Journal of Vertebrate Paleontology 22(2):388±404, June 2002 ᭧ 2002 by the Society of Vertebrate Paleontology

NEW PROBOSCIDEANS FROM THE EARLY MIOCENE OF WADI MOGHARA, EGYPT

WILLIAM J. SANDERS1 and ELLEN R. MILLER2 1Museum of Paleontology, University of Michigan, Ann Arbor, Michigan 48109, [email protected]; 2Department of Anthropology, University of Colorado, Boulder, Colorado 80309

ABSTRACTÐFieldwork conducted since 1981 has greatly increased the sample of proboscidean fossils from Wadi Moghara, Egypt. The Moghara proboscidean assemblage is taxonomically more diverse than previously suspected, comprising four taxa: Gomphotherium angustidens libycum, Afrochoerodon kisumuensis, cf. Archaeobelodon, and Zyg- olophodon aegyptensis, sp. nov. Biochronological analysis of the proboscideans supports previous ®ndings based on the remainder of the fauna that the age of Moghara is early Miocene, approximately 18±17 Ma. The composition of the Moghara proboscidean assemblage suggests complex biogeographic distribution patterns of proboscideans through- out Eurasia and Afro±Arabia during the early Miocene. Moghara and other pene-contemporaneous Afro±Arabian sites were apparently characterized by a relatively high degree of mammalian species-level endemism.

INTRODUCTION tablished an age for Moghara of about 18±17 Ma, and dem- onstrated a strong biogeographic connection between Moghara This paper presents a description and taxonomic interpreta- and contemporaneous faunas in East Africa (Miller, 1996, tion of new fossil proboscideans recovered from Wadi Mog- 1999). Proboscideans, however, were omitted from this faunal hara, Egypt (Fig. 1). Moghara encompasses 40 terrestrial mam- analysis because of uncertainty over their taxonomic af®liations malian fossil localities, spread over a distance of 55 km, in the (Miller, 1999). northeastern portion of the Qattara Depression, approximately The new proboscidean sample (Table 1) is important for a 60 km south of El Alamein (Miller, 1999). Fossils are derived number of reasons: (1) it is valuable for systematic evaluation from the type section of the early Miocene Moghara Formation, of previously collected specimens; (2) it can be used to test which consists of at least 230 m of a clastic sequence with a previous hypotheses about the age of Moghara; and (3) it pro- composite fauna derived from ¯uvio-marine deposits (Said, vides for a more complete assessment of the biogeographic as- 1962a). The Moghara Formation overlies Eocene-Oligocene sociations of the Moghara mammals. More complete informa- marine shales of the Dabaa Formation and sabka (a mixture of tion about the composition and biogeographic relationships of sand, silt, and salt) that form the ¯oor of the depression, and the Moghara fauna is critical for testing hypotheses about the underlies the middle Miocene marine Mamura Formation that timing, nature, and degree of contact between early Miocene crowns the Qattara escarpment (Said, 1962b, 1990; Marzouk, Afro±Arabian and Eurasian faunas (Bernor et al., 1987). Mog- 1969; Omara and Ouda, 1972; Omara and Sanad, 1975; Hantar, hara is one of only a handful of early Miocene Afro±Arabian 1990). localities found outside of East Africa (other exceptions being Geological evidence indicates that in the early Miocene, Gebel Zelten, Libya; As-Sarrar and Ad Dabtiyah, Saudi Arabia; Moghara was part of a slow-moving ¯uviatile, estuarine, and Rotem and Yeroham, Israel; and Arrisdrift, Namibia) (Fig. 1; lagoonal system (Said, 1962b). The majority of sedimentary Hamilton, 1973; Savage and Hamilton, 1973; Hamilton et al., units are comprised of sandstones, siltstones, and calcareous 1978; Thomas et al., 1982; Whybrow, 1984, 1987; Tchernov et shales, and the great number of silici®ed tree trunks at Moghara al., 1987). Moreover, despite its evident ties to East African shows that this system was forested (Bown et al., 1982). This localities (Miller, 1999), Moghara is physically closer to Eu- environmental interpretation is supported paleontologically rope, and so provides an unusual perspective from which to (Said, 1962b). The fauna from aquatic depositional environ- view the biogeography of Miocene mammals. ments is a mix of primarily marine animals (e.g., sharks and Abbreviations Institutional: CGM, Cairo Geological Mu- rays) along with primarily freshwater animals that can tolerate seum; DPC, Duke Primate Center; KNM, National Museums slow-moving water with a high sediment content (e.g., cat®sh). of Kenya; M followed by a number series (e.g., M 15524), The land mammal fauna is numerically dominated by anthra- catalogue number, The Natural History Museum, London; WS, cotheres, a group of artiodactyls that preferred aquatic or aquat- West Stephanie (Buluk), Kenya; hyphenated number series ic margin environments, as they are found elsewhere most com- (e.g., 97-655), year and ®eld number of Duke Primate Center monly in ¯uviatile and lacustrine deposits (Pickford, 1991a). specimen with no assigned catalogue number. Dental: DP, up- Initial paleontological work in the late 1800s and early 1900s per deciduous premolar (for example, DP3 ϭ upper third de- at Moghara produced a modest assemblage of fossils mammals, ciduous premolar); dp, lower deciduous premolar; ET, enamel including a small number of gomphothere specimens originally thickness; H, height; L, length; m, lower molar; M, upper mo- allocated to ``Mastodon spenceri'' and ``Mastodon angustidens lar (for example, M1 ϭ upper ®rst molar); W, width; x,asin var. libyca'' (Fourtau, 1918, 1920). The subsequent taxonomic x3x, denotes a tooth comprised of three lophs with an anterior history of these fossils is complicated and their phylogenetic and posterior cingulum; ؉, indicates a missing portion of a interpretations have remained unresolved (see below). Renewed tooth, and that the original dimension was greater. Chronolog- ®eld efforts at Moghara (1981±1997) greatly enhanced the orig- ical: Ma, million years ago; MN, European Neogene mammal inal collection, and the site is now known to contain a high faunal zone (Steininger et al., 1996); Myr, million years. diversity of proboscideans (Fig. 2) as well as many other taxa De®nitions Dental: abaxial conelet, the outer, main cone previously unknown from the area (Miller, 1999). Recent faunal in each half-loph(id) (Tassy, 1996a); accessory central conules, analysis of non-proboscidean mammalian taxa from the site es- enamel covered pillars situated at the anterior and/or posterior

388 SANDERS AND MILLERÐNEW PROBOSCIDEANS FROM WADI MOGHARA 389

FIGURE 1. Map of major localities discussed in the text.

faces of the loph(id)s or in the transverse valleys, partially (Osborn, 1942); zygodont crests, enamel crests running from blocking them centrally (Tobien, 1973a); adaxial conelet(s), the the apices of the abaxial conelets of the posttrite half-loph(id)s inner, or meso-, conelet(s) in each half-loph(id) (Tassy, 1996a); to the bottom of the transverse valleys, and ending near the anancoidy, alternation of half-loph(id)s, in which lingual half- middle axis of the crown (Tobien, 1975). Taxonomic: amebel- loph(id)s are anterior to buccal half-loph(id)s (Tobien, 1973a); odontine, refers to members of the Amebelodontinae, e.g., Ar- chevroning, the arrangement of half-loph(id)s to form an ante- chaeobelodon, Amebelodon, Protanancus, and Platybelodon; riorly pointing ``V,'' or chevron (in Afrochoerodon and Choer- choerolophodontine, refers to members of the Choerolophodon- olophodon, the pretrite half-loph(id) of the chevron is more an- tinae, e.g., Afrochoerodon and Choerolophodon; mammutid, re- teriorly extended than the posttrite half-loph(id)) (Tobien, fers to members of the Mammutidae, including Eozygodon, 1975); choerolophodonty, molar crowns covered with tubercles Zygolophodon, and Mammut. (Osborn, 1942); crescentoids, enamel crests running from the apices of the abaxial conelets of the pretrite half-loph(id)s to SYSTEMATIC PALEONTOLOGY the bottom of the transverse valleys, and ending near the middle Class MAMMALIA Linnaeus, 1758 axis of the crown (Tobien, 1975); interloph(id), transverse val- Order PROBOSCIDEA Illiger, 1811 ley or space between two loph(id)s; pretrite, refers to the more Suborder ELEPHANTIFORMES Tassy, 1988 worn half of each loph(id), which is buccal in lower and lingual Superfamily ELEPHANTOIDEA Gray, 1821 in upper molars (Vacek, 1877); posttrite, refers to the less worn Family GOMPHOTHERIIDAE Hay, 1922 half of each loph(id), which is lingual in lower and buccal in Genus GOMPHOTHERIUM Burmeister, 1837 upper molars (Vacek, 1877); ptychodonty, plication or infolding GOMPHOTHERIUM ANGUSTIDENS LIBYCUM of enamel borders with grooving of the sides of the molars (Fourtau, 1918) 390 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 22, NO. 2, 2002

FIGURE 2. Diagrammatic representations of elephantoid taxa represented at Wadi Moghara. Abbreviations: ac, anterior accessory central conule; ecto1, ecto¯exus between lophs 1 and 2; mc, mesoconelet; pc, posterior accessory central conule; pcr1, posterior crescentoid of the ®rst loph, found on the pretrite side; po, main cone, posttrite side; pr, main cone, pretrite side; t, lateral tubercle; x, anterior or posterior cingulum; zc, zygodont crest, found on posttrite side; 1, 2, 3, 4, loph(id) number, counted from the anterior. A, 97-1245, left m3, Gomphotherium angustidens libycum. B, DPC 14584, right M3, Afrochoerodon kisumuensis. C, CGM 30892, right M3, cf. Archaeobelodon. D, DPC 12598, right M3, Zygolophodon aegyptensis, sp. nov.

TABLE 1. Dimensions (in mm) and indices of proboscidean teeth from Wadi Moghara. See abbreviations in text.

Loph(id) W/L H/W Taxon and Accession Number Specimen Formula L W H Index Index ET Gomphotherium angustidens libycum DPC 2502 r. dp3 x2ϩ 49.3ϩ 28.0 (1)c worn Ð Ð Ð DPC 12926 r. m2 x3x 116.4 59.4 (3) 42.0 (3)c 51 71 unworn DPC 4581 r. m2 x3x 119.4 60.2 (3) worn 50 Ð 5.0±5.9 97-1245 l. m3 x3x 147.1 68.4 (2) worn 46 Ð 5.3 DPC 2579 r. m3 ϩ3ϩ ϩ113.0ϩ 69.1 (3) 50.0 (3) Ð 72 unworn 97-675 r. M1 x3x 76.0 49.5 (2) worn 65 Ð 3.2 DPC 6442 l. M2 x3x 122.2 68.9 (3) 44.5 (2) 56 65 5.0±5.5 97-655 l. M2 x3x 107.6 61.4 (1) worn 57 Ð 5.0 97-656 r. M3 x4x 149.5 77.7 (2) worn 52 Ð 6.2±7.2 M 14075 r. m3 ϩ4x ϩ148.2 65.0 (2) worn 44 Ð 6.0 Fourtau (1918, 1920), ``Mastodon angusti- l. M2 3x 107.0 68.0 (2) worn 64 Ð Ð dens var. libyca'' l. M3a x3x or x4 143.0 86.0 (1) worn 60 Ð Ð r. m2b 3x 107.0? 68.0? Ð 64? Ð Ð r. m3 4x 175.0 66.0 (1) Ð 38 Ð Ð Fourtau (1918, 1920), ``Mastodon spen- r. m2 x3x 119.0 64.0 (3) worn 54 Ð Ð ceri'' (type) Zygolophodon aegyptensis DPC 9009 r. m3 x4x 148.8 55.8 (1) worn 38 Ð Ð DPC 5932 r. M2 x3x 112.5 68.0 (1) worn 60 Ð 5.2 DPC 12598 r. M3 x3x or x4x 121.1 64.7 (2) 43.0 (2) 53 66 unworn DPC 5920 r. M3 ϩ4ϩϩ116.0ϩ 75.6 (3) (originally 49.0 (3) Ð Ð 5.8±6.8 wider anteriorly) cf. Archaeobelodon CGM 30892 r. M3 ϩ4x ϩ120.0 60.0 (2) Ð 44±46d Ð Ð Afrochoerodon kisumuensis DPC 14584 r. M3 x3x 113.4 72.6 (1) worn 64 Ð 5.0±6.0 aFourtau (1918, 1920) provides a loph formula of x4 for this specimen, but the fourth loph is only incipiently formed. The formula could just as accurately be expressed as x3x. bThe dimensions given for this molar by Fourtau (1918, 1920) were almost certainly erroneously duplicated from the dimensions for an upper M2 in his sample, and therefore are incorrect. cNumbers in parentheses indicate the widest or highest loph(id)s in each molar. dWidth±length index varies according to whether original molar length is estimated to have been 130 or 135 mm. SANDERS AND MILLERÐNEW PROBOSCIDEANS FROM WADI MOGHARA 391

Mastodon angustidens var. libyca Fourtau, 1918:84. half-lophid, a smaller accessory central conule fused to the an- Mastodon spenceri Fourtau, 1918:89. terior face of the second pretrite half-lophid, and a large tuber- Rhynchotherium spenceri, Osborn, 1936:485. cle occupies the ®rst interlophid at the lingual margin of the Trilophodon angustidens var. libycus, Osborn, 1936:260. crown. The anterior cingulum is relatively large and has three Gomphotherium angustidens, Hamilton, 1973:276. conelets, while a narrow cingular shelf rims the buccal side of Gomphotherium angustidens, Tobien, 1973b:214±215. the crown. The crown is supported by a large posterior root Gomphotherium angustidens, Coppens et al., 1978:342±343. beneath lophids 2 and 3 and a smaller root below lophid 1. ``Gomphotherium'' pygmaeus, Coppens et al., 1978:344. There is no cementum. Gomphotherium spenceri, Savage, 1989:594. DPC 12926 (Fig. 3B, C) and DPC 4581 (Fig. 3F) are both Gomphotherium angustidens, Savage, 1989:594. isolated right m2s. DPC 12926 is unworn, though an anterior interproximal facet indicates that it had emerged from its crypt. Original Diagnosis Relative width of M3 intermediate be- No roots are preserved. Lophid formula is x3x. The anterior tween M3 widths of G. angustidens of the OrleÂanais, France cingulum has six small conelets and continues on the buccal and G. angustidens pontileviensis of the BleÂsois, France (Four- side as a narrow shelf. Only a remnant of the lingual cingulum tau, 1918). is present. The posterior cingulum is formed by three large Emended Diagnosis Medium-sized gomphothere. Molar conelets and is more prominent than the anterior cingulum. crowns bunodont; loph(id)s divided by a pronounced median It is continuous with the buccal shelf. The lophids are divis- sulcus; half-loph(id)s dominated by massive, pyramidal abaxial ible into pre- and posttrite half-lophids; each is comprised of a conelets either subdivided super®cially from a much smaller large, bulbous main cone on the outside of the crown, accom- adaxial conelet or unaccompanied. Vestiges of lateral cingulae panied by (except for posttrite half-lophid 2) a smaller meso- present as tubercles blocking the entrances to transverse valleys. conelet. The lophids are separated by V-shaped interlophids. Intermediate molars trilophodont. Cementum poorly expressed Each pretrite half-lophid has a small anterior and larger poste- or absent; enamel thick and unfolded. Anterior and posterior rior accessory central conule. The posterior accessory central accessory central conules generally large and simple, forming conule of lophid 1 is doubled and the anterior accessory central trefoil wear ®gures with pretrite conelets. Differs from more conule of lophid 3 is fused to the mesoconelet. advanced forms of Gomphotherium angustidens by variable de- DPC 4581 is morphologically similar to DPC 12926 but is velopment of fourth loph in M3 and having some m3s with more worn and shows that, with wear, accessory central conules only three lophids. Differs from Gomphotherium steinheimense are incorporated into the pretrite half-lophids to form trefoils. and Gomphotherium browni by smaller size and weaker sub- In addition, the wear on the crown of DPC 4581 reveals that division of adaxial and abaxial conelets, and from ``pygmy'' the enamel is thick and unfolded. The specimen exhibits both gomphotheres of Siwa, Egypt; Ghaba, Oman; Gebel Zelten, anterior and posterior interproximal facets, has a thin anterior Libya; and Kabylie, Algeria by larger size and lesser develop- root below lophid 1, and a stouter root supporting lophids 2 ment of cementum. Differs from gomphotheres of Gomphoth- and 3. Neither specimen shows a trace of cementum. erium ``annectens grade'' (including Gomphotherium sp. from Specimen 97-1245 is a left dentary fragment preserving the Mwiti, Kenya, Gomphotherium cooperi from the Bugti Beds, alveolus for m2, the crown of m3 (Figs. 2A, 4C), and a portion Pakistan, Gomphotherium annectens from the Hiramaki For- of the coronoid process. The dentary measures 91.6 mm in mation, Japan, and Gomphotherium sylvaticum from the early width at m3. The m3 is moderately worn, and is severely frac- Miocene of Western Europe) by greater expression of accessory tured due to weathering. Nonetheless, the crown is largely intact central conules (DepeÂret, 1897; Fourtau, 1920; Forster Cooper, and clearly shows three lophids, the worn remnant of an ante- 1922; Osborn, 1932; Bergounioux and Crouzel, 1959; Aram- rior cingulum, and a prominent distal heel composed of two bourg, 1961; Tassy, 1983a, 1985, 1994, 1996b; Gaziry, 1987; stout conelets. A broad interproximal facet is present anteriorly, Gentry, 1987a; Roger et al., 1994; GoÈhlich, 1998). and the crown is slightly curved longitudinally, convex to the Referred Specimens DPC 2502, right dentary fragment lingual side. Large tubercles are present lingually in the inter- with partial dp3 (Fig. 3A) and alveolus for dp2; DPC 12926, lophids at the crown margin. Enamel is thick and unfolded, and right m2 (Fig. 3B, C); DPC 4581, right m2 (Fig. 3F); DPC there is no cementum. Anterior and posterior accessory central 2579, partial right m3 (Fig. 4A, B); 97-1245, dentary fragment conules are strongly developed and, with attrition, merge with with left m3 (Figs. 2A, 4C); 97-675, right M1 (Fig. 3D); 97- the pretrite half-lophids to form trefoils. Each half-lophid has a 655, left M2; DPC 6442, left M2 (Fig. 3E); 97-656, right M3 small, internal mesoconelet super®cially subdivided from a (Fig. 4D); M 14075, right m3; unnumbered specimens attribut- larger, laterally-placed main cone. ed to Mastodon angustidens var. libyca (Fourtau, 1918), left DPC 2579 (Fig. 4A, B) is an m3 that is morphologically maxillary fragment with M2±3; right dentary fragment with similar to 97-1245 and shows the better part of three unworn m2±3; left dentary fragment with m1, m2±3 developing in lophids, but the specimen is missing its anterior and posterior crypt, and the roots of dp4; two symphyseal fragments (one ends. As in 97-1245, the interlophids are blocked by large tu- with an associated right lower incisor fragment); unnumbered bercles on the buccal side. Also similar is the subdivision of type specimen of Mastodon spenceri (Fourtau, 1918), right den- each half-lophid into a larger abaxial conelet and smaller ad- tary fragment with symphysis and m2. The Fourtau specimens axial conelet, and the presence of large anterior and posterior were originally housed in the Cairo Geological Museum but pretrite accessory central conules. The half-lophids are demar- apparently are now missing. cated by a sharp, narrow median sulcus. The crown is consid- Description DPC 2502 is a right dentary fragment preserv- erably wider basally than at its apex. In lateral view, the lophids ing dp3 (Fig. 3A) and the alveolus for dp2. The dentary is 53.9 are angled mesially. There is no cementum. mm high and 43.3 mm wide at dp3. The presence of an anterior Specimen 97-675 (Fig. 3D) is a small, moderately worn right interproximal facet on the dp3 indicates that it was erupted at M1 preserving most of two roots. The crown is somewhat frac- the same time as dp2. Although the dp3 is broken through the tured and the lingual side of the second loph is broken. Loph second lophid, it is obvious that it originally had either three formula is x3x. There is a trace of an anterior interproximal lophids or a substantial heel. The ®rst lophid is divided into facet. Each posttrite half-loph is formed of a single large cone, pre- and posttrite halves by a median sulcus, with each half- while the pretrite half-lophs are each subdivided into a small lophid composed of a single stout cone. There is a prominent, interior mesoconelet and a larger exterior cone. The pretrite free posterior accessory central conule behind the ®rst pretrite half-lophs are accompanied by strongly expressed anterior ac- 392 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 22, NO. 2, 2002

FIGURE 3. Gomphotherium angustidens libycum specimens from Wadi Moghara. Anterior is to the left. Specimens in B±F to same scale. A, Occlusal view, DPC 2502, right dp3. B, Occlusal view, DPC 12926, right m2. C, Lingual view, DPC 12926, right m2. D, Occlusal view, 97± 675, right M1. E, Occlusal view, DPC 6442, left M2. F, Occlusal view, DPC 4581, right m2.

cessory central conules. Evidence for a small posterior acces- cingular shelf is broken away. The remains of three roots are sory central conule is found in the trefoil wear ®gure of pretrite preserved: two thin ones below loph 1, and a larger one under half-loph 1. A tubercle blocks the interloph at the crown margin lophs 2 and 3. The enamel is thick and unfolded, and there is on the lingual side between lophs 1 and 2. There is no cemen- no cementum. tum. Specimen 97-656 (Fig. 4D) is a heavily worn right M3, pre- DPC 6442 (Fig. 3E) and 97-655 are M2s. Although DPC serving traces of an anterior cingulum, four lophs, and a small 6442 was assembled from a number of pieces, it is more com- posterior cingulum of two conelets tightly appressed to the last plete than 97-655. DPC 6442 has a loph formula of x3x. The loph. Alternatively, the last loph and posterior cingulum can be anterior cingulum is small, most prominent buccally, and con- thought of as a coherent unit forming a prominent distal ``heel,'' nects with a narrow lingual cingular shelf. The posterior cin- in which case the loph formula would be x3x. Tubercles block gulum is low and composed of many small conelets. Strong the interlophs between lophs 1±2 and 2±3 on the buccal side. tubercles are present buccally in the interlophs, at the crown A low, beaded cingulum runs along the lingual side of the margin. The lophs are formed of pre- and posttrite half-lophs, crown. Enamel is very thick and unfolded. There is no cemen- each with a large main cone to the outside of the crown and a tum. The heavy wear on the crown obscures any sign of trefoil smaller mesoconelet. In lophs 2 and 3, the anterior accessory wear patterns, though at least the ®rst two lophs apparently had central conules are large, while the posterior accessory conules anterior and posterior pretrite accessory central conules. The are smaller. The enamel is thick, and there is no cementum. fourth loph lacks both. Specimen 97-655 is a heavily worn left M2, exposing a pre- Remarks The taxonomic history of Moghara Gomphoth- trite trefoil wear pattern showing the greater size of the anterior erium is complicated. Most of the ®rst proboscidean specimens than posterior accessory central conules. Broad interproximal from the site were originally placed by Fourtau (1918, 1920) facets are present anteriorly and posteriorly. Tubercles block the in Mastodon angustidens var. libyca. Osborn (1936) subse- interloph between lophs 1 and 2 on both sides, but the lingual quently revised Fourtau's taxon to Trilophodon angustidens var. SANDERS AND MILLERÐNEW PROBOSCIDEANS FROM WADI MOGHARA 393

FIGURE 4. Gomphotherium angustidens libycum specimens from Wadi Moghara. Anterior is to the left. Specimens to same scale. A, Occlusal view, DPC 2579, partial right m3. B, Lingual view, DPC 2579, partial right m3. C, Occlusal view, 97±1245, left m3. D, Occlusal view, 97±656, right M3.

libycus, re¯ecting then current nomenclature for Gomphother- ton, 1973). However, while Savage (1989) listed G. angustidens ium. Fourtau (1918) did not formally designate a type specimen from both Gebel Zelten and Moghara, he included G. pygmaeus for this taxon, perhaps because he was naming a subspecies. only in the Zelten fauna. Adding to the confusion, Tobien Nevertheless, Osborn (1936:260±261) assumed a maxillary (1973b) suggested that specimens attributed to ``G.'' pygmaeus fragment with M2±M3 (®gured in Fourtau, 1918:84) to be the might not be Gomphotherium, but Choerolophodon (based on type. The M3 of this specimen has three lophs and a large distal the thick cementum covering the crown of the Kabylie speci- heel equivalent to an incipient fourth loph and diminutive pos- men [but see below]). As well, Madden (in Coppens et al., terior cingulum, and prominent accessory central conules form- 1978) believed that they belonged in a genus other than Gom- ing (with the pretrite abaxial and adaxial conelets) trefoil enam- photherium. el ®gures with wear. It is widest at the ®rst loph and low Fourtau (1918, 1920) also erected Mastodon spenceri for a crowned, with a modest lingual cingulum (Fourtau, 1918; Os- partial right dentary with m2 from Moghara. Fourtau's diag- born, 1936). Furthermore, the M3 is of relatively greater nosis was based on the ventral de¯ection of the symphysis of breadth than is typical in Trilophodon angustidens (Fig. 5; Os- this specimen. Osborn (1936) later used this feature to re-assign born, 1936). Despite the width of M3 in this specimen, the M2 it to the New World genus Rhynchotherium. In turn, Osborn's and other specimens of Fourtau are metrically more typical of interpretation was discounted by Tobien (1973b), who argued Gomphotherium angustidens (Fig. 5), and are morphologically that the specimen more reasonably belongs in the Old World inseparable from the more recently collected molars from Mog- taxon Gomphotherium angustidens, as it can be accommodated hara assigned to G. angustidens libycum. within that species' range of variation for symphyseal angula- More than ®fty years after Fourtau, Hamilton (1973) referred tion. This opinion was shared by Coppens et al. (1978). More an isolated m3 (M 14075) from Moghara and m3s from Siwa recently, Savage (1989) revised Fourtau's species designation to G. angustidens. Coppens et al. (1978:343) appeared to have to Gomphotherium spenceri for the Moghara specimen, togeth- accepted this attribution for Fourtau's libyca specimens as well, er with gomphothere material from Siwa, Egypt. As Mastodon although they inexplicably refer ``other material . . . from Wadi is a junior synonym of the American mastodon Mammut (To- Moghara'' to ``Gomphotherium'' pygmaeus, along with fossil bien, 1973b; Gaziry, 1976), Savage was correct in recognizing proboscideans from Gebel Zelten, Libya, and Kabylie, Algeria. that Gomphotherium is the proper generic name for the Mog- The actual specimens assigned to ``G.'' pygmaeus were not hara gomphotheres. speci®ed, and it is possible that their decision was an error A more intriguing view of the specimen is that of Tassy based on acceptance of the idea that the faunas from Moghara (1985), who speculated that the downturn and deep gutter of and Gebel Zelten are closely comparable (Savage and Hamil- the symphysis might instead indicate choerolophodontine af®n- 394 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 22, NO. 2, 2002

FIGURE 5. Bivariate plots of natural log-transformed molar crown length versus width in Gomphotherium. Comparative dimensions are from Fourtau (1920), Forster Cooper (1922), Bergounioux and Crouzel (1959), Arambourg (1961), Hamilton (1973), Gaziry (1976, 1987), Tassy (1983a, 1985), Gentry (1987), Roger et al. (1994), GoÈhlich (1998), and Sanders (pers. obs.; in press). Symbols: ૽, Wadi Moghara, Egypt; Ⅵ, Siwa, Egypt; ࡗ, Gebel Zelten, Libya; ᭢, Ghaba, Oman; ᭡, Kabylie, Algeria; , Mwiti, Kenya; ᭪, Gomphotherium cooperi, Bugti Beds, Pakistan; ⅜, G. cooperi, Ad Dabtiyah, Saudi Arabia; ᭝, G. angustidens (including G.''subtapiroideum''), Sinap Formation and PasËalar, Turkey, Vordersdorf b. Eibiswald, Sandelzhausen, Feldmoching, Babing, Massenhausen, Giggenhausen, and Odelzhausen, Germany, and Simorre, Tournan, En PeÂjouan, and Villefranche d'Astarac, France; #, G. steinheimense, Massenhausen, Hinterauerbach, and Steinheim, Germany; ᭞, G. browni, Chinji For- mation, Siwalik Series, Pakistan. A, m2. B, m3. C, M2. D, M3. SANDERS AND MILLERÐNEW PROBOSCIDEANS FROM WADI MOGHARA 395 ities. The m2 of this specimen, however, does not have such 1987a), they and the Moghara specimens differ from the type diagnostic choerolophodontine features as chevroning of lo- collection of G. cooperi (Bugti Beds, Pakistan) and molars of phids with anterior advancement of pretrite mesoconelets rela- other, primitive G.``annectens group'' gomphotheres in having tive to posttrite half-lophids. Instead, it exhibits the trefoil wear greater in¯ation and independence of both anterior and poste- pattern typical of Gomphotherium (Fourtau, 1920:®g. 60). In rior accessory central conules, and a tendency for better devel- addition, prior assessment of the degree of ventral de¯ection of opment of fourth loph(id)s in M3/m3 (Tassy, 1996b). Instead, the symphysis in this specimen seems exaggerated. While the they are more like the simplest molars of G. angustidens.Itis dorsal surface of the symphysis is downturned relative to the quite possible that the Ad Dabtiyah and Moghara gomphotheres horizontal alveolar plane, there is little angulation of the ventral belong to the same subspecies; if so, the Ad Dabtiyah material margins of the symphysis and corpus of the dentary (Fourtau, would be subsumed into G. angustidens libycum. 1920:®g. 61). In any case, symphyseal angulation differs widely One caution against synonomy is the uncertain attribution of among choerolophodontine species, and the primitive condition an isolated proboscidean lower incisor (M 42941) from Ad for this feature is not known for the subfamily (Tassy, 1985; Dabtiyah (Gentry, 1987a). This specimen is ¯attened, dorsally Tassy et al., 1989; Sanders, in press). Symphyseal angulation concave, and forms an elongate ovoid in cross-section, with a also varies substantially in Gomphotherium angustidens (Tassy, mid-length width of 45.3 mm and height of 28 mm. It has been 1985). argued elsewhere that dorsoventral compression of lower inci- The new Moghara proboscidean material provides substantial sors is primitive for elephantoids, and that if this specimen be- new information which helps to resolve the contentious tax- longs with the molars from the site, the species should be con- onomy of Gomphotherium specimens collected earlier from sidered a ``proto-gomphothere'' (Tassy, 1994, 1996b:90). How- Moghara and elsewhere in the region. Among the new fossil ever, even if this incisor is associated with the molars from the teeth are specimens identi®ed as Gomphotherium by their pre- site, the occlusal morphology of the molars is improbably ad- trite trefoil wear patterns, lack of posttrite accessory central vanced for a proto-gomphothere. Furthermore, among early conules, trilophodonty, and transverse orientation of cones and gomphotheres there is considerable variation in the shape of conelets comprising the loph(id)s. All of the proboscidean mo- lower incisors, and a similar degree of cross-sectional ¯attening lars previously collected from Moghara share a similar occlusal has been observed in some specimens attributed to Gomphoth- arrangement, and are therefore also assigned to Gomphother- erium from the Iberian Peninsula (Mazo, 1996:139). Alterna- ium. tively, the Ad Dabtiyah incisor may belong to an amebelodon- Morphologically, there are no compelling reasons to assign tine (Tassy, 1996b), although Gentry (1987a) considered the Ad the Moghara Gomphotherium specimens to multiple species. Dabtiyah incisor to be too narrow for this assignation. It re- While there is variation in M3/m3 formulae (ranging from x3x mains possible that the incisor belongs to an amebelodontine, to an incipient x4x to fully expressed x4x; see Table 1), the even though no amebelodontine molars are known from Ad molars of the combined sample are consistent in having half- Dabtiyah, because this taxon is present at Moghara. loph(id)s composed of a massive outer cone and small meso- conelet; large, simple anterior and posterior accessory central Subfamily CHOEROLOPHODONTINAE Gaziry, 1976 conules contributing to pretrite trefoil ®gures with wear; and Genus AFROCHOERODON Pickford, 2001 tubercles blocking the lateral edges of interloph(id)s. The de- AFROCHOERODON KISUMUENSIS (MacInnes, 1942) gree of proportional variability in the combined sample is not greater than that observed in other gomphothere species, and Diagnosis Small choerolophodontine. Molar occlusal mor- may be accounted for in part by strong sexual dimorphism ap- phology simple; little or no choerolophodonty or ptychodonty. parently typical of gomphotheres (Tassy, 1985; GoÈhlich, 1998). Pre- and posttrite half-lophs offset transversely; pretrite meso- It is pertinent to mention here that Fourtau's (1918, 1920) di- conelets are anterior to posttrite conelet(s) and pretrite half- mensions for m2 in the ``M. libyca'' right dentary are dubious: lophs are angled obliquely, relative to the long axis of the tooth they are the same as for the ``M. libyca'' M2, and the width is (except for loph 1). Chevroning of lophs not pronounced. Dis- impossibly greater than that of the m3 (see Table 1). tinguished from other choerolophodontine species (except per- The gomphothere status of molars attributed to ``G. pyg- haps Choerolophodon palaeindicus) by small molar crown size, maeus'' (here in quotation marks because it is a nomen dubium simpler occlusal morphology, weaker chevroning of half-loph [Roger et al., 1994]) is also sound; despite their thick cemen- pairs, and fewer lophs in M3 (Tassy, 1985, 1986). tum, none show the characteristic choerolophodontine feature Referred Specimens DPC 14584, right M3 (Figs. 2B, 6A, of advancement of pretrite mesoconelets anterior to the posttrite B). half-lophs. Nonetheless, there is no evidence in the combined Description DPC 14584 is a moderately worn right M3 Moghara molar sample of a ``pygmy''-sized gomphothere spe- with three lophs, a thin anterior cingulum, and a small posterior cies. In contrast, they are considerably larger than the type of cingulum tightly appressed to the last loph, with six small ``G. pygmaeus'' from Kabylie and molars from a group of di- conelets. In occlusal view, the crown is sub-triangular (Fig. minutive Afro±Arabian gomphotheres from Siwa (including an 6A). In lateral view, the lophs are bulbous and apically diver- m3 [M 11964] from near Siwa purchased by The Natural His- gent (Fig. 6B). The interlophs are occupied along their basal tory Museum, London from Lady Moon in 1920), Gebel Zelten margins by low tubercles. There is a broad anterior interprox- (including a specimen from 20 km north of Zelten reported by imal facet. The enamel is thick and unfolded. There is no ce- Hormann, 1963), and Ghaba (Fig. 5). The Moghara specimens mentum. Each posttrite half-loph is composed of a single, mas- also lack the thick cementum found covering the crowns of the sive cone. Although more worn, it is evident that the pretrite molars from Kabylie and Gebel Zelten (DepeÂret, 1897; Ber- half-lophs are composed of two or three conelets. In lophs 2 gounioux and Crouzel, 1959; Gaziry, 1987). Molars of G. and 3, the pretrite mesoconelet is positioned anterior to the cor- browni from South Asia and G. steinheimense from Europe are responding posttrite conelet, and the pretrite half-lophs are an- larger (Fig. 5) and have a more complex crown structure than gled obliquely to the long axis of the tooth. Loph 3 has an those from Moghara. anterior accessory central conule fused with the pretrite meso- In occlusal morphology and proportions, the closest af®nity conelet, and a posterior accessory central conule is wedged in of the Moghara specimens is with the Gomphotherium sample interloph 2 on the pretrite side. The con®guration of pretrite from Ad Dabtiyah, Saudi Arabia (Fig. 5). Although the molars and posttrite half-lophs of loph 3 form a weak, anteriorly-point- from Ad Dabtiyah have been assigned to G. cooperi (Gentry, ed chevron. 396 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 22, NO. 2, 2002

FIGURE 6. Elephantoid molar specimens from Wadi Moghara. Specimens to same scale. A, Occlusal view, DPC 14584, right M3, Afrochoerodon kisumuensis. Anterior is to the left. B, Buccal view, DPC 14584, right M3, Afrochoerodon kisumuensis. Anterior is to the left. C, Occlusal view, CGM 30892, right M3, cf. Archaeobelodon. Anterior is to the left.

Remarks Specimen DPC 14584 is identi®ed as choerolo- phodon specimens from the Kamlial Formation of the Siwalik phodontine by the offset of pre- and posttrite half-lophs, and Series, Pakistan, which have been considered as possibly be- the oblique orientation of pretrite half-lophs 2 and 3. It most longing to C. palaeindicus (Tassy, 1983a), are more derived closely resembles the type M3s of Afrochoerodon kisumuensis, than A. kisumuensis in having four lophs in M3. Although mo- M 15524 (ϭKB-A-100), from Maboko, Kenya (MacInnes, lar morphology of middle Miocene A. chioticus from Chios, 1942; Tassy, 1977a, 1986). In the type M3s, as in the Moghara Greece is simple, the M3 of this species also has four lophs specimen, there are only three lophs, enamel is thick and un- and is considerably larger than DPC 14584 and the type M3s folded, the posterior cingulum is small, occlusal morphology is of A. kisumuensis (Tobien, 1980). Similarly, while exhibiting simple, and chevroning is weak. The type M3s and the Moghara occlusal morphology somewhat similar to that in the Moghara specimen have similar dimensions (Fig. 7), though in occlusal and Maboko specimens, M3s of A. zaltaniensis from Gebel Zel- view DPC 14584 appears more triangular in shape. Also, M ten differ in having four lophs, larger size, and greater choer- 15524 differs in having a trace of cementum in its interlophs. olophodonty (see Gaziry, 1987; contra Pickford, 1991b). Rath- The development of cementum appears to be highly variable in er, the Zelten Afrochoerodon is morphometrically identical to, choerolophodontine molars (Sanders, in press). Absence of ce- and may belong to the same species as, A. ngorora (Fig. 7; mentum in DPC 14584 and poor expression of this feature in Pickford, 2001), the earliest previous record of which is from the Maboko type M3s show that thick cementum is not a choer- the middle Miocene site of Fort Ternan, Kenya (see Tassy, olophodontine synapomorphy, contra Tobien (1973b). 1986). The distinctiveness of M3 in A. kisumuensis, and the Choerolophodon has recently been rede®ned to encompass striking similarity of DPC 14584 and the type M3s of A. kis- only late Miocene African and Eurasian choerolophodontines, umuensis, make it reasonable to assign the Moghara specimen with a new genus, Afrochoerodon, erected to accommodate ear- to that species. ly-mid Miocene African and Mediterranean species, including A. kisumuensis (Pickford, 2001). This division was based on Subfamily AMEBELODONTINAE Barbour, 1927 cranio-dental differences, particularly in the relative length and cf. ARCHAEOBELODON Tassy, 1984 height of the cranium and angulation of the face. Afrochoero- don kisumuensis is distinguished from other subfamilialsÐex- Diagnosis Small amebelodontine. Crown relatively narrow cept possibly Choerolophodon palaeindicus, from Dera Bugti, (Table 1); main cones massive and bunodont; mesoconelets PakistanÐby the plesiomorphic condition of its molars (Forster comparatively tiny; anterior and posterior accessory central Cooper, 1922; Raza and Meyer, 1984; Tassy, 1985, 1986). It is conules small and present on each pretrite half-loph; anterior possible that C. palaeindicus and A. kisumuensis are conspe- accessory central conules slightly larger than posterior central ci®c, as no known features distinguish them (Tassy, 1985). Un- conules; small anterior accessory central conules present on fortunately, only an m3 and a very worn M2 are known for C. each half-loph of the posttrite side. Distinguished from other palaeindicus, so there is no basis for direct comparison with amebelodontines by the small size of its pretrite accessory cen- DPC 14584. In any case, although its cranium is unknown, tral conules, and from Protanancus in particular by lack of an- because of its primitive molars Choerolophodon palaeindicus ancoidy of lophs, lesser development of cementum, and smaller should probably be transferred to Afrochoerodon. Choerolo- size (Tassy, 1983b, 1984, 1986). SANDERS AND MILLERÐNEW PROBOSCIDEANS FROM WADI MOGHARA 397

FIGURE 7. Bivariate plot of natural log-transformed M3 crown length FIGURE 8. Bivariate plot of natural log-transformed M3 crown length versus width in Afrochoerodon and Choerolophodon. Comparative di- versus width in Archaeobelodon and Protanancus. Comparative dimen- mensions are from Viret and YalcËinlar (1952), Gaziry (1976, 1987), sions are from Tassy (1983b, 1986), and Sanders (pers. obs.). Symbols: Tassy (1983a, 1986), Pickford (2001), and Sanders (pers. obs.). Sym- ૽, Wadi Moghara, Egypt (representing an estimate range of L ϭ 130± bols: ૽, Wadi Moghara, Egypt; Ⅺ, Afrochoerodon ngorora, Ft. Ternan 135 mm); ᭣, cf. Archaeobelodon, Rusinga, Kenya; ᭤, Archaeobelodon and Ngorora (Ngorora Formation, Member E), Kenya; ⅜, A. kisumuen- aff. ®lholi, Buluk, Kenya; , Protanancus macinnesi, Maboko and Ft. sis, Maboko and Cheparawa, Kenya; ࡗ, A. zaltaniensis, Gebel Zelten, Ternan, Kenya; , Protanancus chinjiensis, Chinji Formation, Siwalik Libya; ᭡, C. corrugatus, Dhok Pathan and Chinji formations, Siwalik Series, Pakistan. Series, Pakistan; Ⅵ, C. pentelici, Kayadibi and Kizil-Irmak, Turkey, and Maragheh, Iran; ᭝, cf. Choerolophodon, Kamlial Formation, Siwalik Series, Pakistan. 1973b; Tassy, 1985, 1986). Lateral expansion of the lower in- cisors reached its most extreme derivation in Platybelodon,in which they are shovel-shaped and lined with a mass of dentinal Referred Specimens CGM 30892, right M3 (Figs. 2C, tubules (Osborn and Granger, 1931, 1932). While there are no 6C). lower incisors attributable to an amebelodontine in the Moghara Description CGM 30892 is an anteriorly broken M3 that collection, the presence of posttrite anterior accessory conules retains enough of the crown to show that it originally had at indicates an amebelodontine af®nity for CGM 30892. Molars least four lophs and a small but distinct posterior cingulum of Gomphotherium angustidens may also have posttrite acces- (Figs. 2C, 6C). The fourth loph has a complete complement of sory conules, but they tend not to be as regularly developed as cones, conelets and accessory central conules, and is delineated in amebelodontines (Tassy, 1984, 1985). from the third loph by an ecto¯exus. Remnants of a cingulum The more precise identi®cation of CGM 30892 with the ear- are expressed as a narrow shelf on the lingual side; small tu- ly-to-middle Miocene genus Archaeobelodon of Europe and bercles also occupy the buccal exits to the transverse valleys. East Africa is sustained by the small size of its pre- and posttrite The crown is relatively narrow, and the lophs are aligned trans- accessory central conules, absence of posttrite posterior acces- versely relative to the long axis of the tooth. There is no chev- sory central conules, and lack of anancoidy (expressed to vary- roning or anancoidy of lophs, and only a trace of cementum in ing degrees in Protanancus) (Tassy, 1983b, 1984, 1985, 1986). the transverse valleys. The main cones of each each loph are Metrically, CGM 30892 is most similar to KNM-WS 21, attri- bunodont and massive in comparison with their accompanying buted to Archaeobelodon aff. ®lholi, a right M3 from Buluk, mesoconelets. On the pretrite side, the anterior and posterior Kenya measuring 124 mm in length and 61 mm in width (Fig. accessory central conules are tightly af®xed to the main cone 8; Tassy, 1986). Although the Moghara specimen is anteriorly of each half-loph, and embrace the pretrite mesoconelets. The broken, its greatest width is 60 mm and its original length can pretrite accessory central conules are modest in size, except for be estimated at 130±135 mm. Morphologically, it is closest to the ®rst posterior accessory central conule, which is consider- an unnamed species of cf. Archaeobelodon from Rusinga, Ken- ably larger and doubled. The most notable feature of the crown ya (MacInnes, 1942; Tassy, 1979, 1984, 1986). is the presence of anterior accessory central conules in each posttrite half-loph. These are petite but nevertheless conspicu- Family MAMMUTIDAE Hay, 1922 ous. There are, however, no posterior accessory central conules Genus ZYGOLOPHODON Vacek, 1877 on the posttrite side. ZYGOLOPHODON AEGYPTENSIS, sp. nov. Remarks Members of the Amebelodontinae are character- (Figs. 2D, 9A±F) ized by their tendency towards narrow molars, development of secondary trefoils on the posttrite sides of their molars, strongly Etymology Aegyptus (L.), referring to the country of oc- built symphyses, and broad, ¯attened lower incisors (Tobien, currence. 398 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 22, NO. 2, 2002

FIGURE 9. Zygolophodon aegyptensis, sp. nov. specimens from Wadi Moghara. Anterior is to the left. Specimens to same scale. A, Occlusal view, DPC 9009, right m3, type specimen. B, Lingual view, DPC 9009, right m3, type specimen. C, Occlusal view, DPC 5932, right M2. D, Occlusal view, DPC 5920, incomplete right M3. E, Occlusal view, DPC 12598, right M3. F, Buccal view, DPC 12598, right M3.

Holotype Primate Center, Duke University, DPC 9009, and four; from Eozygodon by relative narrowness of the crown right m3 (Fig. 9A, B). and stronger development of fourth loph(id)s in m3/M3; and Type Locality Wadi Moghara, northern Sahara, Egypt. from Mammut by absence of cementum, wider median sulci, Age Early Miocene, ca. 18±17 Ma (Miller, 1999). stronger expression of pretrite crescentoids, stronger expression Hypodigm DPC 5932, right M2 (Fig. 9C); DPC 12598, of cingula, less anteroposterior attenuation of loph(id) apices, right M3 (Figs. 2D, 9E, F); DPC 5920, incomplete right M3 narrower crowns, and smaller size (Anthony and Friant, 1940; (Fig. 9D). Tobien, 1975, 1996; Tassy, 1977b, 1985; Tassy and Pickford, Diagnosis Small mammutid with narrow molars (relative 1983). to length), particularly m3. Pretrite accessory central conules Description DPC 9009 (Fig. 9A, B) is a weathered and transformed into anterior and posterior crescentoids that vary worn right m3, with four lophids and a small, low posterior in shape from blunt pillars to sharp crests; these are most prom- cingulum. It preserves a small anterior root (under lophid 1) inent in the ®rst two loph(id)s. Smaller crescentoids, or ``zyg- and massive posterior root (below lophids 2±4). In occlusal odont crests,'' are formed on the posttrite side of upper molars. view, it is longitudinally curved (concave buccally) and excep- Distinguished from Zygolophodon turicensis (and all other spe- tionally narrow, relative to length (Fig. 10A). Lophids 3 and 4 cies of Zygolophodon) by the extreme narrowness of m3, m3 are strongly convex anteriorly. In lateral view, the lophids are interlophid breadth, and anterior convexity of m3 lophids three well spaced by anteroposteriorly broad, open transverse valleys. SANDERS AND MILLERÐNEW PROBOSCIDEANS FROM WADI MOGHARA 399

The posterior pretrite crescentoids are better developed than the anterior crescentoids, which they contact with wear. The post- trite half-lophids are smaller and simpler in construction than the pretrite half-lophids, and in lateral view these are apically constricted anteroposteriorly to form narrow crests. There are traces of lingual and buccal cingula, but these have been re- duced by the effects of natural sandblasting. DPC 5932 (Fig. 9C) is a nearly complete right M2, with three lophs and signs of low, narrow anterior and posterior cingula. Even with wear, a sharp, distinct median sulcus divides the half- lophs. A narrow cingulum is present on the lingual margin of the crown. The lophs are low and bulbous, and unaccompanied by accessory central conules. Zygodont crests and posterior crescentoids can be observed on posttrite half-lophs 1 and 2 and pretrite half-loph 2, respectively. There is no cementum. The anterior face of the crown is marked by a broad interprox- imal facet. DPC 12598 (Figs. 2D, 9E, F) is a complete right M3, with three lophs, a thin anterior cingulum composed of numerous tiny tubercles, and a modest-sized heel. The heel is composed of three conelets and forms an incipient fourth loph. As in the m3, the crown of DPC 12598 is relatively narrow (Fig. 10B). It is unworn and has incompletely formed roots, suggesting it was unerupted at time of death. In occlusal view, the half-loph pairs are divided by a distinct median sulcus, and the crown narrows considerably at the posterior end. The anterior cingu- lum continues along the lingual margin almost imperceptibly as a very thin ribbon of tiny tubercles. Each half-loph has a main abaxial conelet separated super®cially from a more medially placed secondary or adaxial conelet (with the exception of the ®rst pretrite half-loph). Lophs 1 and 2 exhibit strong posterior crescentoids; on the pretrite side they run medially from the outer main cone, and on the posttrite side zygodont crests de- scend towards the midline from the secondary conelets. The transverse valleys are V-shaped and are unencumbered by ac- cessory central conules. They are bordered by pyramidal- shaped lophs which apically narrow to a point. Transversely, these form crests that are highest laterally. There is little sign of anterior zygodont crests on the posttrite abaxial conelets. DPC 5920 (Fig. 9D) is a right M3 from a larger individual. It is broken at both ends, but preserves parts of four lophs. The specimen is slightly worn and apically its lophs are not as an- teroposteriorly narrow as those of DPC 12598. It also differs from DPC 12598 in having a relatively wider fourth loph com- prised of four conelets. As well, development of crescentoids and zygodont crests is blunter in this specimen than in DPC 12598. As pointed out by Tobien (1975), it can be dif®cult to distinguish between the molars of early Zygolophodon and Gomphotherium, but only the former has zygodont crests. In addition, while accessory central conules (found in Gomphoth- erium) and crescentoids (found in Zygolophodon) appear to be homologous structures, as in DPC 5920 crescentoids are more closely appressed to the pretrite main cones. There is a trace of cementum in the interlophs. The vestiges of a cingulum are present on the lingual margin of the crown. As in DPC 12598, lophs 2 and 3 are anteriorly convex in occlusal view.

←

FIGURE 10. Bivariate plots of natural log-transformed molar crown length versus width in Zygolophodon and Eozygodon. Comparative di- mensions are from Osborn (1936), Tassy (1977b, 1983b, 1985), Tassy and Pickford (1983), GoÈhlich (1998), and Sanders (pers. obs.). Sym- bols: ૽, Wadi Moghara, Egypt; ⅜, Zygolophodon turicensis, numerous early-late Miocene sites, Europe; ⅷ, Eozygodon morotoensis, Moroto, Uganda, and Meswa Bridge, Kenya; ࡗ, Z. atavus, Djilancik, Kazakhs- tan; ᭡, Z. metachinjiensis, middle-upper Chinji formation, Siwalik Se- ries, Pakistan; Ⅵ, Z. gobiensis, Tung Gur, China. A, m3. B, M3. C, M2. 400 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 22, NO. 2, 2002

Remarks The occurrence of Zygolophodon at Wadi Mog- (ca. 15 Ma, Pickford, 2001), it could as easily point to an older, hara is unexpected, as it is otherwise documented in the African early Miocene, age. There are strong similarities between the Miocene only by an upper molar fragment from Cherichira, comparable molars of A. kisumuensis from Maboko and of pos- Tunisia (Tassy, 1985). Nonetheless, its presence at Moghara is sibly conspeci®c Choerolophodon palaeindicus from the older plainly marked by a strikingly well preserved and unworn M3 Bugti Beds of Pakistan (Tassy, 1986), thought to be early Mio- (DPC 12598), which exhibits the characteristic occlusal mor- cene in age (Raza and Meyer, 1984; Barry et al., 1985; Bernor phology of the genus. Although this specimen is small and nar- et al., 1987; see below). In addition, the Afrochoerodon at Ge- row (Fig. 10B), a second M3 from the locality (DPC 5920) bel Zelten is morphologically reminiscent of the more advanced would have been larger when complete and proportionally more species A. ngorora, indicating that Moghara is older than the similar to M3s of Z. turicensis, as is the M2 (DPC 5932) (Fig. estimated age of 16.5 Ma for Gebel Zelten (Pickford, 1991b; 10C). The difference between these specimens in robustness contra Selley, 1969). and size is typical of the polymorphic nature of mammutid spe- The Moghara mammutid is somewhat less informative for cies (Tobien, 1975, 1996; Tassy, 1985). biochronological correlation. The most ancient mammutid, Eoz- More obviously distinctive is DPC 9009, which is extraor- ygodon morotoensis, is from Meswa Bridge, Kenya, with an dinarily slender for its length compared with m3s of Z. turi- age around 23±22 Ma (Bishop et al., 1969; Pickford and An- censis and other species of Zygolophodon (Fig. 10A). It should drews, 1981; Tassy and Pickford, 1983; Pickford, 1986). It is be noted that there is little to differentiate morphologically the also known from Moroto, Uganda, initially dated radiometri- molars of Zygolophodon species (Tassy, 1985), except for the cally to ca. 14 Ma (Bishop et al., 1969) but later revised to large size of the M3/m3s of Z. metachinjiensis. DPC 9009 is Ͼ17.5 Ma and Ͼ20.6 Ma by biochronological correlation and also peculiar in the anterior convexity of its third and fourth 40Ar±39Ar radioisotopic dating, respectively (Pickford et al., lophids. Other mammutids have m3s with transversely straight 1986; Gebo et al., 1997). The temporal distribution of Zygolo- lophids, set obliquely to the long axis of the crown (Tobien, phodon in Europe (including Z. turicensis) encompasses mam- 1996). The specimen also has extremely anteroposteriorly broad mal faunal zones MN4 through MN9 (Tassy, 1985; GoÈhlich, and open transverse valleys, giving it an unusual lateral pro®le. 1998), ca. 18±9.5 Ma (Steininger et al., 1996). The new mam- In this aspect DPC 9009 resembles an m3 of Z. gromovae (No. mutid from Moghara is morphologically and proportionally 3218-5) from the middle Miocene of Ulan Tologoj, Western closer to Z. turicensis than to E. morotoensis. Some justi®cation Mongolian People's Republic (Dubrovo, 1974:®g. 3); however, for an age assessment for Z. aegyptensis, sp. nov. commensu- that molar is considerably longer and would have been rela- rate with the older part of the Zygolophodon chronologic range tively wider when complete. in Europe is its small size of the third molars and strong ex- The new Moghara specimens are valuable in con®rming an pression of zygodont crests. These features did not undergo early distribution of Zygolophodon in North Africa. They clear- strong orthogonal change over time, but are more frequent in ly have closer phylogenetic af®nities with early Miocene mam- earlier than later members of Zygolophodon (Tassy, 1985; To- mutids of Europe than they do with East African forms. This bien, 1996). reinforces the biogeographic connection of Moghara to Europe The identi®cation of the Moghara gomphothere as G. angus- implied by the speci®c identity of its Gomphotherium speci- tidens libycum likewise has limited value for precision in bioch- mens with G. angustidens. It is more dif®cult to gauge whether ronological correlation because G. angustidens libycum is po- Z. aegyptensis is primitive or derived in relation to its European tentially known from only one other Afro-Arabian site, Ad congeners. Despite the persistence of Zygolophodon from the Dabtiyah, which cannot be radiometrically dated. Nonetheless, early Miocene into the Pliocene, the genus exhibited no obvious within Gomphotherium there is close correspondence between evolutionary trends (Tassy, 1985; Tobien, 1996), and the Mog- time and morphological change (Tassy, 1983a, 1994, 1996b), hara specimens are not proportional or structural intermediates so the occlusal organization of G. angustidens libycum molars between Eozygodon and other mammutids. can provide clues to their age. The most primitive gomphoth- eres are in the Gomphotherium ``annectens group'' (Tassy, DISCUSSION 1990, 1994, 1996b), temporally restricted to the early Miocene. This group includes G. cooperi from the Bugti Beds in Paki- The new proboscidean assemblage from Wadi Moghara ex- stan, which are Ͼ18.3 Ma, and possibly older than 20 Ma (Bar- hibits a surprising degree of taxonomic diversity, with four gen- ry et al., 1985; Johnson et al., 1985; Bernor et al., 1987; Tassy, era represented among ®fteen molars (Fig. 2). These fossils, 1996b). If the older age for the Bugti Beds is correct, G. cooperi and the proboscidean specimens from previous Moghara col- is the earliest known species of gomphothere. Other species in lections, can be allocated to Gomphotherium angustidens liby- the group, such as European G. sylvaticum, and G. annectens cum, Afrochoerodon kisumuensis, cf. Archaeobelodon, and Zyg- from the Hiramaki Formation of Mino, Japan, are from locali- olophodon aegyptensis, sp. nov. The co-occurrence of these ties associated with MN4 faunas (Tassy, 1985, 1996b) that date taxa suggests that a more elaborate biogeographic network to 18±17 Ma (Steininger et al., 1996). An additional taxon in linked Moghara with Eurasian fossil localities than was previ- the ''annectens group'' is Gomphotherium sp. from Mfwanga- ously suspected, and supports the idea that species-level ende- no, Kenya, which may be as old as 19 Ma (Pickford, 1986; mism of proboscideans throughout the early Miocene Afro± Tassy, 1986). It also occurs at Mwiti, Kenya, estimated to be Arabian region was relatively high. between 17 and 16 Ma (Pickford, 1986; Tassy, 1986). Molars of G. angustidens libycum are slightly more advanced in occlu- Biochronology sal structure than molars of this basal group. On the other hand, Based on faunal correlation, the Moghara fauna is dated to G. angustidens libycum is clearly more archaic than the middle- the early Miocene, about 18±17 Ma (Miller, 1996, 1999). How- late Miocene species G. steinheimense, from a series of MN7± ever, proboscideans were omitted from Miller's analyses due to MN9 localities in Germany (GoÈhlich, 1998), extending from confusion over their taxonomic af®liations. On balance, the pre- ca. 12.5 to 9.5 Ma (Steininger et al., 1996), and G. browni from sent results support Miller's (1996, 1999) ®ndings. For exam- the Chinji Formation, Siwalik Series, Pakistan (Tassy, 1983a), ple, while the presence of Afrochoerodon kisumuensis at Mog- spanning from slightly younger than 14 Ma to about 10.5 Ma hara marks an af®nity with the middle Miocene East African (Pilbeam et al., 1996). sites of Maboko (ca. 16±15 Ma; Tassy, 1977a; Andrews et al., Morphologically, the closest match for G. angustidens liby- 1981; Pickford, 1986; Feibel and Brown, 1991) and Cheparawa cum is with the earliest, simplest forms of European G. angus- SANDERS AND MILLERÐNEW PROBOSCIDEANS FROM WADI MOGHARA 401 tidens. Gomphotherium angustidens (including G.``subtapiro- cated by replacement in East Africa of an endemic African ideum''), principally from Western and Central Europe, ranges fauna, represented at Meswa Bridge around 22 Ma, by a more chronologically from MN5±MN9 (Tassy, 1985; GoÈhlich, 1998), diverse assemblage containing a number of Eurasian immi- approximately 17.0±9.5 Ma (Steininger et al., 1996; note that grants, such as at Songhor ca. 19 Ma (Bernor et al., 1987). while Mazo [1996] lists G. angustidens from MN 4 localities The geographic origins of the Moghara proboscideans are not in the Iberian Peninsula, these specimens are likely referrable clear, and this complexity hints at an intricate pattern of pro- to G. sylvaticum). Thus, the comparative development of oc- boscidean excursions throughout the early Miocene Old World. clusal structure in the molars of G. angustidens libycum is con- The amebelodont at Moghara, cf. Archaeobelodon, belongs to cordant with a terminal early Miocene age for the Moghara an extremely primitive species which apparently did not dis- fauna. Ad Dabtiyah, where G. angustidens libycum may also perse farther than North Africa from its probable origin in East occur, and Moghara have a low faunal resemblance index of Africa. It was, however, rapidly replaced toward the end of the 27% (three of eleven non-proboscidean mammalian genera early Miocene by the more widespread and advanced A. ®lholi, shared; see Whybrow, 1987; Miller, 1999). This probably re- which did reach Europe. At the same time, choerolophodontines ¯ects a slightly younger geological age for Ad Dabtiyah (Gen- were widely distributed from South Asia to East Africa. Un- try, 1987b, c; Pickford, 1987), and may be also be in¯uenced fortunately, the imprecise dating of the Bugti Beds in Pakistan by ecological and paleogeographic factors. In any case, the in- makes it dif®cult to know whether the distribution of this pro- ferred time difference (probably not greater than 2 Myr) be- boscidean subfamily documents its expansion from an East Af- tween the two sites is not extreme for the chronologic range of rican center of origin around 18.5±17.5 Ma, an earlier Africa± a proboscidean species. Eurasia immigration event, or supports a more vicariant pattern The strongest evidence among the proboscidean sample for of differentiation of elephantoids (see Raza and Meyer, 1984; an early Miocene age estimate comes from the recovery of an Thomas, 1985; Tchernov et al., 1987). Thus, choerolophodon- archaic amebelodont from the Moghara sediments. Salient fea- tines could have dispersed in either direction across an early tures of its occlusal morphology (tiny mesoconelets, small pre- Miocene intercontinental landbridge. Similarly, the Moghara trite accessory conules, lack of anancoidy, absence of posttrite species of Zygolophodon could be part of the group of mam- posterior accessory central conules) mark it as belonging to a mutids descended from East African Eozygodon that penetrated less derived taxon than Archaeobelodon ®lholi or Protanancus. into Europe during the early Miocene ``Proboscidean Datum The former species occurs in European sites associated with Event'' along with Gomphotherium (Berggren and Van Cou- MN4±8 faunas (Tassy, 1984, 1985), approximately 18.0±11.2 vering, 1974; Madden and Van Couvering, 1976; Thomas, Ma. It may also be present near the base of the sequence at 1985; Tassy, 1977b, 1990). Alternatively, it could represent a Mwiti (17±16 Ma), and at Buluk, Kenya (ca. 17.2 Ma) (Pick- fringe population from a subsequent re-invasion of Africa by ford and Andrews, 1981; McDougall and Watkins, 1985; Pick- Eurasian Zygolophodon. The poor fossil record of Zygolopho- ford, 1986; Tassy, 1986). Protanancus macinnesi is known from don in the Miocene of North Africa (Tassy, 1985) and absence middle Miocene sites of East Africa, including Maboko, as well of the genus in East Africa accords more closely with the latter as from middle Miocene sites in southern Africa and the West- idea. Finally, as its dentition is more similar to that of G. an- ern Rift (Tassy, 1985, 1986); its South Asian congener P. chin- gustidens than the G.``annectens-group,'' Moghara Gomphoth- jiensis is essentially con®ned to the Chinji Formation (Tassy, erium was probably not part of the initial radiation of African 1983b, 1985; see above). Rather, the closest af®nitiy of the fauna into Eurasia (``Proboscidean Datum Event''). To the con- Moghara specimen is with cf. Archaeobelodon from Rusinga, trary, Gomphotherium angustidens libycum probably represents Kenya (MacInnes, 1942; Tassy, 1984, 1986), to which it is sim- a reverse incursion, into Africa from Europe, at the close of the ilarly primitive. Rusinga has an estimated age of 17.8 Ma (Pick- early Miocene. Gomphotherium angustidens libycum and Z. ae- ford, 1981, 1986; Drake et al., 1988), and therefore predates gyptensis, sp. nov. are reasonably interpreted as local North other East African amebelodont sites. Biochronologically, Wadi African variants of the European species G. angustidens and Z. Moghara is most comparable to Rusinga, with 75% of its non- turicensis, respectively. proboscidean mammalian genera in common (Miller, 1996, The overall taxonomic uniqueness of such a diverse Moghara 1999). proboscidean assemblage provides an argument for endemism at the species level. This idea has been advanced for the Mog- Biogeography hara fauna as a whole (Miller, 1996), and a ``quasi-spatial iso- lation'' caused by environmental differences has been suggested Recent work on the faunal similarities of Moghara indicated for mammalian faunas of other pene-contemporaneous Afro± a strong biogeographic connection with East Africa (Miller, Arabian sites (Tchnernov et al., 1987:284). For example, the 1996, 1999), and the identi®cation of an archaic amebelodon- taxonomic diversity and endemism of anthracotheres from tine and A. kisumuensis at Moghara reinforces that link. How- Moghara parallel the relative uniqueness of the proboscidean ever, A. kisumuensis is virtually indistinguishable from the assemblage. At Moghara, there are four anthracothere species, choerolophodontine of the Bugti Beds, Pakistan, suggesting that three of which do not occur elsewhere (the fourth is found early Miocene proboscideans used a faunal corridor between outside Moghara only at one other locality, Siwa, Egypt) (Pick- North Africa and southern Eurasia. Similarly, the recovery of ford, 1991a). The case for endemism in the Moghara fauna is Zygolophodon from Moghara and the recognition of strong further strengthened by an examination of the distribution pat- morphological correspondence between the Moghara gom- terns of ``pygmy'' gomphotheres and deinotheres. ``Pygmy'' photheres and Gomphotherium angustidens also demonstrate an gomphotheres are present at a number of sites throughout Afro± early Miocene connection with Europe. Arabia, which are close both geographically and chronologi- The timing and occurrence of Eurasian proboscidean ele- cally to Moghara. The absence of ``pygmy'' gomphotheres at ments in the Moghara fauna is consistent with geological evi- Moghara hints at a degree of regional ecological heterogeneity dence indicating the establishment of an early Miocene (20±18 during the early Miocene. Similarly, deinotheres are some of Ma) landbridge between Afro±Arabia and Eurasia as a result the most common faunal elements in early Miocene Afro±Ara- of the collision of the Arabian peninsula with the Anatolian bian faunas, but so far they are curiously absent from Moghara. plate (Adams et al., 1983; RoÈgl and Steininger, 1983; Thomas, Harris (1978:321), citing Osborn (1936:103) as his source, re- 1985; RoÈgl, 1998, 1999). The opportunity for intercontinental ported that isolated teeth of Prodeinotherium hobleyi from interchange of mammalian faunas during this interval is indi- Moghara were housed at the British Museum (Natural History). 402 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 22, NO. 2, 2002

However, it is likely that Harris mis-read Osborn's comment on letin du Museum National d'Histoire Naturelle, Paris, 2e SeÂrie 12: Fourtau's (1918, 1920) observation about the absence of dein- 449±455. otheres at Moghara. We have found no deinothere material from Arambourg, C. 1961. Note preÂliminaire sur quelques VerteÂbreÂs nou- Moghara either in museum collections or among the enhanced veaux du Burdigalien de Libye. Comptes Rendus Sommaire des Seances de la SocieÂteÂGeÂologique de France 1961:107±109. sample of proboscideans from the site. The fact that Moghara Barbour, E. H. 1927. Preliminary notice of a new proboscidean Ame- shares many generic-level but not many species-level probos- belodon fricki, gen. et sp. nov. Nebraska State Museum, Bulletin cideans with other Afro±Arabian early Miocene localities sug- 13:131±134. gests a scenario of a richly speciose proboscidean fauna dis- Barry, J. C., N. M. Johnson, S. M. Raza, and L. L. Jacobs. 1985. Neo- tributed over diverse landscapes, and engaged in more compli- gene mammalian faunal change in southern Asia: correlations with cated migration events than was previously suspected. climatic, tectonic, and eustatic events. Geology 13:637±640. Berggren, W. A., and J. A. Van Couvering. 1974. The late NeogeneÐ biostratigraphy, geochronology, and paleoclimatology of the last 15 CONCLUSIONS million years in marine and continental sequences. Palaeogeogra- The new proboscidean collection from Wadi Moghara is sub- phy, Palaeoclimate, Palaeoecology 16:1±216. Bergounioux, F. M., and F. Crouzel. 1959. Nouvelles observations sur stantially larger than previous samples from the site. Compar- un petit Mastodonte du Cartennien de Kabylie. Comptes Rendus gative morphometric study of all specimens indicates the fol- Sommaire des Seances de la SocieÂteÂGeÂologique de France 1959: lowing: (1) the Moghara proboscidean fauna is taxonomical- 101±102. ly diverse, containing three gomphotheres, Gomphotherium Bernor, R. L., M. Brunet, L. Ginsburg, P. Mein, M. Pickford, F. RoÈgl, angustidens libycum, Afrochoerodon kisumuensis, and cf. Ar- S. Sen, F. Steininger, and H. Thomas. 1987. A consideration of chaeobelodon, and a mammutid, Zygolophodon aegyptensis, sp. some major topics concerning Old World Miocene mammalian nov.; (2) no evidence at Moghara of ``pygmy'' gomphotheres chronology, migrations and paleogeography. Geobios 20:431±439. or deinotheres; (3) the estimated age of the proboscidean as- Bishop, W. W., J. A. Miller, and F. J. Fitch. 1969. New potassium-argon semblage is consistent with the recent determination of an early age determinations relevant to the Miocene fossil mammal se- quence in East Africa. American Journal of Science 267:669±699. Miocene age, ca. 18±17 Ma for Moghara (Miller, 1996, 1999); Bown, T. M., M. J. Kraus, S. L. Wing, J. G. Fleagle, B. H. Tiffney, E. (4) by comparison, the composition of the proboscidean fauna L. Simons, and C. F. Vondra. 1982. The Fayum primate forest re- at Gebel Zelten, Libya, suggests a slightly younger age (Pick- visited. Journal of Human Evolution 11:603±632. ford, 1991b; Miller, 1999); and (5) the Moghara proboscideans Burmeister, H. 1837. Handbuch der Naturgeschichte. Zum Gebrauch bei reveal complex biogeographic af®nities with East Africa, Ara- Vorlesungen entworfen. Zweite Abteilung. Zoologie. T. C. F. En- bia, Europe, and possibly South Asia, but also agree with a slin, Berlin, 795 pp. tendency for local mammalian species-level endemism, appar- Coppens, Y., V. J. Maglio, C. T. Madden, and M. Beden. 1978. Pro- ently common elsewhere throughout early Miocene Afro±Ara- boscidea; pp. 336±367 in V. J. Maglio and H. B. S. Cooke (eds.), bia. Evolution of African Mammals. Harvard University Press, Cam- bridge. DepeÂret, C. 1897. DeÂcouverte du Mastodon angustidens dans l'eÂtage ACKNOWLEDGMENTS Cartennien de Kabylie. Bulletin de la SocieÂteÂGeÂologique de France, Series 3 25:518±521. We are grateful to Meave Leakey (National Museums of Drake, R. L., J. A. Van Couvering, M. Pickford, G. H. Curtis, and J. Kenya), M. el-Bedawi (Cairo Geological Museum), Elwyn Si- A. Harris. 1988. New chronology for the early Miocene mamma- mons (Duke University Primate Center), Jeremy Hooker (The lian fauna of Kisingiri, western Kenya. Journal of the Geological Natural History Museum, London), Berna Alpagut (Ankara Society, London 145:479±491. University), Ilhan Temizsoy (Anatolian Civilizations Museum), Dubrovo, I. A. 1974. Some new data on mastodonts from Western Mon- and Muhammed Arif (Paleontology and Stratigraphy Branch, golia; pp. 64±73 in N. M. Kramarenko (ed.), Mesozoic and Ce- Geological Survey of Pakistan) for access to specimens in their nozoic Faunas and Biostratigraphy of Mongolia. Izdatelstro ``Nau- care. G. Naim (Egyptian Geological Survey and Mining Au- ka,'' Moscow. Feibel, C. S., and F. H. Brown. 1991. Age of the primate-bearing de- thority) gave permission for ®eldwork at Wadi Moghara. Bon- posits on Maboko Island, Kenya. Journal of Human Evolution 21: nie Miljour provided expert assistance with the ®gures. We 221±225. thank Richard Hulbert, Nicholas Czaplewski, Jeheskel Sho- Forster Cooper, C. 1922. Miocene Proboscidia [sic] from Baluchistan. shani, and Pascal Tassy for their careful efforts to improve the Proceedings of the Zoological Society of London 1922:606±626. manuscript. Fieldwork at Wadi Moghara and museum research Fourtau, R. 1918. Contribution a l'EÂ tude VerteÂbreÂs MioceÁnes de of ERM was supported by a Dissertation Improvement Grant l'EÂ gypte. Survey Department, Ministry of Finance, Egypt, Govern- from the National Science Foundation, The L.S.B. Leakey ment Press, Cairo, 121 pp. Foundation, The Boise Fund of Oxford University, Sigma Xi, ÐÐÐ 1920. Contribution a l'EÂ tude VerteÂbreÂs MioceÁnes de l'EÂ gypte. and a Dean's Dissertation Award from Washington University. Survey Department, Ministry of Finance, Egypt, Government Museum research of WJS was supported by a Scott Turner Press, Cairo, 121 pp. [2nd printing] Gaziry, A. W. 1976. JungtertiaÈre Mastodonten aus Anatolien (TuÈrkei). Award in Earth Science from the Department of Geological Geologisches Jahrbuch 22:3±143. Sciences, University of Michigan, and funding from Terry Har- ÐÐÐ 1987. New mammals from the Jabal Zaltan site, Libya. Senck- rison (Wembere-Manonga Expedition, Tanzania) and Mikael enbergiana lethaea 68:69±89. Fortelius and John Kappelman (Sinap Project, Turkey). Gebo, D. L., L. MacLatchy, R. Kityo, A. Deino, J. Kingston, and D. Pilbeam. 1997. A hominoid genus from the Early Miocene of LITERATURE CITED Uganda. Science 276:401±404. Gentry, A. W. 1987a. Mastodons from the Miocene of Saudi Arabia. Adams, C. G., A. W. Gentry, and P. J. Whybrow. 1983. Dating the Bulletin of the British Museum (Natural History) (Geology) 41: terminal Tethyan event. Utrecht Micropaleontology Bulletin 30: 395±407. 273±298. ÐÐÐ 1987b. Rhinoceroses from the Miocene of Saudi Arabia. Bul- Andrews, P., G. E. Meyer, D. R. Pilbeam, J. A. Van Couvering, and J. letin of the British Museum (Natural History) (Geology) 41:409± A. H. Van Couvering. 1981. The Miocene fossil beds of Maboko 432. Island, Kenya: geology, age, taphonomy and palaeontology. Journal ÐÐÐ 1987c. Ruminants from the Miocene of Saudi Arabia. Bulletin of Human Evolution 10:35±48. of the British Museum (Natural History) (Geology) 41:433±439. Anthony, R., and M. Friant. 1940. Remarques sur le Mastodon borsoni GoÈhlich, U. B. 1998. Elephantoidea (Proboscidea, Mammalia) aus dem Hays et les autres Mastodontes Zygolophodontes de l'Europe. Bul- Mittel- und ObermiozaÈn der Oberen SuÈûwassermolasse SuÈd- SANDERS AND MILLERÐNEW PROBOSCIDEANS FROM WADI MOGHARA 403

deutschlands: Odontologie und Osteologie. MuÈnchner Geowissen- ÐÐÐ 1991a. Revision of Neogene Anthracotheriidae of Africa; pp. schaftliche Abhandlungen (A) 36:1±245. 1491±1525 in M. J. Salem and M. T. Busrewil (eds.), The Geology Gray, J. E. 1821. On the natural arrangements of vertebrose animals. of Libya. Academic Press, New York. London Medical Repository 15:296±310. ÐÐÐ 1991b. Biostratigraphic correlation of the Middle Miocene Hamilton, W. R. 1973. A lower Miocene mammalian fauna from Siwa, mammal locality of Jabal Zaltan, Libya; pp. 1483±1490 in M. J. Egypt. Palaeontology 16:275±281. Salem, O. S. Hammuda, and B. A. Eliagoubi (eds.), The Geology ÐÐÐ, P. J. Whybrow, and H. A. McClure. 1978. Fauna of fossil of Libya (volume IV). Academic Press, New York. mammals from the Miocene of Saudi Arabia. Nature 274:248. ÐÐÐ 2001. Afrochoerodon nov. gen. kisumuensis (MacInnes) (Pro- Hantar, G. 1990. North Western Desert; pp. 293±320 in R. Said (ed.), boscidea, Mammalia) from Cheparawa, Middle Miocene, Kenya. The Geology of Egypt. Elsevier, Amsterdam. Annales de PaleÂontologie 87:99±117. Harris, J. M. 1978. Deinotherioidea and Barytherioidea; pp. 315±332 ÐÐÐ, and P. Andrews. 1981. The Tinderet Miocene sequence in Ken- in V. J. Maglio and H. B. S. Cooke (eds.), Evolution of African ya. Journal of Human Evolution 10:11±33. Mammals. Harvard University Press, Cambridge. ÐÐÐ, B. Senut, D. Hadoto, J. Musisi, and C. Kariira. 1986. DeÂcou- Hay, O. P. 1922. Further observations on some extinct elephants. Pro- vertes recentes dans le sites mioceÁnes de Moroto (Ouganda Ori- ceedings of the Biological Society of Washington 35:97±101. ental): aspects biostratigraphiques et paleoecologiques. Comptes Hormann, K. 1963. Note on a mastodontoid from Libya. Zeitschrift fuÈr Rendus de l'Academie des Sciences, SeÂrie II, 302:681±686. SaÈugetierkunde 28:88±93. Pilbeam, D., M. Morgan, J. C. Barry, and L. Flynn. 1996. European Illiger, C. D. 1811. Prodromus Systematis Mammalium et Avium Ad- MN units and the Siwalik Faunal Sequence of Pakistan; pp. 96± ditis Terminis Zoographicis Uttriusque Classis. Salfeld, Berlin, 301 105 in R. L. Bernor, V. Fahlbusch, and H.-W. Mittmann (eds.), The pp. Evolution of Western Eurasian Neogene Mammal Faunas. Colum- Johnson, N. M., J. Stix, L. Tauxe, P. F. Cerveny, and R. A. K. Tahirkheli. bia University Press, New York. 1985. Paleomagnetic chronology, ¯uvial processes, and tectonic Raza, S. M., and G. E. Meyer. 1984. Early Miocene Geology and Pa- implications of the Siwalik deposits near Chinji Village, Pakistan. leontology of the Bugti Hills, Pakistan. Memoirs of Geological Sur- Journal of Geology 93:27±40. vey of Pakistan 11:43±63. Linnaeus, C. 1758. Systema naturae. Editio decima, reformata. Stock- Roger, J., M. Pickford, H. Thomas, F. de Lapparent de Broin, P. Tassy, holm, Laurentii Salvii 1:1±824. W. Van Neer, C. Bourdillon-de-Gissac, and S. Al-Busaidi. 1994. MacInnes, D. G. 1942. Miocene and post-Miocene Proboscidia [sic] DeÂcouverte de VerteÂbreÂs Fossiles dans le MioceÁnedelaReÂgion du from East Africa. Transactions of the Zoological Society of London Huqf au Sultanat d'Oman. Annales de PaleÂontologie (Vert.-Invert.) 25:33±106. 80:253±273. Madden, C. T., and J. A. Van Couvering. 1976. The proboscidean datum RoÈgl, F. 1998. Palaeogeographic considerations for Mediterranean and event: early Miocene migration from Africa. Geological Society of Paratethys Seaways (Oligocene to Miocene). Annalen des Natur- America Abstracts with Program:992±993. historischen Museums in Wien 99A:279±310. Marzouk, I. 1969. Rock stratigraphy and oil potentialities of the Oli- ÐÐÐ 1999. Mediterranean and Paratethys palaeogeography during gocene and Miocene in the Western Desert. UAR 7th Arab Petro- the Oligocene and Miocene; pp. 8±22 in J. Agusti, L. Rook, and leum Congress, Kuwait 54 (B-3). P. Andrews (eds.), Hominoid Evolution and Climatic Change in Mazo, A. V. 1996. Gomphotheres and mammutids from the Iberian Europe, Vol. 1. The Evolution of Neogene Terrestrial Ecosystems Peninsula; pp. 136±142 in J. Shoshani and P. Tassy (eds.), The in Europe. Cambridge University Press, Cambridge. Proboscidea. Evolution and Palaeoecology of Elephants and Their ÐÐÐ, and F. Steininger. 1983. Vom Zerfall der Tethys zu Mediterran Relatives. Oxford University Press, Oxford. und Paratethys. Annalen des Naturhistorischen Museums in Wien McDougall, I., and R. T. Watkins. 1985. Age of hominoid-bearing se- 85:135±163. quence of Buluk, northern Kenya. Nature 318:175±178. Said, R. 1962a. UÈ ber das MiozaÈn in der westlichen wuÈste AÈ gyptens. Miller, E. R. 1996. Mammalian paleontology of an Old World monkey Geologisches Jahrbuch 80:349±366. locality, Wadi Moghara, early Miocene, Egypt. Ph.D. dissertation, ÐÐÐ 1962b. The Geology of Egypt. Elsevier, Amsterdam, 377 pp. Washington University, St. Louis, 372 pp. ÐÐÐ 1990. Cenozoic; pp. 451±486 in R. Said (ed.), The Geology of ÐÐÐ 1999. Faunal correlation of Wadi Moghara, Egypt: implications Egypt. Elsevier, Amsterdam. for the age of Prohylobates tandyi. Journal of Human Evolution Sanders, W. J. In press. Proboscideans from the Sinap Formation, Cen- 36:519±533. tral Turkey; in J. Kappelman, M. Fortelius, S. Sen, and R. Bernor Omara, S., and K., Ouda. 1972. Review of the lithostratigraphy of the (eds.), Geology and Paleontology of the Miocene Sinap Formation, Oligocene and Miocene in the northern Western Desert. 8th Arab Central Turkey. Columbia University Press, New York. Petroleum Congress, Algiers. Savage, R. J. G. 1989. The African dimension in European early Mio- ÐÐÐ, and S. Sanad. 1975. Rock stratigraphy and structural features cene mammal faunas; pp. 587±599 in E. H. Lindsay, V. Fahlbusch, of the area between Wadi Natrun and the Moghara depression and P. Mein (eds.), European Neogene Mammal Chronology. Ple- (Western Desert, Egypt). Geologisches Jahrbuch 16:45±73. num Press, New York. Osborn, H. F. 1932. Trilophodon cooperi, sp. nov., of Dera Bugti, Ba- ÐÐÐ, and W. R. Hamilton. 1973. Introduction to the Miocene mam- luchistan. American Museum Novitates 585:1±6. mal faunas of Gebel Zelten, Libya. Bulletin of the British Museum ÐÐÐ 1936. Proboscidea. A Monograph of the Discovery, Evolution, (Natural History) (Geology) 22:513±527. Migration and Extinction of the Mastodonts and Elephants of the Selley, R. C. 1969. Near-shore marine and continental sediments of the World, Vol. I. Moerotherioidea, Deinotherioidea, Mastodontoidea. Sirte Basin, Libya. Quaternary Journal, Geological Society of Lon- American Museum Press, New York, 802 pp. don 124:419±460. ÐÐÐ 1942. Proboscidea. A Monograph of the Discovery, Evolution, Steininger, F. F., W. A. Berggren, D. V. Kent, R. L. Bernor, S. Sen, and Migration and Extinction of the Mastodonts and Elephants of the J. Agusti. 1996. Circum-Mediterranean Neogene (Miocene and Pli- World, Vol. II. Stegodontoidea, Elephantoidea. American Museum ocene) marine-continental chronologic correlations of European Press, New York, 828 pp. mammal units; pp. 7±46 in R. L. Bernor, V. Fahlbusch, and H.-W. ÐÐÐ, and W. Granger. 1931. The shovel-tuskers, Amebelodontinae, Mittman (eds.), The Evolution of Western Eurasian Neogene Mam- of Central Asia. American Museum Novitates 470:1±12. mal Faunas. Columbia University Press, New York. ÐÐÐ, and ÐÐÐ 1932. Platybelodon grangeri, three growth stages, Tassy, P. 1977a. PreÂsence du genre Choerolophodon Schlesinger (Pro- and a new Serridentinae from Mongolia. American Museum Nov- boscidea, Mammalia) dans le MioceÁne est-africain. Comptes Ren- itates 537:1±13. dus de l'Academie des Sciences, SeÂrie D 284:2487±2490. Pickford, M. 1981. Preliminary Miocene mammalian biostratigraphy for ÐÐÐ 1977b. Decouverte de Zygolophodon turicensis (Schinz) (Pro- western Kenya. Journal of Human Evolution 10:73±97. boscidea, Mammalia) au Lieu-Dit Malartic a Simorre, Gers (Vin- ÐÐÐ 1986. Cainozoic paleontological sites of Western Kenya. MuÈn- dobonien Moyen); Implications paleÂoeÂcologiques et biostratigrap- chner Geowissenschaftliche Abhandlungen (A) 8:1±151. hiques. Geobios 10:655±659. ÐÐÐ 1987. Miocene Suidae from Ad Dabtiyah, eastern Saudi Arabia. ÐÐÐ 1979. Les Proboscidiens (Mammalia) du MioceÁne d'Afrique Bulletin of the British Museum (Natural History) (Geology) 41: orientale: reÂsultats preÂliminaire. Bulletin de la SocieÂteÂGeÂologique 441±446. de France 21:265±269. 404 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 22, NO. 2, 2002

ÐÐÐ 1983a. Les Elephantoidea MioceÁnes du Plateau du Potwar, dentale. Comptes rendus de l'AcadeÂmie des Sciences, Paris (SeÂrie Groupe de Siwalik, Pakistan. IIe Partie: Choerolophodontes et II) 309:2143±2146. GomphotheÁres. Annales de PaleÂontologie (Vert.-Invert.) 69:235± Tchernov, E., L. Ginsburg, P. Tassy, and N. F. Goldsmith. 1987. Miocene 297. mammals of the Negev (Israel). Journal of Vertebrate Paleontology ÐÐÐ 1983b. Les Elephantoidea MioceÁnes du Plateau du Potwar, 7:284±310. Groupe de Siwalik, Pakistan. Ie Partie: Introduction, Cadre chron- Thomas, H. 1985. The early and middle Miocene land connection of ologique et geÂographique, MammutideÂs, AmeÂbeÂlodontideÂs. Anna- the Afro-Arabian plate and Asia: a major event for hominoid dis- les de PaleÂontologie (Vert.-Invert.) 69:96±136. persal?; pp. 42±50 in E. Delson (ed.), Ancestors: The Hard Evi- ÐÐÐ 1984. Le mastodonte aÁ dents eÂtroites, le grade trilophodonte et dence. Alan R. Liss, New York. la radiation initiale des Amebelodontidae; pp. 459±473 in E. Buf- ÐÐÐ, S. Sen, M. Khan, B. Battail, and G. Ligabue. 1982. The lower fetaut, J.-M. Mazin, and E. Salmon (eds.), Actes du Symposium Miocene fauna of As-Sarrar (Eastern Province, Saudi Arabia). At- paleÂontologique G. Cuvier, MontbeÂliard. lal, Journal of Saudi Arabian Archaeology 5:109±136. ÐÐÐ 1985. La place de mastodontes mioceÁnes de l'Ancien Monde Tobien, H. 1973a. The structure of the mastodont molar (Proboscidea, dans la phylogeÂnie des Proboscidea (Mammalia): hypotheÁses et Mammalia). Part 1: the bunodont pattern. Mainzer Geowissenschaf- ten Mitteilungsblatt 2:115±147. conjectures. These de Doctorat d'Etat eÁs Sciences Naturelles, Univ- ÐÐÐ 1973b. On the evolution of mastodonts (Proboscidea, Mam- ersite Pierre et Marie Curie, Paris, 861 pp. malia). Part 1: the bunodont trilophodont groups. Notizblatt des ÐÐÐ 1986. Nouveaux Elephantoidea (Mammalia) dans le MioceÁne  Hessichen Landesamtes fuÈr Bodenforschung zu Wiesbaden 101: du Kenya. Cahiers de Paleontologie, Editions du Centre de la Re- 202±276. cherche Scienti®que, Paris, 135 pp. ÐÐÐ 1975. The structure of the mastodont molar (Proboscidea, Mam- ÐÐÐ 1988. The classi®cation of Proboscidea: how many cladistic malia). Part 2: the zygodont and zygobunodont patterns. Mainzer classi®cations? Cladistics 4:43±57. Geowissenschaften Mitteilungsblatt 4:195±233. ÐÐÐ 1990. The ``Proboscidean Datum Event'': how many probos- ÐÐÐ 1980. A note on the skull and mandible of a new choerolopho- cideans and how many events?; pp. 237±252 in E. H. Lindsay, V. dont mastodont (Proboscidea, Mammalia) from the Middle Mio- Fahlbusch, and P. Mein (eds.), European Neogene Mammal Chro- cene of Chios (Aegean Sea, Greece); pp. 299±307 in L. L. Jacobs nology. Plenum Press, New York. (ed.), Aspects of Vertebrate History. Museum of Northern Arizona ÐÐÐ 1994. Gaps, parsimony, and early Miocene elephantoids (Mam- Press, Flagstaff, Arizona. malia), with a re-evaluation of Gomphotherium annectens (Mat- ÐÐÐ 1996. Evolution of zygodons with emphasis on dentition; pp. sumoto, 1925). Zoological Journal of the Linnean Society 112:101± 76±85 in J. Shoshani and P. Tassy (eds.), The Proboscidea. Evo- 117. lution and Palaeoecology of Elephants and Their Relatives. Oxford ÐÐÐ 1996a. Dental homologies and nomenclature in the Probosci- University Press, Oxford. dea; pp. 21±25 in J. Shoshani and P. Tassy (eds.), The Proboscidea. Vacek, M. 1877. UÈ ber oÈsterreichische Mastodonten und ihre Beziehun- Evolution and Palaeoecology of Elephants and Their Relatives. Ox- gen zu den Mastodon-Arten Europas. Abhandlungen der Kaiser- ford University Press, Oxford. lich-KoÈniglichen geologischen Reichenstalt 7:1±45. ÐÐÐ 1996b. The earliest gomphotheres; pp. 89±91 in J. Shoshani Viret, J., and I. YalcËinlar. 1952. Synconolophus serridentinoides, nou- and P. Tassy (eds.), The Proboscidea. Evolution and Palaeoecology velle espeÁce de Mastodonte du Miocene supeÂrieur de Turquie. of Elephants and Their Relatives. Oxford University Press, Oxford. Comptes Rendus de l'AcadeÂmie des Sciences 254:870±872. Whybrow, P. J. 1984. Geological and faunal evidence from Arabia of ÐÐÐ, and M. Pickford. 1983. Un nouveau Mastodonte zygolopho- ``mammal'' migrations between Asia and Africa during the early donte (Proboscidea, Mammalia) dans le MioceÁne InfeÂrieur Miocene. Courier Forschungsinstitut Senckenberg 69:189±198. d'Afrique Orientale: SysteÂmatique et PaleÂoenvironnement. Geobios ÐÐÐ 1987. Miocene geology and palaeontology of Ad Dabtiyah, 16:53±77. Saudi Arabia. Summary. Bulletin of the British Museum (Natural ÐÐÐ, S. Sen, J.-J. Jaeger, J.-M. Mazin, and N. Dalfes. 1989. Une History) (Geology) 41:367±369. sous-espeÁce nouvelle de Choerolophodon pentelici (Proboscidea, Mammalia) aÁ Esme AkcakoÈy, MioceÁne superieÂur d'Anatolie occi- Received 13 July 2000; accepted 4 August 2001.