen is not homologous with the supraorbital foramen of artiodactyls, which lies between the orbits (not on the orbital rim) and transmits the frontal vein. Parietal Description. The parietals span most of the dorsal braincase in all tapiromorphs, broadly contacting the frontals anteriorly, the supraoccipital posteriorly, and the squamosal laterally. In a number of taxa, such as Hyrachyus, a small anterior wing of the parietal extends into the orbital wall, where it contacts the alisphenoid and prevents the squamosal from contacting the frontal. This wing is absent in Tapirus, and a squamosaljfronta1 contact is present in this genus. Unfortunately, poor preservation of this area makes it difficult to say much about the distribution of this trait. The greatest extent of the sagittal crest is commonly seen along the contact of the parietals, so a discussion of this structure is most appropriately given here. Many tapiromorphs possess a true sagittal crest, i.e. a single median ridge, formed ontogenetically by the fusion of two parasagittal ridges that migrated to the midline from a more lateral position. Some rhinocerotids, including Subhyracodon, have arrested the migration of the parasagittal ridges, so that these ridges form the limits of a sagittal table. A true sagittal crest is seen in other primitive rhinocerotids, however, such as Teletaceras and Trigonias. The genus Tapirus shows three different sagittal crest morphologies. Tapiruspinchaque possesses a true sagittal crest, that develops in the manner described above, and this also appears to be true for fossil species in this genus (Simpson, 1945; Ray & Sanders, 1984). lbpirus baidii and T. indicus possess a sagittal table, a condition that represents a different point in the same developmental pattern that produces a true sagittal crest. Tapirus terrestris also possesses a single median crest, but the developmental path to this morphology is different in this species from the one followed

16 16 L. T. HOLBROOK in T pinchaque. In T terrestris, a single crest erupts from the dorsal midline at an early age, and increases in height until adulthood, resulting in the high-brow appearance that is characteristic of this species (Holbrook, in prep.). No other taxon analysed in this study appears to possess this highly unusual developmental pattern. Discussion. Possession of a true sagittal crest is primitive for perissodactyls, as it is characteristic not only of tapiromorphs, but also early equids, early brontotheres, and phenacodonts. A sagittal table, however, does occur in later brontotheres. The development of the sagittal crest is a consequence of the development of the temporalis musculature, for which the sagittal crest (and the parasagittal ridges) provide a relatively greater area of origin. Generally speaking, there is a tendency among herbivores, including ungulates, to reduce the size of the temporalis relative to that of the masseter. It might therefore seem surprising to find well-developed sagittal crests in presumably herbivorous early perissodactyls, as well as some living tapirs. Bressou (1961), contrary to Windle & Parsons (1901), noted that the temporalis of Tapirus indicus was well-developed, as it is in T. terrestris (pers. obs.). It would be interesting to know whether this muscle morphology reflects a difference in function between tapirs (and possibly other tapiromorphs) and other herbivorous ungulates. The sagittal table may be a consequence of body size, perhaps reflecting an allometric reduction in the relative size of the temporalis that simply is not significant at smaller body sizes. Suggestive evidence for this hypothesis is the distribution of this trait among large-bodied forms, including rhinocerotids, large brontotheres, and the larger species of Tapirus. Notable exceptions, however, which do not support this idea, include Paraceratheriurn and amynodontids, largebodied tapiromorphs that possess prominent true sagittal crests. Another exception to the hypothesis that the sagittal table is a consequence of body size comes from Coombs (1975) study of sexual dimorphism in Mompus. She noted that larger specimens tended to possess a sagittal crest, while a sagittal table was more characteristic of smaller individuals. This could be explained if one sex possessed a relatively larger temporalis musculature than the other. Sphenoid complex Description. The morphology of the presphenoid, basisphenoid, orbitosphenoid, and alisphenoid is fairly conservative among tapiromorphs. The presphenoid is rarely well-preserved and shows no significant variation. The basisphenoid can usually be discerned from its contact with the basioccipital; it also shows no significant variation. The orbitosphenoid is pierced by the optic foramen, which in all tapiromorphs is clearly well separated from the more posteroventral set of orbital foramina (i.e. the sphenorbital fissure, foramen rotundum, and anterior opening of the alisphenoid canal). These latter foramina pierce the alisphenoid, or, in the case of the sphenorbital fissure, lie on the alisphenoid s contact with the orbitosphenoid. The alisphenoid has a prominent orbital wing that contacts the orbitosphenoid, frontal, palatine, squamosal, and, in most cases, the parietal. In Tapirus, the parietal contact is precluded by the expanded contact with the squamosal. A ventral lamina of the alisphenoid overlies the external face of the pterygoid hamulus. In ventral view, the alisphenoid contacts the squamosal posterolaterally. The foramen ovale is distinct from the foramen lacerum medium in many taxa, including Cardiolophus, Isectolophus, Eornompus, Mompus, Heptodon, Helaletes, Lophiuletes, Hyrachyus, amynodontids, Triplopus, Hyracodon, Forstentooperia, Juxia, Paraceratheriurn, Uintaceras, and early rhinocerotids. These foramina are confluent in Pmtapirus, lbpirus, and later rhinocerotids. Discussion. The position of the optic foramen has been used by MacFadden (1976) to distinguish equids from other perissodactyls. According to MacFadden (1976), equids share a derived condition where the optic foramen is posteriorly placed in the orbit, so that it lies very close to the sphenorbital fissure. In support of his argument, MacFadden cited Simpson s (1952) claim, based on the holotype of Hyracotheriurn, that the optic foramen is confluent with the sphenorbital fissure in this genus, forming a single large foramen. Radinsky (1965a) demonstrated that a similar large foramen in Heptodon transmitted the optic nerve only, and that the foramen rotundum opens into the alisphenoid canal, giving the impression that a foramen is missing. Nevertheless, the large optic foramen of Hyrumtheriurn, as figured by Simpson, does appear to be posteriorly placed. Given the tenuous systematic position of Hyracotheriurn (Hooker, 1989, 1994), a posteriorly-positioned optic foramen may be synapomorphic for the Equidae or a subgroup of that family; the derived condition is present in Mesohippus and later equids, but information is not available on other early equids, such as Omhippus. It is not clear, however, that the derived condition is present in any other perissodactyl. Hooker (1989) claimed that the derived condition is present in Pachynolophus, Palueotheriurn, and Plagiolophus, but

17 later (Hooker, 1994) stated that these same taxa possess the primitive condition. From illustrations and descriptions of Pachynolophus (Savage et al., 1965; Remy, 1972) and Pulaeotherium (Remy, 1992), it would appear that these taxa retain the primitive condition of an optic foramen that is well-separated from the more posterior orbital foramina. Hooker (1994) also claimed that the derived condition is present in the Eocene brontothere Palaeosyops and the phenacodontid Ectocion. Hooker s (1994) claim about the condition of the optic foramen in Palaeosyops comes from an illustration by Osborn (1929: 326, fig. 275). Examination of brontothere skulls in the AMNH failed to clearly identify the optic foramen in any specimen, although one or two posteriorly placed foramina could be identified in the orbit of some specimens. The orbit of most brontotheres is quite anteriorly placed, but it is possible that a groove above the maxillary tubercle transmitted the optic nerve (and possibly other orbital nerves and blood vessels) from a posteriorly placed optic foramen (and other orbital foramina). Nevertheless, there is no clear evidence that brontotheres possess the derived position of the optic foramen. Hooker s evidence for the derived position of the optic foramen in Ectocion comes from Thewissen s (1990) illustration of the skull of Ectocion. In this illustration, there is a foramen visible behind the zygomatic arch which is separated from the more posterior foramina and occupies the same general position as the optic foramen in Phenacodus. It appears that this is the actual optic foramen and that, in Thewissen s illustration, the sphenorbital fissure has been labelled as the optic foramen and the foramen rotundum has been labelled as the sphenorbital fissure. Thus, Ectocion also possesses the primitive condition of the position of the optic foramen. Gazin (1956) described the primitive condition in a specimen of Ectocion from the Almy Formation (late Paleocene) of Wyoming. This reinterpretation of the position of the optic foramen has ramifications for Hooker s (1994) phylogeny of early perissodactyls. In Hooker s phylogeny, tapiromorphs were united with Pachynolophus and Hallensia on the basis of a few reversals and parallelisms. One of these reversals is from the derived position of the optic foramen to the primitive condition. Hooker s analysis made a posteriorly placed optic foramen primitive for perissodactyls, based on the presence of this condition in the immediate outgroup (Ectocion). The interpetation of the orbital foramina of Ectocion given here makes the condition observed in tapiromorphs, Pachynolophus, and Hallensia a primitive retention rather than a reversal. The confluence of the foramen ovale and the foramen lacerum medium is seen in all extant perissodactyls, but, as Edinger & Kitts (1954) demonstrated, this TAPIROMORPH OSTEOLOGY 17 condition has been acquired independently in all three extant lineages. The non-confluent condition is clearly primitive, as it is seen in non-perissodactyl outgroups and brontotheres, although the confluent condition is seen in some other groups of mammals (e.g. xenarthrans, pholidotans, camelids). MacFadden (1976) used the confluence of these foramina as a synapomorphy for the Equidae, as this condition is seen in Hyracotheriurn. Gingerich (1991) claimed that the primitive condition is present in at least one early specimen of Hyracotherium. The derived condition is also seen in Palaeotherium (Remy, 1992), Pachynolophus (Savage et al., 1965), the palaeotheriid Plagiolophus (Hooker, 1989), and Lambdotherium (CM62459). This condition was acquired by later rhinocerotids, but it may be a synapomorphy of tapirids. Squamosal Description. The squamosal forms, in lateral view, much of the posteroventral quadrant of the skull in all of the taxa under study. The dorsal border of this bone is usually a semicircular suture shared with the mastoid portion of the periotic, the supraoccipital, exoccipital, parietal, and alisphenoid. In Tupirus (and Equus), there is also a contact with the frontal, a contact absent in Hyrachyus, but few other taxa preserve the sutures of this region. The zygomatic process of the squamosal extends laterally, then anteriorly. The ventral face of the lateral extension bears part of the glenoid fossa, while the narrower anterior portion meets the jugal ventrally. In rhinocerotids, the blade of the anterior portion is notched posteriorly, giving it an undercut appearance. Several interesting characters occur in the vicinity of the jaw articulation. The glenoid fossa, in primitive tapiromorphs, usually bears a flat articular facet, roughly trapezoidal in outline and tapering posteriorly. The postglenoid process is usually very distinct, and its shape and orientation show interesting variation and distribution. In Lophiodon and chalicotheres, the process is peg-like and oriented anteriorly (Fig. 7B). In Homogalax and Cardiolophus, there may be a slight angling of the process anterolaterally. This condition, however, does not approach the more extreme anterolateral angling of the process (Fig. 7C) seen in Isectolophus, Heptodon, Hyrachyus, tapirids, helaletids, and all rhinocerotoids. The postglenoid process is further modified in some rhinocerotoids. In rhinocerotids, the anterior face of the process is convex and divided by a ridge into medial and lateral portions. A similar morphology is seen in Uintaceras, amynodonts, and indricotheres. The glenoid fossa, as previously stated, is generally flat, but in some tapiromorphs it appears to be divided

18 18 L. T. HOLBROOK A Figure 7. Basicrania of Palaeotheriurn (A), Eomorvpus (B), and Teletacems (C). Arrow indicates postglenoid process. (A modified from Remy, 1992; B modified from Osborn, 1913; C modified from Hanson, 1989.) B into regions. This condition occurs in rhinocerotoids, Hyrachyus, and tapirids. The glenoid fossa of all tapiromorphs extends to some extent onto the zygomatic arch, a feature which is not distinctive of chalicotheres, as claimed by Osborn (1913) and Colbert (1934). The postglenoid foramen, which transmits a vein draining from the squamosal sinuses, is clearly present in skulls of Homogalax, Cardwlophus, Eomorvpus, Isectolophus, Heptodon, Helaletes, Hyrachyus, Triplopus. Among tapiromorphs, the foramen is definitely

19 absent only in Colodon, tapirids, Paraceratherium, Juxia, Forsterrooperia, Uintaceras, rhinocerotids, and amynodontids. The postglenoid and posttympanic processes are separated in most tapiromorph taxa, leaving the external auditory meatus open ventrally. In some advanced rhinocerotids and amynodontids (but not the primitive members considered here), the ventral tips of these processes are fused, enclosing the external auditory meatus ventrally. The posttympanic process of Eomompus and Mompus is much shorter than the paroccipital process of the exoccipital (Fig. 8B), whereas in other tapiromorphs the length of the posttympanic process approaches or equals the length of the paroccipital process (Fig. 8A). The situation in other chalicotheres is less clear. Discussion. The peg-like morphology of the postglenoid process is seen in brontotheres, Palaeotherium (Fig. 7A), Lam bdotherium, phenacodonts and other archaic ungulates, while primitive equids more closely resemble Homogalax and Cardiolophus. Whereas it is not clear whether a more peg-like process or a flatter process is primitive, it is clear that the distinctly anterolaterally-oriented condition is derived and may be a synapomorphy for Isectolophus, Tapiroidea, and Rhinocerotoidea. The ridge dividing up the anterior face of the process may be a synapomorphy of rhinocerotids, amynodontids, and indricotheres. A simple articular surface also appears to be primitive, as evidenced by its presence in phenacodonts and non-tapiromorph perissodactyls. A glenoid surface divided into regions is certainly derived, but the possible influence of body size on this character cannot be ruled out. In contrast to statements by Prothero et al. (1988) and Court (1992), the postglenoid foramen is primitively present in perissodactyls, as seen in the abovementioned tapiromorphs, Palaeotherium (Remy, 1992), Hyracotherium, and the equids Mesohippus and Equus. The foramen is present in a number of other ungulate taxa (including phenacodonts and Pleuraspidotherium [UCMP ) whose close relationship to perissodactyls and each other was thought by Prothero et al. (1988) to be supported in part by the loss of this foramen. A small posttympanic process may be a synapomorphy for a clade including Eomompus, Mompus, and possibly other chalicotherioids. This trait may be related to the development of an auditory bulla that is fused to the skull in chalicotherioids. Periotic Description. The periotic is divided into two regions, the mastoid and the petrosal. The mastoid is exposed A B TAPIROMOFWH OSTEOLOGY 19 PtP I ty mf PtP Figure 8. Posterior skulls of TeZetaceras (A) and Eomompus (B) in lateral view. (A modified from Hanson, 1989 B modified from Osborn, 1913.) Abbreviations: mf: mastoid foramen; ptp: posttympanic process; ty: tympanic. laterally as a narrow triangle between the squamosal and exoccipital in a number of tapiromorphs, including Cardiolophus, Heptodon, Helaletes, Colodon, Hyrachyus, Triplopus, Hyracodon, amynodontids, and Teletaceras. In all of these taxa a mastoid foramen (Fig. 8A) is present. Rhinocerotids, other than Teletaceras, are amastoid, i.e. the mastoid is not exposed externally but instead is covered by the contact of the squamosal and exoccipital. Although they are amastoid, a mastoid foramen is visible between the squamosal and exoccipital in Trigonias and Subhyracodon. In tapirids, an intermediate condition is present, where the mastoid exposure is reduced but not absent. Petrosal morphology figures prominently in mammalian phylogeny, but the petrosal of tapiromorphs and other perissodactyls, where known, is very conservative in morphology. Radinsky (1965a) and Cifelli (1982) described the petrosal of Heptodon in some detail, and this description is generally applicable to all perissodactyls. Rather than attempt an exhaustive (and redundant) description of petrosal morphology for all tapiromorphs, only a few petrosal features will be described here. In Heptodon, the petrosal contacts

20 20 L. T. HOLBROOK the basisphenoid medially, the squamosal laterally, and the basioccipital posteriorly. The ventral surface of the promontorium lacks grooves for arteries, specifically the stapedial and internal carotid arteries (Cifelli, 1982). A number of openings pierce or are associated with the petrosal, including the cochlear aqueduct, a slit-like exit in the ventromedial side of the promontorium that drains blood from the cochlea (Cifelli, 1982). Discussion. The absence of the mastoid foramen was used by Prothero et al. (1988) as a synapomorphy for a clade including living ungulates except artiodactyls, as well as a number of extinct forms. The mastoid foramen is clearly present in primitive tapiromorphs; thus, this synapomorphy does not pertain to perissodactyls. Loss of the foramen is generally correlated with amastoidy, although the foramen is visible in Trigonias, which is otherwise amastoid. The lateral exposure of the mastoid is primitive for tapiromorphs, as it is characteristic of all perissodactyls that are not amastoid. This condition, however, is unusual among mammals; among eutherians, the mastoid exposure primitively faces posteriorly. Webb & Taylor (1980) and Coombs & Coombs (1982) argued that a lateral mastoid exposure is primitive for artiodactyls and that the posterior mastoid exposure seen in many ruminants is derived. A lateral exposure of the mastoid is also found in Radinskya, and this feature was used by McKenna et al. (1989) to ally this genus with perissodactyls. Amastoidy is a condition seen in many eutherians, including pholidotans, dermopterans, cetaceans, suiform artiodactyls, tethytheres, and hyracoids. This feature is certainly derived and has been proposed as a synapomorphy of tethytheres and hyracoids (Novacek, 1986, 1992a, 1992b; Novacek & Wyss, 1986, 1987; Novacek et al., 1988; Tassy & Shoshani, 1988). The common occurrence, however, of amastoidy among apparently unrelated eutherian groups, including some perissodactyls, casts some doubt on the validity of this character as a paenungulate synapomorphy (Fischer, 1989). The intermediate condition observed in tapirids is probably not, as stated by Fischer (1989), transitional between horses and rhinoceroses. Most families of rhinocerotoids are not amastoid, including the most primitive rhinocerotid, Teletaceras. Many later rhinocerotids are amastoid. While the morphology of the petrosal does not vary in any systematic way among tapiromorphs, there are some features seen in tapiromorphs and other perissodactyls that are important for higher-level phylogenetic studies. The ventrally-positioned slit-like cochlear aqueduct, for instance, was considered by Cifelli (1982) to be a derived feature shared by perissodactyls and Pheruxodus. The slit-like condition is present in Hyracotherium and Omhippus, but, according to Court (1990), a more dorsal and rounded (i.e. primitive) condition obtains in Pachynolophus, Plagiolophus, and Zbpirus. Additionally, Cifelli (1982) described hyracoids as lacking the slit-like condition, whereas Court (1990) described hyracoids as possessing that same condition. The cochlear aqueduct is absent in Arsinoitherium, most proboscideans, and most sirenians, but it is present in Numidotherium and Prorastomus, the earliest members of the Proboscidea and Sirenia, respectively. The absence of grooves on the promontorium for the internal carotid and stapedial arteries is probably related to the medial course of the former artery and the absence of the latter. There is some question as to the primitive course of the internal carotid artery in mammals. A medial, or extrabullar, course, as seen in perissodactyls, also occurs in marsupials, monotremes, xenarthrans, and hyracoids, suggesting that this condition is primitive for mammals. Wible (1986), on the basis of ontogeny, commonality, and comparisons with non-mammalian outgroups, argued that a transpromontorial course of the internal carotid artery (i.e. a come that crossed the promontorium) is the primitive condition for mammals. According to Wible, the medial course of the internal carotid is a derived feature and not homologous in monotremes, marsupials and eutherians. Fischer (1986,1989) has used this feature as a synapomorphy of hyracoids and perissodactyls. A third condition is observed in proboscideans (perbullar), where the artery runs within the bulla, and the condition in sirenians is unknown. Arsinoitherium lacks a promontorial groove, but it is not certain whether this indicates an extrabullar or perbullar condition. Even if Wible s (1986) argument is correct, the interpretation of this character is not clear, because hyracoids, perissodactyls, and tethytheres each possess one of two derived conditions, whose relationship to each other and to the primitive condition is not known. It is possible that the perbullar condition is derived from an extrabullar condition (or vice versa) that &-st appeared in the common ancestor of perissodactyls, hyracoids, and tethytheres. Occipital bones Description. The supraoccipital, basioccipital, and paired exoccipitals dominate the posterior aspect of the skull, and the sutures between these bones are often obliterated by age, giving the appearance of a single occipital bone. The foramen magnum is bordered by the basioccipital ventrally and the exoccipitals dorsally and laterally. The supraoccipital does not participate in the border of the foramen magnum. This

21 bone extends dorsally onto the skull roof, contacting the parietal and interparietal and forming the posterior end of the sagittal crest and the lambdoidal crest, which is strong in perissodactyls. The flaring lambdoidal crest provides the attachment site for the large nuchal ligament. The wings of the lambdoidal crest extend laterally and ventrally, onto the squamosal bone, where they join the crest from the dorsal aspect of the zygomatic arch. The ventral side of the basioccipital has, in some tapiromorphs, a median keel. In Paraceratherium, a pair of tuberosities are present at the juncture of the basioccipital and basisphenoid. In all tapiromorphs, the anteromedial corner of the articular surface of the occipital condyle extends anteriorly as a small lappet onto the ventral aspect of the basioccipital. Just anterior to the occipital condyle, the hypoglossal foramen pierces the basioccipital. Each exoccipital possesses a ventral paraoccipital process. The exoccipital contacts the mastoid portion of the periotic, and, in amastoid forms, the exoccipital is joined to the posttympanic process of the squamosal. Discussion. The exclusion of the supraoccipital from the border of the foramen magnum is a derived feature found in all perissodactyls (contra Court, 1992) and shared with hyracoids (contra Court, 1992) and tethytheres. This character has been used as a synapomorphy uniting tethytheres and hyracoids (Shoshani, 1986) or tethytheres, hyracoids, and perissodactyls (Tassy & Shoshani, 1988). Fischer (1989) argued that this feature should not be considered homologous in hyracoids and tethytheres, because the supraoccipital does form part of the border of the foramen magnum in neonatal and juvenile hyraxes, a developmental pattern observed in perissodactyls as well. An alternative explanation is that the ontogenetic transformation itself (from inclusion to exclusion of the supraoccipital) should be considered as the derived feature in this case. Given this, it is likely that the condition seen in adults of perissodactyls, hyracoids, and tethytheres is the result of the same derived ontogenetic transformation and could therefore be considered as a valid synapomorphy uniting these taxa. Elaboration of the ventral basioccipital, whether by means of a median ridge or paired tuberosities, is seen in larger taxa and is certainly related to an allometric relationship between the size of the head and the neck muscles. The anteromedial extension of the articular surface of the condyle was deemed a characteristic of chalicotheres by Borissiak (1946) but is certainly a primitive feature of tapiromorphs. Pmpanic TAPIROMORPH OSTEOLOGY 21 Description. The tympanic (or ectotympanic) is not often preserved in tapiromorph fossils, since it is usually loosely attached to the skull. In Heptodon, the tympanic is a thin horseshoe-shaped ring of bone. In living tapirs, the tympanic is expanded into a platelike bulla, but this bulla is not fused to the skull, Hence, no such bone is preserved on many fossil tapiromorphs. The chief exception to this rule is seen in the chalicotheres, including Litolophus, Eomompus, and the chalicotheriids. In these taxa, the tympanic is fused to the skull and covers the petrosal as a bulla. In some tapiromorphs, there may be an ossification forming the external auditory meatus but not covering the petrosal. This condition is seen in some specimens of!l riplopus, Hyracodon, Amynodon, and Subhyracodon. This may represent the ectotympanic, but, as with extant tapirs, the bulla of extant rhinoceroses is loosely attached to the skull. An entotympanic element is also present and fused to the skull in extant rhinoceroses, but it does not form a bulla, i.e. the entotympanic does not cover the petrosal ventrally. Discussion. The condition observed in Heptodon is the primitive eutherian condition. The fused tympanic bulla of chalicotherioids may be a synapomorphy of this group, because the bullae of other tapiromorphs were apparently not tightly fused to the skull. Mandible Description. The mandibular symphysis of all tapiromorphs is fused and extends posteriorly to at least the level of the first premolar. The symphysis is only slightly tilted dorsally in most taxa, so that the incisors are somewhat procumbent. The angle of the mandible and the coronoid process are well developed in all tapiromorphs. The mandibular condyle is positioned well above the level of the tooth row in all tapiromorphs except Eomompus. One character of systematic significance is the presence of the postcotyloid process, a buttress on the posterior border of the mandible just ventral to the mandibular condyle and running inferomedially. This process is well-developed in rhinocerotids, but smaller versions are seen in other rhinocerotoids, including Hyrachyus, Paraceratherium (AMNH 26172), Juxia (AMNH 20286), Uintaceras, and amynodontids (with the exception of Metamynodon). Hyracodon does not possess a postcotyloid process, nor do Colodon, Heptodon, Helaletes, Schlosseria, Lophialetes, Pmtapirus, or Mompus. Some Tapirus possess a tubercle in a position similar to that of the postcotyloid process, but the orientation and shape of this tubercle do not match that of the more crest-like postcotyloid process. The articular surface of the condyle is generally like

22 22 L. T. HOLBROOK a half cylinder, but in many taxa a lappet extends onto the medial part of the posterior side of the condyle, and a distinct groove often undercuts the lappet ventrally. In rhinocerotoids, this posterior articular lappet appears as a natural but distinct extension of the articular surface, generally convex, whereas in the schizotheriines Mompus and 5locephalonyx this surface is flat or slightly concave and extends at right angles to the main articular surface, giving the impression of a completely separate facet. In lbpirus and Protapirus, the lappet is more medially oriented. No lappet can be discerned in Heptodon, Colodon, Schlosseria, or Lophialetes. Discussion. The relatively long, slender symphysis and somewhat horizontal position of the incisors are features also seen in a number of non-tapiromorph perissodactyls, including Hymtherium, Pachynolophus, and Lambdotherium. In Palueotherium and brontotheriids, the symphysis is shorter, stouter, and turns more dorsally, so that the incisors point more vertically. This difference in jaw morphology between Lambdotherium and brontotheres is one reason that Mader (1991) excluded Lambdotherium from the Brontotheriidae. The presence of a postcotyloid process is a potentially useful character, but there are some difficulties in interpreting it. In their cladogram, Prothero et al. (1986) claimed that this process is not present in Teletaceras (identified as Clarno rhino ), but Hanson (1989) described a prominent postcotyloid process in this genus. Examination of the type of Z&Zetaceras radinslzyi (UCMP ) conhns Hanson s description. A well-developed process may be a synapomorphy of rhinocerotids, and the process in general may be synapomorphic for rhinocerotoids. The process is not found in Hyracotherium or brontotheres, but a similar process is seen in Equus. The posterior articular lappet is present in Hyracotherium and Equus, as well as in brontotheres, who often show a morphology similar to that of schizotheriine chalicotheres. The apparent absence of this feature in some tapiroids may have more to do with the small size and poor preservation of specimens of these taxa than with a shared evolutionary history. POSTCRANIAL, OSTEOLOGY The postcranial skeleton of tapiromorphs is compared on a bone-by-bone basis below, with information presented in the same manner as in the section on cranial osteology. VEmBRAE Atlas Description. Tapiromorphs show two different conditions in the proportions of the atlas, the difference sf tf Figure 9. Atlas of Paraceratherium in dorsal view. (Modified from Cooper [1923].) Abbreviations: an: atlantal notch; sf: spinal foramen; tf: transverse foramen. lying largely in the relative size of the transverse processes. These processes may be mediolaterally narrow or broad. Most tapiromorphs retain the narrow morphology. The only exceptions are among the rhinocerotoids (and, to some extent, large chalicotheres like Mompus); some larger forms, particularly amynodontids, Juxia, and rhinocerotids, possess the %road morphology. Paraceratherium, however, possesses the primitive morphology (Fig. 9). The atlantal notch is an emargination of the anterior edge of the transverse process adjacent to the neural arch, which accommodates the first spinal nerve after its exit &om the spinal foramen. In Tapirus and some living rhinoceroses this notch is bridged anteriorly by bone, so that it forms a foramen. Four foramina piercing the atlas are the right and left spinal foramina and transverse foramina (or intervertebral canals). The spinal foramen pierces the anterodorsal aspect of the neural arch and is the exit of the ht spinal nerve from the neural canal. This morphology of the foramen does not appear to vary among perissodactyls. The intervertebral canal is variably present among the taxa surveyed. This canal has its anterior opening on the ventral side of the transverse process, and its posterior opening may also be on the ventral transverse process or may open on the posterior edge of this process, adjacent to the axial facets. The canal is clearly present in Paraceratherium (Fig. 9), Hymdon, Triplopus, amynodontids, and Trigonias. Some (but not all) rhinocerotids (e.g. Subhyracodon) lack the canal. The atlas of Juxia does not preserve the cranial opening of this canal, but the caudal opening is present, indicating presence of the canal.

23 Discussion. The primitive proportions of this bone are approximately equal transverse width and anteroposterior length (i.e. the narrow morphology), although it appears that the broad transverse processes of brontotheres were already present in Eotitanops (Osborn, 1929). The atlases of Phenacodus (Cope, 1884) and equids (e.g. Mesohippus [Scott, 19411) possess relatively narrow tranverse processes. Kitts (1956) did not describe the atlas of Hyrmotherium, claiming that poor preservation rendered it uninformative. Considering the similar morphology in brontotheres and some large tapiromorphs, the broad morphology may be an adaptation for supporting a large head and neck, by providing large attachments for musculature. A difference between the broad morphology in brontotheres and tapiromorphs is the shape of the atlantal notch. In brontotheres the notch is broad and shallow, whereas in rhinocerotoids the notch is narrow and more pronounced. Prothero et al. (1986) claimed that the prominent atlantal notch is a synapomorphy of rhinocerotoids (exclusive of Hyrachyus), but the morphology of rhinocerotoids is not easily distinguishable from that of other tapiromorphs for which this character is known. Prothero et al. (1986) also used the absence of an intervertebral canal as a synapomorphy of rhinocerotoids exclusive of Hyrachyus. A careful analysis of the distribution of this character, however, reveals that this cannot be the case. Only some rhinocerotids actually lack this canal. It is present in all other rhinocerotoids for whom this part of the atlas is well preserved. In extant rhinocerotids, the vertebral artery does not pierce the transverse process of the atlas. The presence of the caudal opening of the intervertebral canal in the atlas of Juxia indicates that the artery did pass through the transverse process. Axis Description. The axis of perissodactyls differs from that of phenacodonts (and is similar to that of hyracoids) in having a relatively low, anteroposteriorly long spinous process. On the axis of Phenacodus, the spinous process is tall and narrow, not very different from that of a thoracic vertebra. Other vertebrae Description. Other than differences related to size, there are few features of phylogenetic interest on the other cervical, thoracic, lumbar, sacral, and caudal vertebrae of tapiromorphs. This apparent conservatism may be illusory, because relatively complete vertebral series are not known for many taxa. Table TAPIROMORPH OSTEOLOGY 23 Ihble 3. Vertebral numbers in tapiromorphs. Data on extant taxa are from Flower (1885) Genus Thoracic Lumbar Sacral Tapiridae Tapirus Hyrach yidae Hy mchyus Amynodontidae Amynodon Shammynodon Rhinocerotidae Subhyracodon Dicerorhinus Rhinoceros lists the number of thoracic, lumbar, and sacral vertebrae in various tapiromorphs. The centra of cervical vertebrae 3 to 7 are opisthocoelous. The only significant morphological difference seen among the posterior cervical vertebrae is the presence of a prominent lamina descending from the ventral aspect of the transverse process of the sixth cervical vertebra. In Paraceratherium, this lamina or flange is divided into anterior and posterior segments (Granger & Gregory, 1936). Where they are known, the thoracic vertebrae are characterized by tall spinous processes that are inclined posteriorly. In larger taxa, the tips of these processes may be expanded for the attachment of the nuchal ligament. The number of thoracic vertebrae varies among tapiromorphs. The number ranges from 20 in Dicemrhinus to 17 in Sharamynodon. The spinous processes of the lumbar vertebrae of tapiromorphs are shorter and more robust than those of the thoracic series and are inclined slightly anteriorly or not at all. The zygapophyses of the lumbar vertebrae are cylindrical or embracing. Five lumbar vertebrae seem to be the typical number for tapiromorphs. The sacrum consists of six, or in some cases fewer, fused vertebrae, although poor preservation often makes it difficult to be certain of the number. Caudal vertebrae are rarely preserved, and there appears to be little variation in the morphology of these simple vertebrae. Discussion. None of the variation described above appears to signify anything of phylogenetic importance, at least in terms of potential synapomorphies that could be used in tapiromorph systematics. The variation in the number of various kinds of vertebrae does not follow any phylogenetic pattern. In fact, at least

24 24 L. T. HOLBROOK in the domestic horse Equus caballus, the number of vertebrae in any of these regions (with the exception of the cervical region) shows intraspecific variation (Flower, 1885). The descending lamina of the sixth cervical vertebra is a primitive feature, seen in other perissodactyls and non-perissodactyl outgroups. The functional significance of the inclination of the neural spines of the lumbar regions has been discussed by Slijper (1946). According to Slijper, the spinous processes of the lumbar vertebrae are primitively inclined anteriorly. This anterior inclination is reduced in perissodactyls in response to the deemphasis of spinal flexors and stiffening of the back. The embracing zygapophyses of the lumbar vertebrae are a derived feature. In most mammals, these zygapophyses are flat. According to Thewissen & Domning (1992), embracing zygapophyses are also found in artiodactyls and Arctocyon. PECTORAL LIMB Scapula Description. The general proportions of the scapular blade, which is usually long and narrow, are remarkably conserved among early tapiromorphs, even among lineages as postcranially different as cursorial tapiroids like Helaletes, mediportal rhinoceroses like Trigonias and Subhyracodon, and chalicotheres. The most divergent morphology among early tapiromorphs appears to be that of Metamynodon, which has a robust, almost circular shoulder blade. Near the scapular neck, the supraspinous fossa primitively ends in a fairly deep coracoscapular notch. This is seen in Heptodon, Helaletes, and the chalicothere Mompus (Coombs, 1978a). This notch is most exaggerated in lbpirus. In the rhinocerotoids Metamynodon, Hymdon, and rhinocerotids, the notch is distinctly shallower, but still present. The acromion process is present in Helaletes, Lophialetes (Reshetov, 1979), Metamynodon, and possibly Juxia, but not in Tapirus, Mompus, Hyrachyus, Hyracodon, Uintaceras, and rhinocerotids. For all other taxa, the condition of the acromion is not known. In most taxa lacking the acromion, the spine is reduced ventrally (distally) and does not extend far onto the neck, often disappearing dorsal to the neck. The one exception is Hyrachyus, which retains a high scapular spine extending onto the neck. The tuber spinae can be absent, weak, or strong. It is likely that development of this structure is related more to body size than to phylogeny. Coombs (1978a) noted the difference in development of this structure among different-sized specimens of Mompus elatus. Smaller specimens have a weak tuber spinae, whereas a prominent tuber spinae is present on large specimens. Among taxa that preserve this area, the tuber spinae is present but not prominent in Hyrachyus and relatively strong in Helaletes, Metamynodon, Hyracoo!on, Juxiu, and rhinocerotids. Helaletes, which is smaller than Hyrachyus, is the only known exception to the body size rule. A strong, anteromedially-directed coracoid process is present in a number of tapiromorphs, including Homogalax, Helaletes, Hyrachyus, and Triplopus. A number of other taxa reduce the coracoid to a low tubercle on the anterior end of the glenoid. These taxa include Heptodon, lbpirus, hphialetes, Metamynodon, Hymcodon, Juzia, Uintaceras, and rhinocerotids. Discussion. Primitively, the blade of the perissodactyl scapula is relatively long and narrow, as it is in Phenacodus and Hymtherium. Artiodactyls primitively have a more triangular blade. Phenacodus, Hyracotherium, and brontotheres also show the deep coracoscapular notch, indicating that this is a primitive trait. A high spine extending distally well onto the neck is seen in Phenacodus, living hyraxes, the brontothere Eotitamps, and Hymtherium, and appears to be a primitive trait. An acromion process is likewise a primitive feature. The tuber spinae is weak or absent in Hymtherium and Eotitanops, although it is strong in later brontotheres, supporting the idea that this character is related to size. A strong, anteromedially-directed coracoid process appears to be primitive for tapiromorphs. Among outgroups, it is seen in phenacodonts, living and fossil hyracoids (Fischer, 1986), and Hyracotherium. Humerus Description. The humeral head, in all tapiromorphs, is hemispherical and posterodorsally oriented. The morphology of the proximal tuberosities is also fairly conservative in tapiromorph evolution. The greater tuberosity is large and has a prominent anteromedial?look. In lbpirus, the greater tuberosity is divided by a groove into anterior and posterior components, but no other taxon in this study appears to share this morphology. In primitive tapiromorph lineages, the humeral shaft lacks any crests or tuberosities (Fig. 1OA). The chief exceptions to this are among rhinocerotoids. The deltopectoral crest and deltoid tuberosity are conspicuous in rhinocerotids (Fig. 1OB). The crest extends anteromedially from the anterior face of the shaft in Trigonias and Subhyrarodon, as well as in Recent rhinoceroses. A similar morphology is seen in Uintaceras (Holbmk & Lucas, 1997), suggesting rhinocerotid affinities for this genus.

25 TAPIROMORPH OSTEOLOGY 25 Figure 10. Right humeri of Colodon (AMNH 10781) (A) and Uintaceras (CM 12004) (B) in anterior view. Abbreviations: dpc: deltopectoral crest. Scale bar = 1 cm. Among other rhinocerotoids, the situation is less clear. Amynodontids, such as Amynodon, appear to lack a rhinocerotid-like deltopectoral crest, but many specimens are too distorted to be certain that no crest exists at all. Scott (1941) described a deltopectoral crest in Metamynodon but was careful to note that it was not as elaborate as that of rhinocerotids. Scott (1941) also described a distinct crest and tuberosity in Hyracodon, but all specimens of Hyracodon (as well as of Triplopus) that I have examined lack this morphology. Forster Cooper (1923) described a humerus of Paraceratherium that lacked a crest or tuberosity. Granger & Gregory (1936) described a humerus of Paraceratherium that possessed a deltopectoral crest, but the morphology of this specimen (which is much larger than those described by the previous authors) appears to be more similar to that of Metamynodon (as figured by Scott [1941, plate XCIV, figs 3 and 3al) than to those of rhinocerotids. Some chalicotheres (e.g. Mompus) and lophiodontids (Lophiodon) also show some development of the deltopectoral and supinator crests, but the deltopectoral crest differs from that of rhinocerotids in much the same way as that of amynodontids. The humerus of Chulicotherium has a smooth shaft with little or no crest development. The humeri of eomoropids are unknown. The distal humerus is generally mediolaterally compressed in all tapiromorphs, with a reduced entepicondyle. Only chalicotheriids (e.g. Mompus)

26 26 L. T. HOLBROOK possess a well-developed entepicondyle (Coombs, 1978), but, as in other perissodactyls, there is no entepicondylar foramen. A supratrochlear foramen is present in Homogalax and Lophiodon, but in chalicotheriids, Colodon, lbpirus, Lophiuletes, Deperetella, Hyrachyus, Triplopus, Hymcodon, amynodontids, and rhinocerotids, the olecranon and coronoid fossae are separated by a thin lamina of bone. In Heptodon, Helaletes, Uintaceras, and some specimens of Hyrachyus, this lamina is perforated, but the perforation may be artificial. The distal articulations are compressed into what is essentially a single trochlea. The lateral part of this articulation (the capitulum) bears a prominent keel in most taxa. The main exceptions are larger taxa, where the keel is poorly defined or barely discernible, as in amynodonts, rhinocerotids, Lophiodon and Mompus. This keel separates the medial part of the capitulum from a lateral portion that Rose (1996) refers to as the lateral articular shelf. This shelf is generally narrow and tapers distally in most taxa, including Homogalax, Heptodon, Hyrachyus, and Hynu;.odon. In Colodon, Deperetella, and Lophiuletes, the shelf extends more distally and does not taper in its width. The distal condyles are similar, although the medial is usually larger. One unusual feature of Lophialetes is that the lateral condyle extends more distally than the medial (Radinsky, 1965b). Discussion. The humerus of the most primitive perissodactyls displays a generally subcursorial morphology, with a smooth, gracile shaft and a narrow distal extremity. A similar morphology is seen in recent hyracoids and some phenacodonts, but other phenacodonts show a different morphology whose polarity at higher taxonomic levels is uncertain. The ancestors of tapiromorphs, however, apparently possessed the subcursorial morphology. The morphology of the head of the humerus is fairly conservative in perissodactyl evolution, appearing as a hemisphere whose centre points more or less posterodorsally. In phenacodonts, the head is more medially placed. This difference refleds the emphasis in perissodactyls on restricting limb motion to the parasagittal plane, a cursorial adaptation. In primitive perissodactyls like Hyracotherium, the greater tuberosity is almost as large as the head, whereas the lesser tuberosity is much smaller. The anterior hook of the greater tuberosity is a feature commonly seen in other mammals, including artiodactyls and terrestrial carnivorans. The extent of the hook varies among extinct ungulates. As figured by Cope (1884, plate XLIXb, fig. 2a) and described by Kitts (1956), the greater tuberosity of Hymcotherium does not curve anteromedially to cover the bicipital groove, as it does in tapiromorphs. The morphology of Hymcotherium is similar to that of phenacodonts, but it is difficult to say whether the tapiromorph condition is primitive or derived. In Hymtherium, Homogalax, and other primitive perissodactyls, the shaft of the humerus is generally smooth and lacks distinct crests or tuberosities. A similar morphology occurs in Recent hyracoids and the phenacodonts Copecwn and Ectocion (Thewissen, 1990). Other phenacodonts, including TetracZaenodon, Pheenacodus, and Meniscotherium (Thewissen, 1990; Williamson & Lucas, 1992), possess a conspicuous deltopectoral crest. Nevertheless, the absence of a distinct deltopedoral crest and deltoid tuberosity appears to be primitive for tapiromorphs. Indeed, the presence of a distinct crest and tuberosity is found, among tapiromorphs, only in some rhinocerotoids, chalicotheres, and lophiodonts. It might be argued that the variation in deltopectoral crest development is linked to body size. However, Paraceratherium and amynodontids like Metamynodon are larger than some primitive rhinocerotids, yet the crest in these taxa is less developed. The distinctive condition seen in rhinocerotids and Uintaceras should be considered as a derived character and a potential synapomorphy. A mediolaterally compressed distal humerus is a feature common to many sub-cursorial and cursorial taxa, including artiodactyls (Rose, 1985) and a number of archaic ungulates, including phenacodonts. Phenacodonts differ from both artiodactyls and perissodactyls in the possession of the entepicondylar foramen and a prominent entepicondyle. A supratrochlear foramen, which Rose (1985) has described as a cursorial adaptation, is found in phenacodonts and most species of Hyracotherium, but is absent in brontotheres. Presence of the foramen is probably primitive for tapiromorphs, but loss of the foramen must have occurred more than once in perissodactyl evolution. A keeled capitulum is also found in artiodactyls (Rose, 1985); this trait helps to restrict the movement at the elbow joint to the parasagittal plane. Phenacodonts, including Phenacodus (e.g. CM 44857) and Meniscotherium (Williamson & Lucas, 1992), have a smooth spherical capitulum without a distinct keel, but there is a lateral articular shelf that tapers distally. Hyracotherium possesses a very distinct keel and a tapering lateral articular shelf. A tapering shelf is likely to be primitive for tapiromorphs. The distally extended shelf of some tapiromorphs is actually very similar to that of artiodactyls (Rose, 1985). Radius Description. The proximal end of the radius bears two concave facets for the distal humerus, the medial facet generally being larger than the lateral. In a number

27 TAPIROMORPH OSTEOLOGY 27 of taxa (Homogalax, Isectolophus, Mompus, Heptodon, Helaletes, Colodon, npirus, Lophialetes, Hyrachyus, Triplopus, and Hyracodon), a distinct lateral process extends well beyond the thickness of the shaft, deepening the groove bearing the lateral facet. In Deperetella, amynodontids, Juxia, Paraceratherium, Uintaceras, and rhinocerotids, this process is weak or absent. Discussion. In proximal view, the outline of the radial head of perissodactyls and other cursorial taxa is irregular, as opposed to the more uniformly rounded radial heads of non-cursors (Rose, 1990). An irregular head restricts the ability to supinate the forearm. A distinct lateral process on the proximal radius is probably primitive for tapiromorphs, as it is also characteristic of other primitive perissodactyls. The loss of this process may be synapomorphic for amynodontids, indricotheres, and rhinocerotids (and probably convergent in Deperetella), but it might be argued that this trait is related to size. The presence of the process in a large non-rhinocerotoid, Tapirus, weakens this argument. Ulna Description. The ulna is free in most tapiromorphs, but it may be fused to the radius proximally and distally in Tapirus. The olecranon process is short but otherwise well-developed it may be more robust in larger taxa but otherwise varies little among tapiromorphs. The semilunar notch is relatively shallow, especially since the coronoid process is reduced. The humeral articulation is mediolaterally convex and articulates with the distal trochlea of the humerus. The radius articulates proximally with the ulna on its anterior face, just distal to the semilunar notch, so that the humeral articulations of the radius contact those of the ulna. Two facets are present on the anterior face of the ulna for the proximal radius. These facets can be distinct from each other, as in Homogalax, Isectolophus, Colodon, Tapirus, Hyrachyus, Paramynodon, Hyracodon, Trigonias, and Uintaceras, or they may be confluent, as in Mompus, Helaletes, and Metamynodon. The shaft of the ulna is bowed in tapiromorphs, so that it presents a concave surface posteriorly and a convex one anteriorly. The distal ulna articulates with the distal radius medially. The distal ulna bears two facets, for the cuneiform and pisiform. In most tapiromorphs, the pisiform facet lies posterolateral to that for the cuneiform, including in Deperetella, contrary to the statement of Radinsky (1965b), which said that the pisiform facet lies posterior to the cuneiform facet in Deperetella and lateral in other tapiromorphs. Discussion. The proximal radial facets of the ulna are confluent in Phenacodus, but separate in the brontothere Palaeosyops. The distribution of this character among tapiromorphs suggests that the separate condition is primitive, but the distribution of the confluent condition does not appear to be phylogenetically significant. A bowed ulnar shaft is characteristic of all perissodactyls and is probably derived, since a straight ulnar shaft is characteristic of Phenacodus (Cope, 1884), Meniscotherium (Williamson & Lucas, 1992), and early artiodactyls (Rose, 1985). Carpus Description. The carpus of tapiromorphs consists of eight bones that articulate in an alternating fashion (as discussed below; see Fig. 11A). Features relevant to the discussion of tapiromorph phylogeny will be emphasized here. The scaphoid, lunar, cuneiform, and pisiform comprise the proximal row of carpals, which articulates with the radius and ulna proximally and the distal carpal row distally. The scaphoid articulates with the radius proximally and three distal carpals: the trapezium, trapezoid, and magnum. The radial facet of the scaphoid varies in shape among tapiromorphs. In most taxa, this facet is more or less flat and round, and the anterolateral border has a short contact with the lunar (Fig. 11B). In rhinocerotids, amynodontids, and Uintaceras, the facet is saddle-shaped and somewhat square or rhomboid in outline, due in part to the posterior elongation of the lunar contact (Fig. 11D). A third condition is seen in tapirids, where the lunar contact has become emarginated, so that the lunar is embraced by the scaphoid (Fig. 11C). The functional significance of these morphologies is discussed below. The radial facet of the lunar resembles an anteroposteriorly convex half-cylinder. In living rhinoceroses (rhinocerotines), the ulna also articulates with the lunar (Heissig, 1989). In anterior view, the lunar contacts the unciform and, in some taxa, the magnum. Anterior contact between the lunar and magnum is present in Isectolophus, chalicotheres, Lophiodon (Deperet, 1903), Heptodon, Hyrachyus, Paraceratherium, and amynodontids. Anterior contact between the lunar and magnum is reduced but present in Protapirus, Tapirus, Colodon, Uintaceras, and Juxia. The anterior magnum contact is lost in some taxa, including Hyracodon and rhinocerotids. A pair of distal facets on the lunar articulate with the unciform and hump of the magnum. In most tapiromorphs, these facets have an extensive contact with each other (Fig. 11E). In Tapirus, this contact is lost (Fig. 11F). The distal carpal row is comprised of the trapezium, trapezoid, magnum, and unciform. The trapezium is

28 28 L. T. HOLBROOK B A MC 111 E F % I1 Figure 11. Manual elements of some tapiromorphs. (A). Right manus of Uintacerus (CM12004) in anterior view. (Scale bar= 1 cm.) (l3-d). Scaphoids of Heptodon (B), Zbpirus (C), and in proximal view. Radial facet is shaded; solid line indicates extent of contact with lunar. Anterior- is towards the top of the page. B and D are right scaphoids that have been Right lunars of Heptodon ( 3) and Zbpirus (F), in distal view. Anterior is towards the top of the page. Facets for the unciform and magnum are shaded. (A, B, D, E, and F modified from Radinsky, 1965a; C based on left scaphoid of CM ) Abbreviations: cun: cuneiform; lun: lunar; mag: magnum (for A), magnum facet of lunar (for E and F); MC: metacarpal; sca: scaphoid; trd trapezoid; unc: unciform (for A), unciform facet of lunar (for E and F). cj. a small, simple bone that contacts the scaphoid, trapezoid, and, in some taxa, the second metacarpal. Contact between the trapezium and second metacarpal is known to be present in IsectoZophus, most species of Mompus, Heptodon, Pmtapirus, Rpirus, Hyrachws, and Uintacems; this contact is absent in Lophialetes,

29 Schlosseria, Deperetella, Hyracodon, Trigonias, and Subhyracodon. The condition in amynodontids is unknown. Cerdeiio (1995: table 3) scored hyracodontines as unknown for this trait and Trigonias as possessing the contact, but neither of these statements is supported by examination of specimens nor by the observations of Scott (1941). Cerdeiio s (1995) data do show, however, that this contact is present in a number of rhinocerotids, including the extant genus Rhinoceros. In some derived chalicotheres, including one species of Moropus, the trapezium is lost altogether (M.C. Coombs, pers. comm.). The trapezoid, magnum, and unciform all bear distal facets for the second, third, and fourth metacarpals, respectively. The distal facet of the unciform also articulates with the fifth metacarpal, which may be functional or vestigial. Discussion. The carpus of perissodactyls is essentially alternating, i.e. the proximodistal articulations between the carpal rows are such that at least some of the bones articulate with more than one other in the adjacent row. In an alternating arrangement, the scaphoid articulates with the trapezium, trapezoid, and magnum, while the lunar articulates with both the magnum and the unciform. This arrangement has been contrasted with the serial arrangement, where each bone in the proximal row articulates with only one bone in the distal row, e.g. scaphoid and trapezoid (and trapezium, but not magnum), and lunar and magnum. Serial arrangement of the carpals and tarsals is also known as taxeopody. Cope (1884) first recognized these two patterns of carpal arrangement and used them as evidence of phylogenetic relationships. Cope (1884) first described the serial arrangement in Phenacodus and argued that this was the primitive condition for mammals, a point refuted by Matthew (1897), who described the alternating pattern in Tetraclaenodon. Later, Radinsky (1966b) demonstrated that the alternating arrangement is present in Phenacodus uortmani, and argued that both arrangements are present in this genus. This latter remark has proved to be important to arguments that the serial arrangement is not a useful character in phylogenetic studies of ungulates (e.g. Prothero et al., 1988). Thewissen (1990) has demonstrated that the carpus of Phenacodus is indeed alternating. Contact between the scaphoid and magnum as well as between the lunar and unciform exists in all species of Phenacodus for which the carpus is known. Williamson & Lucas (1992) described the carpus of Meniscotherium as serial, yet they stated that the scaphoid contacts the magnum and that the lunar contacts the unciform. As noted by Thewissen (1990), the alternating arrangement appears to be characteristic of all phenacodonts. The condition in Recent hyracoids is quite different TAPIROMORPH OSTEOLOGY 29 from that of either perissodactyls or phenacodonts. For one thing, hyracoids possess a free centrale. This bone lies between the scaphoid, trapezoid, lunar, and magnum, and it is likely that in perissodactyls and in phenacodonts it has fused to the scaphoid to form the process of that bone which contacts the magnum. Furthermore, hyracoids lack a lunar/unciform contact, so that they are considered to possess the serial arrangement. Rasmussen, Gagnon & Simons (1990) have noted evidence of taxeopody in isolated carpals (and tarsals) of Oligocene hyracoids from the Fayum. Taxeopody has been considered to be a synapomorphy of tethytheres and hyracoids by some workers (e.g. Novacek& Wyss, 1986,1987; Novaceket al., 1988), but Fischer (1986, 1989) has argued that taxeopody in hyracoids and tethytheres is not homologous. In any case, taxeopody is clearly a derived condition, which makes it difficult to use hyracoids as an effective outgroup for carpal morphology of perissodactyls. All perissodactyls essentially retain the alternating carpal arrangement. The only notable deviation from this pattern is the reduction or loss of the anterior contact of the lunar and magnum. The primitive condition, where, in anterior view, the lunar appears to rest equally on the magnum and unciform, is seen in equids and brontotheres. The distribution of reduced lunar/magnum contact suggests that it must have evolved more than once in tapiromorph history. Little previous work has been done on phylogenetic aspects of carpal morphology in tapiromorphs. Klaits (1972, 1973) has examined some of the functional changes seen in different lineages of rhinocerotids and in lhpirus. Klaits, however, failed to distinguish between primitive and derived morphologies, or else assumed that Tapirus possessed the primitive condition. Klaits (1972) identified a suite of carpal characteristics that she believed were responsible (at least in part) for the differences one sees between living tapirs and rhinoceroses in the motion of the wrist during walking. Tapirs move the manus in an orthal plane (i.e. the manus only swings anterad and posterad), whereas living rhinoceroses rotate the manus ectally before placement on the substrate and bring the manus back entally as it is lifted. This difference in movement pattern appears to be independent of the number of functional digits, since Klaits (1972) observed a morphology similar to that of living rhinoceroses in the tetradactyl fossil rhinocerotid Aceratherium. Klaits considered the rhinoceros morphology and movement to be derived from that of the tapir, but if one follows the evolution of Klaits suite of characters through the fossil record, it becomes apparent that her interpretation is not entirely correct. The following discussion of tapiromorph carpal morphology will point out some of the evidence relevant to Klaits ideas.

30 ~ ~~ 30 L. T. HOLBROOK The shape of the radial facet of the scaphoid provides an important clue about motion of the manus, because whether or not rotation of the manus takes place during a step is largely determined by the articulation between the scaphoid and lunar. The emarginated lunar contact seen on the radial facet of the scaphoid of npirus allows for the two bones to mutually embrace each other in such a way that the lunar is locked into a position that allows it to move only proximodistally (Klaits, 1972). In rhinocerotids (and, presumably, amynodontids and Uintacems), the extended, straight lunar contact on the radial facet of the scaphoid indicates a different sort of embrace, one that allows the lunar to rotate within a socket formed by the scaphoid (Klaits, 1972). Both of these morphologies are derived; the primitive condition, a flat fade with a short, straight lunar contact, is observed in other tapiromorphs and Hyracotherium. The absence of contact between the distal facets of the lunar is a derived feature of lhpirus. Contact between the trapezium and second metacarpal is probably a primitive tapiromorph feature, but until the condition of this is known in more taxa it will be difficult to assess the significance of loss of this contact. Metacarpals and phalanges Description. Tapiromorphs possess either three or four functional digits (i.e. possessing phalanges). In tridactyl forms, a vestigial fifth metacarpal is usually present. A tridactyl manus is known to be present in chalicotheriids, Colodon, Lophialetes, Schlosseria, Triplopus, Hyracodon, and Paraceratherium. Most rhinocerotids are also tridactyl, but some, including Trigonias and a few specimens of Neogene aceratheres, are tetradactyl. Among the early rhinocerotids considered here, Subhyracodon is known to be tridactyl, and Hanson (1989) has argued for tridactyly in Teletaceras on the basis of the shape of metacarpal IV. The distal articulations of the metapodials are fairly conservative among tapiromorphs, being smooth, somewhat cylindrical articulations with a posterior median keel. The only notable morphological variation in tapiromorph phalanges concerns the ungual phalanges. In all taxa, the ungual phalanges are notched at the tip. In most taxa, the unguals are flattened, as for hooves, but in chalicotheriids they are mediolaterally compressed claws. The ungual phalanges of eomoropids have not been described. Table 4 lists the metacarpal proportions for a variety of tapiromorphs. Discussion. Perissodactyls are sometimes called mesaxonians, referring to the mesaxonic symmetry of their metapodials. In other words, the axis of symmetry able 4. Metapodial proportions of tapiromorphs Genus Metacarpal I11 Metatarsal I11 ~ Hyrachyidae Hymhyus Amynodontidae Amynodon Hyracodontinae 1Fiplopus Hyracodon Indricotheriinae Juxia Pammtherium Rhinocerotidae Subhyracodon Dicerorhinus Rhinoceros Uintaceras qw 4.64 (n=5) 4.91 (n= 1) 6.98 (n=2) 5.33 (n = 2) 5.02 (n = 2) 3.52 (n = 3) 3.11 (n=2) 3.00 (n = 1) 2.99 (n = 2) 3.42 (n = 1) qw 4.46 (n = 7) 2.80 (n = 1) 5.39 (n=2) 4.83 (n = 1) 4.57 (n=4) 3.40 (n = 3) 3.75 (n=2) 2.83 (n = 1) 2.89 (n = 2) 3.14 (n = 1) of the mesaxonic foot passes through the third digit. The third metapodial is commonly enlarged, since it bears the most weight. Mesaxonic symmetry is probably a primitive feature, since a similar pattern is seen in the hands and feet of a number of mammalian groups (including in the hands of humans). The enlarged third metapodial is also seen in phenacodonts, whose fore and hind feet are also mesaxonic. Artiodactyls are paraxonic, i.e. the axis of the foot passes between the third and fourth digits, which are asymmetrical. This condition is seen in the earliest artiodactyls, including Diacodexis. In the fore feet of Diacodexis pukistanensis (Thewissen & Hussain, 1990) and Hypertragulus (Scott, 1940), however, paraxonic symmetry is not as well developed as in other artiodactyls. In these two taxa, the third metacarpal is longer than the fourth, which may be evidence of the primitive mesaxonic symmetry of their ancestors. A tetradactyl manus is probably primitive for perissodactyls. Rose (1996) claims that the medially-bevelled proximal end of the second metacarpal in Homogalax may indicate the presence of a vestigial first digit, but the evidence is not conclusive. Deperet (1903) believed that the second metacarpal of Lophwdon indicated the presence of the first digit, but, again, there is no conclusive proof that the manus of Lophiodon or any other perissodactyl was pentadactyl (Matthew, 1917). Loss of a functional fifth manual digit has occurred several times in perissodactyl evolution. This trait is seen in many equids and almost certainly evolved more than once in the Tapiromorpha, including within the Chalicotheroidea, at least once among tapiroids, and probably two or more times among rhinocerotoids. This pattern of multiple parallelism indicates that number

31 of manual digits is a poor indicator of phylogenetic relationships. Nevertheless, tridactyly has been used as a synapomorphy for the Hyracodontidae semu Radinsky (1966a, 1967a) (Lucas et al., 1981; Lucas & Sobus, 1989; Prothero et al., 1986, 1989). This synapomorphy is further refuted by the presence of a tetradactyl manus in Juxia. Hanson (1989) argued that tridactyly is primitive for rhinocerotids, based on the inferred tridactyly of Teletaceras. Cerdeiio (1995) scored Teletaceras as unknown for this character (tables 2 and 3, character 48: Metacarpal V: functional (0), reduced (l) ), but her results suggest that either tridactyly evolved multiple times within the Rhinocerotidae, or reversals to the tetradactyl condition occurred in Trigonias and the subfamily Aceratheriinae. The presence of four functional digits in the manus of Uintaceras, which is thought to be closely related to rhinocerotids (Holbrook & Lucas, 1997), favours a tetradactyl manus as primitive for rhinocerotids. Another character that has been used to support a relationship between hyracodonts and indricotheres is metapodial proportions. The primitive condition for tapiromorphs (and perissodactyls in general) appears to be relatively long metapodials, which may be further elongated (as in Triplopus) or shortened (as in rhinocerotids) (Table 4). Some workers (Lucas et al., 1981; Lucas & Sobus, 1989; Prothero et al., 1986) have argued that relatively long metapodials are a derived feature shared by the small cursorial hyracodonts and indricotheres. This argument seems tenuous for two reasons. First, the metapodials of indricotheres are not particularly gracile; in fact, their proportions are rather similar to those of other large rhinocerotoids. Second, elongate metapodials, such as those seen in Hyracodon, are not a derived feature but a primitive feature, typical of many early perissodactyls, including Hyracotheriu m and Ho moga lax. Osborn (1913) and Colbert (1934) claimed that the distal articulations on the metapodials of all chalicotherioids were specialized in having distinct sesamoidal and phalangeal surfaces. Radinsky (1964) contested this claim, stating that the distal articular surfaces on the metapodials of eomoropids were not different from those of other Eocene perissodactyls, a position that is supported here. Fischer (1986) claimed that the notched tips of the ungual phalanges of fossil perissodactyls were evidence of a derived hoof structure shared with hyracoids. Novacek (1992b) claimed that Hyracotherium possessed unguals without a notch, but Cope (1884; plate XLIXb) figured a forefoot of Hyracotherium with a notched ungual phalanx on digit 111. In any case, notched unguals appear to be characteristic of perissodactyls, since they are found among tapiromorphs and brontotheres. TAPIROMORPH OSTEOLOGY 31 The clawed unguals of chalicotheres are their most distinctive feature. Although claws are primitive for eutherians, it is more likely that the fissured claws of chalicotheriids are derived from a hoofed ancestor. PELVIC LIMB Innominate Description. The ilium of tapiromorphs possesses a concave anterior iliac crest (Fig. 12B), giving the impression of two wings (the tuber sacrale and the tuber coxae) attached to the shaft leading to the acetabulum. In Colodon, Juxia, and Subhyracodon, a small tuberosity is present on the ventral face of the tuber sacrale. A pair of tuberosities may be present on the medioventral aspect of the iliac shaft, one near the acetabulum for the origin of the rectus femoris. The acetabulum bears a distinct fossa that extends to its ventral margin and accommodated the ligamentum teres in life. Anterior to this fossa, a notch may be present in the ventral margin of the acetabulum, indicating the contact between the ilium and pubis. On the dorsal side of the innominate, another process lies posterior to the acetabulum, the ischial spine. This process may appear as a simple tuberosity, as in Isectolophus, Tapirus, and Hyrachyus, or it may be an elongated crest, as in Colodon, Juxia, Subhyracodon, and Trigonias. The pubic symphysis is relatively short in many tapiromorphs, such as in Heptodon, Tapirus, Hyrachyus, and somewhat longer in large taxa like rhinocerotids. The ischium is a narrow blade posteriorly, but in some taxa it may be expanded dorsally into a tuberosity, as in Colodon. Discussion. A concave crista iliaca is primitive for tapiromorphs, as this condition is also seen in equids, brontotheres, and Hallensia (Franzen, 1990). Franzen (1989) argued that the convex crista iliaca of Palaeotherium was a primitive character that excluded the Palaeotheriidae from the Equidae. Most mammals, including the non-perissodactyl outgroups studied here, possess the convex condition (Fig. 12A), indicating that this condition is indeed primitive for perissodactyls. Franzen s (1990) phylogeny united palaeotheriids with Hallensia and equids, requiring parallel evolution of the concave condition in tapiromorphs. Another possible explanation is that the Palaeotheriidae is the sister-taxon to the rest of the Perissodactyla, a view supported by Butler s (1952) work on dental evolution of perissodactyls. A short pubic symphysis is a primitive mammalian feature and is seen in primitive perissodactyls and phenacodonts. A long pubic symphysis is characteristic of artiodactyls, including Diacockxis (Thewissen &

32 32 L. T. HOLBROOK aic aic tc \ A Figure 12. Innominate bones of Meniscotherium (A) and Subhyracodon ( 3) in dorsal view. (A modified from Williamson & Lucas, 1992; B modified from Scott, 1941.) Abbreviations: aic: anterior iliac crest. Hussain, 1990). The artiodactyl condition is produced by fusion of the symphysis at two points along the midline, usually leaving an elongate depression between the sites of fusion. In tapiromorphs, there is no evidence that the pubic symphysis has been produced by fusion at more than one point. Femur Description. The femoral head is more or less spherical in all of the taxa considered here. The only distinctive feature of the head that varies at all is the position of the fovea capitis. In the vast majority of perissodactyls, the fovea forms an indentation of the margin of the head. In those chalicotheres where it is known, the fovea is positioned more centrally, away from the head margin. This morphology is also seen in Lophiodon (Filhol, 1888). The fovea is weak in Mompus, and in Chalicotherium it is absent (Zapfe, 1979). The femora of Eomompus and Litolophus are unknown. The greater trochanter is roughly conical, with the apex pointing proximally, and is rugose proximally. The height of the greater trochanter relative to the head varies among tapiromorphs. In Homogalax, Isectolophus, and Helaletes, the rugose portion of the greater trochanter lies entirely above the head. In Hyrachyus, Tbpirus, and Heptodon, the distal margin of the rugose area may lie as far down as the distal margin of the head, although the proximal greater trochanter still extends beyond the head. In Hyracodon, amynodonts, and primitive rhinocerotids, the greater trochanter extends no further than the head, and the distal margin of the rugose portion may lie below that of the head. In chalicotheres andhphiodon, the proximal portion of the greater trochanter lies well below the level of the head. The lesser and third trochanters are positioned on tt \ It -h tt Figure 13. Femora of HeZaZetes (AMNH19226; left femur, reversed) (A) and Subhyracodon (right femur) (B) in anterior view. (A modified from Radinsky, 1965a; B modified from Scott, 1941.) Abbreviations: gt: greater trochanter; h head; It: lesser trochanter; tt: third trochanter. the medial and lateral sides, respectively, of the proximal half of the femoral shaft. They primitively appear as broad flattened processes, with the lesser lying slightly more proximal relative to the third. The two trochanters can be approximately equal (e.g. as in Heptodon) (Fig. 13A), or the lesser can be significantly smaller. In rhinocerotids, the lesser trochanter is greatly reduced (Fig. 13B). In Lophiodon, both the B

33 lesser and third trochanters are relatively reduced (Filhol, 1888). Distally, the patellar groove is bounded by a pair of keel-like eminences. In Heptodon, Helaletes, Colodon, Hyrachyus, Triplopus, and Hyracodon, these eminences are symmetrical, but in some other tapiromorphs the medial eminence is enlarged. In Zzpirus, Lophialetes, and Depemtella, the medial keel is simply enlarged relative to the lateral one. In rhinocerotids and Paraceratherium the medial eminence is expanded into a tuberosity (Fig. 13B). Of the distal condyles of the femur, the lateral condyle is somewhat larger than the medial. On the posterior side, the lateral condyle grades into the shaft, whereas the articular surface on the medial side is distinctly separated from the shaft proximally by a groove. Also on the posterior side, a fossa for the gastrocnemius muscle is present proximal to the lateral condyle in Heptodon, Helaletes, Colodon, Lophialetes, Tbpirus, Hyrachyus, Hyracodon, and rhinocerotids. This fossa is absent in Paraceratherium, Metamynodon, Lophiodon, and Mompus. Discussion. The presence and position of the fovea capitis varies among mammals, but the primitive position for the fovea of perissodactyls appears to be on the margin of the articular surface of the head, as this condition is seen in non-tapiromorph perissodadyls. Thewissen & Domning (1992) scored the marginal position of the fovea as a derived feature of perissodactyls and hyracoids. The central position of the fovea in chalicotheres and lophiodonts may be a synapomorphy of these taxa, but data from eomoropids are needed to test this idea. A relatively high greater trochanter appears to be primitive for tapiromorphs, as it is also seen in Hyracotherium, consistent with the idea of cursoriality being primitive for tapiromorphs, if not for all perissodactyls. An argument against the latter case has been made by Franzen (1989), who claimed that the low greater trochanter seen in Palaeotherium and Plagiolophus is a primitive character, as it is similar to the condition in Phenacodus. This is also true of Diacodexis (Rose, 1985), the Paleocene condylarth Pleuraspidotherium (Thewissen, 1991), and brontotheres, although the femur is not known in Eotitanops (or, for that matter, in Lambdotherium). A flat elongated lesser trochanter is characteristic of perissodactyls, as well as Phenacodus and Meniscotherium. In many other mammals, including Diacodexis and Pleuraspidotherium, the lesser trochanter is an unflattened knob placed proximally. In non-perissodactyls, the lesser trochanter is more proximally positioned than the third trochanter. The reduction of the lesser trochanter may be a rhinocerotid synapomorphy. TAF IROMORPH OSTEOLOGY 33 Presence of the third trochanter is probably primitive for eutherians, but the size of the third trochanter may be of phylogenetic significance. Reduction or loss of the third trochanter is characteristic of artiodactyls. The expansion of the medial trochlear ridge may also be a synapomorphy for rhinocerotids and Paraceratherium, but this morphology also occurs in derived equids and bovids. Hermanson & MacFadden (1996) recently discussed the evolution of this morphology in equids, which they believe is associated with the passive stay-apparatus of the hindlimb in equids. Hermanson & MacFadden (1996) claimed that this morphology did not appear in rhinocerotids until the Miocene. Contrary to their statements, the expanded medial trochlear ridge is present in the earliest unequivocal rhinocerotids whose femora are known, including Trigonias and Subhyracodon. Rhinoceroses are not known to possess a passive stay-apparatus, and it is unclear what the functional significance of this morphology is. A distinct fossa for the gastrocnemius muscle is found in all extant perissodactyls, as well as in artiodactyls. This trait is very rare, however, in other mammals and is absent in cursorial members of other orders (e.g. Carnivora, Rodentia). (A similar fossa occurs on the femur of the armadillo Dasypus, but it does accommodates the origin of the superficial digital flexor, not the gastrocnemius. The fossa is absent in at least one other armadillo, Tolypeutes [T. Koneval, pers. comm.].) The fossa is absent in brontotheres, hyraxes, Phenacodus, and Tetraclaenodon, indicating that presence of the fossa is derived. It is likely that the fossa was secondarily lost in Paraceratherium and Metamynodon, but it is less clear whether the condition observed in Mompus and Lophiodon is a primitive retention or a reversal. Patella Description. The patella of most tapiromorphs is roughly teardrop-shaped and is anteroposteriorly thick (Fig. 14A). The two posterior facets for the patellar groove of the femur are subequal in size. In rhinocerotids and Paraceratherium, the patella is broad and flattened and has a process extending from the medial side, a condition also seen in Equus. In Paraceratherium, the medial facet extends onto this process, but this does not occur in rhinocerotids (Fig. 14D,E). Lophialetids and deperetellids also have flattened patellae, but the patella is proximodistally elongate and coffin -shaped in outline (Fig. 14B,C). Discussion. The unflattened condition is primitive and found in Hyracotherium. Heissig (1989) proposed that a flattened patella is a synapomorphy of rhinocerotids

34 34 L. T. HOLBROOK A B Figure 14. Patellae of Uintaceras (CM12004, right) (A), Lophiaktes (AMNH81790, right) (B), Deperetelh (AMNH 81833, right) (C), 12.igonb.s (AMNH 13226, and Paraceratherium (AMNH 26189, left, reversed) (E) in posterior view. Scale bar = 1 cm. and indricotheres, but two facts conflict with this hypothesis. The first is that the condition of the medial facet is different in the two taxa. The second problem is that the primitive condition is seen in Uintaceras, which appears to be closely related to rhinocerotids. The condition of the patella in indricotheres other than Paraceratherium is not known. The condition observed in lophialetids and deperetellids may be a shared derived character uniting these taxa. Tibia and fibula Description. Two flat facets for the femoral condyles dominate the proximal end of the tibia. The two intercondyloid tubercles lie between the facets, separated by a groove. In most tapiromorphs, the lateral tubercle is taller than the medial, but the opposite is true in 1965a). The shaft is generally triangular in cross-section, especially proximally. A smooth groove for the middle patellar ligament lies on

35 the proximal end of the cnemial crest, which is fairly prominent and varies little among tapiromorphs. The distal end bears deep facets for the astragalar trochlea. The medial malleolus is divided into anterior and posterior parts. The fibula, in those taxa for which it is known, is a simple bone with expanded ends for articulation with the proximal tibia and distally with the tibia and astragalus. The fibula is fused to the tibia in Colodon and Deperetella. Discussion. Kitts (1956) noted that the medial intercondyloid tubercle is higher than the lateral in equids later than Mesohippus. Radinsky (1965a) suggested that this change in the evolution of both horses and tapirs was related to enlargement of the lateral epicondyle of the femur and would help prevent lateral dislocation of the femur. Tarsus Description. Seven (or in some cases fewer) elements comprise the perissodactyl tarsus: the astragalus, calcaneum, cuboid, navicular, ectocuneiform, mesocuneiform, and entocuneiform. The proximal trochlea of the astragalus, in tapiromorphs, is formed by two prominent keels separated by a deep groove. In some taxa, the posterodistal portion of the lateral keel has a prominent lateral extension. In many taxa, the trochlea lies almost directly above the neck and distal head (Fig. 15B). In rhinocerotoids, including Hyrachyus, hyracodontids, most indricotheres, amynodontids, rhinocerotids, and Uintaceras, the trochlea is laterally offset (Fig. 15C). This condition is heightened by the presence of a prominent tubercle on the distomedial neck. Juxia is a notable exception, as it possesses the tapiroid condition. On the posterior (plantar) aspect of the astragalus are three facets for the calcaneum: the proximal calcaneal, sustentacular, and distal calcaneal facets. The proximal calcaneal facet is large and concave in all taxa. The sustentacular facet is a flat oval, whereas the distal calcaneal facet takes one of two forms in tapiromorphs. This latter facet can be a small triangular articulation on the distomedial corner of the posterior aspect of the astragalus, as in Homogalax, Eomompus, Mompus, Heptodon, Helaletes, Colodon, Lophialetes, Schlosseria, Deperetella, and Juxia (Fig. 15E). In the second condition, this facet extends laterally as a strip and is confluent with the distal part of the sustentacular facet. This condition is seen in Hyrachyus, hyracodontids, amynodontids, rhinocerotids, Uintaceras, and some Tapirus (Fig. 15D). In Hyracodon and Triplopus, the contact between the two TAPIROMOFPH OSTEOLOGY 35 facets is exaggerated as a ridge (Osborn, 1890). The distal astragalar facet is absent in Paraceratherium. The saddle-shaped navicular facet, the classic perissodactyl synapomorphy (Radinsky, 1966b), varies little among tapiromorphs. There is, however, considerable variation in the presence or absence of an anterodistal contact with the cuboid, i.e. one that precludes contact between the calcaneum and navicular, which is discussed below. Both contacts can in fact be absent, and a distinct anterior astragalocuboid contact is present in Homogalax, Borrisiakia, Chalicotherium, Lophialetes, Deperetella, Tapirus, Hyrachyus, Hyracodon, amynodontids, Juxia, Paraceratherium, and rhinocerotids. This contact is absent in Colodon, Helaletes, Heptodon, Mompus, Eomompus, and Litolophus. The calcaneum possesses a moderately long tuber calcis, the end of which is usually slightly expanded. Two pits can be found on the shaft near the astragalar facets. One pit is situated just proximal to the lateral astragalar facet and accommodates the fibular malleolus in extreme flexion (Radinsky, 1965a). This pit is known in Homogalax, Eomompus, Litolophus, Heptodon, Helaletes, Lophialetes, Schlosseria, Hyrachyus, Triplopus, Hyracodon, Juxia, Teletaceras (Hanson, 1989), and Uintaceras; it is absent in Tapirus, Deperetella, Urtinotherium, Paraceratherium, and rhinocerotids besides Teletaceras. The second pit lies lateral to the first and accommodated the insertion of the short lateral ligament (Radinsky, 1965a). The distribution of this pit matches that of the fibular pit, with the following exceptions: present in Paraceratherium; absent in Lophialetes and Schlosseria. Hanson (1989) did not describe this pit in Teletaceras, but its absence in this taxon cannot be confirmed. On the lateral side of the calcaneal shaft, at the distal end, a low tuberosity, the peroneal tubercle, is present in some taxa, namely Homogalax, Cardiolophus, Tapirus, Deperetella, Hyrachyus, Juxia, Uintaceras, amynodontids, and rhinocerotids. The tubercle is absent in Eomompus, Litolophus, Heptodon, Helaletes, Lophialetes, Schlosseria, Triplopus, Hyracodon, Urtinotherium, and Paraceratherium. Three facets on the calcaneum articulate with the astragalus: the lateral astragalar, sustentacular, and distal astragalar facets. The lateral astragalar facet consists of two parts that meet at approximately right angles and correspond to the lateral calcaneal facet of the astragalus. In Homogalax, the lateral astragalar facet is more rounded, i.e. the parts meet at a more obtuse angle. The sustentacular facet varies little among tapiromorphs; the distal astragalar facet is confluent with a facet for the navicular in those taxa where a calcaneonavicular contact is present. Absence of calcaneonavicular contact has been observed in Homogalax (Rose, 1996), Caidiolophus (Gin-

36 36 L. T. HOLBROOK cal av I sust B mes 2 Figure 15. Pedal elements of some tapiromorphs and other perissodadyls. (A). Tarsus of Hyracotherium. Calcaneum and navicular are shaded. (Modified from Kim, 1956.) Abbreviations: cak calcaneum; nav: navicular. (B,C), Astragali of Tbpirus (E3) and Subhyracodon (C) in anterior view. (E3 after Radinsky, 1965a; C after Scott, 1941.) (D,E), Astragali of and Pmpulaeothrium (E) in posterior (plantar) view. Sustentacular and distal calcaneal facets are shaded. (D modified from Radinsky, 1965a; E modified from Franzen, 1990.) Abbreviations: dcf: distal calcaneal facet; sust: sustentacular facet. (F,G), Entocuneiforms of Deperetelkz 0 and Heptodon (G), in anterior view. Mesocuneiform and navicular facets are shaded. (Both modified from Radinsky, 1963b.) Abbreviations: mes: mesocuneiform facet; nav: navicular facet. gerich, 1991), Lophiodon (Deperet, 1903), Litolophus (Radinsky, 1964), Mompus (Coombs, 1978a), Borissiakia (Borissiak, 1946), Chulicotherium (Zapfe, 1979), Lophialetes (Reshetov, 1979), Deperetella, Rpirus, Hyrachyus, hyracodontids, indricotheres, amynodontids, and rhinocerotids. This contact is present in Heptodon (Radinsky, 1965a), Eomompus (Osborn, 1913; Radinsky, 1964), HeZuktes, and Colodon sust (Fig. 15A). Radinsky (1966b) described evidence or this contact on naviculars of Lophialetes, but, as discussed below, facets alone cannot be used to infer the presence or absence of this contact. The main distal articulation on the calcaneum is the crescent-shaped or oval facet for the cuboid, which varies little in morphology among tapiromorphs. The navicular is fairly conservative in morphology

37 among tapiromorphs, mainly varying in its proximodistal height. The navicular tends to be higher in smaller taxa. Its proximal surface does bear the saddleshaped facet for the astragalus that is characteristic of perissodactyls. The navicular articulates with the cuboid laterally, and, in some cases, described above, with the calcaneum proximolaterally. In Homogalax, Hyrachyus, and Lophialetes, a small facet may be present on the anterolateral corner of the proximal face, that is similar to the facet for the calcaneum seen in taxa possessing a calcaneonavicular contact. In these three taxa, which do not possess a calcaneonavicular contact, this facet does not normally articulate with the calcaneum but with either the cuboid or nothing at all. The distal surface of the navicular bears facets for each of the cuneiform bones. The ectocuneiform is the largest of the three cuneiform bones found in tapiromorphs. Its proximal and distal surfaces bear large facets for the navicular and third metatarsal, respectively, and it articulates via smaller facets with the cuboid laterally and the mesocuneiform medially. The distal facet, like the proximal, is essentially flat, but a small process extends distally from the most posterior point of the distal facet. This latter trait was previously described as a potential synapomorphy of amynodontids, rhinocerotids, and Uintaceras (Holbrook, 1999), but other ceratomorphs, such as Hyrachyus and Colodon, have an ectocuneiform that is slightly concave distally and approaches the condition described seen in the aforementioned rhinocerotoids. The morphology of the mesocuneiform is conservative among tapiromorphs, but there are some phylogenetically interesting features of the entocuneiform. The entocuneiform bears facets for the navicular, mesocuneiform, second metatarsal, and the vestigial first metatarsal. It articulates with these bones in such a way that it projects posterolaterally, behind the tarsus, so that the small first metatarsal lies behind and articulates with the third metatarsal. The facets for the navicular and mesocuneiform are adjacent on the anteromedial aspect, but their exact relationship varies among tapiromorphs. These facets may lie on the anterior edge of the bone, as in Heptodon and Tapirus (Fig. 15F), or the mesocuneiform facet may lie behind (i.e. posterolateral to) the navicular facet (Fig. 15G), as in Deperetella, Hyrachyus, Juxia, rhinocerotids, and Uintaceras. The three functional metatarsals articulate proximally with the distal ends of the mesocuneiform, ectocuneiform, and cuboid. The ectocuneiform bears a distal facet for the third metatarsal, but it may also bear medial and lateral facets for the second and fourth metatarsals, respectively, as in Homogalax, Eomompus, Heptodon, Colodon, some Tapirus, and Paraceratherium. Other taxa lack one or both of these TAF IROMORPH OSTEOLOGY 37 latter contacts. Helaletes, Lophialetes, Schlosseria, Hyrachyus, Amyrwdon, Metamynodon, Hyracodon, rhinocerotids, and Uintaceras all lose the contact between the ectocuneiform and fourth metatarsal, but retain the the contact with the second metatarsal. This latter contact is also lost in Deperetella, Triplopus, andjuxia, as well as extant rhinocerotids. A contact between the cuboid and metatarsal I11 is seen in Metamynodon. Chalicotheriids show variation in the presence or absence of these contacts, but this may be a consequence of their derived foot structure. In Tapirus, an articulation with metatarsals I1 and IV is present on the ectocuneiform of T indicus, but the articulation with metatarsal IV is absent or reduced in the other three species. In fact, in adult I: pinchaque and some specimens of T terrestris, metatarsal I11 has a small articulation with the cuboid. Discussion. The offset position of the astragalar trochlea relative to the neck is a derived character that may be used to unite Hyrachyus with other rhinocerotoid families. Another character of the astragalus that may also be a rhinocerotoid synapomorphy (including Hyrachyus) is the confluence of the sustentacular and distal calcaneal facets. The ridge produced by this confluence in Hyracodon and Triplopus may be a hyracodontine synapomorphy. The main exception to these synapomorphies is Juxia, which possesses the same condition seen in non-rhinocerotoid tapiromorphs, brontotheres, and equids. Calcaneonavicular contact or its absence has been a character of uncertain polarity for some time. The presence of this contact in Hyracotherium and Heptodon suggested that this condition was primitive for perissodadyls, but the absence of contact in Homogalax (Rose, 1996), Cardiolophus (Gingerich, 1991), brontotheres (Osborn, 1929), and phenacodonts (Williamson & Lucas, 1992; Thewissen, 1990) makes it likely that the latter condition is the primitive one. The presence of pits on the calcaneum for the distal fibula and the short lateral ligament are primitive perissodactyl features. The presence of the peroneal tubercle is also a primitive feature. Rose (1996) noted that this feature is present in Hyracotherium, contrary to the statement of Thewissen & Domning (1992). The presence of a false calcaneal facet on the navicular of some taxa may indicate that the calcaneum and navicular can come into contact during certain movements. This may be particularly true of Hyrachyus, since the confluence of the sustentacular and distal calcaneal facets could allow for greater mobility of the astragalus relative to the calcaneum. The presence of a posterodistal process of the ectocuneiform was used by Holbrook (1999) to support a sister taxon relationship between Amynodontidae and Fthinocerotidae (including Uintaceras). The removal of

38 38 L. T. HOLBROOK this character from the analysis results in a strict consensus tree where these two families form a polytomy with Forstemperia, Juxia, and Pamceratherium. The arrangement of the facets on the entocuneiform is an interesting character, but unfortunately it is not known for enough taxa to assess its polarity adequately. The orientation of the entocuneiform and first metatarsal is distinctive among perissodadyls. In pentadactyl taxa, like Phenacodus, the entocuneiform lies at the medial end of the distal tarsal row and does not wrap around posteriorly. The first metatarsal, like the other metatarsals, is oriented proximodistally. Hyracoids lack even a vestige of the first pedal digit, but the entocuneiform of Recent hyracoids is positioned and oriented in a manner similar to that OfPhenacodus. This difference in the orientation of the entocuneiform between hyracoids and perissodactyls suggests that the reduction and/or loss of the first pedal digit happened independently in these two groups. This goes against Fischer s (1986, 1989; see also F rothero et al., 1988) claim that the tridactyl pes of these two groups is synapomorphic. The articulation of three metatarsals with the edocuneiform is primitive for perissodadyls, but the significant variation observed within the genus lbpirus provides some reason to believe that the loss of one or more contacts is likely to have arisen more than once, or, alternatively, to have secondarily reappeared. Metatarsals and phalanges Description. In all tapiromorphs for which the pes is known, only three functional digits are present in the hindfoot. Radinsky (1963b) described a vestigial first metatarsal in a number of perissodactyls, and this vestige appears to be present in most tapiromorphs. This vestige occupies a central position posterior to the proximal ends of the other metatarsals; it articulates mainly with the entocuneiform, but may also contact the third metatarsal. Rose (1996) inferred the presence of a first digit in Homogalax from the medial beveling of the proximal metacarpal 11, and Radinsky (1963b) came to the same conclusion based on a facet on the entocuneiform of this genus. It is not known, however, whether or not this digit was functional. Radinsky (1963b) claimed that the entocuneiforms of Eomompus and Litolophus also bear facets for metatarsal I. Radinsky (1963b) stated that in chalicotheriids and Lophiodon, the first metatarsal and entocuneiform are absent, but Coombs (1978a) provided evidence for the presence of an entocuneiform (but not of a fist metatarsal) in the chalicotheriids Mompus and Ancylotherium. Other tapiromorphs for whom a first metatarsal is known include (from Radinsky, , and pers. obs.): Heptodon, Helaletes, Coloclon, npirus, Schlosseria, Depemtella, Hyrachyus, Hyracodon, and rhinocerotids. In some taxa, namely Coloclon, Hyracdn, and rhinocerotids, the first metatarsal may be fused to the entocuneiform. The entocuneiform of Uintacems bears a facet for the first metatarsal, but no such metatarsal has been identified with certainty. It is possible that an elongate posterior process on the second metatarsal actually represents the fist metatarsal, which has become fused to the second metatarsal. Such a process is not observed on the second metatarsals of other tapiromorphs. The ungual phalanges of the pes are similar to those of the manus. They are notched at the tip and flattened in most taxa, whereas in chalicotheriids they are fissured claws. Table 4 lists the proportions of the metatarsals for a variety of tapiromorphs. Discussion. Wortman (1896) claimed that he had observed evidence of a fifth digit in Systemodon (= Homogalax), and Kitts (1956) claimed that a vestigial fifth digit is present in Hyracotherium, but no actual bone has been identified as a fifth metatarsal in these taxa. There is no direct evidence for the existence of a first metatarsal in early equoids and brontotheres, but the presence of this metatarsal is certainly primitive for tapiromorphs. Whereas the first metatarsal no longer functions as part of a digit in perissodactyls, it does act as a brace in the tarsus and provides attachment for deep flexor muscles of the foot, the contrahentes (Radinsky, 1963b). The hindfeet of most of the non-perissodactyl outgroups studied are functionally pentadactyl, including those of phenacodonts and tethytheres. The pes of hyracoids, however, is tridactyl, and Fischer (1986, 1989; see also Prothero et al., 1988) has used this trait to unite hyracoids with perissodactyls. The presence of a fifth metatarsal in some perissodactyls, as claimed by Kitts (1956) and Wortman (1896), would demonstrate that the condition in most perissodactyls and hyracoids is convergent, but more information is needed on the pes of Hyracotherium and Homogalax. (See also the discussion of the entocuneiform above.) Limb proportions Description. Table 5 compares the limb proportions of various tapiromorphs. Some variation exists in the relative lengths of different limb segments among tapiromorphs. The variation in metapodial proportions has been discussed with other features of the metacarpals and metatarsals; this section will focus on the relative length of long bones. The lengths discussed

39 IBble 5. Limb proportions of tapiromorphs. Data on Paraceratherium are from Granger & Gregory (1936); data on chalicotheres are from Coombs (1983). Indices are defined in the text (under Limb proportions) Genus Brachial index Tapiridae Rpirus 89.9 Helaletidae Helaletes 95.7 Hyrachyidae Hyrachyus 92.4 Amynodontidae Amynodon 94.7 Hyracodontidae Triplopus 134 Hyracodon 105 Indricotheriidae Paraceratherium 124 Rhinocerotidae Subhyracodon 94.8 Dicemrhinus 90.2 Chalicotheriidae Mompus 96.3 Chalicotherium Incertae sedis Uintaceras 80.7 Crural index Intermembral index here are the lengths from proximal to distal articular surfaces, i.e. the measurements for the humeral, radial, and femoral lengths are taken from the most proximal point of the head to the most distal extent of the distal articulation. For the tibia, the measurement is taken from the tip of the intercondyloid tubercles to the distal extent of the malleolus. The ratio of the length of the radius to the length of the humerus, multiplied by 100, is the radiohumeral (or brachial) index. For most tapiromorph taxa, this index ranges between 90 and 100. The only taxa that exceed a value of 100 are Heptodon, Triplopus, Hyracodon, and Paraceratherium (measurements for Paraceratherium from Osborn, 1923). Due to the small sample sizes measured, these values may not be significantly higher than those for other tapiromorphs. The radiohumeral index of Uintaceras does appear to be unusually low for tapiromorphs. The tibiofemoral (or crural) index is the ratio of the tibia length to the femoral length, multiplied by 100. An index greater than 100 is seen in Homogalax (based on measurements of Rose, 1996), Heptodon, Helaletes, TAPIROMORPH OSTEOLOGY 39 and Colodon. Indices less than 100 are characteristic of larger taxa, including chalicotheriids, Hyrachyus, amynodontids, and rhinocerotids. Chalicotherium possesses an unusually low index of 57. Accurate femoral lengths are not available for Uintaceras, but the tibia is clearly much shorter than the femur in this genus. The same is true of Paraceratherium, for which Granger & Gregory (1936) published a tibiofemoral index of 66. The intermembral index is the ratio of the sum of the radial and humeral lengths to the sum of the tibia1 and femoral lengths, multiplied by 100. In most tapiromorphs this index is less than 100, indicating that the forelimb is clearly shorter than the hindlimb. The only exceptions are chalicotheriids, where the index can get as high as 140, as seen in Chalicotherium. Discussion. It is difficult to treat these data in a manner that can be utilized in a phylogenetic analysis, but they still reveal some interesting patterns. Regarding the brachial index, the differences between genera are not great, but a relatively long radius is probably derived. This interpretation would be consistent with the idea that indricotheres are related to the hyracodonts. The crural index of Paraceratherium, however, is more similar to that of rhinocerotids and chalicotheres. Given that the mode of life for this genus was probably very different from that of other tapiromorphs (with the possible exception of chalicotheres), it seems unwise to use isolated limb proportions to support alternative phylogenetic hypotheses. Indeed, greater similarity to Puraceratherium in limb proportions is seen in the bear Ursus and the extinct South American large clawed herbivore Homalodotherium than among other tapiromorphs (Coombs, 1983). One other pattern in the limb proportions of tapiromorphs that can be noted is the increase in the intermembral index that is seen in rhinocerotoids and chalicotheres. Primitively, the fore limb of tapiromorphs is shorter than the hind limb. This is also seen in Hyracotherium. In rhinocerotids and amynodontids, the fore and hind limbs are of a more equal length, and in chalicotheres, the fore limb is actually longer. This increase in the relative length of the fore limb may be a consequence of increased body size, although Paraceratherium and Tapirus do not follow such a pattern. PHYLOGENETIC ANALYSIS Characters of tapiromorph cranial and postcranial osteology are used here as primary data for examining tapiromorph phylogeny.

40 40 L. T. HOLBROOK METHODS Characters were drawn from the observations described above. Characters were scored for a selection of ingroup and outgroup taxa (Table 8). Missing data can provide a source of great uncertainty in computer analyses, and analyses that include taxa that are very incompletely known often give ambiguous results, i.e. numerous equally parsimonious trees. In order to avoid this problem, taxa were chosen for this analysis largely on the basis of completeness. As a consequence, the taxa analysed largely represent North American tapiromorph diversity, because data could be reliably gathered for these taxa from direct observation of material; few taxa from outside of North America were available for direct study. The endemic Asian tapiroid Lophiaktes is known from numerous disarticulated remains from Mongolia, but many important parts of its osteology are unknown. Separate analyses were run excluding and including Lophialetes, in order to examine the effect that this representative of an unusual but poorly known group of tapiromorphs would have on the analysis. Among the taxa analysed, Hyracotherium and Phenacodus were the only definite non-tapiromorphs included. P hendus was designated as an outgroup in the analysis, but Hyracotherium was included in the ingroup, in order to determine the membership of Tapiromorpha. Those taxa that are more closely related to Tapirus than to Hyracotherium are considered to be tapiromorphs. Data were entered into MacClade 3.05 (Maddison & Maddison, 1992); shortest trees were generated with PAUP for the MacIntosh computer (Swofford, 1991), using the Branch and Bound option. Unless otherwise noted, multistate characters were treated as unordered. Polarity was determined by the outgroup method of Nixon & Carpenter (19!43), i.e. trees were rooted according to the designated outgroup (Phenacodus), and the root determined the direction of character change. Two main criteria were used for selecting characters: (1) variability between, but not within, terminal taxa; and (2) heritability. Features that varied within wellestablished (i.e. terminal) taxa were not considered. Features that appeared to be related to other aspects of an organism s biology (e.g. body size) that are poorly correlated with phylogeny were not considered. CHARACTER DESCRIPTIONS Tables 6 and 7 list the characters used in this analysis, with their primitive and derived conditions, and Table 8 provides the data matrix listing the character states observed in all of the taxa included in the analysis. These character states are described below, with the derived condition(s) given first. Note that, for purposes able 6. Cranial characters. The characters used in this analysis are listed with their primitive and derived states, and the scores for derived states. Primitive states are scored as zero (0) Character C1. nasal length long Primitive state (0) Derived state(s) C2. nasal shape posteriorly narrow C3. nasolacrimal present contact C4. premaxilla small, contacts nasals C5. incisive paired foramen C6. maxillary fossa shallow or absent C7. infraorbital over premolars foramen C8. narial incision over canine or P1 C9. supraorbital absent foramen C10. postglenoid present foramen C11. postglenoid facing anterior process C12. ant. face of flat or concave postglenoid and undivided process C13. posttympanic long process C14. postcotyloid absent process short (1) posteriorly broad (1) absent (1) robust, no nasal contact (1); small, no nasal contact (2) single, median (1) well-developed pocket (1); vertical groove (2) over molars (1) over P4 or molars (1); retracted and excavated posteroventrall y (2) present (1) absent (1) facing anterolateral convex with median ridge short (1) present (1) of the analysis, polarity is ultimately determined by tree rooting. Cmnial characters Cl. Nasals short (1). The nasals of perissodactyls primitively extend anteriorly to the point above the anterior

41 able 7. Postcranial characters. The characters used in this paper are listed with their primitive and derived states and the scores for the derived states. Primitive states are scored as zero (0) Character P1. acromion process P2. deltopectoral crest P3. entepicondylar foramen P4. capitulum of humerus P5. lateral process of proximal radius P6. scaphoid radial facet Primitive state (0) Derived state present absent (1) absent or low ridge present unkeeled present short lunar contact P7. iliac crest convex P8. lesser prominent trochanter p9. medial small, same as trochlear ridge lateral of femur P10. gastro- absent cnemius fossa P11. patella unflattened P12. astragalar lined up above trochlea neck P13. sustentacular separate and distal calcaneal facets of astragalus P14. astragalar not saddle-shaped head P15. pes pentadactyl present, hooks laterally (1) absent (1) keeled (1) weak or absent (1) emarginated contact (1); long, straight contact (2) concave (1) weak or absent (1) enlarged (1) present (1) broad and flattened (1); elongate and flattened (2) laterally offset (1) confluent (1) saddle-shaped (1) tridactyl (1) tip of the premaxilla or beyond (Fig. 3A), as in Hyracotherium. Possession of significantly shorter nasals is a derived condition (Fig. 3B). TAPIROMORPH OSTEOLOGY 41 C2. Nasals posteriorly broad (1). The nasals of most mammals are splint- or diamond-shaped, and the posterior portion of the nasals intrudes between the frontal~. This is probably the primitive condition for eutherians. The nasals of perissodactyls are unique in having a triangular shape, where the base of the triangle is a suture with the frontal that does not intrude but, instead, generally runs transversely. C3. Lacrimal small, not contacting nasal (1); or large and not contacting nasal (2). Nasolacrimal contact is primitive for perissodactyls, as demonstrated by presence of this contact in many tapiromorph and non-tapiromorph perissodactyls. This contact is the consequence of the broad posterior portion of the nasals and the prominent facial exposure of the lacrimal. Reduction of the facial exposure lacrimal results in loss of nasolacrimal contact, a derived state. The absence of nasolacrimal contact in Phenacodus is a fundamentally different condition from that seen in some tapiromorphs, such as Heptodon. Phenacodus possesses a prominent facial exposure of the lacrimal, but its nasals are very narrow and do not spread posteriorly (as in perissodactyls) to reach the lacrimals. To reflect this fundamental difference, Phenacodus is scored as 2 for this character. While the condition in Phenacodus may not actually be derived, this score reflects the fact that the condition in this genus is different from other taxa. C4. Remaxilla robust, not contacting nasals (1); or small and not contacting nasal (2). The primitive condition of the premaxilla, seen in many tapiromorphs, non-tapiromorph perissodactyls, and non-perissodactyl outgroups, is a relatively small bone with a prominent ascending process that contacts the nasals. In some tapiromorphs, such as tapirids, the premadla is robust and does not contact the nasals. Despite the fact that its only contact is with the maxilla, the premaxilla is still a well-developed element in condition 1. In rhinocerotids, the ascending process of the premaxilla is reduced and does not contact the nasals; the premaxilla of rhinocerotids is generally not well-developed, and this condition is considered here to be fundamentally different from that of tapirids. C5. Incisive foramen single and median (1). The contact between the premaxilla and maxilla on the palate is pierced by one or two incisive foramina. Primitively, a bilateral pair of foramina are present, as in Hyracotherium. In some tapiromorphs, a single, median foramen is present. C6. Maxillary (preorbital) fossa well-developed pocket (1); or vertical groove (2). The facial portion of the maxillary primitively possesses a shallow fossa or no

42 ~ ~~ ~ ~ 42 L. T. HOLBROOK able 8. Data matrix. The scores for each character assigned to various taxa are listed. Missing entries are indicated by a question mark C?') Homogalax Cardiolophus Isectolophus Eomompus Mompus Heptodon Helaletes Colodon Plesiocolopirus Pmtapirus Tapirus Lophialetes Hyrachyus Rostriamynodon Amynodon T?iplopus Hyracodon Forstemooperia Paraceratherium Trigonias Subhyracodon Uintaceras Hymcotherium Phenamdus Homogalax Cardiolophus Isectolophus Eomompus Mompus Heptodon Helaletes Colodon Plesiocolopirus Protapirus Tapirus Lophialetes Hyrachyus Rostriarnynodon Amynodon z-iplopus Hyracodon Forstemooperia Paraceratheriu m Trigonias Subhyracodon Uintaceras Hyracotherium Phnacodus C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C14 0 1??? 0? 0 0? 0 0?? ?? 1??? 0 0? ?? 1??? 0 0? 1? ? ? ? ? 1 0? 0 1 1? ? 1 0? ? 1 0 0? ?? 0? 0 1 0???? ? ? ? ?? ? 0? ? ? P1 P2 P3 P4 P5 p6 P7 P8 P9 P10 P11 P12 P13 P14 P15 P16? ?? ??????????? O O? l????? ??????? 1????? 0? 0??? ? ? ? ????????????????????? l?????????? ? ?????????????????? ? ? ? 0 0? ???????????????? l o l l? ? 1 0? ? ? 0 0

43 fossa of any note, as in Hyracotherium. In some rhinocerotoids, a prominent fossa is present. It is difficult to say what actually occupied this fossa in life, and it is possible that these fossae are not homologous in different rhinocerotoids. In the absence of evidence to the contrary, these fossae are scored as homologous for this study. Some tapiroids possess another type of fossa, a vertical groove anterior to the orbit. In Tapirus, these grooves accommodate cartilaginous nasal diverticula. C7. Infraorbital foramen positioned over molars (1). The infraorbital foramen of most mammals is positioned over the upper premolars, usually P2 or P3, and this position is found in most perissodactyls. The infraorbital foramen of amynodontids is positioned more posteriorly, over the molars and often within the posterior part of the maxillary fossa. C8. Posterior edge of narial incision retracted to point over P4 or molars (1); or retracted as in state 1 and posteroventrally excavated (2). The narial incision primitively has its posterior border placed no further back than over the first premolar. A number of tapiromorphs have a border positioned much more posteriorly. In Helaletes, Colodon, and Plesiocolopirus, the ventral aspect of the posterior border is deep and rounded, giving the impression of a keyhole shape. This character was treated as ordered, to reflect the fact that state 2 is a special case of state 1. C9. Supraorbital foramen present (1). The postorbital process of the frontal is primitively unpierced, as in Hyracotherium and Phenacodus. In Eomompus and Mompus, a foramen pierces this process (Fig. 6). (210. Postglenoid foramen absent (1). This foramen is present in Phenacodus, Hyracotherium, and a number of tapiromorphs, and its presence is therefore primitive for perissodactyls. Cll. Postglenoid process facing anterolaterally (1) (Fig. 7C). The postglenoid process of the squamosal of Hyracotherium and Phenacodus is small, peg-like, and faces anteriorly; this is the primitive condition. In a number of tapiromorphs, this process is large, flattened, and obliquely oriented. C12. Anterior face of postglenoid process convex with median ridge (1). The anterior face of the postglenoid process is primitively flat or concave. In some rhinocerotoids, this face has become convex and divided into medial and lateral portions by a ridge. TAPIROMORPH OSTEOLOGY 43 C13. Posttympanic process short (1). In Hyracotherium and many tapiromorphs, the posttympanic process of the squamosal is about as long as the postglenoid process. The posttympanic process is significantly shorter than the postglenoid process in Eomompus and Mompus (Fig. 8B). This feature may be related to the presence of an auditory bulla that is solidly fused to the skull in chalicotherioids. C14. Postcotyloid process of dentary present (1). Present in rhinocerotids and Uintaceras, this process is a buttress on the posterior edge of the ascending ramus of the mandible, just below the mandibular condyle. This process is absent in other perissodactyls and nonperissodactyls. Postcranial characters P1. Acromion process of scapula absent (1). A distinct acromion process is present on the scapula of Hyracotherium and many non-perissodactyls, including Phenacodus, and its presence is considered to be primitive. P2. Deltopectoral crest of humerus prominent and hooking laterally (1) (Fig. 10B). The deltopectoral crest of non-tapiromorph perissodactyls and non-perissodactyls is either absent or a low ridge. P3. Entepicondylar foramen of humerus absent (1). The presence of this foramen is considered to be primitive for eutherians (Thewissen & Domning, 1992), and it is present in Phenacodus. P4. Capitulum of humerus keeled (1). The capitulum is primitively a rounded surface articulating with the radius, as in Phenacodus. In perissodactyls, the capitulum is trochleate due to a median ridge or keel. P5. Lateral process of proximal radius weak or absent (1). The radius of Hyracotherium shows the primitive condition, where the lateral articular facet for the humerus extends beyond the shaft laterally on a prominent process or tuberosity. P6. Radial facet of scaphoid with emarginated lunar contact (1) or long, straight lunar contact (2). The lunar contact of the radial facet of the scaphoid is relatively short in Hyracotherium and brontotheres, and this condition is considered to be primitive. P7. Anterior iliac crest concave (1). The anterior crest of the ilium is primitively convex, as in Phenacodus and many other eutherians.

44 44 L. T. HOLBROOK P8. Lesser trochanter of femur weak or absent (1). The lesser trochanter of Phenucodus and Hymtherium is present as a prominent flange. P9. Medial trochlear ridge of femur expanded into tuberosity (1). The medial and lateral trochlear ridges are primitively about equal in size, as in Pheenacodus and Hyracotheriurn. Pl 0. Gastrocnemius (supracondylar) fossa of femur present (1). A distinct fossa for attachment of the gastrocnemius is not present above the lateral condyle on the posterior side of the femur in Phenacodus, Hyracotherium, and brontotheres. P11. Patella broad and flattened (1) or elongate and flattened (2). Primitively, the patella is teardropshaped and anteroposteriorly thick, as in Hymcotherium and Phenacodus. P12. Trochlea of astragalus laterally offset from neck (1). The trochlea of equids and primitive brontotheres like Eotitanops lies more or less directly above the neck of the astragalus. P13. Sustentacular and distal calcaneal facets of astragalus confluent (1) or contluent with a ridge formed at their junction (2) (Fig. 15D). These facets are separate in non-tapiromorph perissodactyls (Fig. 15E). This character was treated as ordered, to reflect the fact that state 2 is a special case of state 1. P14. Navicular facet of astragalus saddle-shaped (1). The astragalar head (navicular facet) of non-perissodactyls is not saddle-shaped. In Phenacodus, this facet is rounded for a ball-and-socket joint. P15. Pes tridactyl(1). Five digits on the pes is primitive for eutherians and is seen in Phenacodus. RESULTS The first analysis (excluding hphialetes) resulted in 3078 shortest trees, each of 49 steps, and each with a consistency index (CI) of 0.73 (0.70 whenuninformative characters are removed) and a retention index of The strict and Adams consensus cladograms derived from these results are illustrated in Figures 16 and 17. The second analysis (including hphialetes) resulted in 7956 shortest trees, each of 51 steps, and each with a consistency index (CI) of 0.73 (0.69 when uninformative characters are removed) and a retention index of The strict and Adams consensus cladograms derived from these results are illustrated in Figures 18 and 19. The strict consensus cladograms of both analyses show a basal polytomy of Hyracotherium, Homogalax, Cardiolophus, chalicotherioids, and a tapiroid/rhinocerotoid clade, which is here referred to as unequivocal Tapiromorpha. Unequivocal Tapiromorpha consists of a polytomy of seven clades: the genera Isectolophus, Heptodon, Hyrachyus, Hyracodon, and Triplopus, as well as a tapiroid and a rhinocerotoid clade. Rhinocerotidae and Amynodontidae are supported as monophyletic in both analyses, and Tapiridae (Protapirus and Tbpirus) is supported as monophyletic when Lophialetes is excluded. Helaletes, Colodon, and Plesiocolopirus group with the tapirids (in what is called here unequivocal Tapiroidea ), with Helaletes coming out as sister-taxon to the rest. Uintaceras is placed as the sister-taxon to Rhinocerotidae, with Amynodontidae and Paraceratheriurn forming a trichotomy with this clade, and Forstemperia is the sister-group to these. Hyrachyus, HyIYZCOdOn, and Triplopus are not unequivocally identified as rhinocerotoids in the strict consensus. The main consequence of including Lophialetes in the second analysis is that relationships among the members of the monophyletic tapiroids become completely unresolved. Adams consensus trees have the advantage of limiting the effects of rogue taxa (i.e. taxa of particularly uncertain position), but many authors avoid them because they can result in a tree with clades that are not found in the constituent trees. Because missing data in fossils often lead to rogue taxa, Adams consensus trees are employed here for heuristic purposes. The Adams consensus cladograms for both analyses are identical in most respects. The basal polytomy for unequivocal tapiromorphs is reduced to three clades, Isectolophus, Tapiroidea, and Rhinocerotoidea. Hyrachyus and a hyracodontid clade (Triplopus and Hyracodon) are placed within the Rhinocerotoidea. Heptodon is placed as the sister-taxon to the rest of the tapiroid clade. The topology of the rest of the tapiroid clade essentially corresponds to that of the corresponding strict consensus. The only exception is that the Adams consensus for the second analysis identifies a monophyletic Tapiridae. DISCUSSION The large number of shortest trees resulting from both analyses could be caused by several factors, including

45 TAPIROMORPH OSTEOLOGY Homogalax Cardiolophus Isectolophus Heptodon Helaletes Colodon Plesiocolopirus Protapirus Tapirus Hyrachy us - Rostriamynodon Amynodon is Paraceratherium Trigonias Subhyracodon Uintaceras Forstercooperia Triplopus ::z Hyracodon Hyracotherium Phenacodus 18 Homogalax Cardiolophus Heptodon Helaletes Colodon Plesiocolopirus Protapirus Tapirus Lophialetes Hyrachyus Rostriamynodon Amy nodon -L- Paraceratherium Trigonias Su bhyracodon Uintaceras Forstercooperia Triplopus I Hyracodon Eomoropus Moropus Hyracotherium Phenacodus I 17 Homogalax Cardiolophus Isectolophus Heptodon Helaletes Colodon Plesiocolopirus Protapirus Tapirus, r Hyrachyus Rostriamynodon Amynodon Paraceratherium I Trigonias Subhyracodon I Uintaceras IForstercooperia Triplopus - Hyracodon Eomoropus Moropui Hyracotherium Phenacodus 19 Homogalax Cardiolophus Isectolophus i Plesiocolopirus Protapirus Tapirus Lophialetes Hyrachyus Rostriarnynodon Amynodon Paraceratheriurn Trigonias Subhyracodon Uintaceras Forstercooperia Triplopus Hyracodon Figures Consensus cladograms. Fig. 16. Strict consensus cladogram of 3078 shortest trees resulting from analysis of tapiromorph relationships, with Laphialetes excluded. Fig. 17. Adams consensus cladogram of 3078 shortest trees resulting from analysis of tapiromorph relationships, with Lophialetes excluded. Fig. 18. Strict consensus cladogram of 7956 shortest trees resulting from analysis of tapiromorph relationships, with Lophialetes included. Fig. 19. Adams consensus cladogram of 7956 shortest trees resulting from analysis of tapiromorph relationships, with Lophialetes included.

46 46 L. T. HOLBROOK missing data and character conflict. Missing data for Lophialetes is certainly an important component for the great increase in the number of shortest trees resulting from the second analysis. Consensus trees are simply summaries of information that is common to all (or, in the case of majority rule, most) of the equally most-parsimonious trees. They should not be viewed as outright rejections of particular hypotheses, but simply as statements of what is and what is not unequivocally supported by the data. Hooker s definition of Tapiromorpha cannot be unequivocally supported by these data. Neither Homogalax, Cardiolophus, nor Chalicotherioidea is placed closer to unequivocal tapiromorphs than is Hyracotherium in any of the consensus trees. Similarly, the accepted concept of Rhinocerotoidea (Prothero et al., 1986) is not unequivocally supported in the strict consensus, although it is supported by the Adams consensus trees. The placement of Heptodon as an unequivocal tapiroid is supported by the loss of nasolacrimal contact, but it is likely that the absence of knowledge of this character in some taxa, notably Isectolophus, is the reason that the strict consensus does not reflect support for this placement (whereas the Adams consensus does). The consensus trees do demonstrate some unequivocally supported topologies that contradict at least some previous hypotheses. Isectolophus is unequivocally placed with the Tapiromorpha, clearly supporting the paraphyly of the Isectolophidae sensu Radinsky (1963a). Additionally, there are a number of interesting results regarding Rhinocerotoidea. The position of Uintaceras agrees with the results of Holbrook & Lucas (1997), but the close relationship of Amynodontidae and Rhinocerotidae (plus Uintaceras) relative to Hyracodontidae disagrees with the phylogeny of Prothero et al. (1986), who placed Hyracodontidae as the sister-taxon to Fthinocerotidae. Furthermore, indricotheres (Forstemoperia and Paruceratherium) do not come out as a monophyletic group, nor are they found within Hyracodontidae, as suggested by several workers (Lucas & Sobus, 1989; Lucas et al., 1981; Prothero et al., 1986). Instead, Paraceratherium and Forstemperia come out as successive sister-groups to the amynodontid-rhinocerotid- Uintaceras clade. This is a novel result. There are few unequivocal synapomorphies for the clade of rhinocerotids, Uintaceras, amynodontids, Paraceratherium, and Forstercooperia, but this is due at least in part to the lack of information on the postcrania of Forstenxmperia. Some of the derived character states influencing relationships within this clade of rhinocerotoids exclusive of hyracodontids and Hyrachyus are C6(1), C10(1), C12(1), C14(1), P2(1), P5(1), p6(2), and P9(1). Characters C14 and p2 mainly concern the sistergroup relationship of Uintaceras and rhinocerotids, but the other characters each lend some support (often equivocal) to a close relationship between two or more of these taxa. The results of the second analysis suggest that Lophialetes is a tapiroid of unknown affinities within the clade, but this is a very tentative inference. The association of Lophialetes with tapiroids is not strongly supported and is based on characters that could be interpreted as superficial similarities due to similar modifications of the skull to (possibly) accommodate a mobile muscular proboscis. Such features are probably not present in other lophialetids, such as Schlosseria (Radinsky, ; pers. obs.) and Eoletes, whose skull was recently described by Lucas, Emry & Bayshashov (1997). Thus, these similarities may eventually be interpreted as homoplasies rather than as synapomorphies. Lophialetids and another enigmatic endemic Asian family, deperetellids, have been classified as tapiroids since their discovery (Matthew & Granger, 1925), but their exact affinities are somewhat uncertain. Dashzeveg & Hooker (1997) provided a recent analysis of deperetellid relationships based on tooth morphology. A better understanding of some of the characters discussed here may help to elucidate the phylogeny of these taxa. In summary, this analysis is meant more to demonstrate work in progress than to argue for weakly supported hypotheses of tapiromorph phylogeny. The point is to present new data for addressing this issue and to demonstrate the hypotheses that they support. TAPIROMORPH OSTEOLOGY AND THE ORIGIN OF PERISSODACTYLS One other important phylogenetic point that can be taken from the preceding description and discussion of tapiromorph osteology is that tapiromorphs tell us something about character states present in the ancestor of perissodactyls. Some of the characters described here provide examples of tapiromorphs clarifying ideas about the primitive state for the order. Primitive features not previously recognized Two such characters are the presence or absence of the postglenoid and mastoid foramina. These two foramina are clearly present in primitive tapiromorphs, as they are primitively in eutherians. Extant perissodactyls and many later fossil perissodactyls lack these foramina, and, partly on the basis of the condition in these taxa, the absence of these foramina has been used to unite perissodactyls with tethytheres, hyracoids, and a number of fossil groups (Court, 1992; Prothero et al., 1988). The presence of these foramina

47 in some tapiromorphs (and in some other early perissodactyls) indicates that perissodactyls possessed these foramina primitively, and thus that perissodactyls retained the primitive eutherian condition at their origin. Therefore, the absence of these foramina cannot be used to unite perissodactyls with the other ungulate groups mentioned above. The best demonstration of the utility of additional data from a larger representation of perissodactyl diversity would be to include these data in analyses of perissodactyl or ungulate interrelationships. To further illustrate the importance of using more of an order s diversity to assess primitive character states, two previously published data sets were reanalysed, first with their original character scores, then with some characters rescored to reflect the data presented here, as well as additional information on other ungulate orders. The following studies were reanalysed Court (1992) and Thewissen & Domning (1992). It should be pointed out that Thewissen and Domning recognized the potential shortcomings of their strategy of representing Artiodactyla, Perissodactyla, and Proboscidea each with a single oldest or most primitive genus (Diacodexis, Hy raco t he riu m, and Moeri theriu m, respectively). The following characters were rescored in the two studies: Thewissen & Domning (1992) Character 4: Mesostyle of upper molars present. The authors note the presence of a mesostyle in some early perissodactyls, but assign the primitive condition to the order on the basis of its absence in the earliest forms. The presence of mesostyles in an early perissodactyl, Lambdotherium, and their persistence in at least some early lineages, namely brontotheriids and palaeotheriids, suggests that a more accurate scoring of this character would be unknown (? ) or polymorphic ( O/l ). The latter scoring is adopted here. Character 9: Metastylids of lower molars present. The authors assigned the primitive condition to perissodactyls on the basis of Hyracotherium sandrae. As argued for character 4, the presence of metastylids in a number of early perissodactyls, e.g. Homogalax, Lambdotherium, and Camliolophus, suggests that polymorphic ( O/l ) is a more accurate scoring for perissodactyls. Hooker (1994) interpreted the metastylids of some perissodactyls as part of a twinned metaconid, and thus not homologous with the metastylids of other mammals. Some perissodactyls retain both a metastylid and a twinned metaconid. Characters 28 and 29: Sulci for promontorial (28) and stapedial (29) branches of internal carotid artery absent. Among the taxa scored as derived for character 28 are the following: Arctocyon, Hyopsodus, Pleuraspidotherium, and Meniscotherium. Cifelli (1982) de- TAPIROMORPH OSTEOLOGY 47 scribes the promontorial sulcus as present in all of these taxa, except Arctocyon, for which the condition is unknown. In this study, these taxa were rescored as primitive ( O), except Arctocyon, which was scored as unknown (? ). Cifelli s descriptions of ungulate petrosals also contradict Thewissen & Domning (1992) by describing the stapedial sulcus as absent in Hyopsodus and present in Phenacodus and Meniscotherium. Thewissen and Domning disputed the presence of a stapedial sulcus in Phenacodus, but Cifelli illustrates the stated conditions in Hyopsodus and Meniscotherium. Williamson & Lucas (1992) do not mention the stapedial sulcus in their description, but their illustration of the petrosal of Meniscotherium does not clearly indicate whether a sulcus is present or absent going to the fenestra ovale. In this study, Hyopsodus was rescored as derived ( l ), and Meniscotheriurn was rescored as primitive ( 0). Character 41: Entepicondylar foramen absent. Court (1994) has described the presence of an entepicondylar foramen in the early proboscidean Numidotherium. In fairness to Thewissen and Domning, this information was not available to them, and it is likely that they might have considered it for scoring Proboscidea. Rescoring this character as 0 for Proboscidea, however, allows us to examine the effects of representing Proboscidea only with Moeritherium, which is arguably still the earliest proboscidean for which we have enough information to score many characters. Character 42: Supracondylar foramen of humerus present. The authors note the presence of primitive and derived states within Hyracotherium, and both states are observed in other early perissodactyls. Oddly, the authors state that the primitive condition was scored for Hyracotherium, but the matrix contains a derived score for this taxon. The variety of observed states within Hyracotherium, let alone perissodactyls in general, suggests a polymorphic state; thus perissodactyls were rescored as O/l. Characters 43 and 52: Less than five functional digits of hand (43) and foot (52). The authors scored Diacodexis as 1 for character 52 and Moeritherium as? for both of these characters. Based on the presence of five digits in the foot of some Paleogene artiodactyls (Scott, 1940) and of five digits in the hands and feet of known proboscideans, these characters are rescored as primitive ( 0). Character 51: Peroneal tubercle of calcaneus absent. Rose (1996) has described the presence of peroneal tubercles in Hyracotherium and Homogalax. On this basis, this character was rescored as 0 for perissodactyls. Court (1 992) Character 21: Postglenoid foramen absent. As mentioned above, this foramen is present in a number of

48 48 L. T. HOLBROOK outgroup Arctocyon Diacodexis Hyopsod us Pleuraspidotherium Ectocion Meniscotherium Phenacodus Moeritherium Sirenia Desmostylia Hyracotherium Hymcoidea outgroup Arctocyon Artwdactyla Hyopsodus Pleuraspidotherium Meniscotherium Proboscidea Sirenia Desmostylia Perissodactyla Hyracoidea b I C Meniscotherium PhedUS Ectocion I B outgroup Arctocyon Artiodactvla Hyopsod& Pleuraspidotherium Meniscotherium Phemcodus Ectocwn Perissodactyla Hyracoidea Desmostylia Proboscidea Sirenia outgroup Pleuraspidotheriurn Meniscotherium Figure 20. Shortest trees resulting from analysis of data of Thewissen & Domning (1992) using their original data (A), a corrected version of their data (B), and after rescoring some characters on the basis of information from other members of the orders Artiodactyla, Perissodactyla, and Proboscidea (C). (See text for explanation of rescoring.) fossil perissodactyls. This character was scored as 0 for Perissodactyla. Character 46: Reduction of the scapula acromion process. The scoring for this character may depend on the definition of reduced. Court s meaning is not clear; in another paper (Court, 1995), he remarks that this process is present in all proboscideans except Moeritherium. Court (1992) considered the proboscidean condition to be reduced. A number of early perissodactyls possess an acromion process, but it is not clear whether this process would be considered to be reduced. Certainly, having no clavicle with which to articulate, the acromion of perissodactyls lacks the elaboration of the distal end for this articulation, but that is likely to be true of any group lacking a clavicle (a condition scored as another derived state in Court [1992]). For heuristic purposes, this character is rescored as 0 for Perissodactyla. Reanalysis of Thewissen & Domning (1992) The results of three separate analyses of the data of Thewissen & Domning (1992) are summarized in Figure 20. All analyses were performed using PAUP

49 A ii 4- I Hyp. Ancestor Artiodactyla Hyracoidea Perissodactyla Sirenia Embrithopoda Proboscidea Hyp. Ancestor Artiodactyla Hyracoidea Sirenia Embrithopoda Proboscidea Perissodactyla Hyp. Ancestor Artiodactyla Hyracoidea Sirenia Embrithopoda Proboscidea Perissodactyla B Figure 21. Shortest trees resulting from Court s (1992) original data set (A) and after rescoring of some characters based on observations in this paper (B). (See text for explanation of rescoring.) Reanalysis of the original data produced two shortest trees (Fig. 20A) and supports a sister-taxon relationship for Perissodactyla (Hyracotherium) and Hyracoidea, as well as paraphyly of Tethytheria by virtue of Desmostylia coming out as the sister-group to Perissodactyla + Hyracoidea. TAPIROMORPH OSTEOLOGY 49 A second analysis was performed after rescoring characters that were considered to be incorrectly scored, but not on the basis of examination of a wider diversity within extant orders. The rescored characters include characters 28,29,42 (where Hyracotherium is scored as 0, to correspond to the character description, not the original matrix), and 51. This analysis also produced two shortest trees (Fig. 20B), which were not substantially different from those of the first analysis. The only difference of note here is that the relationships within the clade including Hyracotherium, Hyracoidea, and Desmostylia are not resolved in the strict consensus. The third and final analysis of these data involved rescoring characters on the basis of observed variation in other representatives of Perissodactyla, Artiodactyla, and Proboscidea. The additional rescored characters include characters 4, 9, 41, 42 (scored as O/l in Perissodactyla), 43, and 52. The result of the analysis was a single shortest tree (Fig. 20C). This tree supported a Paenungulata (sensu Novacek, 1986) clade, i.e. where Hyracoidea is the sister-taxon to a monophyletic Tethytheria. Reanalysis of Court (1 992) Again, all analyses were performed with PAUP The results are summarized in Figure 21. Reanalysing Court s (1992) original data produces a single shortest tree (Fig. 21A), where the sister-group of the monophyletic tethytheres is Perissodactyla, and Hyracoidea is the sister-taxon to that clade. Rescoring characters 21 and 46 produces two shortest trees (Fig. 21B), one identical to the tree from the analysis of the original data, the other switching the relative positions of Perissodactyla and Hyracoidea, thus supporting Paenungulata sensu Novacek. The strict consensus of these two trees is a trichotomy of Hyracoidea, Perissodactyla, and tethytheres. The object of these exercises is not to argue for a (or against) certain schemes of relationships, but to demonstrate the different effects of representing an order with a single genus (or, in the case of Court [1992], with information presumably based on a limited sampling, at least of perissodactyls) and representing it on the basis of a larger diversity of genera within the order. Although the effects of rescoring changes are limited to only a part of the tree derived from either data set, the changes in shortest topology or topologies are distinct and significant, and they are relevant to a particularly controversial question, namely the relationships among tethytheres, hyracoids, and perissodactyls. Part of the route towards a better answer to this question is inferring ancestral character states for these groups from as much information as we can.

50 50 L. T. HOLBROOK This demonstration should not be taken as an indictment of either Thewissen & Domning (1992) or Court (1992). Both of these authors were simply using the data at hand to examine a particular question (relations of Arsinoitherium for Court, phylogenetic position and importance of phenacodontids for Thewissen and Domning) in the context of ungulate phylogeny. In order to avoid the problems confronted by these authors, morphological studies of individual mammalian orders must contribute detailed studies of character systems among a wide range of genera. Such comprehensive databases are and will be important not only for assessing mammalian phylogeny as evidenced by morphology, but also for the inevitable comparisons between morphological studies of mammal phylogeny and similar studies using other types of data, such as biomolecules. Perissodactyl synupomorphies This study also helps to identify potential synapomorphies for the order Perissodactyla. The saddleshaped navicular facet of the astragalus is the one well-established osteological synapomorphy noted in previous literature (e.g. Radinsky, 1966b). Even this classic character is not sufficient to provide strong support for this order, because many other orders possess derived conditions of this facet that could be interpreted by parsimony algorithms as derived from the perissodactyl condition. Fischer & Tassy (1993) considered the presence of nasolacrimal contact in perissodactyls to be a potential synapomorphy for the group. As described earlier, this contact is largely a consequence of the unusual morphology of perissodactyl nasals. Thus, the posterior expansion of the nasals and their roughly transverse contact with the frontals is considered here to be a potential synapomorphy of Perissodactyla. This character state is found in all perissodactyls studied here, including those that lose nasolacrimal contact. Among the features of the postcranial skeleton, several help to distinguish perissodactyls from more archaic ungulates, such as Phenacodus. Some of these features, such as the keeled capitulum of the humerus and the tridactyl pes, are related to cursoriality and may be found in other groups (e.g. hyracoids). The reorientation of the entocuneiform, noted by Radinsky (1963b), is an unusual feature of perissodactyls and is considered here to be a potential synapomorphy for the order. This feature is not found in other ungulates that have also reduced or lost the hallux. ACKNOWLEDGEMENTS This study stems from dissertation work performed in partial fulfillment of the requirements for a doctoral degree at the University of Massachusetts at Amherst. I thank M. Coombs, L. Godfrey, W. Bemis, J. Meng, and E. Brainerd for comments on earlier drafts of this paper. I thank the following individuals for their assistance and allowing me access to specimens in their care: R. Tedford and J. Alexander of AMNH; M. Dawson and A. Tabrum of CMNH; F? Holroyd of UCMP; R. Stucky of DMNH, D. Whistler and S. Macleod of LACM; F. Jenkins and C. Schaff of MCZ; I? Murphy of UC; P. Gingerich, M. Uhen, and W. Sanders of UM; K. Rose of Johns Hopkins University; L. Thomas of the Pratt Museum, Amherst College, Amherst; R. Emry, C. Ray, R. Purdy, L. Gordon, and M. Frank of USNM; J. Lillegraven and J.-I? Cavigelli of UW; M. Turner of YPM. Financial support for this research was provided by the Theodore Roosevelt Memorial Fund of AMNH, a Sigma Xi Grant-in-Aid of Research, and a Smithsonian Graduate Fellowship. REFERENCES Borissiak AA [A new chalicothere from the Tertiary of Kazakhstan.] A M Nauk. SSSR, Trudy Paleontologichsltogo Institutu 13(3): Bressou C La myologie du tapir. Mammalia 25: Butler PM The milk-molars of Perissodactyla, with remarks on molar occlusion. Proceedings of the Zoological Society of London 121(4): Carroll RL Vertebrate Paleontology and Evolution. New York W.H. Freeman and Co. Cerdeiio E Cladistic analysis of the family Rhinocerotidae (Perissodactyla). American Museum Nouitates 3143: Cifelli RL The petrosal structure of Hyopsodus with respect to that of some other ungulates, and its phylogenetic implications. Journal of Paleontology 56(3): Colbert EH Chalicotheres from Mongolia and China inthe AmericanMuseum. Bulletin of the American Museum of Natural History 67(5): Colbert MW, Schoch RM Tapiroidea and other moropomorphs. In: Janis C, Scott K, Jacobs L, eds. Evolution of Tertiary Mammals of North America, Vol. 1: Terrestrial Carnivores, Ungulates, and Ungulatelike Mammals. Cambridge: Cambridge University Press, Coombs MC Sexual dimorphism in chalicotheres (Mammalia, Perissodactyla). Systematic Zoology 24: Coombs MC. 1978a Reevaluation of early Miocene North American Mompus (Perissodactyla, Chalicotheriidae, Schizotheriinae). Bulletin of the Carnegie Museum of Natural History 4: Coombs MC. 1978b. A premaxilla of Moropus elatus Marsh, and evolution of chalicotherioid anterior dentition. Journal of Paleontology Coombs MC Large mammalian clawed herbivores: a comparative study. Transactions of the American Philosophical Society 73(7): 1-96.

51 Coombs MC Interrelationships and diversity in the Chalicotheriidae. In: Prothero D, Schoch R, eds. The Evolution of Perissoductyls. New York: Oxford University Press, Coombs MC, Coombs W, Jr Anatomy of the ear region of four Eocene artiodactyls: Gobiohyus,?Helohyus, Diacodexis, and Homacodon. Journal of Vertebrate Paleontology 2(2): Cooper CF Baluchitherium osborni (?syn. Indricotherium turgaicum, Borrisiak). Philosophical Transactions of the Royal Society of London, B 50: Cope ED The Vertebrata of the Tertiary formation of the West. U.S. Geological Survey of the Territories Report 3: Cope ED The Vertebrata of the Swift Current River, 11. American Naturalist Court N Periotic anatomy of Arsinoitherium (Mammalia, Embrithopoda) and its phylogenetic implications. Journal of Vertebrate Paleontology lo(2): Court N The skull of Arsinoitherium (Mammalia, Embrithopoda) and the higher order interrelationships of ungulates. Palaeovertebrata 22(1): 143. Court N Limb posture and gait in Numidotherium koholense, a primitive proboscidean from the Eocene of Algeria. Zoological Journal of the Linnaean Society 111: Court N A new species of Numidotherium (Mammalia: Proboscidea) from the Eocene of Libya and the early phylogeny of the Proboscidea. Journal of Vertebrate Paleontology 15(3): Dashzeveg D, Hooker JJ New ceratomorph perissodactyls (Mammalia) from the Middle and Late Eocene of Mongolia: their implications for phylogeny and dating. Zoological Journal of the Linnaean Society 120: de Bonis L, Bouvrain G, Koufos G, I)assy P Un crhe de chalicothere (Mammalia, Perissodactyla) du Miocene superieur de Macedoine (Grece): remarques sur la phylogenie des Chalicotheriinae. Palaeovertebmta 24( 1-2) Deperet C Etudes paleontologique sur les Lophiodon du Minervois. Archives, Museum des Sciences Naturelle, Lyon 9(1) Edinger T, Kitts DB The foramen ovale. Evolution Emry RJ A tiny new Eocene ceratomorph and comments on tapiroid systematics. Journal of Mammalogy 70(4) Filhol H Etude sur les vertebres fossiles de Issel (Aude). Memoires, Societe Geologique de France Fischer MS Die Stellung der Schliefer (Hyracoidea) im phylogenetischen System der Eutheria. Zugleich ein Beitrag zur Anpassungsgeschichte der Procaviidae. Courier Forschungsinstitut Senckenberg 84: Fischer MS Hyracoids as perissodactyls. In: Prothero D, Schoch R, eds. The Evolution of Perissodactyls. New York: Oxford University Press, Fischer MS, Tassy P The interrelation between Proboscidea, Sirenia, Hyracoidea, and Mesaxonia: the morphological evidence. In: Szalay FS, Novacek MJ, McKenna TAPIROMORF H OSTEOLOGY 51 MC, eds. Mammal Phylogeny, Volume 2, Placentals. New York: Springer-Verlag, Flower WH An Introduction to the Osteology of the Mammalia, Third Edition. London: MacMillan and Co. Fortelius M, Kappelman J The largest land mammal ever imagined. Zoological Journal of the Linnaean Society 107: Franzen JL Origin and systematic position of the Palaeotheriidae. In: Prothero D, Schoch R, eds. The Evolution of Perissoductyls. New York: Oxford University Press, Franzen JL Hallensia (Mammalia, Perissodactyla) aus Messel und dem Pariser Becken sowie Nachtrage aus dem Geiseltal. Bulletin de l lnstitut Royal des Sciences Naturelles de BelgLque (Science de la Terre) 60: Froehlich DJ Phylogenetic systematics of basal penssodactyls. Journal of vertebrate Paleontology 19(1): Garevski R Beitrag zur Kenntnis der Pikermifauna mazedoniens. Fragmenta Balcanica Gazin CL The upper Paleocene Mammalia from the Almy Formation in western Wyoming. Smithsonian Miscellaneous Collections 131(7): Gingerich PD Systematics and evolution of early Eocene Perissodactyla (Mammalia) in the Clarks Fork Basin, Wyoming. Contributions from the Museum of Paleontology, University of Michigan 28: Granger W, Gregory WK Further notes on the gigantic extinct rhinoceros, Baluchitherium, from the Oligocene of Mongolia. Bulletin of the American Museum of Natural History 72: Gregory WK. 192Oa. Studies in comparative myology and osteology: no. IV - a review of the evolution of the lacrymal bone of vertebrates with special reference to that of mammals. Bulletin of the American Museum of Natural History 42(2): Gregory WK. 192Ob. Studies in comparative myology and osteology no. V - on the anatomy of the preorbital fossae of Equidae and other ungulates. Bulletin of the American Museum of Natural History 42(2) Hanson CB Teletaceras radinskyi, a new primitive rhinocerotid from the late Eocene Clarno Formation, Oregon. In: Prothero D, Schoch R, eds. The Evolution of Perissodactyls. New York Oxford University Press, Hanson CB Stratigraphy and vertebrate faunas of the Bridgerian-Duchesnean Clarno Formation, north-central Oregon. In: Prothero D, Emry R, eds. The Terrestrial Eocene-Oligocene Transition in North America. Cambridge: Cambridge University Press, Heissig K The Rhinocerotidae. In: Prothero D, Schoch R, eds. The Evolution of Perissoductyls. New York Oxford University Press, Hermanson JW, MacFadden BJ Evolutionary and functional morphology of the knee in fossil and extant horses (Equidae). Journal of Vertebrate Paleontology 16(2): Holbrook LT The phylogeny and classification of

52 52 L. T. HOLBROOK tapiromorph perissodactyls (Mammaha). Cladistics 15(3): Holbrook LT, Lucas SG A new genus of rhinocerotoid from the Eocene of Utah and the status of North American Forstercooperia. Journal of Vertebmte Paleontology 17(2): Holland WJ, Peterson OA The osteology of the Chalicotheroidea with special reference to a mounted skeleton of Mompus elatus Marsh, now installed in the Carnegie Museum. Memoirs of the Carnegie Museum Hooker JJ A primitive ceratomorph (Perissodactyla, Mammalia) from the early Tertiary of Europe. Zoologuxll Journal of the Linnean Society 82: Hooker JJ Character polarities in early perissodactyls and their si&cance for Hymcotherium and infraordinal relationships. In: Prothero D, Schoch R, eds. The Evolution of Perissodactyls. New York Oxford University Press, Hooker JJ The beginning of the equoid radiation. Zoological Journal of the Linnean Society Kielandaworowska Z BeIiminary data on Upper Cretaceous eutherian mammals from Bayn Dzak, Gobi Desert. PaZueontologia Polonica Kielandaworowska Z Evolution of the therian mammals in the late Cretaceous of Asia. Part IV. Skull structure in Kennulestes and Asioryctes. Palueontokgza Polonica 42: Kitts DB American Hymcotherium (F erissodactyla, Equidae). Bulletin of the American Museum of Natural History Klaits B The moving mesaxonic manus: a comparison of tapirs and rhinoceroses. Mammalia 36(1): Klaits B Upper Miocene rhinoceroses from Sansan (Gers), France: the manus. Jouml of Paleontology 47(2): Lucas SG, Schoch RM Taxonomy and biochronology of Eomompus and Grangeria, Eocene chalicotheres from the western United States and China. In: Prothero D, Schoch R, eds. The Evolution of Perissodactyls. New York Oxford University Press, Lucas SG, Sobus J The systematics of indricotheres. In: Prothero D, Schoch R, eds. The Evolution of Perissodactyls. New York Oxford University Press, Lucas SG, Schoch RM, Manning E The systematics of Forstemoperia, a middle to late Eocene hyracodontid (Perissodactyla: Rhinocerotoidea) from Asia and Western North America. Journal of Paleontology 550: Lucas SG, Emry RJ, Bayshashov BU Eocene Perissodactyla from the Shinzhaly River, eastern Kazakhstan. Jouml of Vertebmte Paleontology 17(1): MacFadden BJ Cladistic analysis of primitive equids, with notes on other perissodactyls. Systematic Zoology 26: Maddison WP, Donoghue MJ, Maddison DR Outgroup analysis and parsimony. Systematic Zoology 33: Maddison WP, Maddison DR MacClade: Analysis of Phylogeny and Character Evolution. Version 3.0. Sinauer Associates: Sunderland, Massachusetts. Mader BJ The Brontotheriidae: a systematic revision and preliminary phylogeny of North American genera. In: Prothero D, Schoch R, eds. The Evolution of Perissodactyls. New York Oxford University Press, Mader BJ The systematics and phylogeny of North American Eocene brontotheres (Mammalia, Perissodactyla). Unpublished Ph.D. Thesis, Univ. of Massachusetts, Amherst. Matsumoto H Contribution to the knowledge of the fossil Hyraaidea of the Fayum, Egypt, with a description of several new species. Bulletin of the American Museum of Natural History 56: Marsh OC Dinocerata. A monograph of an extinct order of gigantic mammals. US. Geological Survey Monograph 10: Matthew WD A revision of the Puerco fauna. Bulletin of the American Museum of Natural History Matthew WD Absence of the pollex in Perissodactyla. Bulletin of the American Museum of Natural History 37: Matthew WD, Granger W The smaller perissodactyls of the Irdin Manha formation, Eocene of Mongolia. American Museum Novitntes 199: 1-9. McKenna MC Toward a phylogenetic classification of the Mammalia. In: Luckett W, Szalay FS, eds. Phylogeny of the Primates. New York Plenum Publishing, McKenna MC, Manning E Affinities and paleobiogeographic significance of the Mongolian Paleogene genus Phenucolophus. Ghbios, Mkm. Spkc. 1: McKenna MC, Chow M, ing S, Luo Z Radinskya yupingae, a perissodadyl-like mammal from the late Paleocene of China. In: hthero D, Schoch R, eds. The Evolution ofperissoductyls. New York Oxford University Press, Nixon KC, Carpenter JM On outgroups. Cladistics Novacek MJ The skull of leptictid insectivorans and the higher-level classification of eutherian mammals. Bulletin of the American Museum of Natural History 183: Novacek MJ. 199%. Fossils as critical data for phylogeny. In: Novacek MJ, Wheeler QD, eds. Extinction and Phylogeny. New York: Columbia University Press, Novacek MJ. 1992b. Fossils, topologies, missing data, and the higher level phylogeny of eutherian mammals. Systematic Biology 41(1): Novacek MJ, Wyss AR Higher-level relationships of the Recent eutherian orders: morphological evidence. Cladistics Novacek MJ, Wyss AFL Selected features of the desmostylian skeleton and their phylogenetic implications. American Museum Novitates Novacek MJ, Wyss AR, McKenna MC The major groups of eutherian mammals. In: Benton MJ, ed. The Phylogeny and Classification of the!&tmpods, Volume 2. oxford Clarendon Press, Osborn HE Mammals of the Uinta Formation, 4. The evolution of the ungulate foot. %nsactwns of the American Philosophical Society, N.S

53 Osborn HF Eomompus, an American Eocene chalicothere. Bulletin of the American Museum of Natural History 32: Osborn HF' Baluchitherium grangeri, a giant hornless rhinoceros from Mongolia. American Museum Nouitates 78: Osborn HF The titanotheres of ancient Wyoming, Dakota, and Nebraska. U.S. Geological Suruey Monograph Osborn Hl? Amynodon mongoliensis from the Upper Eocene of Mongolia. American Museum Nouitates Owen R Description of teeth and portions of jaws in two extinct anthracotheroid quadrupeds (Hyopotamus uectianus and Hyop. bouinus) discovered by the Marchioness of Hastings in the Eocene deposits on the N.W. coast of the Isle of Wight: with an attempt to develop Cuvier's idea of the classification of Pachyderms by the number of their toes. Quarterly Journal of the Geological Society of London Prothero DR, Manning E, Fischer MS The phylogeny of ungulates. In: Benton MJ, ed. The Phylogeny and Classification of the Tetrapods, Volume 2. Oxford Clarendon Press, Prothero DR, Manning E, Hanson CB The phylogeny of the Rhinocerotoidea (Mammalia, Perissodactyla). Zoological Journal of the Linnean Society 87: Radinsky LB. 1963a. Origin and early evolution of North American Tapiroidea. Peabody Museum of Natural Histoq Yale Uniuersity, Bulletin 17: Radinsky LB. 1963b. The perissodactyl hallux. American Museum Nouitates Radinsky LB Paleomompus, a new early Eocene chalicothere (Mammalia, Perissodactyla), and a revision of Eocene chalicotheres. American Museum Nouitates Radinsky LB. 1965a. Evolution of the tapiroid skeleton from Heptodon to lbpirus. Bulletin of the Museum of Comparative Zoology 134(3): Radinsky LB. 1965b. Early Tertiary Tapiroidea of Asia. Bulletin of the American Museum of Natural History 129: Radinsky LB. 1966a. The families of the Rhinocerotoidea (Mammalia, Perissodactyla). Journal of Mammalogy 47(4): Radinsky LB The adaptive radiation of the phenacodontid condylarths and the origin of the Perissodactyla. Euolution Radinsky LB. 1967a. A review of the rhinocerotoid family Hyracodontidae (Perissodactyla). Bulletin of the American Museum of Natural History 136(1): 146. Radinsky LB Hyrachyus, Chusmotherium, and the early evolution of helaletid tapiroids. American Museum Nouitates Radinsky LB The early evolution of the Perissodactyla. Evolution Radinsky LB Hyrachyus: tapiroid not rhinocerotoid (abstract). Evolutionary Theory Rasmussen DT, Gagnon M, Simons EL Taxeopody TAPIROMORPH OSTEOLOGY 53 in the carpus and tarsus of Oligocene Pliohyracidae (Mammalia: Hyracoidea) and the phyletic position of hyraxes. Proceedings of the National Academy of Sciences, USA. 87: Ray CE, Sanders A Pleistocene tapirs in the Eastern United States. In: Genoways HH, Dawson MR, eds. Contributions in Quaternary Vertebrate Paleontology: a Volume in Memorial to John E. Guilday. Special Publications of the Camgie Museum of Natural History 8: Remy JA Etude du crhe de Pachynolophus lavocati n. sp. (Perissodadyla, Palaeotheriidae) des Phosphorites du Quercy. Paleouertebrata Remy JA Observations sur l'anatomie crlnienne du genre Palaeotherium (Perissodactyla, Mammalia); mise en evidence dun nouveau sous-genre, Franzentherium. Pahovertebrata 21(3-4): Reshetov V Early Tertiary Tapiroidea of Mongolia and the USSR [in Russian]. The Joint Souiet-Mongolian Expedition, Transactions 11: Rose KD Comparative osteology of North American dichobunid artiodactyls. Journal of Paleontology 59(5): 120S1226. Rose KD Postcranial remains and adaptations in early Eocene mammals from the Willwood Formation, Bighorn Basin, Wyoming. In: Bown TM, Rose KD, eds. Dawn of the Age of Mammals in the Northern Part of the Rocky Mountain Interior, North America. Geological Society of America, Special Paper 243: Rose KD Skeleton of early Eocene Homogalax and the origin of Perissodactyla. Palaeouertebrata 25: Savage DE, Russell DE, Louis P European Eocene Equidae. University of California Publications in Geological Sciences 56: Savage DE, Russell DE, Louis P Ceratomorpha and Ancylopoda (Perissodactyla) from the Lower Eocene Paris Basin, France. University of California Publications in Geological Sciences 66: Savage RJG, Domning DP, Thewissen JGM Fossil Sirenia of the west Atlantic and Caribbean region. V. The most primitive known sirenian, Prorastomus sirenoides Owen Journal of Vertebrate Paleontology 14(3): Schoch RM Hyrachyus: tapiroid or rhinocerotoid? (abstract). Euolutionary Theory Schoch RM l ko unusual specimens of Helaletes in the Yale Peabody Museum collections, and some comments on the ancestry of the Tapiridae (Perissodactyla, Mammalia). Postilla, Peabody Museum, Yale University Schoch RM A review of the tapiroids. In: Prothero D, Schoch R, eds. The Evolution of Perissoductyls. New York Oxford University Press, Scott WB Part IV Artiodactyla. In: Scott WB, Jepsen GL, eds. The mammalian fauna of the White River Oligocene. Transactions of the American Philosophical Society 28: Scott WB Part V Perissodactyla. In: Scott WB, Jepsen GL, eds. The mammalian fauna of the White River Oligocene. Transactions of the American Philosophical Society 28:

54 54 L. T. HOLBROOK Shoshani J Mammalian phylogeny: comparison of morphological and molecular results. Molecular Biology and Evolution 3: Shoshani J Hyracoidea-Tethytheria affinity based on myological data. In: Szalay FS, Novacek MJ, McKenna MC, eds. Mammal Phylogeny, Volume 2, Plucentuls. New York: Springer-Verlag, Simpson GG Notes on Pleistocene and Recent tapirs. Bulletin of the American Museum of Natural History 86: Simpson GG Notes on British hyracotheres. Journal of the Linnuean Society, London 42: Sisson S, Grossman JD The Anatomy of the Domestic Animals. Philadelphia: W.B. Saunders Co. Slijper EX Comparative biologic-anatomical investigations on the vertebral column and spinal musculature of mammals. Koninklijke Nederladche Akademie van Wetenschappen 42(288): Swofford DL PAUP: Phylogenetic Analysis Using Parsimony. Version 3.0. Computer program distributed by the Illinois Natural History Survey, Champaign, Illinois. Thssy P Le crbe de Moeritherium (Proboscidea, Mammalia) de l Eocene de Dor el Talha (Libye) et le probleme de la classification phylogenetique du genre dam les Tethytheria McKenna, Bulletin du Mudum National d Histoiw Natuwlle, Paris Tassy P, Shoshani J The Tethytheria: elephants and their relatives. In: Benton MJ, ed. The Phylogeny and Classification of the Tetrapods, Volume 2. Oxford Clarendon Press, Thewissen JGM Evolution of Paleocene and Eocene Phenacodontidae (Mammalia, Condylarthra). Papers on Paleontology, University of Michigan 29: Thewissen JGM Limb osteology and function of the primitive Paleocene ungulate Pleuraspidotherium with notes on Tricuspiodon and Dissacus (Mammalia). Geobws 24(4): Thewissen JGM, Domning DP The role of phenacodontids in the origin of the modern orders of ungulate mammals. Journal of Vertebrate Paleontology 12(4): 49P 504. Thewissen JGM, Hussain ST Postcranial osteology of the most primitive artiodactyl Diacodezis pakistanensis (Dichobunidae). Anutomia, Histokgza, Embryologia Vanderhoof VL A study of the Miocene sirenian Desmostylus. University of California Publications in Geological Sciences 240: Wall WP Cranial evidence for a proboscis in Cadurcodon and a review of snout structure in the family Amynodontidae (Perissodactyla, Rhinocerotoidea). Journal of Paleontology 54(5): Wall WP Systematics, phylogeny, and functional morphology of the Amynodontidae (Perissodactyla: Rhinocerotoidea). Unpublished Ph.D. Thesis, Univ. of Massachusetts, Amherst. Wall WP The genus Amynodon and its relationship to other members of the Amynodontidae (Perissodactyla, Rhinocerotoidea). Jouml of Paleontology 66(2): Wall WP The phylogenetic history and adaptive radiation of the Amynodontidae. In: Prothero D, Schoch R, eds. The Evolution of Perissodactyls. New York: Oxford University Press, Wall WP, Manning E Rostriamynodon grangeri n. gen., n. sp. of amynodontid (Perissodactyla, Rhinocerotoidea) with comments on the phylogenetic history of Eocene amynodonts. Journal of Paleontology 60(4>: Webb SD, Thylor BE The phylogeny of hornless ruminants and a description of the cranium of Amhaeomeryx, Bulletin of the American Museum ofnatural History 167(3): Wible JFL Transformations in the extracranial course of the internal carotid artery in mammalian phylogeny. Journal of Vertebrate Paleontology q4): Williamson TE, Lucas SG Meniscotherium (Mammalia, Condylarthra ) from the Paleocene-Eocene of western North America. Bulletin of the New Mexico Museum of Natural History and Science 1: Windle BCA, Parsons FG On the muscles of the Ungulata. Part I: Muscles of the head, neck, and forelimb. Pmedings of the Zoological Society of London 1901: Witmer LM, Sampson SD, Solounias N The proboscis of tapirs (Mammalia:Perissodactyla): a case study in novel narial anatomy. Journal of Zoology, London Wood HE Revision of the Hyrachyidae. Bulletin of the American Museum of Natural History 67(5): Wood HE Perissodactyl suborders. Journal of Mammalogy 18: 106. Wood HE Cooperia totadentata, a remarkable rhinoceros from the Eocene of Mongolia. American Museum Novitates [with addendum dated 23 February Wortman JL Species of Hyracotherium and allied perissodactyls from the Wahsatch and Wind River beds of North America. Bulletin of the American Museum of Natural History 8(6): Zapfe H Chalicotherium grande (Blainv.) aus der mioziinen Spaltenfiillung von Neudorf an der March (Devinska Nova Ves), Tschechoslowakei. Neue Denkschriften des Naturhistorischen Museums in Wien