<oolo@calJournal ofthe Linnean Socieb (1994), 112: 29-63. With 26 figures Vertebrate Palaeobioloa. Edited ly M. 3.Benton and D. B. Noman The beginning of the equoid radiation J. J. HOOKER Palaeontology Department, Jvatural History Museum, Gromwell Road, London S W7 5BD With the benefit of newly collected material, primitive equoids are analysed cladistically to determine the detailed relationships of the families Equidae and Palaeotheriidae. The most primitive equid is shown to be Pliolophw. Cymbalophus cuniculus and ‘Hyracothm’um’ sandrae are closely related stem equoids. The Pachynolophidae are not equoids, but most closely related to the primitive tapiromorph family Isectolophidae. Hallaria is sister to these two and therefore also not an equoid. Hyracofhm’um is restricted to the type species H. leporinum Owen, part of whose lower dentition is made known for the first time. It is closest to a restricted genus Propachynolophlcs within the family Palaeotheriidae. The original concept of Propachynolophlcs is polyphyletic. Using the cladogram, newly extended stratigraphic ranges and palaeogeography, an attempt is made to reconstruct the very early speciation and biogeographical history of the group. ‘Gmbalophus’ hooka’ Godinot is recombined in the genus Pachynolophus and ‘H.’ pemk in Pliolophu. Hallensia louisi sp. nov. and Propachynolophus levei sp. nov. are described. ADDITIONAL KEY WORDS:-phylogeny palaeobiogeography. Hyracofhm’um ~ Equidae - - Palaeotheriidae CONTENTS Introduction . . . . . . . . . . Dispersal scenarios . . . . . . . . Family definitions . . . . . . . . Newfinds. . . . . . . . . . Cladistic analysis . . . . . . . . . Choice of taxa . . . . . . . . . Choice of outgroup. . . . . . . . Character definitions . . . . . . . Character explanations . . . . . . . Results of the analysis . . . . . . . Stratigraphy . . . . . . . . . Palaeobiogeography. . . . . . . . . Palaeoecology . . . . . . . . . . Systematics. . . . . . . . . . . Acknowledgements. . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 30 30 33 34 34 34 35 38 41 44 46 48 48 60 61 INTRODUCTION Perissodactyls appear abruptly in the fossil record of northern hemisphere continents in the latest Palaeocene (Dashzeveg, 1988; Hooker, 1991; Koch, Zachos & Gingerich, 1992). They are accompanied by the first known members of the Artiodactyla and true Primates. Although the pattern of appearance is widely recognized as a dispersal event rather than a macroevolutionary burst, the area of origin remains elusive (Gingerich, 1989). This is because the nearest relatives in each case in the earlier Palaeocene appear only distantly related 0024-4082/94/090029+35 $08.00/0 29 0 1994 The Linnean Society of London 30 J. J. HOOKER and because Palaeocene faunas are poorly known outside the northern hemisphere. Among these earliest perissodactyls are some that are recognizable as horse relatives. Almost all these are currently placed in the genus Hyracotherium and are restricted to Europe and North America (Kitts, 1956; Savage, Russell & Louis, 1965). The only Asian representative, ‘Hyracotherium’ gubuniai (Dashzeveg, 1979a), appears not to be an equoid (see below). After the earliest Eocene, Europe was isolated from North America by the North Atlantic, which initially divided the two continents east of Greenland (McKenna, 1983). Europe was also isolated in the Eocene from Asia by the Turgai Straits and fi-om Africa by the Tethys Ocean. Horse relatives in Europe underwent evolution in isolation and have long been accepted as remote from the ‘mainstream” evolution to modern horses, which took place in North America (MacFadden, 1976). The horse family, Equidae, is usually restricted to the North American representatives, wherease those that evolved in Europe and became extinct in the Oligocene are usually placed in the family Palaeotheriidae (but see variations in treatment below), the two combined in the superfamily Equoidea (e.g. Prothero & Schoch, 1989). Three questions immediately arise from this scenario. Firstly, if the appearance of equoids represents a dispersal event, was it really synchronous across the Euramerican area? Secondly, are there any European equids or North American palaeotheres? Thirdly are there any equoids that cannot be accommodated in either Equidae or Palaeotheriidae? Dispersal scenerios Different ideas of area of origin for equoids have resulted in the concept of diachronous appearances and different scenarios for dispersal directions. A central American origin implies dispersal from North America to Europe (e.g. Gingerich, 1976). An African (e.g. Gingerich, 1989) or Indian (Krause & Maas, 1990) origin implies dispersal either from Europe to North America or, if Asia were also involved, from North America to Europe or to both independently without crossing the North Atlantic. Relative primitiveness for the earliest representatives in respective continents has also been invoked for determining dispersal direction (e.g. Hooker, 1980; Godinot, 1981, 1982; Gingerich, 1989), although this has not been based on a critical evaluation of character states. Fumib dejinitions How and where to draw the dividing line between equids and palaeotheriids has long been and continues to be a problem, because of conceptual differences as well as a patchy fossil record. Figure 1 shows a distinct difference between those authors who accept a broad concept of the Palaeotheriidae (only omitting Hyrucotherium because it is considered paraphyletic or ancestral to both Equidae and Palaeotheriidae) (e.g. Remy, 1976; Hooker, 1989) and those with a narrow concept including only the highly derived hypsodont genera (e.g. Savage et al., 1965; Franzen, 1990). That of Butler (1952a, b) is intermediate. Whether the authors’ concepts mean that the trees are phylograms (with ancestors at nodes or as grades along branches), or cladograms does not radically affect their L - - - - - - 1 EQUIDAE 1 PALAEOTHERI:IDAE EQUIDAE ] PALAEOTHERIIDAE EPIHIPPUS LOPHIOTHERIUM PACHYNOLOPHUS ANCHILOPHUS PROPALAEOTHERIUM PSEUDOPALAEOTHERIUM PALAEOTHERIUM PARAPLAGIOLOPHUS PLAGIOLOPHUS LEPTOLOPHUS ANCHILOPHUS PACHYNOLOPHUS PROPALAEOTHERIUM LOPHIOTHERIUM PALAEOTHERIUM PLAGIOLOPHUS 1 I;;: , EPIHIPPUS GE 1 EQUIDAE E D PALAEOTHERIIDAE - EQUIDAE PACHYNOLOPHIDAE 1 EUROPEAN 1 HYRACOTHERIINAE EQUIDAE N. AMERICAN HYRACOTHERIINAE HALLENSIA PALAEOTHERIIDAE ANCHILOPHUS PACHYNOLOPHUS ' H ' . SP. (RIANS) CYMBALOPHUS H . TAPIRINUM H. AFF. VULPICEPS H. VULPICEPS H. VASACCIENSE H , LEPORINUM P. MALDANI PROPAL. MESSELENSE LOPHIOTHERIUM P. GAUDRYI PROPAL. HASSIACUM PALAEOTHERIUM PLAG. sp.1 PLAG. ANNECTENS Figure 1. Phylogenetic models of the Equoidea according to: A, Butler (1952a, b); B, Savage et al. (1965), C, Remy (1976); D, Hooker (1989); E, Franzen (1990). N.B., A is based on text as Butler did not provide a branching diagram. Abbreviations: H = Hyracothm'um, 0 = Orohippu; P = Propachynolophus; Propal = Propalaeothrium; Plag = Plagiolophus. C B A OROHIPPUS PROPACHYNOLOPHUS PACHYNOLOPHUS PLAGIOLOPHUS PALAEOTHERIUM ANCHILOPHUS PROPALAEOTHERIUM LOPHIOTHERIUM w P 2 s EM 0 z f + 32 J. J. HOOKER different ideas of group composition, although only A, D and E are implicitly or explicitly based on shared derived characters. The broad concept of Palaeotheriidae within the Equoidea was recently based on two dental synapomorphies (Hooker, 1989, fig. 6.5). These were: (1) presence of an upper molar metaloph and its occlusal partner a lower molar hypolophid; and (2) presence in at least some individuals of an upper molar centrocristal mesostyle. Franzen (1989, 1990) took the view that the palaeothere femur and pelvis are more primitive than those of ~yracotherium in the low trochanter major and short simple rounded ilium, which were rather similar to those of Phenacodus. Those of Equus and various fossil equids back to Hyracotherium have a high trochanter major and expanded ilium with concave crista ilica and complex tuber coxae for attachment of muscles associated with cursoriality. He also considered that the presence of a P2 metacone in Hyracotherium was advanced and that its absence in Palaeotherium (as also in Paraplagiolophus, Plagiolophus and Leptolophus) was primitive. Butler (1952a, b) also considered that Hyracotherium was more advanced dentally than palaeotheres in its premolar and deciduous premolar molarization pattern, but he included in the Palaeotheriidae not only Plagiolophus and Palaeotherium but also Propalaeotherium, Anchilophus and Lophiotherium. He found the order of appearance of cusps to be variable in his palaeothere grouping, but usually different from that of true equids, more closely resembling (especially in Plagiolophus) that of various condylarths. The presence of a metacone on P2 is variable within the Messel assemblage of Propalaeotherium hassiacum Haupt (see Savage et al., 1965: 62, fig. 28), suggesting that for this character the species is intermediate between palaeotheres and equids. In overall dental morphology, P. hassiacum is almost indistinguishable from Propachynolophus gaudy' (see Savage et al., 1965: 26, fig. 7), where the P' metacone is also variable. Franzen (1990) listed a number of autapomorphies for his restricted sense Palaeotheriidae (comprising Palaeotherium, Paraplagzolophus, Plagiolophus, Leptolophus, Cantabrotherium and Pseudopalaeotherium). Notable are: metacarpals longer than metatarsals (rather than the opposite proportions in the Equidae); and length of cervical vertebral series approaching that of the thoracic series. In addition, these derived palaeotheres have moderate to deep narial incisions and relatively elongate braincases, causing anterior shifting of the orbit to at least a central position (Remy, 1992). Autapomorphies cannot be used to relate palaeotheres with other groups, although they are important in assessing the content of the family. Propalaeothrium hassiacum, in addition to being intermediate between Equidae and Palaeotheriidae (sensu Franzen) in presence of P' metacone, has: equal length metacarpals and metatarsals (Haupt, 1925: pl. 20, figs 1, 2); narial incision as far back as P'; and elongate braincase with central orbit (Franzen, 1988: fig. 345). It thus appears that the palaeothere (s.s.) autapomorphies of deepening narial incision and of shortening of the hind limbs are associated with reversals respectively in the premolar and deciduous premolar molarization gradient and in the cursoriality specializations of the ilium and femur. In fact, earlier species of Plagzolophus, with relatively shallow (where known) narial incisions, have more molariform anterior premolars (Hooker, 1986: 37 1) than the advanced P. annectens and P. minor, which were the species mainly used by Butler in his studies. Almost complete premolar molarization in advanced Palaeotheriurn species EARLY HORSE EVOINTION 33 was probably accomplished by subsequent strengthening of the anterior part of the maxilla where it borders the narial incision. Even Propalaeotherium messelerne and Lophiotherium cervulum have slight narial incisions, extending nearly as far as P’ (Franzen, 1988: fig. 344; Deptret, 1917: 78). Moreover, the equid-type ilium appears to be quite widespread among primitive non-equoid perissodactyls: e.g. Hyrachyus (Wood, 1934), Lophialetes (Reshetov, 1979: 83, figs 26, 27), Heptodon (Radinsky, 1965: 92, fig. 14) and Moropus (Holland & Peterson, 1914). It may therefore be primitive for perissodactyls as a whole. Therefore, molar dilambdodonty on a primitive equoid plan, perhaps accompanied by some degree of narial incision (few skulls of primitive members are known), seems the best way of characterizing the Palaeotheriidae. The alternative, to restrict the family to only the derived genera (sensu Franzen), would leave the Equidae paraphyletic (Hooker, 1989). The species Propachynolophus gaud$ and ‘P’. maldani would thus be the most primitive currently recognized members of the Palaeotheriidae. The Equidae are best defined by their mode of molarization of P3. This involves enlargement and lingual shift of the P3 paraconule to become the fourth main cusp (pushing the protocone distally) and distal migration of the P3 metaconid to occupy the distolingual corner of the tooth (being replaced by a neoformed cusp in the position of the metaconid) (character 31 herein). Butler (1952b) has argued that this paraconule and neoformed cusp are really the protocone and metaconid respectively, which have been subjected to delayed development. If this is so, then the distolingual cusp that forms on molariform P3 in palaeotheres cannot be a hypocone, as it forms behind the single lingual cusp of the nonmolariform P3 that Butler has homologized with the hypocone. Similarly, the palaeothere P, entoconid would instead have to be a neoformed cusp behind a mesially positioned entoconid, which is indistinguishable from a metaconid. Whatever the terminology employed, the pattern of molarization of P3 in equids appears to be unique within equoids and thus a valid synapomorphy (see also Van Valen, 1982). Hooker ( 1989) separated Puchynolo~hus,Anchilophus and ‘Hyracot~erium’ sp. from Rians in the family Pachynolophidae, within a nominate superfamily distinct from Equoideea but linked with the latter in the Hippomorpha. These relationships will also be examined below. Hallensia (based on H. matthesi, from the Middle Eocene of Geiseltal, Germany) was originally described as a phenacodontid (Franzen & Haubold, 1986), and later as an equoid (Franzen, 1990). The genus has many characters in common with species attributed to Hyracotherium and is present amongst the Early Eocene assemblages so identified by Savage et al. (1965). It is therefore included in the present study. New jinds For the past three decades, intensive collecting from latest Palaeocene and Early Eocene deposits in the Paris Basin by Messieurs Pierre Louis, Alain Phtlizon and others has augmented the meagre sample of primitive equoids originally available to Savage et al. (1965). These provide bigger samples, greater diversity and stratigraphic range extensions, although assemblages are still dominated by isolated teeth. Nevertheless, the consistent repetition of distinctive 34 J. J. HOOKER morphologies, which can be linked to more complete material of the same or related species, supports the species distinctions recognized here. In some cases, however, small sample size still makes assessment of intraspecific variation difficult (e.g. Propachynolophus levei). Future collecting should provide the necessary test. Godinot (1981) described but did not name a new species of Hyracotheium from Provence as the most primitive yet discovered. This was countered by Gingerich’s (1989) discovery of a tiny equoid in the Clark‘s Fork Basin, Wyoming, earlier than any formerly known from North America, which he considered slightly more primitive than any other in that continent or in Europe. In 1990, Mr John Bruce discovered a lower jaw fragment of Hyracotherium leporinum in the London Clay of the Isle of Sheppey, Kent, England. It is the first to be recognized as belonging to the type species of Hyracotheiium, described by Owen (1841) from a cranium collected over 150 years ago at Herne Bay, Kent. These new finds and controversies encourage a new analysis of the relationships of the most primitive equoids that have usually been included in the genus Hyracotherium, together with the primitive palaeotheres Propachynolophus gaudvi and ‘P’. maldani. As North American species of Hyracotherium are currently under revision by David Froehlich, analysis of species from that continent will be restricted here to two particularly relevant ones recently treated by Gingerich (1 989, 1991). CLJDISTIC ANALYSIS Choice o f taxa This followed the principle of using only the most primitive representatives of well-established groups and their nearest sister taxa (Hooker, 1989). Many of the characters used are dental. This is because so few of the species are known from cranial and postcranial remains and their inclusion would have meant a data matrix with much missing data. Dental characters are combined wherever possible into functional complexes, provided that their co-occurrence either on the same tooth or on occluding partners is consistent. Some characters differ slightly from those used in an earlier study (Hooker, 1989). Their states are tabulated in a data matrix (Table 1). Choice o f outgroup In choosing an appropriate outgroup to polarize the characters, it was regarded as important to avoid using sister orders of the Perissodactyla (e.g. Proboscidea, Sirenia), which are equally derived. The family Phenacodontidae is often regarded as a primitive close relative of the Perissodactyla as well as of several other ungulate orders, recently combined as the Pantornesaxonia (see Thewissen & Domning, 1992). The Equoidea were considered by MacFadden (1976) to be nested within the Perissodactyla because they share a derived posterior position of the optic foramen, unlike other perissodactyls and most other mammals. In this study it was therefore initially felt that primitive nonequoid perissodactyls were an appropriate outgroup for an analysis of the EARLY HORSE EVOLUTION 35 TABLE 1. Data matrix of primitive equoids used in the cladistic analysis. ‘0’ indicates the primitive state. The derived state for binary characters is 1, those for multistate characters are A, B, C, D, (?) indicates missing data. N.B., the postcranial characters attributed to Cardiolophus are derived from the closely related Homogalax. The cranial, postcranial and premolar characters attributed to Halhnsia louisi are derived from H. matthesi. 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 Characters: ~ ~ Propachynolophus gaudryi Propachynolophw leva’ Hyracotherium leponnum ‘Bopachynolophus’ mala’ani Pliolophus pmix Pliolophus vulpiceps ‘Hyracothm’um’ sandrae Cymbalophus cuniculus Pachynolophus hookm Hallensia louisi Cardiolophus Ectocion Phenacodus lABlOBlBClAOBOOBAlOOlllAlBOlllOl??? 1AB???1AB1BO???????OlllA1???110???110???? 1 A B 1 OB 1 A 0 1 AOBO 0 B A 1 1 1 1 1 1 A1 BO 0 1 1 0 1 l ? ? lABlOBlABlAAAAlBAlOOlOlAlBOlllO???? 1 B B 1 OB 1 A 0 0 B A A A l B A O 0 0 0 0 1 A1 A0 0 1 1 1 I ? 1 1 1 BB I O B I A0 OBAAAl B A O O O 1 0 1 A1 A00 1 1 1 1 1 1 1 lBAOlBlAAOBABBlBBlllOOlAlAOOllOl?ll lBAOlBlABOBBBBlBCllllOlAlOOOllOl??? 1BO 1 OAl A C O B A A A I AD1 1 1 10 1 A1 0 10 100 l ? ? ? 1BO1OA1AAOA??B1AC111lOlOlOOOlOOOOll lBAllAlBCOABBBlADlllllOBlOlllOOO?ll 1 AAO 0 0 000 1 OAAAO 00 0 L O 1 0 0 . 4 0 0 0 0 0 0 0 0 I 0 1 000000000O0000000000000000000000000 Equoidea. It was nevertheless noticed that the polarities obtained for many characters were the opposites of those which would be obtained by using phenacodontids (see Thewissen, 1990) as the outgroup. This has been highlighted by varying views on the evolution of bilophodonty and selenolophodonty in perissodactyls (Hooker, 1984, 1989). To resolve this conflict, Phenacodus, considered to be the most distant, reliable and sufficiently well known perissodactyl relative, was used as the outgroup taxon. Ectocion, another phenacodontid, and the tapiromorph Cardiolophus (possibly the most primitive perissodactyl according to Gingerich, 1991: 209), were added to the data matrix. Other potential outgroup taxa, such as Meniscotherium (Meniscotheriidae) and Radinsba (Phenacolophidae), were not used as they were considered to be too autapomorphically derived (McKenna et al., 1989; Thewissen & Domning, 1992). Lumbdotherium, probably the most primitive brontothere (Hooker, 1989), was initially included and formed the basal perissodactyl branch. It was excluded for the final analysis as its highly derived dilambdodonty, combined with many primitive characters, seemed to bias the results. Character dejinitions See Fig. 2 for dental terminology. Character I : Cheek teeth generally bunodont with cresting weak (0); more strongly crested (1). Character 2: Upper molar preparaconule crista directed towards parastyle (0), towards preparacrista, sometimes joining it (A); towards preparacrista, constantly joining it (B). Character 3: Upper molar paraconule mesiodistally distanced from protocone, with oblique preprotocrista joining the paraconule lingually (0); paraconule and protocone in close mesiodistd proximity with less oblique linking preprotocrista (A); paraconule mesiodistally distanced from protocone with near transverse J. J. HOOKER cent & v poproc poproc A C protd I Vd co I lmetd met& \ entd entkl lmetd b entd D B Pmetd 5 1 5 I E 6 j EARLY HORSE EVOLUTION 37 preprotocrista joining paraconule distally, producing a step or re-entrant angle in the protoloph between the paraconule and protocone (B). Character 4: Lower molar metaconid single (0); twinned (1). Character 5: Lower molar protocristid with occlusal edge concave or notched (0); straight (protolophid) (1). Character 6 Lower molar metastylid a distinct cusp (0); a distal projection near tip of metaconid (A); missing (B). Character 7 M'-' metaconule midway between metacone and hypocone, MI-* hypoconulid high and forming part of the arc of cusps demarcating the talonid (0); M1-' metaconule (or its position), mesial of a line drawn between metacone and hypocone, MI-, hypoconulid low and distal to talonid (1). Character 8 MI-' metaconule circular and uncrested (0); with short lophoid mesiobuccally directed premetaconule crista (A); with elongate mesiobuccally directed premetaconule crista, forming part of well developed metaloph (B). Character 9: MI-' metaconule strong, MI-, lacking postentocristid (0); M' metaconule weak, forming slight step in mesiolongual margin of metaloph, MI-? with weak postentocristid which, together with posthypocristid, forms complete hypolophid with concave or notched occlusal edge, in a proportion of individuals (A); state 1 constant (B); M'-' metaconule weak, with metaloph scarcely or not stepped, Ml-2 with complete hypolophid with straight occlusal edge (C). Character 10: Upper molar centrocrista straight or flexed buccally slightly (0); flexed buccally distinctly (1). Character I I : Upper molar mesostyle large, conical, sharply demarcated from centrocrista (0); variably developed by forming a ridge on the buccal wall between paracone and metacone, often with joining low cinplar cusp, in a proportion of individuals (A); missing (B). Character 12: Upper molar buccal and lingual cusp outer walls converge at 100" (0); between 90" and 100" (A); at less than 90" (B). Character 13: Lower molar buccal and lingual cusp outer walls converge at 40" or more (0); at less than 20" (1). Character I 4 Lower molar cristid obliqua attaches to trigonid nearer to metaconid than to protoconid (0); midway between metaconid and protoconid (A); nearer to protonid than to metaconid (B). c Figure 2. Dental terminology. A,B, Phenacodus intenedizu Granger (BMNH.M44732 (A) and M982 1 (B)) from Wyoming; C,D, Cymbulophu cuniculus (Owen) (MNHN-TRY-l7L, from Try (C); IRSNB.EFM107, from Erquelinnes); E, Ectocion osbomianus (Cope) (BMNH.M16471); F, Puchynolophus hooken (Godinot) comb. nov. (University of Montpellier, PAT14). A,C are left MI'*, B,D right M,,,, all in occlusal view. E,F are disto-bucco-occlusal views of left showing wear facets numbered according to Butler (1 95Za) and Hooker (1 984). Abbreviations: b = metaconid butress; cent = centrocrista; co = cristid obliqua; ect = ectocingulum; entd = entoconid; entld = entoconulid; hyd = hypoconid; hyld = hypoconulid; hyp = hypocone; mes = mesostyle; met = metacone; lmetd = simple metaconid; Zmetd = twinned metaconid; met1 = metaconule; metsd = metastylid; mf = mesiolingual metacone fold; pacd = paracristid; pad = paraconid; par = paracone; pal = paraconule; pas = parastyle; phyc = prehypocrista; pmec = premetacrista; pmelc = premetaconule crista; poecd = postentocristid; pohycd = posthypocristid; pomec = postmetacrista; popac = postparacrista; popalc = postparaconule crista; poproc = postprotocrista; ppac = preparacrista; ppalc = preparaconule crista; pproc = preprotocrista; procd = protocristid; prot = protocone; protd = protoconid. B-D are shown reversed. Each scale bar measures 5 mm, the short one for A & B, the long one for C-F. 38 J J HOOKER Character 15: Lower molar cristid obliqua attaches high (0); low (1) on back wall of trigonid. Character 1 6 Lower molar entoconulid a distinct cusp (0); a mesial projection near tip of entoconid (A); missing (B). Character 1 7 Lower molar paracristid, when lightly worn, makes angle to tooth long axis of 50" (0); 40" (A); 30" (B); 20" (C); 10" (D). Chizracter 18: Lower molar paracristid with mesiobuccal angle rounded (0); sharp or bulgmg (1). Character 19: Lower molar trigonid back wall shallow (0); steep (1). Character 20: Upper molar ectocingulum unbroken (0); may be broken at paracone (1). Character 21: M'-* as broad as long (0); broader than long (1). Character 22: Upper molar protocone and hypocone separated no more widely than paracone and metacone (0); more widely (1). Character 23: postmetacrista and lower molar paracristid short (0); long 11). Character 24: Upper molar parastyle small (0); medium (A); large (B). Character 25: M3 smaller than M2, M, not larger than M2 (0); M3 not smaller than M2, M, larger than M, (1). Character 2 6 M, hypoconulid as close to hypoconid as this is to entoconid (0); more distant but closer to hypoconid than this is to protoconid (A); as far or further from hypoconid than this is from protoconid (B). Character 2 7 M, hypoconulid forming distal margin of post-talonid lobe (0); bearing more distal lobe (1). Dashzeveg (1979b: 11) has described a cusp and crest nomenclature for this region. Character 28: M3 hypolophid incomplete (0); complete (1). Character 29: P3 metacone smaller than paracone (0); equal to paracone (1). Character 30: P3 with trigon relatively narrow, protoloph weak and paraconule and P, metaconid very small and poorly defined (0) (Godinot, 1981, fig. 22; 1987, pl. 1, fig. f); trigon broader, protoloph stronger, paraconule (usually) and metaconid much larger and better defined, but smaller than protocone and protoconid respectively (1) (Hooker, 1980, fig. 2a; Savage et al., 1965, fig. 3e). Character 31: P3 paraconule midway between parastyle and protocone, P, metaconid very close to or transversely adjacent to protoconid (0); paraconule mesiolingual of an imaginary line joining parastyle and protocone, metaconid well separated distolingually from protoconid (1) (Cooper, 1932, pl. 50, fig. 3; herein Fig. 23). Character 32: Lower canine to PI diastema short (0); long (1). Character 33: Optic foramen of orbit significantly in front of (0) or close to (1) the posterior group of foramina, viz. anterior lacerate (=orbital fissure or sphenoidal foramen), foramen rotundum and anterior opening of the alisphenoid = alar) canal. Character 34: Navicular facet of astragalus convex (0); saddle-shaped (1). Character 3.5: Astragalar canal present (0); absent (1). z Character explanations 'The protoloph-protolophid complex (characters 2-6) Thc relative mesiodistal position of upper molar paraconule and protocone 13) is normally closed linked to whether the metaconid is single or twinned (4). EARLY HORSE EVOLUTION 39 If the paraconule is distant from the protocone, it occludes between the twinned cusps of the metaconid, producing an extensive buccal phase wear facet-‘a’ (see Fig. 2F and Hooker, 1984: 239). If the paraconule is mesiodistally close to the protocone and more or less aligned with it on the protoloph, the metaconid is simple and facet 2a very restricted (see Fig. 2E). There are exceptions. Cardiolophus and Homogalax have the paraconule aligned with the protocone but the metaconid twinned; here, however, the twinning is oblique not mesiodistal, producing a slightly different occlusal pattern (Hooker, 1984: 239). Phenacodus has well separated paraconule and protocone, but single metaconid; here the difference may be due to the different orientation of the preparaconule crista. The orientation of the preprotocrista is closely linked to the relative positions of paraconule and protocone. The simplest way of making its orientation more transverse is to move the paraconule distally relative to the protocone. This suggests that the separated state of the paraconule and protocone, where the preprotocrista is transverse, is not equivalent to that where the preprotocrista is oblique; hence the sequence of states employed in character 3 and its separation from character 4. The twinned state of the metaconid is closely linked to the primitive notched state of the protocristid, again except in Phenacodus and Cardiolophus. It is therefore treated as a separate character (5). The more distal of the two metaconid cuspules has usually been called a metastylid. This is anomalous, as the paraconule occludes between the cuspules. In other mammals the metastylid occludes with the mesiolongual side of the protocone (e.g. Hooker, 1986: 277278, text-fig. 28). Court & Hartenberger (1992) have demonstrated the same pattern of occlusion in the primitive hyracoid Titanohyax and called into question the homology of paraconule and metastylid in this group. It seems unnecessary to question the homology of the paraconule, but the perissodactyl ‘metastylid’ appears non-homologous with that of hyracoids, primates, meniscotheriids or phenacodontids. ‘Qmbalophus’ hookeri has both metastylid (6) and twinned metaconid cuspules (4) and in comparison with Ectocion demonstrates well the cusp homologies (Fig. 2E & F). The metaconid, twinned or otherwise, can be distinguished by its distolingual butress. T h e metalop~-~po~ophid complex (7, 8, 9) The occlusal relationships are similar to but not identical with those of the protoloph and protolophid (Hooker, 1984: 239). Mesiodistal separation of the metaconule and hypocone and thus stepping of the metaloph is not matched by twinning of the entoconid, but by absence of a postentocristid. The occlusal relationships here, however, are consistent among the taxa, and uppers and lowers are combined into one character (9). Polarization of the different perissodactyl states is, however, difficult. The absence of a postentocristid and mesiodistal separation of metaconule and hypocone are regarded as more primitive than the more bilophodont states for two reasons: (1) although phenacodontid hypoconulid and entoconid cusps are close together, a postentocristid appears to be missing; (2) other mammals progress in the same way (e.g. nyctalodonty to myotodonty in bat lower molars, see Menu & Sigt (1971), even though here the occluding crest in the upper molars is a premetacrista not a metaloph). The crest previously identified in perissodactyls as a premetaconule crista J. J. HOOKER Ficqres 3 5. Hallensia louisi sp. nov.; 3 , mesial \riews of upper preultimate molars; 4, holotype, MNHN.Mu-2 18-L (left) from Mutigny; B, MNHNAV-4770 from Avenay (right tooth rrversed); 4. h i c c a l view of incomplete unworn right M , (reversed) (h1NHN.Mu- 12391) h m hlutigny. showing cntoconulid; 5, dorsal view of frapmentaq mandible with symphysis (AV-841 -Ph, cast BhINH.M.5 I 708) from Avenay. Figure 6, I[vrucvracotherium leporinum Owm, disto-occlusal \iew of lcft h l , IRI\lNH.%f51682) from Sheppey. Scale bar for Figs 3, 4, 6 = 5 m m , for Fig. 5 = 20mm. (Hooker, 1989) appears to be partly homologous in Phenucodus with an enamel fold on the mesiolingual metacone wall unassociated with the metaconule. It links up with a neoformed premetaconule crista, which develops mesiobuccally from the metaconule in primitive perissodactyls and gradually disappears as it becomes bypassed by the lengthening crista (8) (all stages in development of the metaloph). Elongation of the premetaconule crista leading to advanced devdopment of the metaloph is not coupled with development of the hypolophid (9). Bucial crest complex ( 10- 18) There is a complex relationship between flexing of the centrocrista (lo), convergence angles of buccal and lingual cusps (12, 13) and orientations of the cristid obliqua (14) and paracristid (17). The state of each structure is clearly influenced to some degree by the others, but the relationships are not consistent among taxa. Several characters have therefore been employed to describe the different elements. The great variation in upper and lower molar cusp convergence angles (from 0 to B) within Hullensiu louisi (Fig. 3A,B) was impossible to code for any one of the states described. The variation is considered to be rather an autapomorphy of the taxon, and both characters 12 and 13 were coded as ‘?’ to avoid biasing the result. The mesostyle (1 1) is variable in both presence/absence and detailed structure in the perissodactyls studied here, although it appears to be constant in phenacodontids. As has been argued elsewhere (Hooker, 1989: 82-85), this cusp probably rcpresents reduction and loss in perissodactyls. It does not seem possible, however, in practice to distinguish between ‘centrocristal’ and ‘cingular’ types. Its ‘reduced’ state (A) usually consists of a narrow blunt ridge running up the buccal wall midway between the paracone and metacone, and this may merge with (Fig. 14) or be distinct from (Hooker, 1989, fig. 6.3C) a small cusp arising from the ectocingulum. Alternatively the ridge may be stronger but lack the cingular cusp (Fig. 18), or the mesostyle may simply be a small conical cusp in the middle of the centrocrista (Fig. 13). In all the species studied here, EARLY HORSE EVOLUTION 41 where such mesostyles occur, there are some individuals which lack a mesostyle altogether. Therefore the only practical solution seemed to be to lump all these in a state intermediate between constant presence and absence. The angle of the paracristid (17) varies slightly with wear and the character states are scored for the lightly worn state. When unworn, the occlusal edge of the paracristid tends to flare buccally towards its mesial end, reducing overall obliquity by about 10”. In contrast, advanced wear increases the obliquity. The difference between Cymbalophus cuniculus and ‘Hyracotheriunz’ sandrae for this character is subtle, with a small mount of overlap. Six specimens of C. cuniculus and 4 of ‘H:sandrae were examined. P3 paraconule enlargement and Py metaconid distal sh$ (31) The normal development of the metaconid leads it to shift from a position very close to but slightly distolingual to the protoconid in its incipient state, to one transversely opposite the protoconid as it enlarges and occludes with the stronger, more transverse protoloph usually associated with a broader trigon (Butler, 195210). The distolingual position of the large metaconid in the 1 state of character 10 is clearly related occlusally to the position and saliency of the paraconule. Optic foramen (33) This has been discussed by MacFadden (1976: 8, fig. 6), who recognized that the derived state was restricted to horses and palaeotheres and the primitive state was widespread in other perissodactyls (e.g. Radinsky, 1965: 72, fig. 2) and non-perissodactyl mammals. His figure shows the position of the optic foramen changing in relation to both the posterior group of foramina and the more anterior ethmoid foramen, whose positions remain constant. Whereas in the derived state this is true of more primitive horses (e.g. Mesohippus, Hyracotherium-Savage et at., 1965, fig. 23), in modern Equus, the ethmoid foramen has migrated posteriorly too (Sisson & Grossman, 1953: 71, fig. 48). The significance of the character is also less clear in view of some seemingly anomalous occurrences. Perhaps it is related to the position of the orbit. A specimen of the brontothere Palaeoyops has been shown to have a posterior optic foramen, although the ethmoid foramen was apparently not visible (Osborn, 1929: 326, fig. 275). Hallensia matthesi and Pachynolophus lavocati appear to have the primitive condition (Franzen, 1990: 190, fig. 7; Remy, 1972, fig. 8). In Pachynolophus liuinierensis (Savage et al., 1965: 44, fig. 20) the state seems intermediate, although this might be influenced by crushing. Pa~aeotherium(Remy, 1992, fig. 24) and Plagiolophus (pers. obs.) also have the primitive condition. A strange occurrence of the derived state is in Ectocion (Thewissen, 1990, fig. 55A), whilst the primitive state occurs in Phenacodus (Thewissen, 1990, figs 53A, 54A). Results of the anabsis The computer program used was PAUP 3.0 (Swofford, 1990). The search for the shortest ‘tree’ was conducted using the branch and bound method. Multistate characters were ordered. Rooting was by outgroup (Phenacodus) with all character states at zero. The analysis produced three topologically different 42 J. J. HOOKER maximum parsimony cladograms of 92 steps, with a consistency index excluding uninformative characters of 0.544 and a retention index of 0.653. The three cladograms differ only in the interrelationships of three terminal taxa. Hyracotherium Lepoiinum, Propachynolophus leuei and P. gaudyi. The relationship which fitted best with stratigraphic occurrence was preferred (i.e, P. leuei and P. gaudyi more closely related to each other than to H. Leporinum). Both ACCTRAN and DELTRAN optimizations were employed, and the characters which showed different distributions according to each (3, 4, 9, 11, 14, 17, 20, 26, 28, 33) are listed in the explanation to Figure 7. Choices between the equally parsimonious but different patterns of character state changes between the two optimizations are discussed below. The choice for characters 3 and 4 is to ensure that equivalent changes appropriate to their occurrence and occlusal relationships took place at the same nodes. In character 9 it is to avoid a 2-step reversal to the primitive statc in H. leporinum. That for character 11 favours reversal in line with dilambdodont trends. That for character 26 avoids unnecessary reversal. The primitive state for character 28 in ‘Propachynolophus’ aff. maldani (see systematic descriptions), which occurs earlier than in typical members of the species, indicates independent acquisition of the derived state in this species and P, gaudyi. If a stem perissodactyl position for brontotheres is correct, acquisition of character 33 in Ectocion plus perissodactyls and subsequent reversal in tapiromorphs is the likely alternative. Major conclusions from the analysis are: ( 1) Hyracotherium leporinum, the type species, nests within the Palaeotheriidae. (2) ‘Hyracotherium’ penzix and ‘H.’ uulpiceps are sister taxa and equids, for which the genus Pliolophus is available. (3) Equidae and Palaeotheriidae are sister groups. (4) CymbaLophus cuniculus and ‘H.’ sandrae are closely related stem group equoids. N.B. ‘H.’ sandrae Gingerich, 1989, is not a Hyracotherium. It may belong in the genus Systemodon Cope, 1881, because of a similar degree of bilophodonty to that of the type species-S. tapirinus (Cope, 1875) (see Radinsky, 1963: 10; Gingerich, 1991: 183, 185, for discussion of the genus, including early incorrect referral of species now included in Homogalax). (5) Pachynolophus hookeri (recombined from Cymbalophus) and thus the family Pachynolophidae are sister groups to Cardiolophus in the Isectolophidae and therefore not equoids. (6) Hallensia louisi (and thus the genus) is stem group to the Pachynolophidae plus Isectolophidae and therefore not an equoid. Pigurt. 7. Maximum parsimony cladogram of 92 steps, best fit with stratigraphy and using combination of ACC’IRAN and DELTRAN options (see text), generated using PAUP 3.0 from ddtd matrix in Table 1. Broad bar = synapomorphy; narrow bar = normal polanty Iiomoplas): X = revwsal. Reversal to 0 state of multistate character indicated by character number in brackets. Optimization differences per numbered node: ACCTRAN: I 3A, 14A, 33; 2 4, $)A, I1A: 14B, 17A-B, 20; 3 l I B , 26A; 4 3B, 14AR, 17AR, 20R; 5 9B, IIAR, 26B, 28; 6 j14)R; 8 9C; 10 9A- (9)R, 20, 28R; I 2 (9)R; 15 4R; 17 SB, 17C, (26)R; 18 (3)R, 17C, 33R; I 9 9B-C, 17D; 20 I I B , 14AR; 21 3A, 28; DELTRAN: I 14A, 2 9A, 11A, 17A, 3 3A, 4 3B, 4, 26A, 33; 5 26B: 6 (14)R; 7 9B; 8 9C, 28; 10 (9)R, 20; 11 9B, 28; 12 (9)R, 11B; 15 l l B , 14B, 1 7 K , 20; 16 26A; 17 9B, 17C; 18 4, 17B-C, 20; 19 9B-C, 17D; 20 11B; 21 3A, 14B, 28; 22 14B; 23 3A, 33. W W W W W W N L d l e Q I - P U u l w N W I - I - - I - N N W W 4I 1 I ; I I I f IM m p Wm W W ? WpwPLnLn.!wowLnwa - " m m 0 2 4" W I - O I O N ""W , 2 ; I A l ' W N 9 a m c . 1 m m W I- ::! v v -. -. - w W m I - W W N W N 8 s 1 I I A A W l e C I W C I N N A1 I W n W O Ww OmI N N N w N A v m N m W P N - P W SANDRAE ECTOCIOIY HALLENSIA L O U I S I CARDIOLOPHUS PACHYNOLOPHUS HOOXERI 'HYRACOTHERIUM' PLIOLOPHUS VULPICEPS PLIOLOPHUS PERNIX I 'PROPACHYNOLOPHUS ' MALDANI N m N N N # 1 0 - n IN II m 1 0 10100 - -. Em 0 3: 1-4 J J HOOKER The Equoidea are supported as a monophyletic group so long as Hallensza and the Pachynolophidae are excluded. If C. cuniculus and ‘H.’ sandrae are included, then it is shown to be defincd by three synapomorphies. Of these, the incipient molarization of P3 (30) is the most important. The other two (loss of metastylid and entoconulid) are paralleled by non-equoid perissodactyls more advanced than those analysed here. Equoids may not be defined by the posterior position of the optic foramen, if the derived state is really present in brontotheres. This region is not yet known for C. cuniculus or ‘H.’ sandrae and the state is reversed in more advanced palaeotheres. The Palaeotheriidae are defined principally by greater elongation of the M3 hypoconulid (26B) and by upper molar characters that are either reversals or paralleled elsewhere: incipient secondary breakage of the link between preparaconule crista and preparacrista, flexing of the centrocrista and probably incipient reacquisition of a mesostyle [although this may never have been completely lost); all three are minor trends in dilanibdodonty which become accentuated in more derived members of the family. An Adams Consensus of a total of thirty 91, 92 and 93 step cladograms iafter removal of P. leuei) shows consistency of relationship between H. ~ ~ p o r i ~ u ~ and P. g a u d y and between P. pernix and P. uulpiceps; ‘P.’maldani, however, is equally related to these two clades and to a fourth, an unresolved polychotomy of ‘H.’ sandrae, C. cuniculus, P. hookeri, H. louisi and Cardiolophus. Stratigraphy The order of branching can be readily compared with that of stratigraphic occurrence only in northern Europe, where time correlation by other organisms is fairly reliable. If one adds the ‘aff taxa (see systematics section) to the ranges of the definitely identified species, the correspondence between the order of stratigraphic appearance and that of the cladogram branches is quitc close, although ‘P.’maldani appears too late (Fig. 8). ‘Propachynolophus’ sp. 1 may be closely related to the latter, and when better known may resolve the problem. In the Clarks Fork Basin, ‘H.’ sandrae certainly occurs below P. pernix (Gingerich, 199l), but lack of detailed correlation with Europe prevents the stratigraphic testing of these branches relative to the rest of the cladogram. l i p r e 8. Chart showing Stratigraphic ranges of ~irimitiveequoids in the Clarks Fork 8~7\’yomingJ, L.ondon and Paris plus Belgian Basins. Rangcs ending in arrows indicate continuation of the or clade; those with T-bars indicate extinction. Transverse arrows indicate suggcsted dispersal. Lorahtics in capitals, lilhostratigraphy in l o w r case (lithostratigraphy of‘ Frcrich and Belgian localities can be found in Systematics st=ction). Lt’a0-4 arc mammal z o n e (Gingerich, 199 I ) . Probable position of Palaeocenr-Eo~.ene boundary shown with heavy line. Peckzchara charophy~r~,, Tt’etzeliella (dinocyst) and NP (,calcareous nannoplankton) are biostratigraphic indicator5 (Hooker. 1991). Anomalies 24A arid B help date the opening of the Denmark Strait and the Fhglish Channrl. Levels brlow 24B in the Clarks Fork and London Basins all Ilelong to the i-~.\-ersedi n t e n d of Cliron 21 (Aubry, 1985). itlthough the Pans and Belgium basins are 1-oml,int.d, rhr perissodactyl record in the latter Is restricted to Dormaal and Erquelinnes. Brackcts indicate rcu orkrd fixsils; the Suffolk Pebble Reds fauna is reworked almost certainly from the Reading Krds, which contain P. dz.wvza;. NR: no time or thickness dimension is signified by widths of atratigmphic units; morrovcr, mammal rangcs s h v n for Europe represent only one or a few spot (recurrences per unit, not a continuous record. ‘l’he North American record is considrrilhly more rstensivc than shown, with a succession of species in !Val-3 that appear closely related to ‘Ff.’ m i d m e (see Gingerich, 1991). EARLY HORSE EVOLUTION I I I I m I3A37 S 'I3NNVH3 I dOTONXh'3VdO8d HSI? -J 45 16 J. J HOOKER The dating of the southern French locality of Palette containing P. hookeri is problematic. It was thought by Godinot et al. (1987) to be coeval with the Dormaal reference level (MP7), although no restricted range taxon was in common with any from Dormaal or a well correlated locality (Brunet et al., 1987). An MP7 age is essentially based on the rodent Pseudoparamys cezannei Hartenberger (in Godinot et al., 1987), which is judged to be of similar evolutionary grade to an undescribed species of Pseudopnramys from Dorrnaal. However, P. hookeri may also occur at Fournes and Monze, which fit within calcareous nannoplankton zones NP10 (high) or NP11 (low) (Marandat, 1991) which almost certainly postdate North Sea Basin MP7 localities, which correlate with NP9 (Hooker, 1991). There are several taxa which are closely related to those analysed here but are too poorly known to be analysed themselves. Some are described herein. Pliolophus aff. uu@ceps from Soissons and Sinceny appears transitional between typical P. uulpiceps and Cymbalophus cuniculus, as single M”‘ and MI,, have the B state of character 9 (the metaloph-hypolophid). Unfortunately, variation cannot be established and the two teeth could simply be members of a variable assemblage displaying state 9A. If not, then C. cuniculus and ‘H.’ sandrae might branch successively below the palaeothere-equid split, rather than form a clade of their own. O n present information, however, this would mean a less parsimonious cladogram. Mexican ‘Hyracothm‘um’ seekinsi Morris, with its large upper molar parastyle and well developed lower molar hypolophid, yet short premetaconule crista and centrally attached cristid obliqua (Novacek et al., 1991: 48-50), is also poorly known but would fit well as a sister taxon to Cardiolophus and thus not be an equoid (cf. Hooker, 1989: 98). It seems clear that it is in deposits of MP7 age or older that we must search to improve the record of primitive equoids in order better to resolve their relationships. PALAEOBIOGEOGRAPHY The cladogram is plotted on a latest Palaeocene-earliest Eocene palaeogeographic map of North America and Europe (Fig. 9). The earliest equoid in each continent is shown by the analysis to be also the most primitive. As the most primitive members of the nearest sister group occur in Europe, this continent is the current best candidate €or the equoid centre of origm. From there, dispersal to North America would have been either via Greenland or via Asia in the latest Palaeocene prior to the palaeothere-equid split. The choice depends on the authenticity of the Asian Eocene equoid record. This is based on isolated teeth and a few jaw fragments of Hyracotherium gabuniai Dashzeveg, 1979a, and Gobih$pus menneri Dashzeveg, 1979b, from Mongolia, and Propalaeotherium sinense Zdansky, 1930, and P. henaangense Young, 1944, from China. The Hyracotherzum has molars with strong bilophodonty , the uppers having a large parastyle; these characters remove it from the Equoidea and ally it with tapiromorphs. The other three species show dilambdodonty typical of either brontotheres or fairly advanced palaeotheres. The close proximity of the M, hypoconulid to the hypoconid (primitive state of character 26) occurs in brontotheres, whereas the derived distant (26B) state occurs in palaeotheres. EARLY HORSE EVOLUTION 47 Figure 9. Maximum parsimony cladogram from Fig. 7, minus Cadiotophus, Ectocioii and Phenacodus, superimposed on palaeogeographic map of North America and Europe around the PalaeoceneEocene boundary (mainly after Tiffney, 1985, map 3). Continuous thin lines indicate ancient shorelines. Thicker lines encircle geographic ranges. Opposed arrows indicate position of opening of Denmark Strait (D) and English Channel (C). 1: Propachynolophus gaud& 2: Propachynolophus ha’; 3: Hyracothmum leporinum et aff.; 4: ‘Propachynolophus’ maldani; 5: Pliolofihus uulpiceps; 6: Pliolophus perniq 7 : Qmbalophus cunuulus; 8: ‘Hyracothenum’ sandrae; 9: Pachynolophus hookni, 10: Hallensia louisi. Moreover, they have an oblique M3 hypolophid, like brontotheres and unlike palaeotheres, where this crest is nearly always transverse. Although meagre, the weight of evidence favours the identification of these three taxa as brontotheres, and no characters make them unequivocal equoids. In view of the abundance of Asian Eocene mammal localities recording a diversity of tapiromorph perissodactyls (Russell & Zhai, 1987), the absence of equoids appears real and supports the westward Greenland route of equoid dispersal to North America. Speciation of palaeotheres in the north European area seems to have been endemic, there being no authenticated records of the family in North America. Land bridges existed between North America and western Europe via Greenland at this time (McKenna, 1983). However, it seems that by magnetic Anomaly 24 time, rifting east of Greenland had all but changed from subaerial to submarine (Anderson, 1988) and cut off the intercontinental land route. At the same time the London Clay sea (division B, calibrated to zone NP11 and Anomaly 24B) had reached its maximum extent (King, 1981; Aubry, 1985; Townsend & Hailwood, 1985). This was when the first influx of nummulites and planktonic foraminifera into the North Sea Basin took place, an event usually taken to represent the opening of the English Channel and consequent separation of the British and continental west European landmasses (Curry et a k , 1978: 10; Murray & Wright, 1974). P. pernix appeared abruptly in the Clarks Fork Basin, Wyoming, just above the 2000 m level but still within Chron 24R (Gingerich, 1991; Butler, Gingerich 48 J. J. HOOKER & Lindsay, 1981). As P. vulpiceps is slightly more primitive, it is likely that dispersal took place a second time from Europe. It could not have been long after this that interchange of terrestrial faunas between North America and Europe ceased. The opening of the English Channel may have caused the isolation of Hyrarotherium leporinum in Britain and Propachynolophus in France. P. uulpiceps apparently became extinct in France earlier than in England. Because of a poor mammal record in England in the very late Palaeocene (Soissons-Meudon time), it is not certain whether P. aff. uulpiceps or H. aff. leporinum existed there then as well as in northern France. H. aff. leporinum therefore either dispersed there or differentiated into typical H. leporinum in England and into the Proparhynolophus levei-gaud$ lineage in France. It is interesting that, overall, more primitive morphotypes survived longer in southern England than in northern France. In contrast, none of the isolated English species appears to have given rise to latcr equoids, whereas differentiation in northern France led to more derived palaeotheres, which radiated through the remaining Eocene. PA1AEOECOLOGY ‘The dental variation in bilophodonty, selenodonty and bunodonty in these primitive perissodactyls reflect adaptations to diets of leaves, fruits or complex mixtures of the two. The pattern of frequent parallelisms and reiersals in these adaptations suggests fluctuating conditions and may reflect the climatic changes documented around the Palaeocene-Eocene boundary (Koch et al, 1992), pcrhaps influencing length of fruiting seasons. Trends in dilambdodonty, producing a seleriolophodont or bunoselenodont dentition, may have allowed more efficient mastication of both fruit and leaves by specialization in different areas of the tooth. Low crowned selenolophodont ‘Propalaeotherium’ messelense (Haupt) had a diet of leaves and occasionally fruits, according to gut contents preserved at Messel (Franzen, 1988). The different early perissodactyls may also have occupied different microhabitats. A given European site tends to be dominated by one species: e.g. P. uulpiceps at Abbey Wood and Pourcy; Hallensia louisi at Mutigny. SYSTEMATICS This section contains descriptions of the new material, new species and new combinations. For the new material, length and width measurements in millimetres are given in that order, separated by a ‘ x ’ in the ‘Material’ sections; in figures brackets are estimates. Abbreviations are as follows: BMNH = Natural History Museum, London, MNHN = Muskurn National d’Histoire Naturelle, Paris; the suffix Ph attached to a specimen number indicates the private collection of M. A. Phklizon. The synonymy format follows Matthews (1973). Suborder Tapiromorpha? family uncertain Genus Hallensia Franzen & Haubold, 1986 l j p e sfieries. H. matthesi Franzen & Haubold, 1986, from the Middle Eocene of Geiseltal, Germany. EARLY HORSE EVOLUTION 49 Included species. H. parisiensis Franzen, 1990, Sables a Unios et TCrtdines, Monthelon (Marne), France, H . louisi sp. nov. Hallensia louisi sp. nov. (Figs 3-5, 12-15) v. 1922 v? vp. vp. 1922 1965 1965 Propachynolophus nov. sp.; Teilhard de Chardin, p, 69, fig. 33C, pl. 3, fig. 30. Hyracotherium; Teilhard de Chardin, p. 52, fig. 26B. Hyracotherium; Savage, Russell & Louis, pp. 6-8, figs 2d, f, i, j. Hyracotherium larger form; Savage, Russell & Louis, pp. 10 & 11, 13, figs 4f-h. Egmology. After M. Pierre Louis, the collector of most of the material known of this species. Holoppe. Left M”‘ (MNHN.Mu-218-L), from the Argiles a Lignites d’Epernay, Mutigny (Marne), France (Figs 3A, 13) (8.0 x 10.2). Parappes. Right M3 (MNHN.Mu-12371) (Fig. 14) (7.9x9.3); 2 left M3s (MNHN.Mu-201-L, Savage et al., 1965, fig. 2c and Mu-221-L, Savage et al., 1965, fig. 29; fragment of right dentary with MI-, (MNHN.Mu-12303, Fig. 15) (6.8 x 5.6; 7.8 x 6.5); 2 right M I p (MNHN.Mu-6283, Savage et al., 1965, fig. 2d; Mu-12391, Fig. 4); all from Mutigny. Referred material. Right MI’‘ (MNHN.Av-4770, Savage et al., 1965, fig. 4f); part of left and right dentaries with symphysis, worn P,-M,, right P,-, and left and right P, alveoli (Av-841-Ph) cast BMNH.M51708); 2 left M3s (MNHN.Av65519, Savage et al., 1965, fig. 4g; Av-(61)14685 (broken mesially)); from the Argiles a Lignites d’Epernay, Avenay (Marne), France. Right P3-M1 (Louis Colln, MNHN) (7.9 x 8.4; 7.6 x 9.3; 8.1 x 9.6) from the Sables de Guise, Condten-Brie (Aisne), France. Buccal half of left MI-‘ (MNHN.Py-16-CN) from the Falun de Pourcy, Pourcy (Marne), France. Left DP3 (Lepage Collection, cast BMNH.M44757, Fig. 12) from the Sands of Dormaal, Brabant, Belgium. Doubful material. Right M, minus trigonid (Laboratoire de Gtologie, Universitt de Marseille; Teilhard de Chardin, 1922, fig. 26B) from the ConglomCrat de Meudon, Meudon (Hauts-de-Seine), France. Diagnosis. Small species of Hallensia (MIp 8mm long); upper molars with distinct but small parastyle on MI-’, variable on M3; upper molar premetaconule crista, short, joining prominent mesiolingual metacone fold. D~erential diagnosis. H. parisiensis is slightly larger and H. matthesi larger still. Both have the upper molar parastyle so reduced that it is scarcely differentiated from the cingulum that almost encircles the tooth. The extent of the cingulum, used to differentiate H. parisiensis from H. matthesi, seems to be too variable to be a useful character. H. matthesi has a longer premetaconule crista. Description and discussion. The keynote of the dental characters of Hallensia is variability. This is particularly true of the convergence angle of buccal and lingual cusps of both upper and lower molars (Fig. 3). This varies between the primitive and most advanced states of characters 12 and 13 (above). The upper molar centrocrista may be straight or slightly flexed buccally, with or without mesostyle at the crest of the flexure (cf. Fig. 14 with Savage et ak, 1965, fig. 2f,g). The enamel surface may be strongly rugose or smooth. The very variability appears to be an autapomorphy of the genus rather than an incipient 50 J. J. HOOKER trend towards the states of any of the other taxa treated here, where convergence angles and other characters are intraspecifically constant. Some of the more characteristic dental features are: the lingual position of the metaconule and the notch separating it from the hypocone on upper molars; and the relatively high lingual cusps of both upper and lower molars. If the upper molar buccal and lingual cusp convergence is high, as in the holotype and Mu-12303, the cusp tips approach one another rather closely (Figs 13, 15). l’he notch between the metaconule and the hypocone is therefore also distinctly lingual. This can be seen well in the M’s from hlutigny and the Dorrnaal DP3. Unfortunately, a distinct metaconule with a lingual notch is not a constant feature, although apparently the small size of the parastyle on 311-2 . 1s. ?‘he broken mandible (Fig. 5) has teeth that are both very worn and corroded ,perhaps by a predator’s stomach juices). It shows a short symphysis broken immediately in front of the P, alveoli, suggesting a very short postcanine diastema, as in H. matthesi (see Franzen & Haubold, 1986). The post-P, diastema is of moderate length. Few diagnostic features can be gleaned fr-om the tooth crowns, except the relatively broad inflated outline of MI-*, which compares well with those in Figure 15. Fami& Pachynolophidae Pavlow, 1888 Discussion. Hooker (1989) placed the genera Pachynolophus and Anchilophus in this family and in superfamily Pachynolophoidea, as sister group to the Equoidea comprising Equidae plus Palaeotheriidae. The new analysis shows Pachynolophidae not to be equoids and makes the Pachynolophoidea unneccssary. Genus Pachynolophus Pomel, 1847 73’pe species. Pachynolophus duualii Pomel, 1847. Included species. P. livinierensis Savage, Russell & Louis, 1965; P. cesserasicus Ger\,ais, 1849; P. garimondi Remy, 1967; P. lauocati Remy, 1972; P. boixedaten.ri,s Crusafont & Remy, 1970; P. hooken (Godinot, 1987); P. sp. nov. (‘Hyracotheriurn’ aff. cunzculus Owen (Godinot, I98 I)). Emended diagnosis. Primitive perissodactyl characters: P3 with narrow trigon, weak protoloph and essentially no paraconule; P, with very small, poorly differentiated metaconid; upper molars with preparaconule crista confluent with preparacrista and with unstepped protoloph; lower molars with twinned metaconid cuspules, concave protocristid, cristid obliqua joining back of trigonid midway between protonid and metaconid, and steep protoconid backwall; buccal and linLpal cusps converge at 90” on upper molars; and at 40,‘ on lower molars. Derived characters: upper molars with unstepped metaloph, without or with weak metaconule and no mesostyle; lower molars with straight complete hypolophid and postcingulid extending lingual of hypoconulid on M, :a; postcaninc diastema long. Discussion. The advanced bilophodonty of the metaloph-hypolophid complex relates Pachynolophus and other genera in the family with primitive tapiromorphs. EARLY HORSE EVOLUTION 51 Figures 10 & 11. Puciynolophus hooken’ (Godinot) comb. nov. from Palette, occlusal views. 10, left P’-M3 (FSL.2641); 11, left MI-, (UM.PAT14). Scale bar = 5mm. Photographed from casts coated with ammonium chloride. Pachynolophus hookeri (Godinot et al., 1987) comb. nov. (Figs. 2B, 10, 11) v. 1965 %. ? 1987 1991 Hyracotherium cf. cuniculum Owen or Propachynolophus sp.; Savage, Russell & Louis, pp. 14-15, fig. 6. Cymbalophus hookeri Godinot et al.: pp. 281-283, pl. 1, figs f-1. cf. Cymbalophus hookeri Godinot ou Propachynolophus sp.; Marandat, pp. 113-115, pl. 7, figs 1-4. Dzstribution. Early Eocene of Palette (Provence) (type locality) and probably Fournes (Minervois) and Monze (Corbieres); all southern France. Emended diagnosis. Very small Pachynolophus (M‘ 5.43-5.72 mm long). Primitive features are: upper molars with straight centrocrista, width only slightly greater than length, premetaconule crista short joining prominent mesiolingual metacone fold, no flattening of metacone buccal wall, ectocingulum broken at paracone, M3 not longer than M’ or M2; lower molars with entoconulid high on mesial entoconid wall, metastylid retained, longitudinal paracristid with mesial bulge and M, with incomplete hypolophid. Discussion. Godinot (1987) has thoroughly described this species, recognizing its mosaic of primitive and advanced features. However, the analysis here shows that it does not belong to Cymbalophus. A closely related species still unnamed from the Early Eocene of Rians (Provence), but referred by Godinot (1981) to Hyracotherium aK cuniculus and by Hooker (1989) to ‘Hyracotherium’ sp., is slightly more advanced than P. hookeri in a number of characters: it is slightly larger; the upper molars are distinctly broader than long and their premetaconule crista is longer; the lower molar paracristid is oblique; and M3 has a complete J. J. HOOKER 32 hypolophid. The lower molars are too worn to detect if they retain metastylid or entoconulid. The grade of molarization of P, (very small metaconid) fits that of P3 of P. hookeri (narrow trigon with weak protoloph). P, is not yet known for P. hookeri and part of the assessment of character 31 above was based on the species from Rians. Suborder Hippomo?pha Superfarnib Equoidea Gray, 1821 Farnib Equidae Gray, 1821 Genus Pliolophus Owen, 1858 q p e species. Pliolophus vulpiceps Owen, 1858. Included species. ‘Hyracotherium’ pernix Marsh, 1876, and probably also ‘H.’ uasacciense (Cope, 1872) should be placed in Pliolophus. Pliolophus vulpiceps Owen, 1858 (Figs 22 & 23) (For earlier synonymy, see Hooker, 1980, under H. vulpiceps and H. aK. z‘u@zceps). vp. v. v. \,. 1965 1980 1980 1989 Hyracotherium; Savage, Russell & Louis, pp. 8-10, figs 3c, e, f. Hyracotherium vulpiceps (Owen); Hooker, pp. 104, 106, 108, fig. 3. Hyracotherium aK. vulpiceps (Owen); Hooker, pp. 106-1 08, fig. 2. Hyracotherium aff. vulpiceps (Owen); Hooker, pp. 83-84, figs 6.3L, 6.41-K. HoloQpe. Partial skeleton (see Hooker, 1980) (BMNH.44 1 1.J, 44 1 15a & M10657-61), from the Harwich Stone Band, Harwich Member, London Clay Formation, Harwich, Essex, England. Xeu material. Left P2 (MNHN.PY-157 15) (5.2x 4.5); left P4 (MNHN.unnumbered) (6.3 x 7.5); 1 left and 3 right M’% variously damaged (MNHN.PY2431, PY unnumbered (Fig. 22) (7.1 X-), PY-42-L, PY-72-L); 1 right and 2 left M’s (MNHN.PY-41-L (Savage et al., 1965, fig. 3c), PY-39-L ( 7 . 4 9.0), ~ PY-40-1, (ibid. fig. 3d)); left D F (MNHN.PY-46-L, ibid. fig. 3b); 2 left P,s (MNHN.PY-45-L, ibid. fig. 3e; PY-34-Cn) herein, Fig. 23, 6.4 x 3.6:; 2 left P,s (MNHN.PY-3-L (Savage et al., 1965, fig. 3f); PY-20-Ph, 6.2 x 4.3); left lower molar trigonid fra<pent (MNHN.PY-11-L). All from the Falun de Pourcy, Pourcy (Marne), France. Description and discussion. This species was previously definitely recorded only from the London Clay of the London Basin. Two P3s from Pourcy show variation for position of the inetaconid. One shows this cusp lingual of the protoconid (Savage et al., 1965, fig. 3e), the other shows it distinctly distolinLgual of the protoconid (Fig. 23), the advanced state of character 31 as in the holotype (Cooper, 1932, pl. 51, fig. I). The distolingual position indicates a P3 with a large mesiolingual paraconule, as in the holotype and other specimens from the London Clay. The only P’ and P, from the Blackheath Beds of Abbey Wood (referred to Hyracotherium aff. vulpiceps by Hooker, 1980) have the primitive state for character 3 1, but the Pourcy specimens indicate a variability Figures 12-15. Hallensia louisi sp. nov.; 12, left DP3 (original Lepage Collection, cast BMNH.M44757) from Dormaal; 13, holotype left M”‘ (MNHN.Mu-218-L) from Mutigny; 14, paratype right M’ (reversed) (MNHN.Mu-12371) from Mutigny; 15, paratype right M, (reversed) (MNHN.Mu-12303) from Mutigny. Figures 16 & 17. Pliolophus aff. uulpiceps Owen; 16, right MI,, (reversed) (MNHNSN-34-BN) from Soissons; 1 7, left MI’* (MNHNSY-4-De) from Sinceny. Figure 18. ‘Propachynolophus’ aff. maldani (Lemoine), left MI/* (MNHN.AV-1012-L) from Avenay. Figure 19. ‘Pfpachynolophur’ sp. 1, left MIr2 (MNHN.PY-133-L) from Pourcy. Figure 20. Hyrucotha’um aK leporinum Owen, damaged right M’” (reversed) (MNHN.Mu-12329) from Mutigny. Scale bar = 5mm. All occlusal views, photographed from casts coated with ammonium chloride. w Ln 54 J. J. HOOKER which it is impossible to test in the other assemblages because so few specimens exist. For this reason, the Abbey Wood assemblage is here referred to this species. It is possible that acquisition of the derived state of the P3s spread through time from only some individuals to the entire population. It is significant that the largest P3 paraconules (e.g. Cooper, 1932, pl. 50, figs 1, 3) belong to specimens occurring in phosphatic nodules (reworked into the Pliocene Red Crag), which typift- levels of the London Clay above division 123 (C. King, pers. comm.) and are thus younger than the holotype or the Pourcy or Abbey IYood assemblages. The other teeth show variation for such features as presence/absence of upper molar mesiolingual metacone fold, but are constant in their near straight upper molar centrocrista, strongly stepped metaloph, with distinct metaconule, and strongly convergent lower molar buccal and lingual cusps. One M3 (PY39-L) and the upper molars of BMNH.38801 (see Cooper, 1932: pl. 50, fig. 1; Hooker, 1980) have a small mesostylar ridge on the buccal wall; all other specimens examined lack one, and the species is coded state 2 in Table 1. Pliolophus aff. uulpiceps Owen, 1858 (Figs 16 & 17) ,thzteriul. Left MI/' (MNHN.SY-4-De) (7.3 x 8.9) from the Sables de Sinceny, Sinceny (Aisne), France; right (MNHN.SN-34-BN) (7.2 x 5.1) and left M, lacking hypoconulid (MNHN.Braillon Colln unnumbered) (- x 5.5) from the Argiles a Lignites du Soissonnais, Seminaire Pit, Soissons (Aisne), France. Description. The M"' shows a fairly weakly stepped metaloph with short premetaconule crista, joining mesiolingual metacone fold, and weak metaconule. It is otherwise typical of P. uulpiceps. The M,/' has a complete notched hypolophid, but is otherwise typical of P. uulpiceps. Only one of 6 individuals from Abbey Wood has MI,? with a complete hypolophid (Hooker, 1989, fig. 6.41). Discussion. These teeth cannot reliably be identified with P. uu@ceps because of certain primitive traits. They are nevertheless apparently very close and extend the range of the genus Pliolophus down to the latest Palaeocene in the Paris Basin. Fami& Pulueotheriidue Bonupurte, 1850 Genus Hyrucotherium Owen, 1841 7jpe species. Hyracotherium leporinum Owen, 1841. As the cladistic analysis shows this species to be nested within the Palaeotheriidae, the concept of the genus is restricted to the type species, other described species being distributed in other genera (Pliolophus, Cymbalophus, Systemodon, Pachynolophus, Hallensiu). Hyrucotherium leporinum Owen, 184 1 (Figs 6, 21) Hologpe. Facial part of cranium with alveoli or roots of both canines, both P's, left P2 3, left M7 and crowns of right P2-M3 and left F-M2 (BM(NH) no Figure 21. Hyrucotherium leporinum Owen, left dentary fragment with M2-3 and damaged M , (BMNH.M51682) from Sheppey; A, occlusal view; B, buccal view; C, lingual view. Figures 22 & 23. Pliolophus uulpiceps Owen, from Pourcy, occlusal views; 22, right MIr2 (reversed) (MNHN.unnumbered); 23, left P, (MNHN.PY-34-CN). Scale bar = 5mm. Coated with ammonium chloride; figs 22-23 are casts. J. J. HOOKER 56 M16336) from division B of the London Clay Formation, Studd Hill, near Herne Bay, Kent. Topo&pe. Palate with left and right P2-M3 (Sedgwick Museum, Cambridge no C2 1361). .New material. Part of left dentary with M2-3, the hypoconid part of M I and parts of the roots of P, from the London Clay Formation (not in situ, beach collected) probably division D or E (which form the adjacent cliffs), near Barrows Brook, west of Warden Point, north coast, Isle of Sheppey, Kent, BMNH.M51682. M,: 8.9x6.3; M,: 10.8x5.9. Description. The Sheppey specimen (M51682) has a lightly naturally worn M3 with pro<gressively heavier wear forwards to M I . M3 is complete but M2 has much of the lingual side of the metaconid broken away. This makes it difficult to judge whether there were twinned metaconid cuspules, although mesiodistal extent of this cusp complex suggests that they were present. In contrast, the mesiodistally shorter metaconid of M, lacks distinct twinning (often a feature of equoid M,s). The most striking feature of the teeth is the narrowness of the talonid basins. The cristid obliqua is straight and joins the protolophid just lingual of the midline of the tooth and extends high up it. This contrasts with H. uulpiceps and H. aff. uulpiceps, where the cristid obliqua is lower mesially and joins the protolophid at the midline. There is no postentocristid and the posthypocristid is deeply concave distobuccally. These features plus narrowness of the teeth makes them relatively elongate and restricts the area of the talonid basins. The cingulum is complete buccally. The M, hypoconulid terminates a long narrow lobe. A weak, mesially bowed crest links the hypoconid and entoconid on M, and M,. Di.rcussion. Despite the high convergence angle of the buccal and lingual cusps of the upper molars, that of the lowers is much lower. There are considerable similarities with Propachynolophus gaudvi (see Teilhard de Chardin, 1922; Savage et al., 1965) and P. levei sp. nov. (Fig. 24), the main differences being the lingually broken lower molar hypolophid and strongly stepped upper molar metaloph. These similarities, as well as reasonable occlusion obtained by manipulating the lowers and uppers, strongly support attribution of the new dentary to Hyracotherium leporinum. The large mesially situated metaconule and mesially bowed hypoconid-entoconid crest may be autapomorphies of the species. A notable skull feature of H. leporinum is the shallow narial incision, which extends almost to the front of Pi (Owen, 1841, pl. 21, fig. 3). Admittedly, P' has a more anterior position than in more advanced palaeotheres because of the presence of a post-PI diastema, but the incision is significantly deeper than in P. uu@ceps, where it scarcely reaches the canine (Owen, 1858, pl. 2). This is further support for the inclusion of Hyracotherium in the Palaeotheriidae. Hyracotherium aff. leporinum Owen, 1841 (Fig. 20) v. v. 1922 1965 Hyracotherium; Teilhard de Chardin, p. 52, fig. 26C. Hyracotherium; Savage, Russell & Louis, p. 9, fig. 3g. Muterial. Right (Laboratoire de Gkologie, Universiti. de Marseille) from EARLY HORSE EVOLUTION 57 the Conglomkrat de Meudon, Meudon (Hauts-de-Seine), France (Teilhard de Chardin, 1922, fig. 26C). Right M"' (MNHN.PY-146-LX) (7.2 x 8.4) and left M, (MNHN.Py-43-L, Savage et al., 1965, fig. 3g), from the Falun de Pourcy, Pourcy (Marne), France. Right MI/', broken mesiobuccally (MNHN.Mu-12329; Fig. 20) from the Argdes Lignites d'Epernay, Mutigny (Marne), France. Description. These few broken and worn teeth have the diagnostic characters of H. leporinum but are smaller. They extend the record of the now restricted genus Hyracotherium (possibly the species H. leporinum) back to the latest Palaeocene (Conglomkrat de Meudon) in the Paris Basin. The upper molars show the large mesial metaconule, strongly convergent buccal and lingual cusps and incipient dilambdodonty. The lowers show the obliquely orientated cristid obliqua, which extends high up the back of the trigonid, lingual of its midline. Both upper molars also show a mesostyle, variably present in typical H. leporinum. Genus Propachynolophus Lemoine, 1891 l j p e species. P. gaudvi (Lemoine, 1878) from the Sables a Union et Ttrtdines of the Epernay area, France. Propachynolophus levei sp. nov. (Fig. 24) Epmology. After Dr J. Levt, who collected the holotype. Holoppe. Left maxilla with P3-M3 (Levt Collection, casts in MNHN and BMNH.M49399) from the Sables de Cuise, Condt-en-Brie (Aisne), France. P3: 6.6 x 7.5; p: 7.1 ~ 8 . 8 M'. ; 8.1 ~ 9 . 7 M2: , 9.5 x 11.3; M3: 8.7 x 10.0. Referred material. Two right M'/'s (MNHN.AV-842-Ph) (8.5 x 10.3) from the Argiles a Lignites d'Epernay, Avenay (Marne), France; and (MNHN.SZ-7337) (8.2 x 10.5) from the Sables A Unios, Stzanne-Broyes (Marne), France. Lhagnosis. Small Propachynolophus, M' 8.1 mm long. Upper molars with metaloph weakly to strongly stepped, with metaconule of variable saliency; premetaconule crista may be elongate mesiobuccally or shorter, but the mesiolingual metacone fold is largely missing; centrocrista slightly flexed buccally but mesostyle absent in the three available individuals. Description. Only upper teeth are known but they are mainly very similar in morphology to the type species, although lacking the constancy of certain features. P. hei thus shows generally more primitive traits. The P3 shows a strong protoloph detached from the preparacrista but with a weak but double paraconule; postprotocrista and metaconule are completely lacking; the protocone is in a relatively mesial position opposite the paracone and its lingual wall interrupts an otherwise encircling cingulum. P" differs from P3 in being more transversely elongate, the cusps forming an isosceles triangle, having a continuous cingulum, large paraconule and small metaconule. On the preultimate upper molars of all three specimens, the protoloph and metaloph are detached at their buccal ends from more buccal cusps or crests. There is, however, variation in length of the premetaconule crista; it is longer in the Avenay and SkzanneBroyes specimens than in the holotype, its elongation being associated with weakness of the metaconule. The well marked metaconule of the holotype is Figure 24. Propachynofophus fevei sp. nov. holotype left P”-M’ (orignal Lcvk Collection. cast RMNH.RT49399) from CondP-cn-Brie. 1Figures 25 & 26. ‘Propachynolophus’ sp. 1 ; 25, right P’ ~ M ”(reversed) (Ipswich Museum, unnumbered\ from ?Fdlkenham: 26, right R I ’ (rwerscdj (MNHN.Collii Berton, 196-1. innurnbered) from Sincrny. Scale bar = 5 mm. Occlusal views, coated with ammonium chloride, figs 2.1 and 26 arc casts. EARLY HORSE EVOLUTION 59 nevertheless smaller and less mesially salient than in Hyracotherium leporinum. The Avenay specimen contrasts with the holotype in having a large parastyle. M3 in the holotype is only slightly contracted distally but differs from M’-2 mainly in having no metaconule and in the protoloph contacting the preparacrista. Discussion. The slightly more primitive appearance and smaller size, together with earlier occurrence, suggests that P. levei evolved into P. gaudryi. The constant mesostyle absence (1 1B) may reflect the inadequate sample and it may really have been as variable as P. gaudryi. ‘Propachynolophus’ aff. maldani (Lemoine, 1878) (Fig. 18) vp. 1965 Hyracotherium; Savage, Russell & Louis, pp. 6, 8, figs 2a,e. Material. Right P (MNHN.Mu-5919, Savage et al., 1965, fig. 2a); left M3 (MNHN.Mu-104-Ph) (7.6 x 8.6); and two left M,s broken mesially (MNHN.Mu6567, Savage et al., 1965, fig. 2e; Mu-1210) (-x 5.2) from the Argiles a Lignite d’Epernay, Mutigny (Marne) France. Right P4 (MNHN.AV-42-Ph) (5.7 x 6.8) and left M”‘ (MNHN.AV-1012-L) (7.4 x 9.1) from the Argiles a Lignite d’Epernay, Avenay (Marne), France. Left M,12 (MNHN.SZ-7336) (7.9 x 5.4) from the Sables a Unios, SCzanne-Broyes (Marne), France. Description. Size is very similar to that of ‘P.’ maldani (see Savage et al., 1965: 27-34, figs 14 & 15). The M”’ is nearly identical in morphology to their fig. 14d, but lacks a premetaconule crista. The M,s on the other hand show a lingually broken instead of a complete hypolophid (cf. their fig. 2e with their fig. 15a-f). The MI/‘ in contrast has a complete hypolophid, as in typical ‘P.’ maldani. Discussion. These specimens occur earlier than typical assemblages of ‘P.’ maldani from the Sables i Unios near Epernay. The cladistic analysis clearly shows Propachynolophus to be polyphyletic, if ‘P.’ maldani is included. ‘P.’ maldani has a unique combination of advanced characters, even though they are individually shared with other taxa. Its trends in dilambdodonty (characters 10 and 11) as well as high convergence angle of lower molar buccal and lingual cusps (character 13), while maintaining the distinctness of the upper molar metaconule, link the species with ‘Propalaeothenum’ parvulum (Laurillard) and ‘Propalaeothenum’ messelense (Haupt) (if distinct) and with Lophiothenum Gervais. The latter differ mainly in more convergent upper molar buccal and lingual cusps, constant mesostyles and trends in premolar molarization. They also tend to have very long postcanine diastemata (unknown for ‘P.’ maldani). Savage et al. (1965: 67-72) argued against ‘P,’ messelense belonging to Lophiotherium as originally assigned by Haupt (1925). It seems, nevertheless, together with ‘P.’ parvulum, to be very close to Lophiotherium. As Lophiotherium currently comprises a lineage of three ‘chronospecies’, the possibility of enlarging its concept to include ‘P,’ parvulum, ‘P.’ messelense and F.’ maldani would be seriously worth investigating. J. J. HOOKER ti0 ‘Propachynolophus’ sp. 1 (Figs 19, 25 & 26) 1‘. 1965 Hyracotherium (cf. cuniculum Owen-the Russell & Louis, pp. 10-13, figs 4a-e. small animals); Savage, New material. Left MI’* (MNHN.PY-133-L, Fig. 19) (6.4 x 7.8) fi-om the Falun de Pourcy, Pourcy (Marne), France. Left P3 (MNHN.unnumbered, Cahuza Colln) (5.4 x 5.8), left P4 (MNHNSY-5-De) (5.9 x 6.6) and right M3 jMNHN.unnumbered, Berton Colln 1984, Fig. 26) (6.7 x 7.5) from the Sables de Sinceny, Sinceny (Aisne), France. Crushed cranium, with abraded left and right P” M”, largely embedded in phosphatic nodule (Ipsuich Museum, unnumbered, Fig. 25), Red Crag (reworked from London Clay, probably above division A3), ?Falkenham, Suffolk, England; right P3: 5.4 x 6.0; P’: 5.8 x 7.0; h f ’ : (6.6)x 7.3-t; M’: (6.7) x -; M3: 6.6 x 7.2+. Description. The teeth described by Savage et al. (1965), although small, clearly do riot belong to Cymbalophus cuniculus. The upper molar (no lower teeth known) shows trends in dilambdodonty (including mesostyle), convergence of buccal and lingual cusps and distinct stepping of protoloph and metaloph, with discrete metaconule, very similar to ‘P.’ maldani, although the teeth are distinctly smaller. The new teeth are similar except that they show little (Pourcy) or no (Sinceny) sign of centrocristal flexing and no development of a mesostyle. The damage that the teeth of the Falkenham cranium suffered during reworking has removed most useful features of the molars, particularly buccally. However, the premolars are better preserved and P3 shows a broad trigon with strong protoloph and paraconule and slightly distally placed protocone. The worn upper molar metalophs show stepping but no clear metacanule swelling and the specimen may not be conspecific with those from the Paris Basin. Discussion. The material indicates at least one distinct species of \7ery small equoid, probably closely related to P ’ maldani and therefore a palaeothere. It is neither named nor included in the cladistic analysis here, as at least part of the lower dentition is required before its characters can be assessed accurately. .\CKNOLVLEDGEMENTS I would like to thank the following for access to collections in their care: Dr D. E. Russell (MNHN, Paris), Dr P. D. Gingerich (University of Michigan, Ann Arbor), Mr R. A. D. Markham (Ipswich Museum), Drs B. Engesser and F. Wiedenmeyer (Naturhistorisches Museum, Basel), Dr M. Hugueney (Facultt des Sciences, Lyon) and Dr M. Godinot (USTL, Montpellier 11). Special thanks t o illr J. Bruce for making known to me the lower jaw of Hyracotherium leporinum that he collected and for presenting it to the Natural History Museum, London (BMNH). I am grateful to Dr P. L. Forey for help with thc computing. Professor P. M. Butler, Dr D. E. Russell, Dr N. Court, Dr P. Tassy, Dr M. Godinot and Dr J. A. Remy provided stimulating discussion; and Dr N. Court, Mr D. Froehlich, Dr D. E. Russell and Dr. J. G. M. Thewissen critically read the manuscript. Mr P. Hurst of the Photo Unit, NHM, produced the photographs. EARLY HORSE EVOLUTION 61 REFERENCES Anderson MS. 1988. Late Cretaceous and Early Tertiary extension and volcanism around the Faeroe Islands. In: Morton AC, Parson LM, eds. Ear4 Tertiaty Volcanism and the Opening of the NE Atlantic. Geological Society Special Publication 39: 115-122. Aubry M-P. 1985. Northwestern European Paleogene magnetostratigraphy, biostratigraphy, and paleogeography: calcareous nannofossil evidence. Geology 13: 198-202 Bonaparte C-L. 1850. Conspectus Systematum. 4 charts, Leiden. Brunet M, Franzen JL, Godinot M, Hooker Legendre S, Schmidt-Kittler N, Vianey-Liaud M, coordinators. 1987. European reference levels and correlation tables. Munchner Geowksmchaftliche Abhandlungen (A)lO: 13-31. Butler PM. 1952a. The milk-molars of Perissodactyla, with remarks on molar occlusion. Proceedings of the <oohgical Society of London 121: 777-817. Butler PM. 195213. Molarization of the premolars in the Perissodactyla. Proceedings of the <oological Sociep .f London 121: 819-843. Butler RF, Gingerich PD, Lindsay EH. 1981. Magnetic polarity stratigraphy and biostratigraphy of Paleocene and lower Eocene continental deposits, Clark's Fork Basin, Wyoming. Journal of GeoloQ 89: 299-3 16. Cooper CF. 1932. The genus Hyracotheriurn. A revision and description of new specimens found in England. Philosophical Transactions of the Royal Society (B)221: 43 1-448. Cope ED. 1872. On a new genus of Pleurodira from the Eocene of Wyoming. Proceedings of the Amerkan Philosophical Society 12: 1-6. Cope ED. 1875. Systematic Catalogue of the Vertebrata of the Eocene of N m Mexico collected in 1874. Report to the Engineer Department, U S . Army, G.M. Wheeler in charge. Washington 5-37. Cope ED. 1881. A new genus of Perissodactyla diplarthra. American Naturalist 15: 1018. Court N, Hartenberger J-L. 1992. A new species of the hyracoid mammal Ztanohyrax from the Eocene of Tunisia. Palaeontolou 35: 309-3 17. Crusafont Pair0 M, Remy JA. 1970. Les Equoidea (Perissodactyla) de 1'Eoctne prkaxial pyrtneen espagnol. Bulletin du Musium National d'Histoire Natumlle (2)42: 428-434. Curry D, Adams CG, Boulter MC, Dilley FC, Eames FE, Funnell BM, Wells MK. 1978. A correlation of Tertiary rocks in the British Isles. Geological Sociep of London Special Report 12: 1-72. Dashzeveg D. 1979a. pyracotherium from Mongolia]. PaleontologicheskG <hurnal 13: 108-1 13. (in Russian.) Dashzeveg D. 1979b. [An archaic member of the Equidae from the Eocene of Central Asia.] Tm+ Soumestngya Souetsko-Mongokkaya Paleontologheskya Ekrpeditsiya 8: 10-22. (In Russian.) Dashzeveg D. 1988. Holarctic correlation of non-marine Palaeocene-Eocene boundary strata using mammals. Journal of the Geological Society, London 145: 473-478. Deperet C. 1917. Monographie de la faune de mammiferes fossiles du Ludien inferieur d'Euzet-les-Bains (Gard). Annales de I'Uniuersitt! de Lyon (NS,I) 40: 1-290. Franzen JL. 1988. Messeler Paradepferde und andere Unpaarhufer. In. Schaal S, Ziegler W, eds. MesselEin Schaufenster in die Geschichte der Erde und des Lebem. Frankfurt-am-Main: Verlag Waldemar Kramer, 241247. Franzen JL. 1989. Origin and systematic position of the Palaeotheriidae. In: Prothero DR. Schoch RM, eds. n2e Evolution of Pmsodactyls. New York: Oxford University Press, 102-108. Franzen JL. 1990. Hallensia (Mammalia, Perissodactyla) aus Messel und dem Pariser Becken sowie Nachtrage aus dem Geiseltal. Bullet& de l'lnstitut Royal des Sciences Naturelles de Belgzque (Sciences de la Terre) 60: 175-201. Franzen JL, Haubold H. 1986. Ein neuer Condylarthre und ein Tillodontier (Mammalia) aus dem Mitteleozan des Geiseltales. Palaeouertebrata 16: 35-53. Gervais P. 1849. Recherches sur les mammiferes fossiles des genres Palaeotherium et Lophiadon, et sur les autres animaux de la meme classe que l'on a trouvts avec eux dans le midi de la France. Comptes rendus hebdomadaires des Siances de I'Acadhie des Sciences, Paris 29: 381-384. Gingerich PD. 1976. Cranial anatomy and evolution of early Tertiary Plesiadapidae (Mammalia, Primates). Papers in Paleontolou, Museum .f Paleontology, Uniuersig of M u h g a n 15: 1-140 Gingerich PD. 1989. New earliest Wasatchian mammalian fauna from the Eocene of northwestern Wyoming: composition and diversity in a rarely sampled high-flood plain assemblage. UnWerszty of Michigan Papers on Paleontology 28: 1-97. Gingerich PD. 1991. Systematics and evolution of early Eocene Perissodactyla (Mammalia) in the Clarks Fork Basin, Wyoming. Contributions j a m the Museum .f Paleontology, the Uniuersity of Mich%an 28: 181-213. Godinot M. 1981. Les mammiferes de Rians (Eocene inferieur, Provence). Palaeouertebrata 10: 43-126. Godinot M. 1982. Aspects nouveaux des tchanges entre les faunes mammaliennes #Europe et d'amtrique du Nord a la base de I'Eoctne. Giobios (mkmoire sptcial) 6: 403-412. Godinot M, Crochet J-Y, Hartenberger J-L, Lange-BadrC B, Russell DE, Sigh B. 1987. Nouvelles donnees sur les mammiferes de Palette (Eocene inftrieur, Provence). Munchner Geomissenschaftliche Abhandlungen (A)lO:273-288. Gray JE. 1821. On the natural arrangement of vertebrose animals. London Medical Repodoy 15: 296-310. n, 62 J. J. HOOKER Haupt 0. 1925. Die Palaeohippiden der eocanen Susswasserablagerungen von Messel bei Darmstadt. Ahhandlungen der hessischen geologkchen Landesanstalt zu D a m t a d t 6: 1-1 59. Holland WJ, Peterson OA. 1914. The osteology of the Chalicotheroidea, with special reference to a mounted skeleton of Moropus elatus Marsh, now installed in the Carnegie Museum. Memoirs of the Carnqp ‘Miseum 3: 189-406. Hooker JJ. 1980. The succession of Hyracotherium (Perissodactyla, Mammaliaj in the English early Eocene. Bulletin OJ the British A4useum (Natural History) (Geologyj 33: 101-1 14. Hooker JJ. 1984. A primitive ceratomorph (Perissodactyla, Mammalia) from the early ?TertiaT of Europe. +ilogical Journal OJ the Linnean Sociep 82: 229-244. Hooker JJ. 1986. Mammals from the Bartonian (middle/late Eocene) of the Hampshire Basin, southrrn England. B u l l t h Df the British Museum (Xatural Hzstory) (Geology) 39: 191-478. Hooker JJ. 1989. Character polarities in early perissodactyls and their significance for Hyracotheriuni and infraordinal relationhsips. In: Prothero DR, Schoch Rhf, eds. 7 h e Evolution of Peiwdac&Lr. New York: Oxford University Press, 79-101. Hooker JJ. 1991. The sequence of mammals in the Thanetian and Ypresian of the London and Belgalan Basins. Location of the Palaeocene-Eocene boundary. .Meewrletters on Strat@aphy 25: 75-90. King C. 1981. The stratigraphy of the London Clay. Tertiary Research Special Papers 6: 1-158. Kitts DB. 1956. American Hyracotherium (Perissodactyla, Equidae). Bulletin .f the American M u m m 4 Jatural /fi\t(Jly 110: 1-60. Koch PL, Zachos JC, Gingerich PD. 1992. Correlation between isotope records in marine and continental carbon reservoirs near the Palaeocene-Eocene boundary. Natuw, London 358: 3 19-322. Krause DW, Maas MC. 1990. The biogeographic origins of late Paleocene-early Eocenc mammalian immigrants LO the Western Interior of North Amrrica. Geological SocieQ of America, Sperzal P a p s 243: 7 I l0.i. Lemoine V. 1878. Communication sur les ossements fossiles des terrains tertiaires inferirures des rnlirons de Reims faitr a la Socitti. d’Histoire Naturellc de Reims. Bulletin de la Sociiti d’HiJtozre ~%zturelle de Reims. 2: 90-1 1 3 . Lemoine V. 1891. Etude d’ensemble sur les dents des mammiferes fossiles des environs de Reims. Bulletin de In Sociiti Giolo~gique de France (3)19: 253-290. MacFadden BJ. 1976. Cladistic analysis of primitive equids, with notes on other perissodactyls. Gstematir <O<J~OQ 25: 1-14. McKenna MC. 1983. Cenozoic paleogeography of north Atlantic land bridges. In. Britt MHP, Saxov S, Tal\vani M, Thiede J, eds. Structure and development of thp Greenlan&Scotland ridge. New York & London: Plenum Press, 351 399. McKenna MC, Chow M-C, Ting S-Y, Luo Z-X. 1989. Radinskya yupzngae, a perissodartyl-like mammal kom the late Paleocene of China. In: Prothero DR, Schoch RM, eds. 7 h e Evolution ( J PerirsodacQh. Ntw York: Oxford University Press, 24 36. Marandat B. 1991. MammiRrrs de 1’Ilerdirn moyen (Eocene infkrieur) des Corbit-rcs rt du Xlinervois (Bas-Languedoc, France). Palaeovertebrata 20: 55- 144. Marsh OC. 1876. Notice of new Tertiary mammals. V. American Journal of Science (3112: 401 404. Matthews SC. 1973. Notes on open nomenclaturr and on synonymy lists. Palaeontoloa 16: 713-719. Menu H, Sig6 B. 1971. Nyctalodontie et myotodontir, importants caract6res de gradcs evolutifs chez lea c h i r o p t h s entomophages. CompteJ rendus hebdomadazres de.r Siances de l’ilcadhie ded Sczenced, Park (Dj272: 1735 1738. Murray JW,Wright CA. 1974. Palaeogene Voraminiferida and palaeocology, Hampshirr and Paris Basins and the English Channel. Special Papers in Palaeontoloa 14: 1-129. Novacek MJ, Ferrusquia-Villafranca J, fly^ JJ, Wyss AR, Norell M. 1991. It‘asatchian (Early Eocene! mammals and other vertebrates from Baja California, Mrxico: the Lomas L,as 7‘etas de Cahra fauna. Bulletin OJ the American Museum OJ ,Vaturn/ Hictory 208: 1-88. Osborn HF. 1929. The titanotheres of ancient LVyoming, Dakota, and Nebraska. Monographs q/- /he C‘ni/ed States Geolopkal Sum9 55: 1-953. Owen R. 1841. Description of the fossil remains of a mammal (Hyracotherium leporinum) and of a bird (Lithornis culturinus) from the London Clay. Transactions Df the Geologzcal Socie$ of London 6: 203-208. Owen R. 1858. Description of a small lophiodont mammal (Pliolophus uulpiceps, Owen), from the London Cla), near Hanvich. Quarterb Journal .f the Geological Sofie& of London 14: 54-7 1. Pavlow M. 1888. Etudes sur l’histoire paleontologique des ongulks. 11. Le dkveloppement des Equiddc. Bulletin de la Sociiti des Naturalistes de itloscou (NS)2: 135-182. Pomel A. 1847. Notes sur les mammiferes et reptiles fossiles des terrains tocenes de Pans, infirieurs du depot gypeux. Archives des Sciences Physiques et Naturelles de Geniue 4: 326-330. Prothero DRY Schoch RM, eds. 1989. The Evolution of Perissodac&lr. New York Oxford University Prrss, 1-537. Radinsky LB. 1963. Origin and early evolution of North American Tapiroidea. Bulletin of /he Peabodp Aluseum 4 Natural Histoy, Yale Uniuersi& 17: 1-106. Radinsky LB. 1965. Evolution of the Tapiroid skeleton from Heptodon to Tapirus. Bulletin ft the Museum of Comparatiw <oolo~gy, Harvard 134: 69-1 06. EARLY HORSE EVOLUTION 63 Remy JA. 1967. Les Palaeotheridae (Perissodactyla) de la faune de mammiferes de Fons 1 (Eocene supkrieur). Palaeovertebrata 1: 1-46. Remy JA. 1972. Etude du c r h e de Pachynolophus lauocati n. sp. (Perissodactyla, Palaeotheriidae) des Phosphorites du Quercy. Palaeovertebrata 5: 45-78. Remy JA. 1976. Etude comparahue des structures dentaires chez les Palaeotheriidae et diuers autres periisodac@es fossiles. (These presentite pour le doctorat de 3eme cycle en sciences odontologiques, Faculte de Chirurgie dentaire, Universiti. Louis Pasteur.) Strasbourg. Remy JA. 1992. Observations sur l’anatomie cranienne du genre Palaeotherium (Perissodactyla, Mamrnalia): mise en evidence d’un nouveau sous-genre Franzenitherium. Palaeouertebrata 21: 103-224. Reshetov VYu. 1979. [Early Tertiary Tapiroidea of Mongolia and the USSR.] Trudy Sovmestnaya SouetskoMongolskaya Paleontol(~gichesh~~a EkspeditsGa 11: 1-141. (In Russian). Russell DE, Zhai R-J. 1987. The Paleogene of Asia: mammals and stratigraphy. Mhoires du Muslum National d’tlistoire .Nature& (C)52: 1-488. Savage DE, Russell DE, Louis P. 1965. European Eocene Equidae (Perissodactyla). Universip of Cal$mza Publications in Geological Science 56: 1-94. Sisson SB, Grossman JD. 1953. The anatomy .f the domestic animals. 4th ed. Philadelphia & London: W. B. Saunders Company. Swofford DL. 1990. PAUP, Phylogenetic Ana!~szs Using Parsimov, Version 3.0: 1-162. User’s manual, Illinois Natural History Survey. Champaign. Teilhard de Chardin P. 1922. Les rnammifkres de 1’Eockne infirieur franGais et leur gisements [2nd part]. Annales de Pal~ontologie 11: 1-108 (9-116). Thewissen JGM. 1990. Evolution of Paleocene and Eocene Phenacodontidae (Mammalia, Condylarthra). Universi& of Michigan Papers on Paleontology 29: 1-107. Thewissen JGM, Domning DP. 1992. The role of phenacodontids in the origin of the modern orders of ungulate mammals. Journal of Vertebrate pale onto lo^ 12: 494-504. T i 5 e y BH. 1985. The Eocene North Atlantic land bridge: its importance in Tertiary and modern phytogcography of the northern hemisphere. Journal of the Arnold Arboretum 66: 243-273. Townsend HA, Hailwood EA. 1985. Magnetostratigraphic correlation of Palaeogene sediments in the Hampshire and London Basins, southern UK. ,j$~urnal of the Geological Sociep of London 142: 957-982. Van Valen L. 1982. Homology and causes. Journal ofMorpholo~ 173: 305-312. Wood HE. 1934. Revision of the Hyrachyidae. Bulletin of the American Mtlseuni of Natural Hi.rto?y 67: 181295. Young C-C. 1944. Note on the first Eocene mammal from South China. American Museum Nouitates 1268: 1-4. Zdansky 0. 1930. Die alttertiaren Saugetiere Chinas stratigraphischen nebst Bemerkungen. Palaeontologia Sinica (C)6: 1-87.
© Copyright 2026 Paperzz