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

Home , Cornwallius, Desmostylia, Desmostylus, Paenungulata, Tethytheria

AMERICANt MUSEUM Novitates PUBLISHED BY THE AMERICAN MUSEUM OF NATURAL HISTORY CENTRAL PARK WEST AT 79TH STREET, NEW YORK, N.Y. 10024 Number 2870, pp. 1-8, figs. 1-3 April 6, 1987

Selected Features of the Desmostylian Skeleton and Their Phylogenetic Implications

MICHAEL J. NOVACEK1 AND ANDRE R. WYSS2

ABSTRACT According to several standard descriptions, des- pertaining to these traits are either in error, or have mostylians lack certain specializations shared by alternative phylogenetic implications. Hence, proboscideans, sirenians, and hyracoids. These comparisons of these conditions do not exclude specializations are amastoidy and the serial ar- desmostylians from the superordinal group Teth- rangement ofthe carpals with the concomitant loss ytheria (proboscideans and sirenians) or the more of contact between the lunar and unciform. We inclusive Paenungulata (tethytheres and hyra- argue that original descriptions of desmostylians coids). INTRODUCTION The Desmostylia are an extinct order of be more conservative than sirenians (and, by mammals with specializations ofthe skeleton implication, proboscideans) with respect to suitable for an amphibious mode of life. Al- the reduction ofmastoid exposure (Hay, 1915; though this group was formerly associated Abel, 1922; VanderHoof, 1937) and the se- with the aquatic Sirenia (Simpson, 1945), and rial arrangement ofthe carpal elements (Shi- is now placed within a superordinal category kama, 1966). These differences have influ- Tethytheria, which also includes the Probos- enced more recent students of the problem cidea and Sirenia (McKenna, 1975), these (e.g., Tassy, 1981), who have opted for a close avowed relationships present some prob- association ofProboscidea (including the late lems. According to standard descriptions, Eocene-early Oligocene genus Moeritheri- desmostylians lack certain skeletal special- um) and Sirenia to the exclusion of Desmo- izations shared by Proboscidea and Sirenia. stylia. Specifically, desmostylians were claimed to Of significance is the fact that the above ' Chairman and Associate Curator, Department ofVertebrate Paleontology, American Museum ofNatural History. 2Graduate Student, Department of Vertebrate Paleontology, American Museum of Natural History; Graduate Student, Department of Geological Sciences, Columbia University. Copyright ©) American Museum of Natural History 1987 ISSN 0003-0082 / Price $ 1.00 2 AMERICAN MUSEUM NOVITATES NO. 2870 cited specializations, though supposedly ofthe ambiguities discussed below. The char- lacking in desmostylians, are found in hy- acters below are of note. racoids as well as proboscideans and sirenians. Such features have been used as evidence for the monophyly ofthe Paenungulata, MASTOID ExPosuRE a superordinal group comprising tethytheres Typically (and primitively) within mam- and hyracoids (Gregory, 1910; Shoshani et mals the mastoid portion of the petrosal is al., 1978; Novacek, 1982, 1986). Hence, the well exposed on the exterior surface of the lack ofcertain apomorphies in desmostylians skull, as it forms a broad flange in the occiput not only presents a difficulty for recognition lateral to the exoccipital and medial and pos- of the Tethytheria but also for the more in- terior to the squamosal (Novacek and Wyss, clusive Paenungulata, if Desmostylia are to 1986). However, the mastoid is not exposed be retained within these groups. in this fashion in the occiput ofcertain mam- Here we offer arguments why published de- mals, including pholidotans, cetaceans, der- scriptions of desmostylians with respect to mopterans, advanced artiodactyls, probos- the characters in question are either incorrect, cideans, sirenians, and hyracoids. This, the or are subject to alternative phylogenetic in- "amastoid condition" is clearly derived for terpretations. Our emendations allow assign- mammals (Novacek and Wyss, 1986). ment of the Desmostylia to the Tethytheria. Sirenians show exposure ofthe mastoid via They are also consistent with a paenungulate a large fenestra in the dorsal occiput, but this grouping for tethytheres and hyracoids. seems a uniquely specialized condition. Al- though the mastoid is inflated and essentially ACKNOWLEDGMENTS fills this large fenestra, it does not extend around the base of the cranium to form a We thank Malcolm C. McKenna, John flange on the ventral occiput. Thus, sirenians Flynn, Daryl Domning, and Clayton Ray for do not show the continuous mastoid expo- discussion. Daryl Domning provided a par- sure between the horizontal basicranium and ticularly useful review ofthe manuscript. We ventral (vertical) occiput that is characteristic also thank Clayton Ray for access to the im- of most mammals. The opening of the oc- portant collection of desmostylian material cipital fenestra for dorsal exposure of the in- at the United States National Museum flated mastoid is therefore likely a secondary (USNM), Smithsonian Institution. Figures feature, and one that could well derive from were prepared by Lisa Lomauro (figs. 1B, 3) the amastoid condition exemplified by hy- and Chester Tarka (fig. 2). This work was racoids and proboscideans. supported by the Frick Laboratory Endow- The broad occipital exposure of the des- ment Fund in Vertebrate Paleontology. mostylian mastoid was explicitly described by Hay (1915, p. 387), figured but not described by Abel (1922, fig. 3), and figured CHARACTER ANALYSIS and discussed by VanderHoof(1937, figs. 1, Desmostylians include the nominal genera 12). These descriptions are based on a skull Desmostylus, Paleoparadoxia, Cornwallius, ofDesmostylus hesperus (USNM 8191). Van- VanderhooJius, Kronokotherium, and Be- derHoof's (1937) figure 12 of this specimen hemotops (Hay, 1915; VanderHoof, 1937; shows a mastoid element only on the left side Pronina, 1957; Reinhart, 1959; Ijiri and Ka- ofthe skull. Although he did not label sutures, mei, 1961; Shikama, 1966; Domning et al., it appears from his figure 12 that the mastoid 1986; Kronokotherium and Vanderhoofius on the left side of the skull is isolated on its may be junior synonyms of Desmostylus, D. lateral border by a short, curved suture and Domning, personal commun.). Ofthese, only on its medial border by a long, vertically ori- Desmostylus and Paleoparadoxia are well ented suture that runs to the lambdoidal rim represented by nearly complete skulls and (fig. 1A herein). However, the "lateral su- skeletons, and even in such cases damage ture" is more likely a contact between the during preservation has contributed to some occipital exposure of the squamosal and the 1987 NOVACEK AND WYSS: DESMOSTYLIAN SKELETON 3 exoccipital (fig. IB). Contrary to Vander- Hoof's figure, this suture runs all the way to A the occipital foramen. The suture is also preserved on the right side of the skull where it appears to extend beyond the occipital foramen and join the oblique supraoccipital- exoccipital suture either at the lambdoidal crest or at a somewhat more medial junction (fig. 1B). The "medial suture," which VanderHoof (1937) and others seem to have taken as the mastoid-exoccipital suture, is an artifact, a crack in the bone. It is clearly not present on the right side of USNM 8191. It crosses, rather thanjoins, the "lateral suture" (cf. fig. 1A, 1B). It trends lateral to the occipital foramen; this foramen typically lies within the mastoid or exoccipital or in a su- B ture separating these elements from the su- Focc praoccipital. The "medial suture" also marks the lateral edge ofa damaged area where surface bone has been removed. We conclude, then, that earlier descriptions of this important specimen are in error and there is clearly no exposure of the mastoid on the occiput in Desmostylus hesperus. Moreover, published figures and descriptions of Desmostylus hesperus japonicus (a senior synonym of Desmostylus mirabilis Nagao, Ijiri and Kamei, see Shikama, 1966) and Pa- leoparadoxia tabatai (Ijiri and Kamei, 1961, pl. 1, fig. 4 and pl. 3, fig. 5) do not indicate a mastoid in the occiput. Instead, these fig- Exocc Cocc ures correspond closely with the geometry of elements reconstructed in figure 1 B. Finally, immature specimens (USNM 181744) of 4 cm Desmostylus hesperus (fig. 2) and Cornwallius Fig. 1. Posterior views of occipital region in sp. (USNM 181788) show neither mastoid USNM 8191, Desmostylus hesperus. (A) Recon- exposure nor VanderHoof's (1937) "medial struction from VanderHoof(I 937, fig. 12). (B) Re- suture" in the occiput. construction favored in this paper. Symbols are The broad invasion of the squamosal on Exocc, exoccipital; Focc, occipital foramen; Ma, the occipital surface in Desmostylus (figs. 1 B, mastoid process (of the petromastoid); Cocc, oc- 2) is a curious feature. This arrangement, cipital condyle; Socc, supraoccipital; Sq, squa- however, bears striking resemblance to that mosal. Dashed line indicates crack in the bone in Sirenia where the occipital exposure ofthe surface. Hatching indicates damaged area. squamosal is even more marked. It is clear, then, that the ventral mastoid is concealed in these taxa by the expansive paroccipital ly derived feature. It probably represents the apophysis formed by broad contact between marked expansion of the occipital foramen, the squamosal and exoccipital. Only in si- which in Desmostylus is also atypically large renians does a large fenestra expose, more (figs. 1A, 1B, 2). The amastoid condition not- dorsally, the inflated mastoid process. As ed for tethytheres and hyracoids is thus at- noted above, we regard this fenestra as a high- tained in desmostylians. 4 AMERICAN MUSEUM NOVITATES NO. 2870

Fig. 2. Posterior view of the occipital region in USNM 181744, immature skull of Desmostylus cf. hesperus. Abbreviations defined under fig. 1. Note that sutures were not painted, but merely cleaned. 1987 NOVACEK AND WYSS: DESMOSTYLIAN SKELETON 5

CARPAL ARRANGEMENT We believe that the putative alternating pattern of the carpals in desmostylians is A general condition in many amniote either questionable or a highly specialized ar- groups is the alternating arrangement be- rangement that is clearly not equivalent to tween the proximal and more distal carpal the primitive condition for mammals. The elements. This arrangement allows an oblique critical material to consider here is the post- contact between several elements, including cranial skeleton ofthe best represented taxon, the lunar and unciform. The alternating con- Paleoparadoxia tabatai. Despite Shikama's dition is characteristic of therapsids, mono- (1966, fig. 108; 1968, pl.4; and fig. 3Aherein) tremes, marsupials, edentates, rodents, in- astute analysis ofthis material, he had to con- sectivores, creodonts, carnivores, pantodonts, tend with certain ambiguities, due to distor- tillodonts, primates, various condylarths, no- tion of elements in the preserved specimens. toungulates, tubulidentates (fide Shikama, The bones of the carpus in the Izumi speci- 1966, but contra Gregory, 1910), perissodac- men of Paleoparadoxia are not closely im- tyls, and artiodactyls (see survey in Gregory, bricated and there are alternative interpre- 1910). tations oftheir natural arrangement (Shikama, A more derived condition is represented 1966, pp. 34-38). Shikama (1966, fig. 15) by a serial arrangement ofthe carpals, where- claimed that the lunar (bone 34 in his fig. 15) in the proximal and distal elements do not in the Izumi specimen makes a distolateral strongly overlap, and oblique contact be- contact with the unciform (bone 36). This is tween these elements, including the lunar and an odd contact because the overlap of these uniciform, is lost. Although the serial ar- elements in the Izumi specimen lies along an rangement is found in a few extinct lineages oblique, proximal-distal axis rather than the (e.g., phenacodontid condylarths, some litop- mediolateral axis characteristic of the alter- terns) the distribution ofthis specialized con- nating arrangement (fig. 3B). It is noteworthy dition at a higher level is notably restricted. that, except for the lunar, the proximal car- The only mammalian orders that can be char- pals do not strongly overlap in contact with acterized by this feature are the Hyracoidea, the distal carpals (Shikama, 1966, pp. 135- Proboscidea, Sirenia (contra Gregory, 1910), 138; and fig. 3A herein). Cetacea, and the extinct Embrithopoda. (It is The cuneiform and unciform are basically likely that embrithopods have a special re- aligned, although the lunar intrudes to con- lationship to proboscideans, sirenians, and tact the medial facet of the latter element. hyracoids-see McKenna, 1975, p. 42.) Of The lunar is enlarged in a peculiar fashion to these orders, the cetacean condition, at least, the extent that it nearly contacts metacarpal clearly seems an independent derivation. The III (fig. 3A). Also the trapezoid lies distal to putative relatives of the cetaceans, the mes- the scaphoid, but it does not make close con- onychids (Van Valen, 1966; McKenna, 1975) tact with more central proximal elements (the have alternating carpals and retain the lunar- lunar or the centrale), as it does in figure 3B. unciform contact. This is, then, the most In fact, the relationship of the scaphoid-lu- plausible condition for the ancestor ofwhales nar-trapezoid-magnum quartet in Paleopar- and their nearest sister taxon. adoxia is more reminiscent of that in sireni- Because of its limited distribution, the se- ans (fig. 3C) than in the primitive mammals rial carpal arrangement of sirenians, probos- (fig. 3B). cideans, and hyracoids represents a potential Shikama (1966, p. 139) stressed that the synapomorphy for Paenungulata (Shoshani desmostylian carpal arrangement differs from et al., 1978; Novacek, 1982). However, des- the usual conditions where a lunar-unciform mostylians are claimed to retain the more contact is present. For example, in Perisso- conservative alternating pattern and a strong, dactyla and Artiodactyla, there is an intimate overlapping contact between the lunar and contact between the scaphoid and magnum unciform (Shikama, 1966, p. 139, fig. 108). not seen in desmostylians. Moreover, Shi- This contradiction poses a major difficulty kama (ibid.) noted that the proximodistally for allocation of desmostylians to the paen- compressed scaphoid, cuneiform, and "platy" ungulates. unciform in desmostylians suggested a pos- 6 AMERICAN MUSEUM NOVITATES NO. 2870

A B C CE M

I V III V 11 IV I IV III III I

Fig. 3. Left manus of (A) Paleoparadoxia tabatai (after Shikama, 1966, figs. 15, 17, 18, 108; 1968, pl. 4); (B) Pantolambda bathmodon (American Museum ofNatural History [AMNH 2546] after Matthew, 1937, fig. 42, p. 180) and (C) Trichechus sp. (American Museum of Natural History, Comparative Anatomy Collections [AMNH-CA 30]). Symbols are C, cuneiform; CE, centrale; L, lunar; M, magnum; S, scaphoid; T, trapezoid; Tz, trapezium; U, unciform. Not to scale. (B) is taken to represent a generalized condition for eutherians, although the centrale has been lost in many lineages. Note that in (B) the lunar- unciform contact lies along the mediolateral axis separating the proximal and distal carpal elements. Overlap in (A) is largely effected by the distal enlargement of the lunar. sible relationship with Proboscidea and Em- characterize a grouping at a higher level than brithopoda. Our examination of a cast ofthe those for which it has often been applied. Stanford skeleton ofPaleoparadoxia corrob- Tassy (1981, fig. 12) argued that the reduction orates the above assessment of the material of the mastoid apophysis was a synapomor- described by Shikama. (This specimen, Pa- phy uniting proboscideans with sirenians and leoparadoxia, University of California Mu- excluding desmostylians. He retained des- seum of Paleontology (UCMP) 81302, was mostylians within the tethytheres, citing the discovered in 1965, but has never been for- following characters as synapomorphies for mally described.) the group; (1) forward placement ofthe orbit, It is difficult to assess the phylogenetic im- (2) position of the infraorbital foramen di- plications of these comparisons. The ques- rectly below the orbit (obviously correlated tion of transformation of the mammali- with character 1), (3) the strong zygomatic an carpus is in need of modern treatment. process formed by the laterally directed squa- Nevertheless, the desmostylian condition mosal, and (4) the bilophodont-bunolopho- seems to us an anomalous one in several re- dont molars. Tassy (1981) noted a problem spects, and the condition might be considered with the remote position of desmostylians, a potential autapomorphy for the group. As- because he thought this fossil group showed suming that the lunar-unciform contact in some special similarity to sirenians (see Tas- Paleoparadoxia is properly reconstructed, this sy, 1981, fig. 12, characters 14, 25, and dis- articulation is effected by a distally expanded cussion therein) that did not apply to pro- lunar element that nearly contacts metacar- boscideans. He was left with two mutually pal III. It is conceivable that this unique sys- contradictory groupings one favoring a tem derived from a more basic serial carpal closest relationship between sirenians and arrangement. At the very least, the desmo- desmostylians on the basis ofthe above cited stylian condition is not comparable to the characters, and one favoring a closest rela- typical alternating pattern in mammals (cf. tionship between proboscideans and sireni- fig. 3A, 3B). ans on the basis of the amastoid condition. as we is also common INTERPRETATIONS Since, argue, amastoidy PHYLOGENETIC to desmostylians, this contradiction is elim- The specialized amastoid condition noted inated. Clearly, amastoidy is not useful for above is shared by hyracoids, desmostylians, defining tethytheres (including desmostyli- proboscideans, and sirenians. It thus helps to ans) because the character is common to hy- 1987 NOVACEK AND WYSS: DESMOSTYLIAN SKELETON 7

racoids. The trait does seem a potentially use- Hay, 0. P. ful synapomorphy for Paenungulata if it can 1915. A contribution to the knowledge of the be argued that amastoidy in a few other mam- extinct sirenian Desmostylus hesperus malian lineages (pholidotans, dermopterans, Marsh. Proc. U.S. NatL. Mus., 49:381- advanced artiodactyls) was independently 397. derived (Novacek and Wyss, 1986). Ijiri, S., and T. Kamei 1961. On the skulls of Desmostylus mirabilis There is little doubt that the serial carpal Nagao from south Sakhalin and of Pa- arrangement also describes a more inclusive leoparadoxia tabatai (Tokunaga) from group than Tethytheria. The arrangement in Gihu Prefecture, Japan. Earth Sci. 53: hyracoids is closely similar to that in teth- 1-27. [In Japanese; English translation ytheres (Shikama, 1966) and the condition available to us.] seems a potentially useful synapomorphy for Matthew, W. D. Paenungulata (Shoshani et al., 1978; Nova- 1937. Paleocene faunas ofthe San Juan Basin, cek, 1982, 1986; Novacek and Wyss, 1986). New Mexico. Trans. Am. Phil. Soc., 30: Desmostylians still pose a problem for this 1-372. generalization. Here we maintain that the re- McKenna, M. C. construction of the alternating carpal ar- 1975. Toward a phylogenetic classification of rangement in desmostylians is either (1) open the Mammalia. In W. P. Luckett and F. S. Szalay (eds.), Phylogeny of the Pri- to question or (2) the result of a uniquely mates. New York: Plenum, pp. 21-46. specialized pattern involving the hypertro- Novacek, M. J. phy of the lunar element. Thus the carpal 1982. Information for molecular studies from evidence does not unambiguously preclude anatomical and fossil evidence on higher the allocation of desmostylians to Paenun- eutherian phylogeny. In M. Goodman gulata. (ed.), Macromolecular sequences in sys- In closing we note that this paper addresses tematic and evolutionary biology. New directly only the validity ofcertain characters York: Plenum, pp. 3-41. attributed to desmostylians and the impli- 1986. The skull of leptictid insectivorans and cation that such characters would exclude the higher-level classification of euthe- desmostylians from either Tethytheria or rian mammals. Bull. Am. Mus. Nat. We Hist., 183:1-111. Paenungulata. have elsewhere (Novacek Novacek, M. J., and A. R. Wyss and Wyss, 1986) favored the inclusion ofdes- 1986. Higher level relationships ofthe Recent mostylians within Tethytheria on the basis eutherian orders: morphological evi- ofadditional characters cited by Tassy (1981) dence. Cladistics, 2:257-287. and others. Left open is the question of des- Pronina, I. G. mostylian relationships within Tethytheria. 1957. Novyi predstavitel desmostylid Kron- We do not argue here whether desmostylians okotherium brevimaxillare gen. nov. iz are the closest relatives of sirenians or pro- Miostenovykh ottozhenii na Kamchat- boscideans or are the most remote ofthe teth- ka. Dokl. Akad. Nauk. USSR, 117:31 0- ythere clades. 312. Reinhart, R. H. 1959. A review of the Sirenia and Desmo- LITERATURE CITED stylia. Univ. Calif. Publ. Geol. Sci., Abel, 0. 36(1): 1-146. 1922. Desmostylus: ein mariner Multituber- Shikama, T. culate aus dem Mioziin der nordpazi- 1966. Postcranial skeletons of Japanese Des- fischen Kiistenregion. Acta Zool., 3: mostylia. Palaeontol. Soc. Japan, Spec. 361-394. Pap., 12:1-202. Domning, D. P., C. R. Ray, and M. C. McKenna 1968. Additional notes on the postcranial 1986. Two new Oligocene desmostylians and skeletons ofJapanese Desmostylia. Sci. a discussion of Tethytherian System- Rep., YokohamaNatl. Univ., 14:21-26. atics. Smithsonian Contr. Paleobiol., 59: Shoshani, J., M. Goodman, and W. Prychodko 1-56. 1978. Cladistic analysis of the Paenungulata Gregory, W. K. by computer. Am. Zool., 18:601. 1910. The orders ofmammals. Bull. Am. Mus. Simpson, G. G. Nat. Hist., 27:1-524. 1945. The principles of classification and a 8 AMERICAN MUSEUM NOVITATES NO. 2870

classification of mammals. Bull. Am. VanderHoof, V. L. Mus. Nat. Hist., 85:1-350. 1937. A study of the Miocene sirenian Des- Tassy, P. mostylus. Bull. Dept. Geol. Sci., Univ. 1981. Le crane de Moeritherium (Proboscid- Calif. Publ., 24:169-262. ea, Mammalia) de l'Eocene de Dor el Van Valen, L. Talha (Libye) et le probleme de la clas- 1966. Deltatheridia, a new order ofmammals. sification phylogenetique du genre dans Bull. Am. Mus. Nat. Hist., 132:1-126. les Tethytheria McKenna, 1975. Bull. Mus. Natl. d'Hist. Nat., Ser. 4, sec. C, 3:87-147.

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