Two New Oligocene Desmostylians and a Discussion of Tethytherian Systematics DARYL P. DOMNING, CLAYTON E. RAY, and MALCOLM C. McKENNA SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY . SERIES PUBLICATIONS OF THE SMITHSONIAN INSTITUTION Emphasis upon publication as a means of "diffusing knowledge" was expressed by the first Secretary of the Smithsonian. In his formal plan for the Institution, Joseph Henry outlined a program that included the following statement: "It is proposed to publish a series of reports, giving an account of the new discoveries in science, and of the changes made from year to year in all branches of knowledge.'" This theme of basic research has been adhered to through the years by thousands of titles issued in sehes publications under the Smithsonian imprint, commencing with Smithsonian Contributions to Knowledge in 1848 and continuing with the following active series: Smithsonian Contributions to Anthropology Smithsonian Contributions to Astrophysics Smithsonian Contributions to Botany Smithsonian Contributions to the Earth Sciences Smithsonian Contributions to the (Marine Sciences Smithsonian Contributions to Paleobiology Smithsonian Contributions to Zoology Smithsonian Folklife Studies Smithsonian Studies in Air and Space Smithsonian Studies in History and Technology In these series, the Institution publishes small papers and full-scale monographs that report the research and collections of its various museums and bureaux or of professional colleagues in the world of science and scholarship. The publications are disthbuted by mailing lists to libraries, universities, and similar institutions throughout the world. Papers or monographs submitted for series publication are received by the Smithsonian Institution Press, subject to its own review for format and style, only through departments of the various Smithsonian museums or bureaux, where the manuscnpts are given substantive review. Press requirements for manuscript and art preparation are outlined on the inside back cover. Rotjert McC Adams Secretary Smithsonian Institution SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY • NUMBER 59 Two New Oligocene Desmostylians and a Discussion of Tethytherian Systematics Daryl P. Domning, Clayton E. Ray, and Malcolm C. McKenna SMITHSONIAN INSTITUTION PRESS City of Washington 1986 ABSTRACT Domning, Daryl P., Clayton E. Ray, and Malcolm C. McKenna. Two New Oligocene Desmostylians and a Discussion of Tethytherian Systematics. Smith- sonian Contributions to Paleobiology, number 59, 56 pages, 23 figures, 1986.— A new genus, comprising two new species of desmostylians, is de- scribed from marine Oligocene deposits of the Pacific Northwest. Behemotops proteus, new genus, new species, is based on an immature mandibular ramus and apparently associated skeletal fragments from the middle or (more likely) upper Oligocene lower part of the Pysht Formation of Clallam County, Washington. A related new species, Behemotops emlongi, is founded on a mandibular ramus of an old individual and a mandibular fragment with canine tusk from the uppermost Oligocene (early Arikareean equivalent) Yaquina Formation of Lincoln County, Oregon. The two new species are the most primitive known desmostylians and compare favorably with the primitive Eocene proboscideans Anthracobune and Moeritherium, and to the still more primitive tethythere Minchenella from the Paleocene of China. For many years the Desmostylia were widely regarded as members of the mammalian order Sirenia before being accepted as a taxon coordinate with the Sirenia and Proboscidea (Reinhart, 1953). On the basis of cladistic analysis we go a step further and regard the Desmostylia as more closely related to Proboscidea than to Sirenia because the Desmostylia and Proboscidea are interpreted herein to share a more recent common ancestor than either order does with the Sirenia. This analysis also suggests that the common ancestor of the Proboscidea and Desmostylia (but not the Sirenia) had suppressed P5 and the original last molar. These characters may be conver- gent with some other mammals. The Superorder Tokotheria McKenna, 1975, was originally thought to be characterized by loss of both P5 and M3. However, because early sirenians do not show these losses, they may have occurred independently in the common ancestor of proboscideans and des- mostylians and in various other tokotheres. The late Paleocene genus Minchenella Zhang, 1980, from China, is a suitable candidate to be the common ancestor of both the Desmostylia and the Proboscidea. It possesses a small entoconid II on M:<. The Eocene genus Lammidhania Gingerich, 1977, from Pakistan, and the late Paleocene and/ or early Eocene Chinese and Mongolian phenacolophids had not acquired an entoconid II on M^ but are otherwise similar to Minchenella and the anthra- cobunids. The Asiatic occurrence of phenacolophids, Lammidhania, Min- chenella, and anthracobunids suggests an Asian origin for the Proboscidea and is in accord with the exclusively Pacific distribution of the Desmostylia. We believe that desmostylians were amphibious herbivores that fed on marine algae and angiosperms, and that at least the earlier taxa depended to a large extent on plants exposed in the intertidal zone. OFFICIAL PUBLICATION DATE i.s handstamped in a limited number of initial copies and is recorded in the Institution's annual report, Smithsonian Year. SERIES COVER DESIGN: The trilobite Phacops rana Green. Library of Congress Cataloging in Publication Data Domning, Daryl P. Two new Oligocene desmostylians and a discussion of tethytherian systematics. (Smithsonian contributions to paleobiology ; no. 59) Bibliography: p. Supt. of Docs, no.: SI 1.30:59 1. Behemotops proteus. 2. Beheniotopsemlongi. 3. Desmostylia. 4. Proboscidea, Fossil. 5. Paleontology—Oligocene. 6. Paleontology—Washington(State)—Clallam County. 7. Pa- leontology—Oregon—Lincoln County. 1. Ray, Clayton Edward. II. McKenna Malcolm C. 111. Title. IV. Series. QE701.S56no. 59 [QE882.D45J 560 s [569'.6] 85-600322 Contents Page Introduction 1 Acknowledgments 4 Abbreviations 5 Class MAMMALIA Linnaeus, 1758 5 Order DESMOSTYLIA Reinhart, 1953 5 Family Uncertain 5 Behemotops, new genus 6 Behemotops proteus, new species 6 Behemotops emlongi, new species 23 Relationship between Behemotops proteus and B. emlongi 31 History of Desmostylian Systematics 33 Characters Used in Phylogenetic Analysis 37 Comparisons with Early Proboscidea and Minchenella 38 Moeritherium 38 Anthracobune 43 Minchenella 44 Implications for Eutherian Dental Homologies 45 Status of the Tethytheria 46 Desmostylian Lifestyle 47 Conclusions 48 Literature Cited 49 ill Two New Oligocene Desmostylians and a Discussion of Tethytherian Systematics Daryl P. Domning, Clayton E. Ray, and Malcolm C. McKenna Introduction Douglas Emlong's Promethean prowess in dis- covery of unprecedented vertebrate fossils, alike in beds where many, few, or no collectors pre- ceded him, is well known to specialists having personal knowledge of his activities (Ray, 1977). Only a handful of his specimens have thus far been described (Coombs, 1979; Emlong, 1966; Munthe and Coombs, 1979; S.L. Olson, 1980, 1981; Olson and Hasegawa, 1979), but many are under study. Tragically, the flow from the wellspring of these riches ended abruptly on 8 June 1980 with Emlong's death (Ray, 1980), but his already towering reputation as a fossil finder will be progressively and justifiably widened with every added publication of the results of studies in progress on the "Emlong Collection." The purpose ofthe present communication is to make known several of his more remarkable and pro- vocative discoveries: putative desmostylians, much more primitive than any previously known and forging hitherto "missing links" (E.C. Olson, 1981) with primitive proboscideans. Brief men- Daryl P. Domning, Department of Anatomy, College of Medicine, Howard University, Washington, D.C. 20059. Clayton E. Ray, Department of Paleobiology, National Museum of Natural His- tory, Smithsonian Institution, Washington, D.C. 20560. Malcolm C. McKenna, Department of Vertebrate Paleontology, The Amer- ican Museum of Natural History, New York, N.Y. 10024. tion of these fossils was made by Barnes et al. (1985). The three specimens to be described herein are from marine Oligocene deposits ofthe Pacific Northwest (Figures 1-3). The first to be found, from the Yaquina Formation of coastal Oregon, consists of a massive tusk with a bit of poorly preserved bone at the anterior end of the man- dible (USNM 186889; Figures 16 and 18). At the time of its discovery in 1969 it was regarded by Emlong (field list and pers. comm. to Ray) as possibly representing a land mammal but more likely a "new and very aberrant desmostylian." The second specimen to be found, discovered in 1976 in the Pysht Formation on the Olympic Peninsula in Washington, is an immature half mandible with apparently associated postcranial fragments (USNM 244035; Figures 4-11, 12A- D, 14A,c, 15A,B,E,F). It was thought by Emlong to be a desmostylian or possibly a land mammal, although he also believed that the one molar exposed in the field resembled those of sirenians (field list; pers. comm. to Ray, 1976). The third specimen found by Emlong, a half mandible of an old animal with only Ms preserved (USNM 244033; Figures 12F, 16, and 17), collected in 1977 from the Yaquina Formation of Oregon, was described in Emlong's field list as a desmos- tylian or a land mammal and as elephant-like. In 1976 1 SMITHSONIAN CONTRIBUTIONS TO PALEOBIO OGY -PYSHT FM.,TYPE SECTION- -REFERENCE SECTION,UPPER MEM6ER,TWIN RIVER FM. STRAIT OF JUAN DE FUCA REFERRED,LOWER PARTOF TWIN RIVERS FM GETTYSBURG —JUANIAN STAGE. TYPE SECTION LIRACASSIS ><^7VI ZONE.TYPE SECTION UPPER PART OF | TWIN RIVERS FM. LIRACASSIS /feXZONE, "^CANADA UNITED STATES' MILES 0 10 20 30 40 50 a letter to Ray of 27 March 1977, two days after the discovery, Emlong commented as follows: I stopped at Seal Rock . . . and found the most interesting thing of all—a giant desmostylian-like mandible, nearly complete, with teeth [only one as it turned out] that seem FIGURE 1.—Index maps of Pacific Northwest: A, sketch map of part of western Canada, Washington, and Oregon, show- ing some major place names, localities mentioned in text, and location of enlarged area shown in B; B, detail of part of Twin Rivers and Disque 7.5-minute quadrangles, USGS, along north shore of Olympic Peninsula, showing type- locality of Behemotops proteus, and other relevant localities and boundaries discussed in text; data primarily from Dur- ham (1944:113, fig. 6), Brown and Cower (1958:2502, fig. 5), Rau (1964:027, pi. 1), AddicoU (1976a:98, fig. 3), Snavely et al. (1978:A118, fig. 8), and Moore (1984a:719). Spelling "Twin Rivers Formation" is that of Durham only. to resemble those of the specimen from the Twin River [USNM 244035]. This Oligocene specimen is far larger and heavier and I am sure it is a great find, whether desmostylian or land mammal. It came from the Corn- wallius horizon, but is not Cornwallius. It may be related to that giant tusk [USNM 186889] from the Yaquina Formation, and is not far from that area. I am afraid to expose much of the specimen, so I am largely guessing. Emlong's instant intuitions of affinities, al- though based on unprepared specimens, virtually no literature or comparative material, and almost no formal training, proved in this case as in many others to be uncannily perceptive and to fore- shadow our own more belabored conclusions. However, it should be mentioned that his and our views have not been universally accepted by colleagues who have examined these specimens between 1969 and now. NUMBER 59 FIGURE 2.—Index maps of coastal Oregon: A, sketch map of part of northwestern Oregon, showing some major place names and location of enlarged area shown in B; B, detail of part of Yaquina and Waldport 15-minute quadrangles, USGS, along central west coast of Oregon, showing location of Seal Rock State Wayside and other localities discussed in text, including location of enlarged area shown in c; C, detail of area including Seal Rock State Wayside, showing localities for Behemotops emlongi and Arretotherium. SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY FIGURE 3.—Correlation of beds containing Behemotops proteus and Behemotops emlongi with some relevant systems of chronology; after Armentrout et al. (1983, chart). Wavy lines at top and bottom indicate continuation of unit beyond limits of chart. ACKNOWLEDGMENTS.—Of course we are in- debted above all to the late Douglas Emlong, whose perseverance in the field gave us the fas- cinating fossils considered herein. The specimens were skillfully prepared by Ar- nold D. Lewis, and the photographs made by Victor E. Krantz. Figures 1, 2, 5-11, and 14-19 were made by Lawrence B. Isham, including the exquisite drawings for Figures 5-7. Figures 22 and 23 were made by Lisa Lomauro. The re- maining figures were made by Mary Parrish. For the loan of comparative specimens or pro- vision of casts from their respective institutions, we thank Lawrence G. Barnes, the Natural His- tory Museum of Los Angeles County; Philip D. Gingerich, the Museum of Paleontology, Univer- sity of Michigan; J. Howard Hutchison, the Mu- seum of Paleontology, University of California; Charles R. Schaff, the Museum of Comparative Zoology, Harvard University; Richard H. Ted- ford, the American Museum of Natural History; Mary Ann Turner, Peabody Museum of Natural History, Yale University; and Robert M. West, then of the Milwaukee Public Museum. NUMBER 59 Warren O. Addicott and Parke D. Snavely, Jr., both of the United States Geological Survey, Menlo Park, California, have shared generously their special knowledge of the source beds of the fossils and of biostratigraphy and correlation in the Pacific Northwest in general. Kristin Mc- Dougall, also of the USGS at Menlo Park, pro- vided foraminiferal analysis of rock samples, as did William A. Berggren of the Woods Hole Oceanographic Institution. For the privilege of examining Japanese desmostylian material, in- cluding many unpublished specimens, as well as for valuable discussions and superb hospitality, Domning is particularly grateful to Yoshikazu Hasegawa, Yokohama National University; No- rihisa Inuzuka, University of Tokyo; Masaichi Kimura, Hokkaido University of Education; Os- amu Sakamoto, Saitama Prefectural Museum of Natural History; and Yukimitsu Tomida, Na- tional Science Museum, Tokyo. James M. Clark generously gave access to his manuscript on a new species of Paleoparadoxia from Point Arena, California. Jeheskel Shoshani generously shared with us his data and ideas on paenungulate char- acters and relationships. We have also profited greatly through discussion of sirenians, probos- cideans, and desmostylians in general and of our particular specimens with Lawrence G. Barnes, Kishor Kumar, Earl Manning, Adele Panofsky, Roy H. Reinhart, Charles A. Repenning, R.J.G. Savage, and Andrew Wyss. Further, Barnes, Re- penning, and Savage have read the manuscript critically and have supplied us with important unpublished information. Although the end product has benefitted greatly from the assist- ance of all ofthe above-mentioned colleagues, its remaining deficiencies are of course our respon- sibility. Financial support, especially for field work, was provided in part by the Smithsonian Institu- tion through the Smithsonian Research Foun- dation and the Walcott and Kellogg funds. Domning's visit to Japan was generously financed by the Yamagata Prefectural Museum. The sequence of authorship was determined by lot. ABBREVIATIONS.—The following abbrevia- tions are used to identify the institutions listed: AMNH Department of Vertebrate Paleontology, American Museum of Natural History, New York, N.Y. AMNH CA Department of Mammalogy, Comparative Anatomy Collection, American Museum of Natural History, New York, N.Y. LACM Natural History Museum of Los Angeles County, Los Angeles, Cal. MCZ Museum of Comparative Zoology, Harvard University, Cambridge, Mass. NMNH National Museum of Natural History, Smith- sonian Institution, Washington, D.C. NSM National Science Museum, Tokyo, Japan UCMP Museum of Paleontology, University of Cali- fornia, Berkeley, Cal. USGS United States Geological Survey USNM former United States National Museum, col- lections in the National Museum of Natural History, Smithsonian Institution, Washing- ton, D. C. YPM Peabody Museum of Natural History, Yale University, New Haven, Conn. Class MAMMALIA Linnaeus, 1758 Order DESMOSTYLIA Reinhart, 1953 Family Uncertain DISCUSSION.—The family-level taxonomy of desmostylians is at present unsatisfactory. The original Family Desmostylidae Osborn, 1905, was supplemented by the Cornwalliidae Shikama, 1957 (emended from the original spelling "Corn- walliusidae" by Shikama, 1966:153), which was created to accommodate Cornwallius Hay, 1923. When Reinhart (1959:94) generically separated "Cornwallius" tabatai from the type-species C. sookensis and assigned the former to a new genus, Paleoparadoxia, he also erected a third family, Paleoparadoxidae (emended from "Family Pa- leoparadoxia" Reinhart, 1953); however, he re- tained Cornwallius, sensu stricto, in the Desmos- tylidae. Shikama (1966) instead placed both Cornwallius and Paleoparadoxia in his Cornwalli- idae. Apart from these inconsistencies already in the literature, the few and dissimilar genera mak- ing up the order could arguably be arranged in almost any arbitrary number of groups, from a single, all-embracing Desmostylidae to mono- typic families for each genus. However, until desmostylian diversity and phylogeny are better understood, we prefer to reserve judgment on the familial assignment of the taxa described herein, and we suggest a temporary moratorium on new family-level arrangements within the or- der. Behemotops, new genus TYPE-SPECIES.—Behemotops proteus, new spe- cies. INCLUDED SPECIES.—Behemotops proteus and B. emlongi, new species. DIAGNOSIS.—Desmostylian differing from other members of the order in having seven lower postcanine cheek-teeth, without marked diastemata; Pi (or DP]) large, caniniform, pro- cumbent, single-rooted; P9 large, procumbent, with root partially or completely divided; molars brachydont, bunodont, with four principal cusps neither cylindrical nor appressed, and forming a square; metaconids of lower molars not twinned; all permanent teeth in use together at maturity; and lingual surface of mandible lacking swelling at rear of tooth row. ETYMOLOGY.—From the biblical (Job 40:15- 24) Hebraic (and Greek, Latin, and English by adoption) b'hemoth, plural, "great beast," thought by many etymologists and zoologists (including Linnaeus, 1758:74, whose latinized Behemot is used herein) to refer to the Nilotic hippopotamus (and, by others, including Maglio, 1973:2, to the elephant, mammoth, etc.); or b'hemah, singular, "beast," conjectured by some to be derived from a (pos.sibly artificial) Coptic term p-ehe-mau "water-ox"; plus -ops, Greek, suffix, masculine, like or similar aspect; in allusion to the Egyptian source of many specimens of tethytheres and to the hippopotamian habitus and proboscidean af- finities ofthe fossil animal. In any case, we agree with Lydgate (1412-1420, volume 2: page xvii) that the animal's name "doth in latin playne expresse A beast rude full of cursednesse." SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Behemotops proteus, new species FIGURES 4-11, 12A-D, 14A,C, 15A,B,E,F HOLOTYPE.—USNM 244035, right mandibu- lar ramus of immature individual with DP4, Pi (or DPi), and P-^-Ms; and, probably from the same individual, the distal half of the right femur, a proximal fragment of the right tibia, and two phalanges. Field number E76-14, collected II March 1976 by Douglas Emlong. DIAGNOSIS.—Lower canine probably smaller than in B. emlongi (described below); premolars large, high, not molariform; P3 with one domi- nant cusp (protoconid), P4 with two (metaconid and protoconid); metaconid of P4 not twinned; P:, and P4 double-rooted, unlike B. emlongi; DP4 trilobate. ETYMOLOGY.—From the Greek sea-god Pro- teus, son of the sea-goddess Tethys and sea-god Oceanus; old man of the sea and herdsman of the sea-calves (seals) of Poseidon; also able to assume different forms; Latin, masculine, noun in apposition. In allusion to (among other things) its tethytherian affinities and marine habitat, and the pronounced ontogenetic changes in its dental morphology. TYPE-LOCALITY.—South side of Strait of Juan de Fuca, on north shore of Olympic Peninsula, Clallam County, Washington; some 34 km (21 miles) west of Port Angeles and approximately 3.6 km (2.2 miles) east of mouth of East Twin River; 1.6 km (1.0 miles) east and 290 meters (950 feet) north of SW corner. Sec. 19, T. 31 N, R. 9 W, Twin Rivers Quadrangle, 7.5-minute series, USGS; 48°09'38"N, 123°53'55"W; wave-cut bench 15 meters (50 feet) north of cliff face (Figure 1). The postcranial elements came from the same bedding plane as the mandible but were sepa- rated from it horizontally by approximately 6 meters (20 feet). HORIZON.—In place in northwesterly dipping, concretionary, silty, gray mudstone, within the lower part of the type section of the Pysht For- mation (Snavely et al., 1978:A118, Al 19). AGE AND CORRELATION (Figures IB, 3).—The NUMBER 59 5CM J I FIGURE 4.—Behemotops proteus, holotype, USNM 244035, from lower part of Pysht Formation of Washington, diagrammatic representation of right mandibular ramus in labial aspect, showing interpretation of dental loci. rocks from which the holotype of Behemotops proteus was collected are by definition within the type section of the Pysht Formation (lower part), explicitly assigned to the Echinophoria rex Zone (now Liracassis rex Zone). The latter is coeval with the Matlockian and "lower" Zemorrian stages. We regard the lower part of the Pysht Formation as middle or (more likely) late (but not latest) Oligocene in age. In terms of North American Land Mammal Ages, this would imply that B. proteus is Orellan or Whitneyan in age. Although the formational assignment and gen- eral age of the holotype are certain, the nature of deposition, internal conflicts in the biostrati- graphic literature, limited exposures, complexity of the geology (including folding and faulting), and continuously evolving concepts of the rele- vant biostratigraphy and correlation all recom- mend a more extended discussion of the subject than would otherwise be warranted. An intro- duction to the broad regional biostratigraphic framework can be obtained from the recent pub- lications by Addicott (1981), Armentrout et al. (1983), Marincovich (1984), and Moore (1984a, 1984b). For a discussion of the position of the boundary between the Oligocene and the Mio- cene in both marine and continental deposits, see Berggren, Kent, and Flynn (in press). The Twin River Formation (now Twin River Group) was named by Arnold and Hannibal (1913:584, 585) for rocks exposed on the coast from "about three miles east of Twin River west nearly to Pysht Bay," thus including the type- locality of Behemotops proteus. The locality lies SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY FIGURE 5.—Behemotops proteus, holotype, USNM 244035, from lower part of Pysht Formation of Washington, right mandibular ramus, in lingual aspect. DP4 in original position as found, prior to preparation of specimen; P^ and P4 shown in present state of specimen with bone removed dorsal and lateral to teeth; compare Figure 8. Note opening of coronoid canal above crypt of M3. Scale approximately 2 cm. between Durham's (1944:113, figs. 5 and 6) lo- calities A 3683 and A 3684, in a section that he explicitly assigned to the lower part of the Twin River Formation (Twin Rivers formation of Dur- ham) and to his Echinophoria rex Zone. (The Echinophoria rex Zone should now be called the Liracassis rex Zone, and the Echinophoria apta Zone, the Liracassis apta Zone, according to Moore, 1984a:719. In the following discussion we retain the older notation of the references cited.) However, Brown and Gower (1958:2502, fig. 5) redefined the Twin River Formation so that the reference section for their upper mem- ber, between East Twin River and Murdock Creek, includes our locality. Nevertheless, they noted (1958:2510) without disagreement that their upper member included strata assigned by Durham (1944) both to his Echinophoria apta and to his E. rex zones. Our locality lies approximately 366 meters (1200 feet) east of USGS foramini- feral collecting locality f 11802 and 91 meters (300 feet) west of f 11803 (= Durham's 1944, locality A 3684), both of which localities were assigned to the upper member ofthe Twin River Formation and their faunas to the upper part of the Zemorrian Foraminiferal Stage by Rau (1964:G9, G27, table 6, pis. 1 and 4). On the basis of molluscan biostratigraphy of another rock unit, the Lincoln Creek Formation of southwestern Washington, Armentrout (1975:25-29) established the Matlockian mollus- can stage, which he divided into a lower and an upper zone, equivalent, respectively, to the Echinophoria rex and overlying E. apta zones of NUMBER 59 FIGURE 6.—Behemotops proteus, holotype, USNM 244035, from lower part of Pysht Formation of Washington, right mandibular ramus, in labial aspect. DP4 in original position as found, prior to preparation of specimen; bony ramus shown at stage of preparation of Figure 9; P3, P4, and M3 shown as visible in present state of specimen with bone removed to show teeth in labial aspect. Scale approximately 2 cm. FIGURE 7.—Behemotops proteus, holotype, USNM 244035, from lower part of Pysht Formation of Washington, right mandibular ramus, in occlusal aspect. DP4 shown adjacent; Pi (or DPi), P;(, P4, and M3 shown as if fully erupted and in occlusal position; compare Figures 4, 5, and 8- 10, for true position of these teeth. Opening of coronoid canal, adjacent to posterolabial corner of M3, shown only to indicate its presence, as it would have shifted in position through remodeling of bone as M3 erupted. See Figure 5 for true location as preserved. Scale approximately 2 cm. 10 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY FIGURE 8.—Behemotops proteus, holotype, USNM 244035, from lower part of Pysht Formation of Washington, right mandibular ramus, in lingual aspect. Specimen photographed prior to removal of any detectable bone in preparation, and prior to exposure of mandibular foramen and additional preparation on opening of coronoid canal. the Olympic Peninsula. Addicott (1976a) estab- lished molluscan stages for the Neogene of Ore- gon and Washington. His earliest stage, the Ju- anian, shares its type section with the E. apta Zone, and has its base in the sea cliff approxi- mately 2 km east of the mouth of East Twin River (not 3 km as stated erroneously in Addi- cott, 1976a:97, 98; Addicott, pers. comm. to Ray, 21 Sep 1978), between UCMP localities A 3680 and A 3691 (see Durham, 1944, fig. 6). Addicott's (1976b, fig. 2) generalized section of the upper member (= Pysht Formation) of the Twin River Formation erroneously includes the lower half of the upper member of the Twin River Formation in the Echinophoria apta Zone (Addicott, pers. comm. to Ray, 8 Mar 1982). The section from locality A 3691 downward should be referred to the E. rex Zone, including all localities numerically between A 3681 and A 3691, inclusive, and geographically between A 3691 on the west and A 3690 on the east (Dur- ham, 1944, fig. 6). Our locality lies some 1.6 km east of locality A 3691 in westerly dipping strata, thus well below the base of the Juanian Stage and within the reference section of the E. rex Zone. A sample of the enclosing matrix from the holotype of 5. proteus, USNM 244035, was pro- cessed for Foraminifera by Kristin McDougall of the USGS, yielding one planktonic and 27 benthic taxa, indicative of an Oligocene, late Zemorrian age. She stated, "This fauna is quite similar to that found by Rau (1964) in samples f 11801 and f 11802. The faunas suggest cool temperatures in a protected upper bathyal to outer neritic environment" (USGS Report on Referred Fossils, shipment number 0-76-8M, sample Mf 3256, 24 Aug 1976). Snavely et al. NUMBER 59 11 FIGURE 9.—Behemotops proteus, holotype, USNM 244035, from lower part of Pysht Formation of Washington, right mandibular ramus in labial aspect. Specimen photographed prior to removal of any detectable bone in preparation. (1980, fig. 15) have presented a paleogeographic map showing the distribution of late Eocene and Oligocene shelf and deep-water marginal seas in the Pacific Northwest during the time of depo- sition of the Makah, Pysht, Alsea, and other broadly correlative formations of the region. The regional geology of the Olympic Penin- sula, including major, previously unpublished re- sults of mapping by Snavely and associates in the area of concern herein, has been portrayed in a map at the scale of 1:125,000, with a synopsis and bibliography (Tabor and Cady, 1978). The relevant strata are therein mapped as the upper member of the Twin River Formation. Snavely et al. (1978) raised the Twin River to group rank and named formations for its upper members, of which the Pysht Formation is uppermost. The type-locality of the Pysht Formation is the section exposed for some 18 km in the cliffs and on the shore from Pillar Point State Park eastward to 3.5 km west of Low Point. This eastern limit is in the SE corner (not SW as stated erroneously in Snavely et al., 1978:A118), Sec. 19, T. 31 N, R. 9 W, whereas our fossil locality is in the SW corner, in the lower part ofthe Pysht Formation. Armentrout (1977; 1978; 1981:140) re- stricted his Matlockian Stage to the E. rex Zone, leaving the succeeding E. apta Zone for the Ju- anian Stage, thus eliminating the overlap be- tween the two stages. As now restricted, the Matlockian Stage is wholly beneath the Juanian Stage. As originally proposed, the Juanian repre- sented the basal Neogene, lower Miocene mol- luscan stage in the Pacific Northwest, with its base at the Oligocene-Miocene boundary, at 23- 24 million years ago, within the upper part of the Zemorrian Benthic Foraminiferal Stage (Ad- 12 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY FIGURE 10.—Behemotops proteus, holotype, USNM 244035, from lower part of Pysht Formation of Washington, right mandibular ramus in occlusal aspect. Specimen photo- graphed prior to removal of any detectable bone, and prior to additional preparation on opening of coronoid canal. dicott, 1976a:98). Subsequently, however, the upper boundary of the Juanian has been re- garded as almost or exactly coeval with that of the Zemorrian Stage, with the bases of both at 29 or more million years ago. Thus, the Juanian Stage is now usually regarded as late Oligocene (pre-Aquitanian) in age (Addicott, 1977:163, fig. 3; Allison, 1977:876; Armentrout, 1981:140, 142, 145; Armentrout et al., 1983), although Allison (1976; 1978), Allison and Marincovich (1981:4), and Moore (1984b:4) regard the Ju- anian as mostly late Oligocene and partly early Miocene in age. Armentrout's restricted Mat- lockian molluscan stage (= E. rex Zone, = below the uppermost part of the Zemorrian Foramini- feral Stage), representing time from some 29 to 32 million years ago, is thus earlier in age than latest Oligocene (Armentrout, 1981:145), or is in fact early Oligocene, 33-38 million years old (Armentrout et al., 1983, chart; our Figure 3). ASSOCIATED FAUNA.—Douglas Emlong discov- ered numerous specimens of vertebrate fossils in the Pysht Formation. The majority of these are skulls of archaic cetaceans under study by R. Ewan Fordyce and others and as yet alluded to only briefly in the literature (Whitmore and Sanders, 1977:310, 311, fig. 2; Fordyce, 1981:1028, 1033). A single specimen from the same area has provided the basis for a new genus and species of penguin-like pelecaniform bird, Tonsala hildegardae Olson, 1980. This material is thought to be similar in age to B. proteus, but FIGURE 11.—Stereophotographs of casts (whitened) of some postcanine teeth of Behemotops proteus, holotype, USNM 244035, from lower part of Pysht Formation of Washington: A-D, RP:, in anterior (A), medial (B), po.sterior (c), and lateral (D) aspects. E-H, RP„ in anterior (E), medial (F), posterior (G), and lateral (H) aspects. i-L, RM? in anterior (i), medial 0), posterior (K), and lateral (L) aspects. Scale 1 cm. NUMBER 59 13 14 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY 235 56.5 (8.4) (5.5) (16.3) (26.8) 10.2 11.8 (15.1) 39.9 (19.7) (12.2) (23.3) (14.2) 26.0 20.8 14.4 21.5 23.1 15.8 14.5 24.1 (18.0) (17.5) 21.2 31.7 23.7 23.6 (15.5) faults introduce an element of uncertainty, as the direction of throw of the faults and the relation- ship of the residual accumulation of fossils to the bedrock and the faults are unresolved. There is in any case no basis to suppose that any of the material postdates the Pysht Formation. MEASUREMENTS.—The following measure- ments of the holotype right mandibular ramus of Behemotops proteus (USNM 244035), are in millimeters. Those in parentheses are approxi- mate, based on incomplete, damaged, or incom- pletely accessible parts. Those for all alveoli or dental crypts are as preserved, in all cases at least somewhat below the real alveolar border. See "Description, Dental Formula" for identification of dental loci. Maximum length of specimen as preserved Depth of ramus below M) DP2 Maximum diameter of crypt (alveolus?) Minimum diameter of crypt (alveolus?) DP3 Maximum anteroposterior dimension of combined alveoli DP4 Anteroposterior length of crown Width of anterior lobe of crown Width of medial lobe of crown Width of posterior lobe of crown PI (or DP,) Maximum height of crown, measured on P2 medial surface Maximum diameter of crown Minimum diameter of crown at same level Maximum diameter of crypt Minimum diameter of crypt (alveolus?) at P4 Ml M, M, edge Maximum height of crown Anteroposterior diameter of crown Transverse diameter of crown Maximum height of crown Anteroposterior diameter of crown Transverse diameter of crown Maximum height of crown (metaconid) Length of crown Anterior width of crown Posterior width of crown Maximum height of crown (metaconid) Length of crown Anterior width of crown Posterior width of crown Maximum height of crown (base of crown incompletely formed), as preserved (me- taconid) Length of crown (31-7) Anterior width of crown (19.0) Posterior width of crown (21.0) DESCRIPTION.—The holotype, USNM 244035, is generally well preserved, with sharply defined surfaces on the bones and all teeth. For the teeth the only exception is some fracturing in the posterior lobe of DP4 (Figure 10) and a single, major, anteroposterior fracture with some offset and compression in the crown of M] (Fig- ures 10, 12c). For the bones, the lingual surface of the horizontal ramus of the mandible (Figure 8) and the caudal surface of the femur (Figure I 4A) and much ofthe tibial surface are curiously eroded (chemically?) and pitted. This destruction of bone apparently removed the thin lingual walls of the crypts of the unerupted P3, P4, and M3, creating windows through which the initial prep- aration was done to expose these teeth. The phenomenon seems to have affected exactly the parts of the specimen that remained covered by the remnant of a primary concretion of slightly different color and texture than the enclosing secondary concretion. Further preparation was done to expose more of the crowns of the unerupted teeth, by remov- ing much of the labial and dorsal walls of their crypts. The unerupted teeth have been main- tained in their original positions as found, and casts have been prepared of their crowns to en- able viewing in direct occlusal aspect. Dental Formula (Figure 4): Considerable phy- logenetic significance attaches to the identifica- tion of dental loci in Behemotops proteus, in part because of divergent specializations and emphasis or de-emphasis of given loci in related taxa. For example, inferior tusks are developed from sec- ond incisors in moeritheres and other probosci- deans, but from canines in desmostylians. Siren- ians are less critical for such comparisons because their retention of five premolars in the known Eocene taxa, unique among Tertiary placental mammals (Domning, Morgan, and Ray, 1982), sets them phylogenetically apart from desmosty- NUMBER 59 15 Hans and proboscideans, which primitively re- tained only four premolars. The interpretations of B. proteus presented herein rest in part on evidence from B. emlongi from the Yaquina For- mation of Oregon, described later in this paper. Some aspects ofthe dental formula and succes- sion in the holotype of Behemotops proteus seem certain. There is no reason to question the iden- tification of Ml-Ms as conventionally understood (but see below, "Implications for Eutherian Den- tal Homologies"). Mi is considerably smaller than M2 and M3, and is already substantially worn, whereas M^, although fully in occlusal position, is virtually unworn, and the crown of M3 is in- completely formed and remains in its crypt (a small window to which had opened in the bone above it). The early emplacement and attrition of a relatively small M] is a common feature in bunodont herbivores, and is strongly pro- nounced in the desmostylian genus Paleopara- doxia (Figures 12E, 19). The identification of DP4 is also definite, because it was found, deeply worn, in place immediately anterior to and in contact with Mj. Its posterior root, although broken, is still in place anterior to Mi. Moreover, DP4 is trilobate, again as in many herbivorous mammals, for example artiodactyls (see Figure 20), phenacodonts (West, 1971, fig. 1), some specimens of Moeritherium (YPM 34764) but not especially so in the specimen figured by Andrews (1906:110, fig. 43), Phiomia serridens (Andrews, 1906, pi. 18), and Deinotherium and Gomphother- ium (Lartet, 1859, pi. 13: fig. 4c; pi. 14: fig. 4c. See also Frick, 1926; 1933). If DP4 and M1-M3 have been identified correctly, then the tooth occupying the crypt directly under DP4 (and the anterior part of M]) must be P4 and the tooth in the crypt immediately anterior to it must be P3. The two alveoli, subequal in size and circular in cross section, lying one each anterior and poste- rior to the principal cusp of P3, are for the two roots of DP3. Between and slightly lingual to these alveoli is what appears at first sight to be a third alveolus but which may be in reality the apex ofthe P3 crypt below (Figure 10). A similar resorption window can be seen in a specimen (AMNH CA 2423) of the living pygmy hippo- potamus, Hexaprotodon liberiensis, that died at a comparable stage in its ontogeny. Thus, the DP3 of Behemotops proteus was probably double- rooted. Only the deepest apical parts ofthe DP3 alveoli are preserved, because a considerable, although unknown, amount of bone is missing anterior to DP4 along the dorsal, alveolar-mar- ginal edge of the horizontal ramus. Anterior to the P3 locus the situation becomes less clear because of the loss of bone and the absence of all teeth except one (interpreted be- low as P] or DPi). We regard the large, simple alveolus anterior to the P3 locus as in fact the ventral remnant of the crypt for a large, fully formed, but probably unerupted P2. There is at depth a slight vertical crest in the bone on each side of the alveolus, suggestive of incomplete subdivision into anterior and posterior moieties of the root that occupied it. On the anterolabial border of this large crypt or alveolus is an apical remnant of a much smaller alveolus, probably for the anterior root of DP2. The anteriormost tooth present (forming the anteriormost pre- served part of the specimen) is a large, procum- bent, essentially caniniform tooth, with its crown and at least much of its root fully formed but lying in its crypt deep within the mandible at the time the animal died. This tooth, which we re- gard as Pi or DP], looks as if it would also fit the crypt of Pj reasonably well, suggesting that the P9 also was a relatively simple tooth. Unfortu- nately, the holotype of B. proteus retains no ves- tige of the canine or incisive loci. Enough bone of the horizontal ramus remains beneath Pi (or DPi) to demonstrate that the canine or any other enlarged anterior tooth could not have extended to the rear past Pi (or DPi) at the stage of ontogeny represented by the holotype. There is apparently no room for such a tooth, and, fur- ther, there is no hint of growth ofthe mandibular ramus in anticipation of the eventual emplace- ment of such a tusk. Rather, the symphyseal region may have resembled that of Hexaprotodon, with the Pi (or DPi) of B. proteus occupying the position ofthe canine in Hexaprotodon. 16 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY NUMBER 59 17 Dentition: The terms developed by Tobien (1978 and elsewhere) for the description of mas- todont molars, including Moeritherium, can be applied readily to the molars of Behemotops, and are noted herein throughout the description. All teeth except M3 show finely to coarsely crenulated enamel, wrinkled or pustulose, except where smoothed by wear. The only deciduous tooth preserved is the deeply worn, trilobate DP4 (Figures 4-10). Wear has reduced the occlusal plan to a series of three interconnected, subcircular rings of enamel, in- creasing in size posteriorly. The cusps of the upper deciduous premolars of Moeritherium (Schlosser, 1911, pi. 13: fig. 9) have a similar round outline at the stage of wear demonstrated; an only minimally trilobate lower deciduous pre- molar has been reported by Andrews (1906:110, fig. 43). However, DP4 of YPM 34764 is trilo- bate. The anterior lobe of DP4 in B. proteus is bordered anteriorly by a distinct precingulid. Pi (or DPi) (Figures 4-10) is represented by a well-formed but unerupted tooth. The tooth is simple, unicusped, essentially caniniform, single- rooted, robust, and has a bluntly rounded tip reminiscent of the principal cusp of P3 in the same specimen. There are faint longitudinal crests on opposite sides of the crown from ap- proximately 7 to 13.7 mm down from the tip of the tooth, from which point the crests each give way to a single row of small cuspules or coarse crenulations of subequal size, best exposed on the medial surface of the crown where the cus- pules are some six in number. These crests divide FIGURE 12.—Stereophotographs in occlusal aspect of some inferior postcanine teeth of Behemotops proteus from lower part of Pysht Formation of Washington, Paleoparadoxia ta- batai from Izumi locality, Japan (Shikama, 1966:12), and Behemotops emlongi from lower part of Yaquina Formation of Oregon: A-D, Behemotops proteus, holotype, USNM 244035: A, RP,; B, RP4; c, RMi-RM^; D, RM3. A, B, and D are photographs of casts (whitened), because original teeth cannot be viewed in exact occlusal aspect; compare Figures 4-10. E, Paleoparadoxia tabatai, cast of neotype, USNM 26375, LP3-M3. F, Behemotops emlongi, holotype, USNM 24033, LM3. Scale 1 cm. the crown into two unequal sections, a narrower, flattened, posteromedial one and a broader, rounded, anterolateral one. Later in ontogeny this tooth almost certainly would have rotated clockwise (as viewed dorsally) more than 45° about its longitudinal axis, so that the crests would have become anterior and posterior, the smaller, flatter surface lingual, the larger, more convex surface labial, and the greatest diameter anteroposterior. It would have stood much higher than P3 (and perhaps somewhat higher than P2) when fully erupted. P2 (Figures 4, 7, 10) is represented only by its crypt (alveolus). The crypt is similar enough to the shape of Pi (or DPi) to suggest a very similar tooth. However, its orientation, if that of Pj (or DPi) resembled it, would imply rotation of Pi (or DP|) as suggested above. A single root was pres- ent, but traces of fusion of two roots, perhaps separated at some earlier phylogenetic stage, are indicated on the alveolar walls. P3 (Figures 4-8, 1 1A-D, 12A) is dominated by a single, high, bluntly conical cusp (protoconid), with a weak posterior crest. Low on its anterolin- gual slope is a well delimited, subcylindrical, somewhat recurved cusp (paraconid); on its pos- terolingual slope is a weaker, lower, less inde- pendent cusp (metaconid); at its anterolabial base, labial to the paraconid, is a small cingular cusp, which, with a meager shelf at the anterior base of the paraconid, represents a precingulid. Small cusps on the posterior slopes of the proto- conid and metaconid near their bases may rep- resent the hypoconid and entoconid, respec- tively. Posterior to that and spanning almost the breadth of the tooth is a strong postcingulid consisting of some five small cusps, the labialmost three of which are largest and subequal to one another in size. A tiny, marginal, basal cusp oc- cupies the base of the crease delimiting the pro- toconid and possible hypoconid. The anterior and posterior roots are separate, in contrast to those of B. emlongi. P4 (Figures 4-8, 1 1E-H, 12B) is slightly larger than P3, but not so tall. It is dominated by two bluntly conical cusps of almost exactly equal SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY height. The lingual of these (metaconid) lies slightly posterior to a point directly medial to the labial cusp (protoconid), and is simple, with only two minor subsidiary cusps, one each at the base of its anterolingual and posterolingual slopes; the metaconid is not twinned. The metaconid is em- braced labially by the more complex protoconid, with its strong, cuspidate paracristid and proto- cristid curving about the anterolabial and poster- olabial parts of the base of the metaconid. The paracristid terminates in a poorly differentiated and somewhat bifid paraconid near the middle ofthe anterior wall ofthe crown. A well-marked and cuspidate precingulid lies below the para- cristid along the anterolabial base of the crown. Posterior to the metaconid and protoconid are two low cusps (entoconid and hypoconid), the latter the larger, lying posterior to the termina- tion of the protocristid. The entoconid and hy- poconid together are bordered posterolingually, posteriorly, and posterolabially by a strong postcingulid consisting of five to six low cusps of varied size, shape, and height. The anterior and posterior roots are apparently still separate, in contrast to those of B. emlongi. Ml (Figures 4-10, 12c) is by far the smallest ofthe molars, being only slightly larger in crown area than P., but more nearly rectangular. The basic plan of the bunodont, brachydont tooth is simple; it consists of four independent, subequal, major low conical cusps, the protoconid, meta- conid, hypoconid, and entoconid, each occupy- ing a quadrant of a square, to which attach strong pre- and postcingulids. The cingulids are not continued on the lingual and labial walls of the tooth, although suggested by a bulbous expan- sion of the anterolabial part of the base of the protoconid and by a pustulose shelf at the labial outlet of the valley between the protoconid and hypoconid. The conical shape ofthe protoconid is modified slightly by a very weak paracristid (anterior crescentoid of first pretrite, a. cr. 1) that extends anterolingually but lacks an identi- fiable paraconid. The metaconid is not twinned. A cristid obliqua (a. cr. 2) occupies an analogous position on the anterolingual slope of the hypo- conid. A low, worn, independent cusp occupies the center of the occlusal surface of the crown. This cusp lies adjacent to the anterolingual ter- mination of a. cr. 2 and on the anterior slope of the transverse valley. It may represent a conelet subsidiary to the metaconid but, if so, it is pos- teriorly displaced. A low, broad hypolophid ex- tends down the labial slope of the entoconid. The presence of the hypolophid, together with the nature of the wear on the crown, excessive on the anterior walls of the hypoconid and en- toconid and on the posterior walls of the proto- conid and metaconid, impart a mildly lophodont character to this tooth, not clearly evident in the little-worn Mj and even less so in the unerupted M;,. The crenulated postcingulid rises from its lingual and labial extremities to a central hypo- conulid. The anterior and posterior roots are well separated labially, but are closer together lingually. M^ (Figures 4-10, 12c) is almost an exact but enlarged replica of M]. It is almost unworn and therefore reveals the features of the crown clearly. The base of the crown forms a bulbous collar, especially distinct on the anterior half of the tooth. The cingulid, although variably devel- oped, is interrupted only lingual to the entoconid and labial to the hypoconid. The metaconid is not twinned. Anterior to the postcingulid and side by side posterolabial to the entoconid and posterolingual to the hypoconid are two small cusps on either side of the midline of the crown (at least the worn labial member of the pair is identifiable in Mi, and the lingual member could be present but obscured by wear and faulting of the tooth). These cusps are similar to the conelets of Gomphotherium (Tobien, 1978, fig. 1) but are not placed quite so far forward. The cristid ob- liqua (a. cr. 2) of M- supports two small cuspules separated by a notch representing the transverse valley. These cuspules are similar to those of Mi, but unworn. The anterior of the two lies at the base of the posterolabial slope of the metaconid, and might be regarded as its conelet, analogous in position to that ofthe entoconid. The postcin- gulid rises to a median apex as in M|. Insofar as NUMBER 59 19 can be seen, the roots are as in Mi. M3 (Figures 4-8, 1 li-L, I 2D) is basically simi- lar in construction to Mi and M2 insofar as its incompletely formed enamel cap reveals. Its roots had not formed at the time ofthe animal's death. However, the crown does show some uniquely interesting features. For example, all four principal cusps are simpler, higher, more nearly cylindrical than bluntly conical, and thus more widely separated from one another. The metaconid is not twinned. There is no evidence of a paraconid or paracristid (a. cr. 1). The cristid obliqua (a. cr. 2) is sharply defined, as is a com- plementary crest (hypolophid?) extending anter- olabially from the entoconid. The anteriormost ofthe two low cuspules seen on the cristid obliqua is a low, trihedral cusp, similar in position to a more rounded cuspule in the same position on M2, and to the worn cuspule on Mi, but less clearly tied to the metaconid. The postcristid is sharply defined, and posterolabial to the ento- conid it possesses a low, sharply pointed conelet. A small crest descends along the posterolingual base of the hypoconid. The crenulated postcin- gulid shelf is broader anteroposteriorly than in Ml or M2, and lends an angular, V-shaped outline to the posterior end of the tooth. The crown is not sufficiently formed to show the condition of the cingulid elsewhere. Apparently, the odd "bunostylodont" nature of the M3 of Behemotops proteus must be attrib- uted to its incomplete ontogenetic development; at maturity it would presumably have had more swollen cusps like those ofthe other molars. This is indicated by two pieces of evidence. First, M3 contrasts with all other teeth of the specimen in not having wrinkled or crenulated enamel; this suggests that the outer layers of enamel had not yet been deposited. Second, at least in Desmostylus it is certain that the columns of the molars at- tained nearly their full height long before reach- ing their final diameter. The developing molar, therefore, consisted of a group of high, slender columns, initially not connected at their bases and often found isolated. Only in later stages of development was sufficient enamel deposited on each column to fill the spaces between the col- umns. The clearest example of this that we have seen is in NSM 5600, the skull and mandible of Desmostylus described by Yoshiwara and Iwasaki (1902). On plate 2 of their paper an unerupted right upper molar ("M2") is shown. Examination of the actual specimen, now further prepared, discloses the condition described above, the tooth consisting of separate, very slender columns (Fig- ure 13). Thus, there is no reason to doubt that a similar although less extreme process of cusp thickening took place during the dental ontogeny of B. proteus and no reason to suppose that the difference in M3 cusp thickness between the lat- ter and B. emlongi is of any major taxonomic importance. Due allowance for additional enamel deposited on the sides (and to a lesser extent, on the tips) of the cusps of B. proteus would bring the dimensions of its M3 into tolerable agreement with those of B. emlongi. Osteology: The mandible of the holotype, USNM 244035, is robustly proportioned; its hor- izontal ramus was certainly thicker than imme- diately suggested by the specimen as preserved, in view ofthe postmortem loss of bone over much of its lingual surface. A remnant of bone below the posterolingual corner of M2 and the thickness ofthe ramus at Pi (or DPi) and P2 indicates that the horizontal ramus was originally several mil- limeters thicker over much of its expanse (Fig- ures 8 and 10). The ventral margin ofthe man- dible is essentially straight as far as it is preserved. Two small mental foramina open adjacent to the crypt of P3 (Figure 9); undoubtedly others were present more anteriorly. The ostensible foramen midway below DP4 (Figure 9) is in fact a window adjacent to the tip of the crown of P4. The ascending ramus of the mandible has its anterior and (as far as preserved) posterior margins nearly vertical. The rounded articular condyle is ele- vated well above the plane of occlusion; the an- terior margin of the ascending ramus is nearly straight and approximately perpendicular to the plane of occlusion. The coronoid process is broad, smoothly curved, and has a posterior hook. The mandibular foramen lies midway be- 20 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY FIGURE 13.—Desmostylus hesperus, NSM 5600, from Togari, Gihu Prefecture, Honshu, Japan, originally described by Yoshiwara and Iwasaki (1902); right upper molars in labial aspect. Note slender and distinctly separated enamel columns of unerupted and incompletely formed posterior molar, in contrast to thick, appressed columns of fully formed and erupted anterior molar. tween the anterior and posterior margins of the ascending ramus and directly posterior to the crypt for Mi (Figure 5). The mandibular canal passes forward from the foramen, under the ventrolateral edge ofthe developing M3, and less than 4 mm from it. A coronoid foramen (coro- noid canal) pas.ses through the base of the coro- noid process at the rear of the developing alveo- lus of Ml (Figures 5 and 8). Very little can be said about the postcranial skeleton, represented only by a few pieces thought to be part of the holotype. The distal half of the femur (Figure 14A,C) displays an an- teroposteriorly narrow shaft expanding distally into a broadly flattened extremity with a broad patellar facet; the narrowness of the shaft is strongly reminiscent of the condition in Paleo- paradoxia. Its incompleteness and the erosion of the bone on its caudal surface make precise as- sessment of its proportions impossible. The distal epiphysis was not coossified with the diaphysis, as it is slightly displaced. The tibia is represented NUMBER 59 21 FIGURE 14.—Behemotops proteus, holotype, USNM 244035, from lower part of Pysht Formation of Washington, distal part of right femur, in lateral (A) and cranial (c) aspects; Paleoparadoxia tabatai from Izumi locality, Japan (Shikama, 1966:12), cast of neotype, USNM 26375, distal part of right femur, in lateral (B) and cranial (D) aspects. 22 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY -f^s: V*% 5 CM FIGURE 15.—Behemotops proteus, holotype, USNM 244035, from lower part of Pysht Formation of Washington, phalanges, lacking proximal epiphyses, in dorsal (A, B) and palmar (E, F) aspects; Paleoparadoxia tabatai from I/umi locality (Shikama, 1966:12), cast of neotype, USNM 26375, phalanges, in dorsal (c, D) and palmar (G, H) aspects. NUMBER 59 23 by a proximal fragment with separate proximal epiphysis. The broken cross section of the dia- physis, some 60 mm from its proximal end (some 80 mm with the epiphysis in place), indicates a broad, anteroposteriorly flattened tibia with a low tibial crest and very shallow fossae. The two phalanges lack proximal epiphyses. They are strikingly broad, flat, and splayed distally (Figure 15A,B,E,F), also reminiscent of Paleoparadoxia. Behemotops emlongi, new species FIGURES 12F, 16-18 HOLOTYPE.—USNM 244033 (Emlong field no. E77-21), nearly complete left mandibular ramus, considerably damaged in region of sym- physis and posterior border of ascending ramus and condyle; only M3 present, but with partial or complete alveoli of all adult teeth; collected by Douglas Emlong, 25 March 1977. DIAGNOSIS.—Lower incisive alveoli three in number, subequal in size, with round cross sec- tions; lower canine (probably unlike that of B. proteus) greatly enlarged, tusk-like, procumbent, laterally compressed; longest diastema between P2 and P3, shorter than anteroposterior diameter of either tooth; P3 and P4 alveoli simple, housing single-rooted teeth, unlike B. proteus; anterior half of mandible massive; symphyseal region broad, shovel-like. ETYMOLOGY.—For the late Douglas R. Em- long, collector extraordinary. TYPE-LOCALITY.—183 meters (200 yards) south-southeast of Elephant Rock (Figure 2c), Seal Rock State Wayside, Lincoln County, Ore- gon; Sec. 25, T. 12 S, R. 12 W, Waldport Quad- rangle, 15-minute series, USGS (Figure 2); inter- tidal bench 76 meters (250 feet) west of the cliff face. REFERRED SPECIMEN.—USNM 186889 (Em- long field no. 555), much fragmented, poorly preserved anterior part of right mandibular ra- mus with root of P3 and essentially complete canine tusk; collected by Douglas Emlong, April 1969. Seal Rock State Wayside, Lincoln County, Oregon; Sec. 25, T. 12 S, R. 12 W. Waldport Quadrangle, 15-minute series, USGS (Figure 2). Same horizon as USNM 244033, Yaquina For- mation, from the foot of the sea cliff, approxi- mately 43 meters (140 feet) north ofthe isthmus joining Tourist Rock to the mainland. The isth- mus to Tourist Rock lies at latitude 44°29'50"N, longitude 124°05'00"W. HORIZON.—In coarse grit layer of lower part ofthe Yaquina Formation (Snavely et al., 1976). AGE AND CORRELATION (FIGURE 3).—The hol- otype and referred specimens of Behemotops em- longi, USNM 244033 and 186889, are from the Yaquina Formation of western Oregon, assigned to the Juanian Stage (Addicott, 1976a:99; Ar- mentrout, 1981:141) and are somewhat younger than the holotype of B. proteus from the lower part of the Pysht Formation of northwestern Washington. The area has been mapped by Snavely et al. (1976), who noted that the only formation exposed in the area is the lowermost part ofthe Yaquina Formation. Most of the Yaquina Formation is usually as- signed to the Zemorrian Stage, with only the uppermost part possibly referable to the Sauce- sian Stage (Snavely et al., 1969:38; Rau, 1981:81; Armentrout et al., 1983, chart). The exposures in the vicinity of Seal Rock have been identified explicitly as the lower part of the Ya- quina Formation (Emlong, 1966:2; based on pers. comm. from Snavely) and thus pertain to the Zemorrian part of the formation. ASSOCIATED FAUNA.—The Yaquina Forma- tion in Lincoln County, Oregon, has produced, in addition to the two specimens of Behemotops emlongi, a fairly rich fauna of pinnipeds, desmos- tylians, and cetaceans (Ray, 1977:428, 429). One of the cetaceans, Aetiocetus cotylalveus Emlong, 1966, was collected from the upper part of the Yaquina Formation, approximately 0.8 km (0.5 mile) north of Seal Rock State Wayside, and additional skulls are now known from the Ya- quina Formation but are not as yet described (Whitmore and Sanders, 1977:317; L.G. Barnes, pers. comm.). Other Desmostylia from the Ya- quina Formation include specimens of Cornwal- lius, under study by Reinhart (1975; 1982:550, 24 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY FIGURE 16.—Diagrammatic representation of mandible of Behemotops emlongi from lower part of Yaquina Formation of Oregon, mature individual, in occlusal (A) and left lateral (B) aspects; based on both the holotype, USNM 244033, and referred specimen, USNM 186889. Degree of posterior divergence of mandibular rami and length and shape of incisors (but not the fact of their presence as indicated by alveoli) are based primarily on analogy with Paleoparadoxia. NUMBER 59 25 FIGURE 17.—Behemotops emlongi, holotype, USNM 244033, left mandibular ramus from lower part of Yaquina Formation of Oregon, in lingual (A), labial (B), and occlusal (c) aspects. 26 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY fig. 1), collected from the same area and horizon (lower part of the Yaquina Formation) near Seal Rock as the specimens of Behemotops emlongi. The original material of Cornwallius sookensis is from the Sooke Formation on Vancouver Island, in large part at least coeval with the Yaquina For- mation (Durham, 1944:113; Addicott, 1976a: 99). Also from the same horizon in the lower part of the Yaquina Formation is the only land mam- mal known thus far from the formation, a frag- ment of a maxilla with deeply worn M'""' and M^ (USNM 187125; Emlong field no. 291) of an anthracothere (Ray, 1977:431). Richard H. Ted- ford has examined this specimen and has pro- vided us with the following statement (7 Apr 1983). The teeth, although worn and broken, retain the diag- nostic divided mesostyle and loss of paraconule found in only the genus Arretotherium among known North Amer- ican anthracotheres. The dimensions of the M^ of USNM 187125 are: length, 24.6 mm; width across the mesostyle, 25.8 mm. All three of the described species of Arreto- therium[A. acrirf^ni Douglass, 1902; A. leptodus{Maixhevi, 1909); and A. fricki Macdonald and Schultz, 1956] have upper molars of similar size and proportions, so it is not possible to determine the precise affinities ofthe Yaquina anthracothere without further evidence. It is closest in size to the holotype of A. leptodus from the late Arika- reean. As now recognized (Macdonald, 1956, 1963), the genus is confined to the early Miocene (late Arikareean through early Hemingfordian) and seems to succeed the closely related late Oligocene (Whitneyan through early Arikareean) genus Elomeryx. However, an extension of the geological range ofthe genus, in the form of a species like A. leptodus, into the early Arikareean seems indicated by unpublished material from Nebraska, South Dakota, and Wyoming contained in the Frick Collection at the American Museum of Natural History. These records would push the range zone of Arretotherium into the late Oligocene, in agreement with the assignment of a late Oligocene age for most of the Zemorrian and approxi- mately equivalent Juanian stages. MEASUREMENTS.—The following measure- ments of the holotype and referred specimen of Behemotops emlongi are in millimeters. Those in parentheses are approximate, based on incom- plete or damaged parts of specimens. Those for all alveoli are as preserved, in all cases at least somewhat damaged. See "Description, Dental Formula" for identification of dental loci. USNM 244033, holotype, left mandibular ra- mus: Maximum length of specimen as preserved 396 Maximum height of mandible at coronoid process 222 Depth of horizontal ramus below M) 114.1 Depth of horizontal ramus behind M.s 128.3 Maximum width of jaw at symphysis as preserved (119.3) Breadth of jaw behind M3 (35) Canine tusk alveolus, maximum dorsoventral (75.9) diameter Canine maximum transverse diameter (29.4) Alveolar length P, (or DP,)-M:, (208?) P.-M:, (176) P:,-M, (135) M,-M, 87.8 P] (or DPi) (alveolus strongly inclined forward) (36.7) Pj 24.9 P:i 21.5 P4 16.1 M, 20.3 M, 23.0 Ms 40.6 Alveolar width P, (or DP,) (18.6) PL' 15.6 P. (12.3) P4 11.4 M;, (anterior) 6.9 M:i (posterior) 18.0 Diastema between P-^ and P^ 17.0 M:, Maximum height of crown (metaconid) 18.0 Length of crown 37.6 Anterior width of crown 24.2 Posterior width of crown 28.6 USNM 186889, referred anterior fragment of right mandibular ramus: Combined breadth of alveoli ofthe three incisors (63) Maximum depth (as preserved) of I^ alveolus (55) Maxinmm mediolaleral diameter of f, alveolus (17.9) Maximum anteroposterior diameter of I3 alveolus (H.8) Maximum depth (as preserved) of F, alveolus (33.5) C>anine tusk maximum length (as preserved) 267 maximum dorsoventral diameter (63.5) (near alveolar margin) maximum transverse diameter (43.1) (near alveolar margin) NUMBER 59 27 (23.8) (15.7) 43.1 Ps root anteroposterior diameter transverse diameter maximum length (as preserved) DESCRIPTION.—The holotype and referred specimen of Behemotops emlongi, USNM 244033 and 186889, as well as specimens of other taxa from the same horizon in the Yaquina Forma- tion, exhibit a peculiar preservation. The bone is weak and, especially in USNM 186889, has grit from the enclosing matrix pressed into its surface as if to become almost an integral part of the bone (Figure 18). The alveolus ofthe canine tusk of USNM 244033 may well have been com- pressed laterally after burial. A part ofthe lingual wall of the alveolus had disintegrated; for this reason during preparation the mandible was strengthened in that area with fiberglass (Figure 17A,C). The poor preservation of surface detail in both specimens leaves some doubt as to the detailed character of the anterior premolar al- veoli (Figures 17c, 18c), and makes illustration of the incisive alveoli impractical. Dental Formula (Figures 4 and 16): The hol- otype and referred specimen of Behemotops em- longi are compatible with the interpretation of the dental formula of B. proteus and add infor- mation on the permanent incisors and canine. USNM 244033 has its well-worn Ms in place. Its broadly exposed roots consist of a transversely widened anterior root and a larger posterior root, triangular in cross section with the apex of the cross section posterior, under the talonid. Some of the weak, thin-walled bone of the alveo- lar margins of all teeth has been lost in preser- vation or preparation, making their size and character somewhat conjectural. The alveoli of My and Mi indicate transversely widened ante- rior and posterior roots subequal in size but with the anterior root longer in M2 and the posterior longer in Mi. The alveoli of the roots of Mi are comparatively small, shallow, and convergent ap- ically, indicating a small, possibly senescent Mi. The P4 alveolus is shallow, simple, and ovoid in cross section; that of P3 somewhat deeper, more elongate anteroposteriorly, and with the sugges- tion of crests on each side. There is a short diastema anterior to P3, perhaps resulting from the progressive forward tilting of P2 and Pi (or DP]) in accommodation to their position dorso- medial to the alveolus ofthe massive canine tusk. The alveolus of P2 is larger than those of P3 and P4, anteroposteriorly elongate, simple, and has crests on its lingual and labial walls, reflecting indentations in the root of P2. The alveolus of Pi (or DPi) is similar in size to that of P2, strongly inclined forward, and simple, with no indication of subdivided roots. The alveolus of the canine tusk is very large, laterally compressed (possibly in part postmortem), and extends posteriorly to a point at least below Mi. There are vestiges of the deepest parts of the simple alveoli of Ii and I2, and possibly of I3, but the amount and quan- tity of the bone preserved in this region would be inadequate for secure interpretation were it not for the existence of USNM 186889. If cor- rectly interpreted, the apices of the alveoli of these three teeth converge in a triangular ar- rangement. The bone in USNM 186889 is poorly pre- served and meager, but is just sufficient to pro- vide a reliable basis for establishment of the anterior mandibular dental formula. P3 is repre- sented by a remnant of a robust, forwardly tilted, simple root, 43.1 mm long as preserved, and transversely subdivided into subequal anterior and posterior moieties by slight lateral indenta- tions. There is no indication of a diastema ante- rior to P3, but the adjacent shattered, poorly preserved bone is very likely displaced lingually. If this bone and the tooth were swung labially into line with the Pi (or DP]) and P2 alveoli, a diastema would open anterior to P3. The alveoli of P2 and P] (or DPi) are essentially similar to their counterparts in USNM 244033, as far as preserved. Each lies anteriorly inclined along the dorsolingual surface of the canine and separated from it only by a thin alveolar wall. The alveolus of P2 is very incomplete, and inadequate to re- flect subdivision of the root if such is the case. The alveolus of Pi (or DPi) indicates a much smaller, shorter root than in Pi (or DPi) of 28 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY lOCM 1 I FIGURE 18.—Behemotops emlongi, referred anterior fragment of right mandibular ramus from lower part of Yaquina Formation of Oregon, USNM 186889, in lingual (A), labial (B), and dorsal (c) aspects. NUMBER 59 29 USNM 244033. Perhaps the most important fea- ture of USNM 186889 is the presence of alveoli of three incisors, strongly indicating that the tusk is a canine as in Paleoparadoxia (Figure 19). The alveolus of the first incisor is much fractured and obscured by matrix; that of I2, although highly incomplete, is 55 mm deep as preserved and suggests a straight, robust tooth of rounded cross section. The alveolus of I3 is similar, but its preserved part is only some 33.5 mm deep, in- dicating a shorter tooth. Thus, Behemotops emlongi from the lower part of the Yaquina Formation of Oregon retained a complete, inferior, adult dentition of three inci- sors, one canine, four premolars, and three mo- lars. We infer that B. proteus from the lower part of the Pysht Formation of Washington did so as well. Paleoparadoxia, a more derived desmosty- lian, had one less premolar. Dentition: As in Behemotops proteus, all pre- served teeth of both specimens of Behemotops emlongi show the enamel finely to coarsely cren- ulated, wrinkled, or pustulose, except where smoothed by wear. The inferior incisors are known only from their incomplete alveoli in USNM 244033 and 186889. These indicate three incisors, approxi- mately similar in size, straight and subcylindrical in shape, perhaps 15-20 mm in their maximum diameters. The inferior canine, actually preserved in USNM 186889 and represented by its incom- plete alveolus in USNM 244033, is a greatly enlarged, strongly procumbent, laterally com- pressed tusk. It is widest dorsally (anatomically, posteriorly) and narrowest ventrally (anatomi- cally, anteriorly). The ventral narrowing is ef- fected largely by development of single, comple- mentary, broad longitudinal channels on either flattened surface, deeper on the medial side. The tusk is essentially straight in dorsal or ventral aspect, but gently curved in an open S-shape in lateral or medial aspect, with the alveolar end downturned and the extruded end upturned (Figures 16 and 18). The base ofthe tusk is open, with a deep, conical pulp cavity, indicative of persistent growth, maintained at least into old adulthood. The tusk has a thin (perhaps 0.5 mm) sheath of enamel extending entirely around its circumference, from the worn tip proximally for some 95 mm. The exact proximal limit of the enamel crown is difficult to define because the deposition of enamel apparently terminated ir- regularly in streaky continuations of longitudinal ribs and wrinkles, which are apparent on the crown wherever wear facets or polishing have not removed them. There is a large, subplanar wear facet truncating the crown obliquely. This facet would have been produced, not by occlusal wear, but by wear against a substrate, presumably in feeding. There is on the dorsal part of the medial surface of the crown at its widest part what appears to be a facet of occlusal wear (pre- sumably produced by shearing action with a su- perior incisor), recognizable over a length of some 47.5 mm and a maximum width of 7.5 mm. There is no evidence of cementum on the tusk. Except for the root of P3 in USNM 186889, P1-M2 are represented only by alveoli in Behem- otops emlongi. For information on the size of these teeth, see the discussion of the dental formula and the measurements (pages 26, 27). A rather heavily worn M3 is the only tooth preserved in USNM 244033 (Figure 12F). It is essentially similar to the molars of the type spec- imen of Behemotops proteus, especially Mi and M2, but differs from the incompletely formed M3 of that specimen in having less stylodont principal cusps. As is demonstrated by Mi ofthe holotype of B. proteus, USNM 244035, thickened enamel can be seen in M3 of B. emlongi, USNM 244033. Its cingulid is continuous labially, being especially strong adjacent to the hypoconid. Two small conelets are present near the posterolingual base of the hypoconid and the posterolabial base of the entoconid, respectively, but are heavily worn. Before wear, the heel was evidently a transverse crest composed of several small crenulations. The metaconid apparently is not twinned. Osteology: The remnant of the bony ramus of USNM 186889 is osteologically useful primarily in revealing the procumbency of the canine tusk 30 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY FIGURE 19.—Paleoparadoxia tabatai, cast of neotype, USNM 26375, from Izumi locality (Shikama, 1966:12), middle Miocene of Japan, left mandibular ramus, in lingual (A), labial (B), and occlusal (c) aspects. NUMBER 59 31 and the incisors, and the broadly scoop-like sym- physeal region. Both of these characteristics are confirmed in USNM 244033, which preserves most of the bony ramus. The most conspicuous character ofthe ramus is the massively expanded anterior end (Figure 16), reflecting the relatively huge canines and battery of six incisors. The under-surface of the symphyseal region is broadly flattened, almost planar. If this surface is oriented perpendicular to the sagittal plane, as it almost inevitably was in life, then the canine was compressed almost exactly in the vertical plane and the postsymphyseal body of the jaw was canted strongly inward dorsally, as in Hex- aprotodon. There is a single, large, mental fora- men adjacent to the canine alveolus and lying ventral to the P2 alveolus (Figures I 6B, 17B). The postsymphyseal ventral margin of the horizontal ramus is nearly straight, but the angular margin is missing, as is the thin posterior margin of the ascending ramus. If one assumes that little of this latter margin is lost, the anterior and posterior margins are essentially parallel and inclined slightly forward. The condyle is incomplete but its position well above the plane of postcanine occlusion is clear. The mandibular foramen opens just above the level of the dorsal border of the bony horizontal ramus and approximately midway between the anterior and posterior mar- gins of the ascending ramus. The coronoid canal (which descends from the rear of the tooth row to open above the mandibular foramen), if pres- ent, is tiny and obscured by poor preservation. The coronoid process is broad, flattened, smoothly rounded in profile, and inclined some- what anteriorly. The bony ramus posterior to the base ofthe canine tusk is relatively thin, and the postcanine dentition of modest size, in contrast to the massive, broad, scoop-like muzzle with large teeth indicated anterior to that point. Relationship between Behemotops proteus and B. emlongi Unfortunately, there are few points of anat- omy on which the present specimens of Behemo- tops proteus and B. emlongi can be directly com- pared. However, these include the size and mor- phology of M3, the postcanine dental formula, the form ofthe ascending ramus, and to a certain extent the form of the symphyseal region of the mandible. As discussed at length above, the den- tal formulae ofthe two species are not demonstr- ably different except for the fusion of the roots of P3 and P4 exhibited by B. emlongi (and by more advanced desmostylians). However, the peculiar M3 morphology of B. proteus is attributable at least in large part to incomplete development; the ascending rami are not significantly different; and the symphyseal region of B. proteus could well have been broad and scoop-like as in B. emlongi. The length of the incomplete M3 crown in B. proteus is 31.7 mm, compared with 37.6 mm for the complete M3 crown in B. emlongi; the difference is even within reasonable limits of intraspecific variation. Only four differences, two of which are in- ferred rather than clearly demonstrated by the specimens at hand, suggest to us that specific distinction is warranted: greater adult mandible size, size and position of the canine tusk, and fusion of the roots of P3 and P4 in B. emlongi. In each of these characters B. emlongi is more de- rived than B. proteus, but the matter is clouded by the immature condition of the only known individual of B. proteus. We base our conclusions about projected man- dible size in part on growth of this bone in Recent Hippopotamidae, the closest living morphologi- cal analogs of desmostylians. An immature Hex- aprotodon (USNM 271019, Figure 20) with M2 partly erupted and DP4 heavily worn (therefore dentally slightly younger than the type of Behem- otops proteus) has a mandible 20 cm long, com- pared to 27.5 cm in an adult (USNM 302054) with worn M3. An immature Hippopotamus (USNM 162976) with M. completely erupted and DP4 heavily worn (comparable to the type of B. proteus) has a mandible 46 cm long; the largest adult mandible we measured (USNM 123387) was 59 cm long. If a similar relationship between growth and tooth eruption existed in B. emlongi 32 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY 0 5CM J I I L B FIGURE 20.—Hexaprotodon liberiensis, modern pygmy hippopotamus, USNM (Division of Mam- mals) 271019 from the National Zoological Park, mandible of immature individual, in occlusal (A) and right lateral (B) aspects. NUMBER 59 33 (an animal intermediate in size between Hexa- protodon and Hippopotamus), its juveniles should have reached 75%-80% ofthe adult mandibular length by the stage of tooth eruption seen in the type of B. proteus. However, the mandible of the latter is only about 55% as long as that of the adult B. emlongi. As noted in our description of Behemotops pro- teus, there does not appear to have been space in the mandible below Pi (or DPi) for an enlarged canine tusk like that of B. emlongi. Rather, any canine or incisor tusks that were present must have been medial to Pi (or DPj). It seems highly unlikely that enough tusk growth or mandibular remodelling could have taken place in the time remaining until eruption of M3 for B. proteus to take on the form of B. emlongi. Principally for the latter reason, we prefer to regard the Washington and Oregon animals as representing separate species of a single genus, pending additional knowledge of their anatomy and ontogeny. As Hirota (1981), Reinhart (1982:554), and Shikama (1966:131) suggested, and as Reinhart (1959:92) hinted but then denied, some or all desmostylians probably were sexually dimorphic. Writing about the proboscidean allies of the des- mostylians, Osborn (1936:183) claimed the same. In all the known Proboscidea there is a marked disparity between the male and female incisive tusks both in length and in diameter. The adult female tusks never fully attain the length ofthe adult male tusks, but a still more striking difference is their slenderness of proportion and diame- ter. Frick (1933:507, 574, 581, 632, 650) also mentioned sexual dimorphism in gomphotheres and mammoths. Nevertheless, in spite of these examples in proboscideans, we believe that the observed differences in morphology between Be- hemotops proteus and B. emlongi are too great to be of sexual origin alone. History of Desmostylian Systematics VanderHoof (1937:170-177) thoroughly re- viewed the desmostylian literature, and his work should be consulted for details to that date. We present only a synoptic coverage here. The desmostylians were first made known by O.C. Marsh (1888) on the basis of some material of Desmostylus itself from marine Neogene de- posits of Alameda County, California. Marsh re- ferred Desmostylus to the Sirenia. Flower and Lydekker (1891), on the basis of Marsh's work, next placed Desmostylus in the Halicoridae (= Dugongidae) and from then until 1953 Desmos- tylus was generally regarded, sometimes with a query, as a sirenian. The second major find of a desmostylian fossil was reported from Japan by Yoshiwara and Iwa- saki (1902), who described and figured the an- terior part of a skull and both lower jaws of a specimen of Desmostylus. They believed their find to be some sort of proboscidean, based in part on a letter from H.F. Osborn. Osborn had ex- amined photographs of the specimen and had read a brief description of the skull, sent to him by Yoshiwara and Iwasaki. In their paper Yoshi- wara and Iwasaki made no mention of Marsh's description of Desmostylus, so presumably they were unaware of it. They believed their find to represent a new genus, but did not supply a new name. Although Osborn had "informed" them that the skull belonged to a proboscidean, Yosh- iwara and Iwasaki demonstrated that it was not like deinotheres or elephantids and therefore would have to represent a branch from the prim- itive proboscideans, near the origin of that order from among the other ungulates. They also men- tioned some similarities to Sirenia. In the same year, however, both Osborn and J.C. Merriam recognized that the Japanese specimen was ref- erable to Marsh's Desmostylus (Osborn, 1902). Schlosser (1904) regarded the Japanese specimen as definitely sirenian. Osborn (1905:109) placed Desmostylus in a monotypic family Desmostylidae and stated that it belonged in either the Sirenia or the Probos- cidea. Merriam (1906; 1911:412) regarded Des- mostylus as a sirenian, possibly requiring its own family, and reinforced the suggestion of relation- ship between sirenians and proboscideans. 34 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Abel (1914:213; 1919:830; 1920:445), in con- trast to his later work, concluded that Desmostylus was more closely related to the Proboscidea than to the Sirenia, although it is not clear why he did so. Later Abel (1922:381; 1923) abandoned his view of proboscidean affinities of Desmostylus in favor of a bizarre notion that it belonged to the mammalian subclass Allotheria (= Multituber- culata). He persisted in this belief even after examining a skull of Desmostylus at the NMNH (Abel, 1926); then, and later (in Weber, 1928:xiii, 44, 85), he placed the family Desmo- stylidae in the Monotremata, suggesting a possi- ble relationship between them and multituber- culates. Still later Abel (1933:875) elevated them in rank, naming an order Desmostyloidea within the subclass Multituberculata. He evidently re- garded them as multituberculates to the end of his career (Abel, 1944). Although it has priority over Desmostylia Reinhart, 1953, Abel's ordinal name apparently has been overlooked com- pletely during the more than 50 years since its creation. No useful purpose would be served by resurrecting it. Adherence to priority in names of suprafamilial taxa is not required, and in this case it would not be in the interest of stability. Therefore, we strongly recommend retention of the shorter, more euphonious, and entrenched name Desmostylia. Hay (1915), followed by Matsumoto (1918), placed the Desmostylidae in the Sirenia, although he emphasized that the Desmostylidae were very different from other (true) sirenians. Later Hay (1923:109) created a suborder Desmostyliformes to contain the Desmostylidae alone among sir- enians, placing all other sirenians in a suborder Trichechiformes. In the same paper and in a succeeding one (Hay, 1924:7) he corrected in detail the misinterpretations of cranial sutures on which Abel's assertion of multituberculate affinities were largely based. Winge (1924:187, 188; 1942:214) regarded Desmostylus as "undoubtedly a lateral offshoot of the oldest manatids." Winge also provided some pithy comments about Abel's theory of multitu- berculate affinities of the desmostylians. VanderHoof (1937) countered Abel's interpre- tations in detail and supported inclusion of des- mostylians in the Sirenia as the suborder Des- mostyliformes. Sickenberg (1938), however, ar- gued strongly against a desmostylian-sirenian re- lationship. Gregory (1951:428, 801-803) re- tained Desmostylus in the Sirenia but gingerly suggested "remote derivation from such a prim- itive proboscidean as Moeritherium." According to Shikama (1966:151), an earlier publication by H. Kishida (1924) assigned Des- mostylus to the Marsupialia. We have not seen Kishida's work. Ijiri (1939) considered Desmostylus to be an ungulate "in the broadest sense" but not a mon- otreme, multituberculate, marsupial, or sirenian. Reinhart (1953) proposed the order Desmo- stylia, essentially an elevation of Osborn's Des- mostylidae and Hay's Desmostyliformes to still higher taxonomic rank. Reinhart's (1959) revi- sion of the Desmostylia led him to believe that the desmostylians, sirenians, and proboscideans are closely related paenungulates, but that the desmostylian stem separated from the other two orders in the Paleocene. On the basis of postcran- ial evidence, Reinhart noted that the desmosty- lians could not be descended from known siren- ians because desmostylians were still capable of locomotion on land. Similarly, dental evidence led him to conclude that Moeritherium (Figure 21) was already too advanced along the probos- cidean path to have been a desmostylian ances- tor. Thus, the Desmostylia were shifted from their former status as a sirenian subdivision and were given taxonomic equality with both the Proboscidea and the Sirenia. McKenna (1975:42) later dubbed this unresolved trichotomous group of paenungulate orders the mirorder Tethyth- eria. The cladistic analysis given below (Figure 22) resolves the trichotomy and indicates that FIGURE 21.—Stereophotographs in occlusal aspect of infe- rior postcanine teeth of Moeritherium trigodon from Djebel- el-Qatrani Formation of Egypt: A, YPM 18181, RP^-M^; B, YPM 18098, RM.-M^; c, AMNH 13437, RP., P3, Mj-Mj. Scale 1 cm. NUMBER 59 35 36 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY within Tethytheria the Desmostylia are more closely related to the Proboscidea. Sera (1954) believed that Desmostylus, sireni- ans, and proboscideans were derived from an- omodontine therapsids. Thenius (in Thenius and Hofer, 1960:189, 190, 196, 197), on the basis of meager similarities and geographic separation between probosci- deans and sirenians on the one hand and des- mostylians on the other, placed the latter in the superorder Protungulata in the sense of Simpson (1945). Later (Thenius, 1969:584-589, 631) he did not employ the group Protungulata, but con- tinued to regard the Desmostylia as "Huftiere" excluded from the superorder "Subungulata" (Sirenia, Proboscidea, Hyracoidea, and Embrith- opoda). Ijiri and Kamei (1961:27) thought that des- mostylians were especially close to perissodactyls and artiodactyls, as did Shikama (1966:153). TETHYTHERIA SIRENIA PROBOSCIDEA DESMOSTYLIA ./ / / '37<],?81,83-92 80,781,82 73-77 <1 = parallelism t t f ^^ ^r ^ '^ 'b" t / S" 69-71 67,? 68 100,13-15 / x