JSirmm ' 4 %,... ** — . >'-• >* • ... w. North American Eocene Sea Cows (Mammalia: Sirenia) ft* . . Ili HP '%$&?& • •;/ on pppglip M ■&Sv Pm wuswm .!|fflau.., d 4%f-/ > >, frr ' 'v ; ❖ Sip saSflH j&MSfc, !«&2g| ^fffigP Ww& ¥% SSii$ SyKM ,3 Asci , TW. mm DARYL P. DOMNING, GARY S. MORGAN, CLAYTON E. RAY , ^Y^sHi ■ * x 'k''v 4„ .-„ » "t, .Vw O V >v# JGB WW ; rw\\\ A V M WZ&P! \ \ \V > i^MPSShrSI ,> & . O'* ’ . x -A. m. TMm UK _.. r % I * n v 52 ^ K kVWH 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 series 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 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 distributed 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 manuscripts are given substantive review. Press requirements for manuscript and art preparation are outlined on the inside back cover. S. Dillon Ripley Secretary Smithsonian Institution SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY • NUMBER 52 North American Eocene Sea Cows (Mammalia: Sirenia) Daryl P. Domning, Gary S. Morgan, and Clayton E. Ray SMITHSONIAN INSTITUTION PRESS City of Washington 1982 ABSTRACT Domning, Daryl P., Gary S. Morgan, and Clayton E. Ray. North American Eocene Sea Cows (Mammalia: Sirenia). Smithsonian Contributions to Paleobiology , number 52, 69 pages, 34 figures, 4 tables, 1982.—The record of Eocene sea cows in North America is reviewed in detail, and that of the world is summarized. The North American record includes some 20 localities, mostly yielding fragments identifiable only as sirenian. Of these, the most extensive materials are a partial skeleton from the Cook Mountain Formation of Texas, numerous isolated elements from the Avon Park Formation of Florida, and a partial skeleton and other specimens from the Castle Hayne Formation of North Carolina. The materials from North Carolina and Florida are middle Eocene in age and are referred to Protosiren species. These specimens provide further confirmation of the fact that Eocene sirenians had a 3.1.5.3 dental formula and were the latest eutherians known to exhibit five premolars. The implications of this for the higher classification of mammals are discussed. The distribution of sirenians suggests a homogeneous middle Eocene Tethyan fauna and also seems to be a more useful guide to the former distribution of seagrasses than are the distributions of Foraminifera. Eocene sirenians have potential value in intercontinental biostratigraphic correlation. Official publication date is 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 Paul North American Eocene sea cows (Mammalia: Sirenia) (Smithsonian contributions to paleobiology ; no. 52) Bibliography: p. 1. Sirenia, Fossil. 2. Paleontology—Eocene. 3. Paleontology—North America. I. Morgan, Gary Scott. II. Ray, Clayton Edward. III. Title. IV. Series. QE701.S56 no. 52 [QE882.S6] 560s 82-3252 [569'.5] AACR2 Contents Page Introduction . 1 Acknowledgments . 1 Eocene Sirenian Records . 2 Erroneous Records . 2 New World Records . 3 Old World Records . 18 Florida Eocene Sirenians . 18 Description . 18 Comparisons . 35 North Carolina Eocene Sirenians . 39 Description . 39 Comparisons . 55 Sirenian Dental Formulae and the Cladistic Classification of Mammals . 59 History, Biogeography, and Correlation . 60 Paleoecology of Eocene Sirenians and Seagrasses . 61 Conclusions . 62 Literature Cited . 64 iii North American Eocene Sea Cows (Mammalia: Sirenia) Daryl P. Domning, Gary S. Morgan, and Clayton E. Ray Introduction World knowledge of Eocene sea cows began spectacularly in 1855 with the description by Owen of Prorastomus sirenoides, based upon a skull, jaws, and atlas vertebra from Jamaica, still one of the oldest and certainly the most primitive known sirenian. Since then a great deal has been learned about Eocene Sirenia in the Mediterra¬ nean region, where discovery and study continue actively at present (Savage, 1977:344-346). In contrast, virtually nothing of consequence has been added for the Western Hemisphere. Thus recent discoveries of assessable, though fragmen¬ tary, materials in Florida and North Carolina seem noteworthy. Our purposes here are to make this new mate¬ rial known; to review other New World Eocene records, published and unpublished (and a few erroneous), in order to bring up to date the meager information on the subject; and to com¬ ment upon these records in relation to sirenian relationships, paleontology, biogeography, and paleoecology, in general. Daryl P. Domning, Department of Anatomy, College of Medicine, Howard University, Washington, D.C. 20059. Gary S. Morgan, Florida State Museum, University of Florida, Gainesville, Florida 32611. Clayton E. Ray, Department of Paleobiology, National Mu¬ seum of Natural History, Smithsonian Institution, Washington, D.C. 20560. Sequence of authors determined by alphabetical order of surnames. The following abbreviations are used through¬ out. BM(NH) CGM ChM MCZ TMM TRO TU UF UF/FGS UGV USGS USNM YPM British Museum (Natural History) Cairo Geological Museum, Egypt Charleston Museum, South Carolina Museum of Comparative Zoology at Harvard College Texas Memorial Museum, University of Texas Timberlane Research Organization, Lake Wales, Florida Tulane University, Vertebrate Paleontology Collection Florida State Museum, Gainesville Former Florida Geological Survey collections de¬ posited in the Florida State Museum, Gaines¬ ville University of Georgia, Department of Geology, Vertebrate Fossil Collections United States Geological Survey former United States National Museum collec¬ tions deposited in the National Museum of Natural History, Smithsonian Institution Peabody Museum of Natural History, Yale Uni¬ versity Acknowledgments. —First and foremost we wish to thank Peter J. Harmatuk, Robert Armi- stead, and John Waldrop, whose alert and dedi¬ cated field work resulted in the recovery of the most useful Eocene sirenian material from the Western Hemisphere in well over a century, dem¬ onstrating once again that advancement in ver¬ tebrate paleontology begins in the field. 1 2 We wish also to thank the following donors, friends, and colleagues who made specimens available to us: Wayne F. Canis, Rick Coffey, David Cramer, Paul Drez, Freeman Foote, Steve Heaton, Markes E. Johnson, Joshua Laerm, Wann Langston, Jr., John A. MacFadyen, Jr., David Mason, Grant E. Meyer, Roy H. Reinhart, Albert E. Sanders, Charles R. Schaff, Vincent Schneider, A1 Simons, Elwyn L. Simons, Erich Thenius, John T. Thurmond, Charles E. Tucker, Michael R. Voorhies, John S. Waldrop, and S. David Webb. We also thank Mrs. Sue Pitts for her efforts to locate additional materials. R.J.G. Savage has shared freely his unparal¬ leled knowledge of Old World fossil sirenians and of Prorastomus sirenoides. E. Allen, Peter van Bree, P.H. de Buisonje, Ragi Eissa, Baher el-Khashab, Matthew Freudenthal, Alan Gentry, Paul F. Huddlestun, Miklos Kretzoi, Jeheskel Sho- shani, and Chris Smeenk have provided useful information and access to specimens. Clair R. Ossian sent us his unpublished manuscript (Ralph S. Kerr, co-author) on Cretaceous seagrass records for North America and permitted us to mention them herein. William A. Deiss assisted in locating information in the Smithsonian ar¬ chives, as did Alta Copeland in the Remington Kellogg Library of Marine Mammalogy. Druid Wilson and L.W. Ward have provided invaluable insight into local and regional prob¬ lems in stratigraphy and correlation. We also thank Laurel M. Bybell, Norman Frederiksen, Thomas G. Gibson, Joseph E. Hazel, and Robert L. Meyer for paleontological analyses. R.J.G. Savage, Roy H. Reinhart, and Malcolm C. McKenna have reviewed the manuscript in whole or in part and are responsible for consid¬ erable improvement, and none of the remaining deficiencies. Several specimens, including the mandible of USNM 214596, were in part prepared or reha¬ bilitated by Arnold D. Lewis. The illustrations were prepared by Lawrence B. Isham, and the photographs by Victor A. Krantz. Donald J. Ortner made X-ray photographs of the mandible from North Carolina. SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Financial support for these studies has been provided in part by the Remington Kellogg Me¬ morial Fund, the Walcott Fund, the Smithsonian Research Foundation, and the Office of the Di¬ rector, National Museum of Natural History. Eocene Sirenian Records Erroneous Records In 1869 Cope described Hemicaulodon effodiens , based on an incomplete tusk, thought to be an upper incisor of an Eocene sea cow, from a marl pit at Shark River, Monmouth County, New Jersey. The specimen has been redetermined as the basal part of the tusk of an advanced, prob¬ ably Pleistocene, walrus (Ray, 1975). In 1887, in his catalog of fossil mammals in the British Museum (Natural History), Lydekker (p. 13) listed under Prorastomus sirenoides , “M. 3565. Shaft of a humerus; from a Tertiary deposit in the island of Sombrero, near St. Kitts.” Kellogg (1966:65) noted the record and remarked that the rocks of the island more recently were regarded as Miocene in age. In 1968 one of us (Ray) examined this specimen and found an unsigned note in its tray, reading as follows: “Probably part of femur of Testudo sombrerensis (Leidy).” The handwriting looked like that of Dr. Ernest E. Williams, though he does not now recall writing the note (pers. comm., 1976), but in any event the redetermination appears to be correct. The extinct tortoise of Sombrero Island, Geochelone (Chelonoidis) sombrerensis , is regarded as ?Late Pleistocene in age by Auffenberg (1974:150). In 1917 Palmer reported a neural arch of a sirenian found “freshly fallen, under a cliff of the Calvert Miocene on the western shore of Mary¬ land” and “suggested that the species was living during the period following the first erosion of the Cretaceous and the deposition of the Eocene, as all the specimens so far found in the Miocene were clearly redeposits from an earlier age.” Kel¬ logg (1925:59) cited Palmer’s report but remarked that Palmer’s “view that it was redeposited from an earlier, probably Eocene, formation does not NUMBER 52 3 appear plausible.” There is to date no indication of redeposited Eocene sirenian remains among the fairly abundant and in part well-preserved sirenian material from the Miocene Calvert For¬ mation (Kellogg, 1966). In the course of describing two archaeocete caudal vertebrae (USNM 13856) collected in 1921 by John Navratil and Mark Francis on the J.H. Giessenschlag property, Burleson County, Texas, Kellogg (1936:261, 271) mentioned that they had been “found near 17 vertebrae belong¬ ing to an Eocene sirenian and of these 15 were articulated.” Stenzel (1938:156) repeated this rec¬ ord on Kellogg’s authority, noting that the spec¬ imens came from the Mount Tabor Member of the Crockett Formation of the Claiborne Group. Kellogg had indicated the source bed as the Yegua Formation, immediately overlying the Crockett Formation, based on correspondence from Stenzel, who apparently changed his mind prior to publication in 1938 (this and most of the discussion of these specimens are based on the Kellogg-Stenzel correspondence preserved in the Smithsonian Institution Archives). Renick and Stenzel (1931:98) referred to this material, to the locality, and to Kellogg’s communications, but only in connection with the cetacean identifica¬ tion, reflecting the fact that only the two archae¬ ocete caudal vertebrae had been studied by Kel¬ logg at that time. These two specimens had been freed by weathering, whereas the 17 associated vertebrae were still embedded in a limonitic ma¬ trix. Ball’s (1931:110) mention of “a number of vertebrae of Zeuglodon ” from Burleson County undoubtedly applies to these specimens. In 1935 Kellogg had some of these vertebrae prepared and decided that they were sirenian but excep¬ tionally large for the Eocene. In 1937 Kellogg (letter to Stenzel of 2 November) reiterated his conclusion that the vertebrae represented a large Eocene sirenian, based on his study that year of the Eocene material in the British Museum (Nat¬ ural History). In the same letter he referred to illustrations of some of the vertebrae, but these have not yet been found among his papers. His file on fossil sirenians includes detailed discussion and description of the specimens, in part under an unpublished cetacean name. Clearly he vacil¬ lated in his ordinal assignment of these tantaliz¬ ing but unsatisfactory specimens, but they are catalogued as archaeocete vertebrae under the number USNM 13857 and are stored with the Archaeoceti, not the Sirenia. Dr. Kellogg devoted much attention during his retirement (1962— 1969) to curation of the fossil marine mammals, and so the physical location and catalogue place¬ ment of the specimens probably reflect his defin¬ itive opinion as to affinities. In any case, these very poorly preserved specimens do indeed ap¬ pear to be archaeocete. In 1969 Voorhies reported as sirenian an iso¬ lated cheek tooth from the upper Eocene Ocala Formation in a kaolin mine east of Huber, Twiggs County, Georgia. These beds are now generally referred to the Tivola limestone of the Jacksonian Stage (Huddlestun, Marsalis, and Pickering, 1974:2-3, 2-6). Dr. Voorhies kindly lent the deeply worn tooth, UGV-41, for study, and after consid¬ erable difficulty in interpretation, we concluded that it compares most favorably to upper right first or second molars of entelodonts (including Achaenodori). Dr. Voorhies tentatively concurred in this and pointed out that at the time of his original identification there were otherwise no land mammals known from the locality and that the tooth was found in place in beds containing a rich marine fauna, but that subsequently a small terrestrial mammalian fauna, including at least one probable entelodont tooth, of Duches- nean age has been recovered from the Twiggs clay overlying the Tivola limestone at this locality (Voorhies, pers. comm., 1975). New World Records Our intent here is to discuss all of the handful of valid occurrences known to us of Eocene sea cows in the Western Hemisphere in the hope that this will serve as a stimulus to further field efforts and as a baseline for future investigations. A few occurrences of uncertain, but possibly Eocene, age are included as well. The locality numbers 4 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Table 1.—World distribution of Eocene sirenians (locality numbers correspond to numbers on maps, Figures 1 and 2) Local¬ ity no. Taxon Locality Material Formation Age Sources and remarks 1 Prorastomus sirenoides Owen, 1855 Jamaica: river bed “near Freeman’s Hall Estate, between the parishes of St. Eliza¬ beth and Trelawney” (Owen, 1855:541) skull, mandible, atlas “Ixiwcr Limestone of the Yellow Limestone formation” (Savage, 1977:346) middle Eocene, Lutetian Owen, 1855, 1875b; Savage. 1977 9 Sirenia indei Mexico Chiapas, between Tumbala and Yajalon 2 rib fragments Eocene or Oligocenc Mullerried, 1932 3 Sirenia indct Texas: Zapata County, 0.8-1.6 km above Falcon Dam on Rio Grande partial skeleton, including articulated vertebrae and ribs Cook Mountain middle Eocene, Claibornian this report 4 Sirenia indct. Alabama: Clarke County, Little Stave Creek rib fragment Gosport middle Eocene, Claibornian Arata and Jackson, 1965 5 Sirenia indet Alabama Monroe County, 3.2 km N of Claiborne rib fragment Gosport middle Eocene, Claibornian Siler, 1964 6 Sirenia indct. Alabama: Choctaw County, Puss Cuss Creek rib fragment Gosport middle Eocene, Claibornian this report 7 Sirenia indct Alabama: Geneva County, Samson rib fragment Moodys Branch or Ocala late Eocene, Jacksonian this report 8 Sirenia indct. Florida: Levy County, Gulf Hammock quarry rib fragments Avon Park middle Eocene, Claibornian Reinhart, 1976 9 Sirenia indet. Florida: Levy County, New Lebanon Dolomite pit near Lebanon rib fragments Avon Park middle Eocene, Claibornian Vernon, 1951; Reinhart, 1976 10 Protosiren species Florida: Levy County, Waccasassa River skull and mandible fragments, teeth, and postcranial fragments Avon Park middle Eocene, Claibornian this report 11 Sirenia indet Florida: Citrus County, Florida Lime Works quarry near Inglis rib fragment Inglis or Avon Park middle Eocene, Claibornian this report 12 Protosiren species Florida. Citrus County, Dunnellon Phosphate Co. pit no. 5 near Hernando two skullcaps Inglis middle Eocene, Claibornian this report 13 Protosiren species Florida: Marion County, Withlacoochec River 4A skullcap Inglis middle Eocene, Claibornian this report 14 Sirenia indet Georgia: Houston County, Clinchfield rib fragments Clinchfield Sand? late Eocene, Jacksonian this report 15 Sirenia indet. Georgia: Washington County, Sandersvilie rib fragments, vertebral fragment Tobacco Road Sand, Sandersvillc Limestone Member late Eocene, Jacksonian Flower and Garson, 1884; this report 16 Sirenia indct. South Carolina: Dorchester County, Giant Portland Cement Co. quarry near l larleyville rib fragments Santee middle Eocene, Claibornian this report; also a skullcap, now lost (Sanders, 1974) 17 Sirenia indet. South Carolina: Berkeley County, Mazyck plantation rib fragments Santee? greensand? Eocene? this report 18 Protosiren species North Carolina New Hanover County, Martin Marietta Co Castle Hayne quarry skullcap Castle Hayne middle Eocene, Claibornian this report 19 Sirenia indct. North Carolina: New Hanover County, Ideal Cement Co. quarry rib fragments Castle Hayne middle Eocene, Claibornian this report 20 Sirenia indct North Carolina: Pender County, 6.4 km SE of Maple Hill rib fragment Castle Hayne middle Eocene, Claibornian this report 21 Protosiren species North Carolina: Jones County, North Carolina Lime Co. Comfort quarry mandible, teeth, and postcranial material Castle Hayne middle Eocene, Claibornian this report 22 Sirenia indct North Carolina: Craven County, near New Bern rib fragments Castle Hayne? middle Eocene? Emmons, 1858 (reprinted 1969), I860 NUMBER 52 5 Table 1.—Continued Local¬ ity no. Taxon Locality Material Formation Age Sources and remarks 23 " Halit her ium sp." Spain: Ebro basin vertebrae and ribs middle-late Eocene; Lutetian, ?Ludian, Barlonian Bataller, 1956; Crusafont-Pairo, 1973 24 ■ J Protosiren minima (Desmarcst, 1822); Protosiren species; Eot hern ides species France: Gironde estuary N of Bordeaux skullcap, teeth, ribs middle-late Eocene; upper Lutetian, upper Bartonian, lower Ludian Sicken berg, 1934; Richard, 1946 25 Eot hero ides species France: Basses-Alpes, Taulannc skulls, mandibles, teeth, postcranial material middle-late Eocene; upper Lutetian- Priabonian Frcudcnthal, 1970, pers comm., 1977; Heal. 1973; Savage, 1977 26 Protuthenum reran ease de Zigno, 1875 Italy: Veneto; Monte Zucllo and other Iocs. skulls, mandibles, teeth, postcranial material late Eocene; middle Auversian, Priabonian De Zigno, 1875, 1880, 1881, and others; Sickenberg, 1934, Piccoli, 1966; Bartolomei, 1969; Savage, 1977; Bi/.zarini ct al., 1977 27 Sirenia indet. Hungary Dudar rib and vertebra fragments early Eocene, Ypresian Kretzoi, 1953 ■.’a Sirenavus hungancus Kretzoi, 1941 Hungary: Felsogalla partial skull, mandible, Ma fragment middle Eocene, Lutetian Kretzoi, 1941 29 A ittsustren pannontea Kordos, 1979 Hungary: Vertcs Mountains left maxilla with P 5 -M :i and isolated P middle Eocene, Lutetian Kordos, 1979 no Protosiren cf .fraasi Abel, 1907 Hungary. Felsogalla mandible fragment with Ma middle Eocene Kordos, 1978 31 Paralitfienum larkanyense Kordos, 1977 Hungary Felsotarkany mandibles, M 3 , vertebrae, ribs late Eocene Kordos, 1977 32 Eolheroides species Hungary Balinka coal mine mandible and rib fragments Eocene Kordos, 1980 33 cf. Eolheroides , “ J Halitherium" Romania: Transylvania (Siebenburgen) skullcap, humerus, rib and vertebra fragments middle-late Eocene; Lutetian, Priabonian Sickenberg, 1934, Grigorescu, 1967; Fuchs, 1970, 1973 34 Sirenia, new genus and species Libya: Bu el Haderait skulls, mandibles, postcranial material middle Eocene. Lutetian Savage, 1971, 1977; Savage and While, 1965; Heal, 1973 35 Sirenia indet Libya: Dor el Talha ribs late Eocene Savage, 1969, 1971, 1977; Heal, 1973 36 Protosiren fraasi Abel, 1907; Eolheroides acgyptiacum (Owen, 1875a); E. abeli (Sickenberg, 1934); " Eolherium ” majus Zdansky, 1938 Egypt Gebel el-Mokattam skulls, mandibles, teeth, postcranial material Mokattam (= Lower Mokattam) middle-late Eocene, upper Lutetian-lowcr Bartonian Sickenberg, 1934; Zdansky, 1938; Said, 1962, 1963, 1965; Savage, 1977 37 Eolheroides hbycum (Andrews, 1902); “ Eolherium" slromen Sicken berg, 1934 Egypt Fayum skulls, mandibles, teeth, postcranial material Qasr el-Sagha late Eocene; upper Bartonian Sickenberg, 1934; Reinhart, 1959; Said, 1962, 1963, 1965; Simons, 1968; Savage, 1977 38 Sirenia indei. Somalia: 25 km SE of Berbera ribs Lower Daban Series middle Eocene; Lutetian Macfadyen, 1952; Savage and Tewari, 1977 39 Sirenia indet. Somalia: 20 km SW of Callis rib Carcar Series middle Eocene Savage and Tewari, 1977 40 cf. Prolothenum Somalia: Mogadishu teeth Carcar Scries middle Eocene Savage, 1969, 1977; Savage and Tewari, 1977 41 /shalhenum •mbathuensis Sahni and Kumar, 1980 India: Simla Hills, Subathu teeth, vertebra, bone fragments Subathu early Eocene, Ypresian (probably middle Eocene: see “Old World Records") Sahni and Kumar, 1980; Sahni, Kumar, and Tiwari, 1980; sirenian identity doubtful (see “Old World Records”) 42 Protosiren fraasi Abel, 1907 India: Kutch, Harudi, Matanomadh fragments of innominate, vertebra Babia Stage middle Eocene, Lutetian Sahni and Mishra, 1975; Savage and Tewari, 1977 43 Sirenia indet. Java: Nanggulan rib fragment late Eocene von Koenigswald. 1952 6 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Figure 1.—Occurrences of North American Eocene sirenians. (Numbers on map (3-22) corre¬ spond to numbered localities listed in Table 1 and discussed in text (localities 1 and 2 are plotted on Figure 2); minimum transgression of Eocene seas (after Cook and Bally, 1975:251) indicated by line with stippling on landward side.) NUMBER 52 7 Figure 2.—Worldwide occurrences of Eocene sirenians. (Numbers on map (1-43) correspond to numbered localities listed in Table 1 (both New and Old World) and to those localities (New World only) discussed in text.) correspond to those in Table 1 and Figures 1 and 2. Localities 10 and 21, from which the most extensive new materials are known, are noted briefly here and discussed at greater length under separate headings. Locality 1.—The holotype and only known specimen of Prorastomus sirenoides Owen, 1855, con¬ sists of a skull, mandible, and atlas vertebra, BM(NH) 44897, preserved in a calcareous nodule found in a river bed in west-central Jamaica, “near Freeman’s Hall Estate, between the Par¬ ishes of St. Elizabeth and Trelawney” (Owen, 1855:541). Owen, dissatisfied by the lack of atten¬ tion given this unusual specimen in the literature, prepared and described it further (1875b). Only recently, however, has the mandible been sepa¬ rated from the skull, and the specimen fully prepared for study (Savage, 1977:346, 347; pers. comm., 1975-1979). According to Owen (1855:541) the river course in which the specimen was found “is composed of red conglomerate and sandstone, overlaid by limestone, differing from the general tertiary car¬ ious limestone of the Island and beneath it.” According to Reinhart (1976:265) the specimen “was collected in the 1400 ft. thick Richmond Formation, from which invertebrate studies . . . are lacking,” but according to Savage (1977:346), “it came from the Lower Limestone of the Yellow Limestone formation, which on foraminiferal ev¬ idence is of Lutetian (Middle Eocene) age.” Locality 2.—Miillerried (1932) reported two fragmentary sirenian ribs from a locality just south of the Rio Hidalgo, about halfway between Tumbala and Yajalon, Chiapas, Mexico. The beds from which the specimens came were said to contain Eocene pectens and foraminifers, but the sirenian was thought nevertheless to pertain to the Oligocene. Maldonado-Koerdell (1953:146) reported the second occurrence of a fossil sirenian in Mexico, a rib fragment from middle or upper Oligocene rocks of the Palenque region, Chiapas, and noted that the geology of northern Chiapas was poorly known in 1932 and that Miillerried himself seemed uncertain about the age ascribed to his specimens. J.W. Durham (pers. comm.) regards the source beds of Maldonado-Koerdell’s specimen as Eocene. The primary importance of these fragments is in demonstrating the presence of sea cows along 8 the shores of the Middle American seaway when the Caribbean and Pacific undoubtedly were di¬ rectly and broadly confluent across southern Mexico (Weyl, 1973) or at least across southern Central America (Woodring, 1966). Well-pre¬ served material of known geologic age from the Eocene or Oligocene of this region would be of great value in clarifying the early history of the Sirenia not only in the Caribbean but also in the North Pacific, where Paleogene sirenians are so far unknown (Domning, 1978). Locality 3.—Sometime in or near 1950, Mr. Glen Evans collected a partial sirenian skeleton from the middle of the Cook Mountain Forma¬ tion (Claibornian, Eocene) on the Rio Grande, 0.8-1.6 km (V 2 to 1 mile) upstream from the Falcon Dam, Falcon Village 7.5 minute quadran¬ gle, USGS, southern extremity of Zapata County, Texas. This specimen, TMM 41843-1, consists of a series of thoracic, lumbar, and caudal vertebrae and some ribs, apparently in or near articulation. Much of the specimen remains largely unpre¬ pared in a block of indurated limestone not seen by us; five vertebrae (similar to those of Eotheroides in size and shape) and a partial rib (somewhat quadratic in cross-section, 43X39 mm) have been lent for study through the courtesy of Dr. Wann Langston, Jr., who also provided the data on the specimen and permitted our reporting it (pers. comm., 1976). Although no cranial or appendicular skeletal elements were found, this specimen represents the most extensive part of an Eocene sirenian re¬ ported thus far in the Western Hemisphere. Un¬ fortunately there is as yet no basis on which to identify a specimen of this sort more precisely, but its discovery does indicate that associated material of good quality may be expected in North America. Locality 4.—Arata and Jackson (1965) re¬ ported the proximal half of a sirenian rib collected by Dr. Harold Vokes in 1964 from the base of the Gosport Formation (middle Eocene, Claibornian) on Little Stave Creek, approximately 6.4 km (4 miles) north of Jackson, Clarke County, Ala¬ bama, where the section was illustrated and de¬ SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY scribed by Rainwater (1955:455-459). The spec¬ imen (TU 1004), in the Tulane University Ver¬ tebrate Paleontology collection, has been made available to us for study through the cooperation of Drs. Arata and R.H. Reinhart. This rib is oval in cross-section, 31X22 mm. There is also a rib fragment, said to be Eocene and from Clarke County, Alabama, in the British Museum (BM(NH) 35644). Locality 5.—Siler (1964) reported a sirenian rib fragment from the Gosport Formation in the NE 1/4, SW 1/4, Sec. 24, T. 7 N, R. 5 E, Monroe County, Alabama. This locality is some 35 km (22 miles) due east of locality 4, above, and approximately 3.2 km (2 miles) north of the town of Claiborne. The specimen (TMM 40628-1) has not been examined by us. Locality 6.—Mr. David C. Mason collected several partial sirenian ribs from a gravel bar in Puss Cuss Creek, SW 1/4 of Sec. 36, T. 10 N, R. 5 W, Choctaw County, Alabama. Two of these ribs have been donated to the Smithsonian Insti¬ tution and are catalogued as USNM 244363. These sirenian fossils are almost certainly derived from the Gosport Sand of middle Eocene (Clai¬ bornian) age, as this is the only formation exposed along Puss Cuss Creek (Dr. Wayne F. Canis, 1978, in litt.). Locality 7.—Early in 1979 the USNM re¬ ceived from Rick Coffey, David Cramer, and Steve Heaton a small suite of fossils collected by them along a tributary of the Pea River, near Samson, Geneva County, southeastern Alabama. In addition to the characteristic Eocene inverte¬ brates, Chlamys deshayesi and Periarchus lyelli, the collection included the major part, lacking only the head, of a well-preserved sirenian rib. This specimen, now USNM 321931, is remarkable for its robust, somewhat flattened, banana-like form, tapering abruptly at its sternal end to an ir¬ regularly rounded termination some 13 mm in diameter for articulation with the costal cartilage. It is distinctly quadrangular in cross-section (maximum diameters 77.7X49.6 mm). The only suitable source beds for the specimen in this area are the Jacksonian Moodys Branch NUMBER 52 9 or Ocala Formations, undifferentiated in Geneva County because of deep weathering (Scott, 1966). Locality 8.—Reinhart (1976:262) reported sirenian rib fragments from the Gulf Hammock quarry of the Dixie Lime Products Company, a dolomite pit in the Avon Park Limestone (middle Eocene, Claibornian), 1.6 km (1 mile) southeast of Gulf Hammock, Levy County, Florida, in the SE 1/4, Sec. 21 and the NE 1/4, Sec. 28, T. 14 S, R. 16 E, Lebanon Station 7.5 minute quadrangle, USGS. Reinhart’s (1971:7) mention of middle Eocene sirenians from Florida was based on the rib fragments from this and the following locality. In addition to numerous ostracods and other microfossils, fossil plant material is abundant in the same strata from which the sirenian ribs were derived. Preliminary investigation of these fossil plants by Dixon (1972) indicates that they are closely related to, if not congeneric with, the modern marine angiosperm Thalassia (Hydro- charitaceae). The co-occurrence of sea cows and seagrass in the Avon Park Limestone is notewor¬ thy and is discussed in more detail under “Paleoecology of Eocene Sirenians and Sea- grasses.” Locality 9.—Vernon (1951:110; repeated by Puri and Vernon, 1959:41) noted a “Manatee” rib in the Avon Park Formation at the New Lebanon dolomite pit in the SW 1/4, NE 1/4, Sec. 12, T. 16 S, R. 16 E, of the Yankeetown 7.5 minute quadrangle, USGS. This abandoned pit lies less than 1.6 km (1 mile) west of the town of Lebanon (not Lebanon Station). Reinhart (1976:262, 264, 265; fig. 26) reported additional sirenian ribs from this locality and illustrated two of them (UF 4503). Locality 10.—Over a period of several years, beginning in 1968, Robert Armistead accumu¬ lated a large collection of vertebrate fossils from the bottom of the Waccasassa River in west-cen¬ tral Florida. Armistead’s collection, now depos¬ ited in the Florida State Museum, was collected along an 8 km segment of the Waccasassa River north of the town of Gulf Hammock, Levy County, Florida. The exact location of this por¬ tion of the Waccasassa River is as follows: Secs. 5, 8, and 17, T. 14 S, R. 16 E; Secs. 20, 29, and 32, T. 13 S, R. 16 E, Bronson SW 7.5 minute quadrangle, USGS. A second collection of fossils from this same part of the river was made in 1971 by John Waldrop, Michael Frazier, and several of Waldrop’s students. These specimens are now housed at the Timberlane Research Organization (TRO), Lake Wales, Florida. At first glance these two collections appear to comprise a typical Rancholabrean (late Pleisto¬ cene) fauna, a very common vertebrate fossil assemblage in Florida’s fossil-rich rivers. The Waccasassa River assemblage includes mastodon, mammoth, horse, tapir, peccary, camel, bison, deer, several varieties of large edentates, alligator, an unidentified crocodilian, and a large sample of sirenian material. A few fossils indicative of a late Hemphillian (late Miocene to early Pliocene) mammalian assemblage are also present. The most common sirenian element is the dense, thick¬ ened skullcap, of which some 30 have been re¬ covered. Ribs and vertebrae are also common. In addition, four isolated teeth, three mandibular fragments, and several cranial fragments are among the identifiable sirenian remains. Com¬ parisons reveal, however, that most of the sirenian fossils are decidedly primitive in structure and quite unlike corresponding elements in the man¬ atee Trichechus manatus, the only sirenian known from Florida’s Pleistocene and Recent faunas. They are equally unlike Metaxytherium, a dugongid common in Florida’s Hemphillian faunas. Fur¬ thermore, the preservation of these sirenian fossils is very different from that of the late Pleistocene and Hemphillian faunal elements, and one of the skullcaps, a vertebra, and several ribs are either enclosed in limestone or have limestone matrix adhering to them. Mixture of fossils of widely different ages is characteristic of Florida’s river- bottom assemblages, and therefore the possibility that the sirenian fossils were derived from an older rock unit was suspected. The geologic map compiled from Vernon’s (1951) study of the geology of Citrus and Levy Counties, Florida, shows that Citrus and Levy counties are underlain almost entirely by carbon- 10 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY ate rocks of middle and late Eocene age. In particular, the entire segment of the Waccasassa River in which the vertebrate fossils in question were collected flows on carbonate bedrock of the Avon Park Limestone of middle Eocene (Claibor- nian) age. Field observations by Morgan confirm that the Waccasassa River in the area of concern does in certain places flow directly on the Avon Park Limestone. Also, numerous outcrops of the Avon Park Limestone up to 1 m in thickness occur along the banks of the river (Vernon, 1951). The presence of limestone matrix on several of the sirenian fossils, identical in lithology to that of the Avon Park Limestone on which and through which the Waccasassa River flows, con¬ firms that these fossils are in fact eroding from the underlying limestone and therefore are Eocene rather than Mio-Pliocene or late Pleisto¬ cene. The existence of the Eocene sirenians from the Waccasassa River has been alluded to previ¬ ously by Domning (1974). On the basis of mollusks (Richards and Palmer, 1953) and Foraminifera (Applin and Applin, 1944; Applin and Jordan, 1945), the Avon Park Limestone has been correlated with formations of middle Eocene (Claibornian) age. The limestone matrix adhering to one of the sirenian fossils contained fragments of the distinctive scutellid echinoid Periarchus lyelli. Included in Waldrop’s vertebrate fossil collection from the Waccasassa River were fragments of Periarchus lyelli and three complete specimens of the echinoid Cassidulus globosus, both of which are characteristic of the Inglis Formation, which overlies the Avon Park Limestone in Citrus and Levy counties (Fischer, 1951). Although the Inglis Formation has been considered to be late Eocene (Jacksonian) by most workers, Richards and Palmer (1953) and Meeder (1976) noted the strong affinities of Inglis mol¬ lusks with those of Claibornian age, as well as with mollusks of Jacksonian age. According to these two papers, both the Avon Park and Inglis faunas are Tethyan in affinity and notably dif¬ ferent from the faunas of the overlying Williston and Crystal River Formations, of late Eocene age. Meeder (1976) suggested that on the basis of worldwide faunal affinities, the Inglis Formation probably should be considered late Claibornian rather than early Jacksonian. The upper part of the Avon Park Limestone and the Inglis Formation, as exposed in Citrus and Levy counties, are very similar in lithology and, in fact, were originally described as a single rock unit, the Gulf Hammock Formation, by Ericson (1945). According to Cole and Applin (1964:14), As the major break in foraminiferal faunas occurs at the top of the Inglis, and as we cannot discover any criteria in continuous sections of well samples whereby the Inglis could be separated from the Avon Park, in the authors’ opinion, the Inglis should be abandoned and these strata included in the Avon Park Limestone. Cole and Applin (1964:14) commented further that “we consider that the evidence is such that this entire section which we refer to the Avon Park Limestone is middle Eocene in age.” Hunter (1976) noted a similar identity in the megafossils of these two formations, with the major faunal change coming at the top of the Inglis Formation. Hunter assigned both the Avon Park Limestone and the Inglis Formation to the same biostrati- graphic unit (Gulf Hammock local stage) and correlated them with the Claibornian Stage of the Gulf Coast region. These two formations ap¬ parently represent a single regressive-transgres¬ sive sequence of very shallow-water limestones and dolomites (Randazzo and Saroop, 1976). The unconformity between the two formations, de¬ scribed by Vernon (1951), appears to be very local in extent and most likely represents a short period of emergence and subsequent erosion dur¬ ing the peak of the regressive phase. Although this is not the place to revise stratigraphic nomen¬ clature, it appears probable that the Avon Park Limestone and Inglis Formation represent a sin¬ gle lithologic unit of late middle Eocene age. A middle Eocene age for these two formations is compatible with the primitive nature of their sirenian remains, which bear a close resemblance to those from the Mokattam Limestone (Lutetian = Claibornian) of Egypt. From the above discussion, it is clear that the NUMBER 52 11 Waccasassa River assemblage represents an ad¬ mixture of late Hemphillian and late Pleistocene fossils, probably derived from surficial sands and clays along the banks of the river, to middle Eocene fossils eroded from the Avon Park Lime¬ stone on which the river flows. Several dentary fragments and vertebrae of an undescribed crocodilian are also represented, which, like some of the sirenian fossils described above, have ad¬ herent matrix of Avon Park Limestone. These fossils, apparently of a marine crocodilian of mid¬ dle Eocene age, are of unknown affinity and are under study at the present time. Locality 11.—A partial sirenian rib (USNM 256687) was collected by G.S. Morgan on 21 February 1978 from the base of the Inglis For¬ mation at the Florida Lime Works quarry, 2.5 km south of Inglis, Sec. 11, T. 17 S, R. 16 E, Yankeetown 7.5 minute quadrangle, USGS, Cit¬ rus County, Florida. The majority of the strati¬ graphic section at the Florida Lime Works quarry is composed of massive, porous, tan, unfossilifer- ous dolomite of the Avon Park Limestone. The top 2-3 meters of the quarry wall consist of chalky, cream-colored limestone with the typical lithology and echinoid fauna of the Inglis For¬ mation. Although there is considerable difference in lithology between the Avon Park Limestone and Inglis Formation in this quarry, there is no definite unconformity between them, suggesting that perhaps the underlying Avon Park Lime¬ stone may have been secondarily dolomitized. The sirenian rib was collected from the basalmost Inglis Formation and is extensively abraded and bored, suggesting that it may have been deposited originally in the underlying Avon Park Limestone and subsequently reworked into the base of the Inglis Formation. It is oval in cross-section, 26X15 mm. Locality 12.—In the vertebrate fossil collec¬ tions of the Florida Geological Survey (now housed at the Florida State Museum, Gainesville) are two sirenian skull caps (UF/FGS V2282, V2288) collected on 13 February 1913 by Her¬ man Gunter from the Dunnellon Phosphate Company pit no. 5, near Hernando, Citrus County, Florida. Although vertebrates of Hem¬ phillian age (late Miocene) have been collected from the phosphatic rock exposed and formerly mined in this quarry (Vernon, 1951), the sirenian skullcaps were undoubtedly derived from the In¬ glis Formation, which unconformably underlies the phosphate deposits of the Alachua Formation at the Dunnellon phosphate pit. Locality 13.—A single sirenian skullcap (UF 24786) in the Florida State Museum collections from the Withlacoochee River 4A site is the only demonstrably Eocene fossil in an otherwise strictly Hemphillian vertebrate fauna. The With¬ lacoochee River 4A site, an underwater fossil locality in the Withlacoochee River, is located about 12.9 km (8 miles) southeast of Dunnellon in the NW 1/4 of Sec. 30, T. 17 S, R. 20 E, Stokes Ferry 7.5 minute quadrangle, USGS, Marion County, Florida. The fossil vertebrates from the Withlacoochee 4A site were, for the most part, preserved in situ in a massive, green clay of late Miocene age Filling a sinkhole developed in the Inglis Formation of late middle Eocene age. Al¬ though the majority of fossils from this site were collected from the green clay, some fossils were found loose on the river bottom. The sirenian skullcap was almost certainly derived from the underlying Inglis Formation and subsequently mixed on the river bottom with the Hemphillian- age fossils. Locality 14.—In the collections of the Smith¬ sonian Institution is a single lumbar vertebra, USNM 13883, identified as Zygorhiza kochii by Remington Kellogg and received in 1935 from the quarry of the Pennsylvania-Dixie Cement Corporation at Clinchfield, Houston County, Georgia. In the tray with the cetacean vertebra was an uncatalogued fragment of a sirenian rib. Examination of the registrar’s records reveals that there were several fragments of sirenian ribs (identified by Kellogg) in the collection from Clinchfield at the time of receipt. Thus this spec¬ imen is undoubtedly one of them and is now catalogued as such under the number USNM 244031. It is oval in cross-section, 31X22 mm. Also in the collections of the Smithsonian In- 12 stitution are three fragmentary pieces of sirenian rib, two probably representing a single rib, all now catalogued under USNM 244029, found in a tray with two other fragmentary specimens, not identified as yet, but apparently not sirenian. The largest rib is subquadratic in cross-section, 41X39 mm. In the tray are notes obviously typed by Remington Kellogg, including a detailed geologic section extracted from a letter from Philip E. LaMoreaux, dated 30 July 1945, indicating that the specimens came from the Clinchfield quarries of the Pennsylvania-Dixie Cement Corporation in Houston County, Georgia, and that they were collected apparently in place in the section some 80 feet below ground level, well down in beds assigned by LaMoreaux (letter of 1945) to the Ocala Formation. Beds then assigned to the Ocala Formation in this area would now be assigned by many authors in large part to the Tivola Formation; however, the bed from which the fossils came was described as, “Gray, medium to coarse sand with scattered granules of sub- angular sand. Fossiliferous. Limy in places.” This lithology seems much like that described for the Clinchfield Sand, including that at or near the sirenian locality in the section at the Medusa Cement Company west quarry, Clinchfield (Huddlestun, Marsalis, and Pickering, 1974:2-6, 2-33, 2-34, 2-35). Dr. Paul Huddlestun (pers. comm., 1977) confirms that the Medusa quarry is one and the same as the Pennsylvania-Dixie quarry and, based on examination of La- Moreaux’s detailed section, concludes that the sirenian ribs indeed came from the Clinchfield Sand. The Clinchfield Sand, regarded as earliest Jacksonian in age, overlies the Lisbon Formation (Claibornian) and underlies the Tivola Forma¬ tion. Locality 15.—In January 1846 Sir Charles Lyell (1850:8, 9) travelled by railway handcar from Savannah to Macon, Georgia, making ob¬ servations on the geology en route, among which was the following: “Near Sandersville I saw a limestone from which Eocene shells and corals are procured, as well as the teeth of sharks and the bones of the huge extinct cetacean called SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Zeuglodon.” Flower and Garson (1884:527) listed as Halitherium under the number 2725 in the museum of the Royal College of Surgeons, “Transverse process of vertebra and eleven por¬ tions of ribs. From Sanderville [sic], Georgia, U.S.A. Presented by Sir Charles Lyell, 1868” and (p. 547) under number 2838, “Body of a dorsal vertebra, probably of a Zeuglodon. From Sander¬ ville [sic], Georgia, U.S. America. Presented by Sir Charles Lyell, 1868.” Unfortunately these specimens apparently were among the many lost in the destruction of that museum during World War II (E. Allan, pers. comm., 1981). There is, however, neither reason to doubt that specimens listed as sirenian by W.H. Flower were correctly identified, nor that rocks ascribed to the Eocene by Lyell were correctly assigned. Further, Sandersville, Washington County, east-central Georgia, later became the type-locality of the Sandersville Limestone Member, first of the Barnwell Formation (Cooke, 1943:61-63, 65), then of the Tobacco Road Sand of the Barnwell Group (Huddlestun and Hetrick, 1979:27, 46, 48). The Sandersville Limestone was deposited during late Eocene, Jacksonian time. Locality 16.—Sanders (1974:8) reported a sir¬ enian skullcap from a quarry of the Giant Port¬ land Cement Company, 3.2 km (2 miles) NNE of Harleyville, Dorchester County, South Carolina. The specimen was found on the surface of the Santee Formation after the overlying Cooper Marl had been scraped away in preparation for mining; thus the exact stratigraphic provenience of the specimen is unknown. Further, the speci¬ men unfortunately is not now to be found in the collections of the Charleston Museum. Nevertheless, the same field work produced another sirenian record, of known horizon, from this locality. According to Albert E. Sanders, the specimen, ChM GPV1306, consists of several fragments apparently of a single rib, found in place in the Santee Formation some 6 meters below the Cooper Marl. This rib is quadratic in cross-section, 46X40 mm. Ward et al. (1979) have recently studied the stratigraphy of rocks exposed in this quarry, NUMBER 52 13 where the Santee Formation is represented by their (upper) Cross Member, regarded by them as middle Eocene (Claibornian), and directly overlain by their (lower) Harleyville Member of the Cooper Formation, regarded as late Eocene (Jacksonian) in age; however, Baum et al. (1980) have raised the Cross Member to formational rank, and suggested a late Eocene age. Locality 17.—In the Museum of Comparative Zoology are three sirenian rib fragments probably of Eocene age. All are indicated as coming from greensand and from the Santee Canal, South Carolina. One of these, MCZ 8634, is attributed to R. Mazycks; the other two, MCZ 8767, to R.W. Gibbes. Gibbes (1845:254), in describing the archaeocete Dorudon serratus, states that the specimens (now MCZ 8763) were found in March last, in a bed of Green sand near the Santee Canal, in South Carolina. The locality is on the plantation of R.W. Mazyck, Esq., about three miles from the entrance of the canal from the head waters of Cooper river. The deposite of Green sand is from four to eight feet thick near the surface, lying on a solid yellowish limestone .... Thus there is every reason to suppose that the sirenian rib fragments were obtained at the same place and time and are of the same age as Dorudon serratus. If so, then the sirenian specimens are actually from marl diggings, as is D. serratus , not from the canal itself. The Santee Canal, long since abandoned, ap¬ parently was excavated shortly after 1800, as the first (Pleistocene) vertebrate fossils were recorded from it in 1802 (see Hay, 1923:119, 156, 162, for a discussion of these occurrences). The canal ex¬ tended from the head of the West Branch of the Cooper River (approximately 2.4 km (1.5 miles) ENE of Moncks (or Monks) Corner, Berkeley County) NNW for some 32.2 km (20 miles) to the Santee River, just downstream from Lake Mar¬ ion. Most of its length is mapped on the Chicora 15 minute quadrangle of the USGS, and its south¬ ern extremity on the Cordesville 7.5 minute quad¬ rangle. With the assistance of C. Wythe Cooke, Kel¬ logg (1936:178) determined that the type-locality of Dorudon serratus lay five and three-fourths miles (9.3 km) north of Moncks Corner and one mile (1.6 km) west of MacBeth. This location is now beneath Lake Moultrie. Kellogg (1936) regarded the source bed of D. serratus to be the Santee Formation, but this may be open to some question. As regards the probably associated sirenian rib fragments, their lack of diagnostic morphology, their recovery so long ago from marl diggings, and the opportunity for de¬ rivation from younger deposits, together dictate caution in age assignment. Locality 18.—On 11 April 1981 Vincent Schneider collected a sirenian skullcap from spoil of the Castle Hayne Formation at the Martin Marietta Company’s Castle Hayne quarry, in the northern corner of New Hanover County, North Carolina, approximately 3.3 km (2.1 miles) ENE of Castle Hayne, and lying in the NW corner of the Scotts Hill 7.5 minute quadrangle, USGS. This quarry was designated the type-locality of the Castle Hayne Limestone of Baum et al. (1978), and the type-locality of the New Hanover Member and a reference locality of the Comfort Member of the Castle Hayne Formation of Ward et al. (1978). The specimen is described below, under “North Carolina Eocene Sirenians.” Locality 19.—In September 1977 a Smithson¬ ian field party including Ralph Eshelman and Peter J. Harmatuk visited the quarry of the Ideal Cement Company located in the northern ex¬ tremity of New Hanover County, North Carolina, immediately south of the junction of Island Creek with the North East Cape Fear River. The quarry lies at approximately 34°22'30"N, 77 o 50'00"W, straddling the boundary between the Mooretown (on the north) and Scotts Hill (on the south) 7.5 minute quadrangles, USGS. This quarry has been designated the type-locality of the Castle Hayne Formation by Ward et al. (1978) and reference locality of their New Hanover and Comfort Mem¬ bers of that formation. Here they were shown a group of nine fragmentary sirenian ribs and given one of the fragments by Mr. Al Simons. These associated specimens apparently were collected in 14 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY place in the Castle Hayne Formation. The bone is compact throughout, hard, and stained uni¬ formly black except for a softer, gray superficial layer approximately one mm thick. Some frag¬ ments are quadratic in cross-section, and the maximum shaft diameters are 48X41 mm. Mr. Simons has subsequently presented all of the specimens through Mr. Harmatuk to the Smith¬ sonian, where they are catalogued under USNM 256688. Locality 20.—In July 1975 Paul Drez col¬ lected a partial sirenian rib in an abandoned quarry approximately 6.4 km (4 miles) southeast of Maple Hill, Pender County, North Carolina (Maple Hill 15 minute quadrangle, USGS). The specimen, donated to the Smithsonian Institution and now catalogued as USNM 244030, is em¬ bedded in a lump of limestone of the Castle Hayne Formation (Claibornian). It is subquad¬ ratic at its larger end, 44X43 mm thick. Locality 21.—For the past several years, at least since 1970, Peter J. Harmatuk has been prospecting for and collecting fossils in a limerock quarry developed in the Castle Hayne Formation near Comfort, Jones County, North Carolina. This quarry, operated until recently by the North Carolina Lime Company under lease from the land owner, lies a few hundred meters north of State Road 41, approximately 6.4 km (4 miles) by road west of the post office at Comfort, and immediately west of Tuckahoe Church, at ap¬ proximately 35°01'N, and 77°34.5 / W, on the Comfort 7.5 minute quadrangle, USGS. Access to the quarry has been granted by Messrs. A.C. Palmer and Harry E. Jones of the North Carolina Lime Company and by Mr. Linsey Vance Ma- ness, owner of the land. As is usually the case in commercial quarries, most of the vertebrate remains have been re¬ covered through diligent combing (by Mr. Har¬ matuk and a few others) of overburden and other residue of mining. In contrast to the macroinver¬ tebrates, vertebrate remains are not sufficiently common to be found often in place or complete under mining conditions. Further, each find of a vertebrate fossil usually represents but a single small part of the total complement of hard parts of the animal, whereas in many situations essen¬ tially whole skeletons of marine macroinverte¬ brates are found in abundance. Hence the neces¬ sity for vertebrate paleontologists to content themselves frequently with float and with frag¬ ments and to revisit given sites time and again if they are to accumulate meaningful samples rep¬ resenting even part of the major elements of the skeleton and dentition. Fortunately, at least some of the vertebrate remains at the Comfort quarry have been found essentially in place. Most important among these are various elements, mostly fragmentary, of what must have been an associated, and at least in part articulated, skeleton of a sea cow. These elements have been recovered by Mr. Harmatuk in the course of numerous visits over a period of several years from a cut bank near the north side of the quarry, through assiduous inspection after rains and through screening, yielding some fragments of teeth, for example, no more than a few mm in maximum dimension. The best specimen re¬ covered to date, a partial mandible, has been reconstituted from many tens of fragments col¬ lected in just this manner. Much of the appar¬ ently associated skeleton undoubtedly was lost to the bulldozer and dragline. The section above pond level at this place, excluding irregular hillocks of overburden piled on the original ground surface, is as follows: Thickness Unit in meters 1. Gray-green, sticky, glauconitic, sandy 1-1.3 clay 2. Tan-yellow, oxidized sand 3 3. Buff-white, shell marl (definitely Castle unknown Hayne Formation) The sea-cow skeleton occurred in the upper¬ most unit. The surface of the bones is off-white, with some superficial dark organic staining. The chalky-white weathered layer typically extends into the bone some 2 mm, beyond which it is replaced abruptly (at least not obviously at any osteologically controlled point such as the demar¬ cation between outer lamellar bone and inner NUMBER 52 15 vesicular bone or marrow cavity) by a dark-brown to gray, hard, mineralized core. The bones are not waterworn or rolled, and retain sharp surface detail in spite of the extremely soft surface that is pasty when wet. There is no indication of boring or encrustation by bottom dwelling invertebrates. Very delicate teeth retain sharp surface detail. The apparent articulation (largely disturbed, it seems, only in quarrying), preservation of deli¬ cate structures, lack of boring and encrustation, and deep weathering combined with retention of surface detail all point toward primary deposi¬ tion, not reworking. The clearly Eocene aspect of the sirenian skeleton thus would militate against the possibility that the enclosing clay (and its underlying sand) is a much younger unit than the underlying marl and merely contains the reworked debris from earlier episodes of deposi¬ tion. Thus, in spite of the contrasting lithology, it would seem that these superficial elastics are a part, possibly condensed in place through ex¬ treme weathering and leaching, of the same dep- ositional cycle as the underlying marl. This quarry is the type-locality for the Comfort Member of the Castle Hayne Formation as re¬ defined by Ward et al. (1978:F8, 20, 21), who regarded our uppermost unit in which the siren¬ ian skeleton occurred as (?) Pleistocene. Thus, we are left with somewhat of a dilemma to be re¬ solved only by additional fossils and/or future stratigraphic studies. The lithology of the beds from which the sea cow came is anomalous for the Castle Hayne Formation, but the sea cow seems inescapably Eocene in character. This section is very similar to that exposed in bluffs of the Trent River approximately 1.6-2.4 km (1-1.5 miles) to the northeast, described by Kellum (1926:10) as USGS locality 10630, three- quarters of a mile NW (not SW as indicated by Kellum) of Comfort on the farm of Miss Sally Simmons (now Eagle’s Nest Farm). Kellum, ap¬ parently correctly, assigned all of this section (excluding only a surficial sand), from the sticky clay downward, to the Castle Hayne Formation. Although the Castle Hayne Formation has been assigned by many authors to the Jacksonian (upper Eocene) stage, the most recent students of the problem have reviewed the evidence and history of study, conducted their own field work, and assigned the Formation to the Claibornian (middle Eocene) stage (Ward et al., 1978; Baum et al., 1978). Although the sirenian collection is far too meager to be definitive, its similarity to Egyptian material from beds of the Lutetian stage is more compatible with the Claibornian assign¬ ment. Isolated sirenian elements (other than those apparently pertaining to the associated skeleton) recovered from the spoil of the Comfort quarry include specimens of somewhat different preser¬ vation, some of which are of yellowish tinge sug¬ gesting derivation from the oxidized sand. No sirenian specimens have been recovered that were definitely preserved in the typical marl. Virtually all specimens reveal the peculiar preservation, with a thin, whitish, chalky outer layer and a thick, dark, hard inner core. Other vertebrates from the quarry include ar- chaeocetes and bony and cartilaginous fishes, vir¬ tually all collected by Mr. Harmatuk. The ar- chaeocete remains consist mostly of isolated teeth (some 20-30 in number), thought to be referable to Zygorhiza kochii and under study by Walter Wheeler. Half of a large tooth was collected by Mrs. Gene Mapes and very kindly donated to the Smithsonian Institution, where it has been cata¬ logued under USNM 244043. Some archaeocete cranial fragments were discovered by Mr. Har¬ matuk and by Yoshikazu Hasegawa on the floor of the quarry in place in the typical Castle Hayne marl, at least 7.6 meters below ground level. A large suite of fish remains was studied by Robert L. Meyer, then a postdoctoral fellow (1975-1976) with the Paleontology and Stratig¬ raphy Branch of the USGS, who has provided the following report, quoted in part here with permission from him and the USGS. Preliminary faunal list: Selachii: Pristis curvidens Leidy—14 rostral spines. Rhinoptera sp. 1—worn medial tooth with pronounced 16 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY anterior shelf, narrow root; does not differ markedly from the extant R. bonasus. Rhinoptera sp. 2—worn medial tooth lacking anterior shelf, wide root with very narrow grooves. Similar to R. sherborni White, but has narrower grooves. Rhinoptera or Myliobatis —worn medial or first lateral teeth (two); broad root with broad grooves, moder¬ ately developed anterior shelf. Myliobatis sp.—two teeth with broad, thick root with wide grooves; possibly upper teeth of Rhinoptera or Myliobatis referred to above. Dasyatidae—incomplete caudal spines. Procarcharodon angustidens (Agassiz)—12 teeth; form very similar to variety called Sokolowi Jaekel by Leriche (1942). Isurus americanus (Leriche)—primitive isurid in the /. deson lineage; placed in the genus Macrorhizodus by Glikman, 1964; species as yet not securely diagnosed, may be I. praecursor (Leriche). Anomolodon nova (Winkler) — (new combination) three teeth. Stnatolamia macrota —26 teeth. Odontaspis sp.—four incomplete anterior teeth. Odontaspis ^?) koerti (Stromer) as defined by White (1934)—seven teeth. “Eugaleus” denticulata ? Gibbes—two teeth lacking roots; generic reference very uncertain. Osteichthyes: Xiphiorhynchus antiquus Leidy—incomplete rostral com¬ plex. Istiophoridae, genus uncertain—incomplete rostral complex. Cylindracanthus rectus —incomplete rostral spines. Carangidae gen. and sp. indet.—incomplete premaxil¬ lary bone; also vaguely similar to Pomatomidae and primitive Scombridae without produced snouts. Indicated Age: Large Striatolamia macrota are indicative of upper Eocene; similar sized teeth are found in the Bartonian of England (based on comparison with USNM collec¬ tion). Odontaspis (?) koerti and Anomolodon nova have imprecisely known range zones but are most charac¬ teristic of middle Eocene. The Procarcharodon and Isurus are like those from the Jackson Group of Mississippi and Alabama reported by Leriche (1942). These fos¬ sils are commensurate with a Bartonian age (Lower Upper Eocene). Environmental indications: Relative abundance of pristids indicates rather shallow, warm water; the rest of the constituents of the fauna are widely ranging; nothing indicative of cool or deep water is present. Matrix samples from both the upper clay at the sirenian skeletal site and the immediately underlying sand have been examined for fora- minifers by Thomas G. Gibson and for silicofla- gellates by Laurel Bybell, with negative results. Norman Frederiksen analyzed a sample for pollen and spores which yielded inconclusive results but suggested an unexpectedly high level of modern contamination. The sirenian material from Comfort is enu¬ merated and described in detail under “North Carolina Eocene Sirenians.” Locality 22.—The only published report known to us of a fossil from North Carolina pertaining to an Eocene sirenian is that by Em¬ mons (1858:212, fig. 34, reprinted 1969; figure only, reproduced by Emmons, 1860:213, fig. 181). This report is repeated here in its entirety and the figure reproduced (Figure 3). The oldest specimen of fossil belonging to the whale or cetacean family, belongs to the genus Physeter, and is re¬ garded as the P. antiquus, (fig. 34). It occurs in the eocene of Craven county. The size of the teeth prove that they belonged to the largest of the class. The largest tooth mea¬ sures six inches in circumference, and is five and a half inches long, though a portion has been broken from the base. Its form is quadrangular, and presents a curve in front, but is rather straight behind. It shows no conical cavity, but is solid throughout. It shows a tendency to exfoliate con¬ centrically. Many fragments more or less rolled and other¬ wise defaced, have been seen in the miocene beds upon the Tar River.—It is probable they may have been removed from a lower to an upper formation. Meaningful comment can scarcely be made on the rolled specimens from the Tar River or on the bulk of those said to be from the Eocene of Craven County; however, the appearance of the one il¬ lustrated and characteristics such as large size, “quadrangular” form, “curve in front,” “straight behind,” “no concavity,” “solid throughout,” and “tendency to exfoliate concentrically” all seemed equally or more applicable to the distal portion of a rib of a sea cow, rather than to a tooth of a cetacean. Fortunately the illustrated specimen (subsequently broken and in part missing) was preserved in the Emmons collection of Williams College, Williamstown, Massachusetts, and has now been transferred permanently to the Smith- NUMBER 52 17 sonian Institution (USNM 329063). It proves in deed to be the distal part of a sirenian rib, and thus, if its Eocene age is correct, as seems likely, it constitutes the first record for the Eocene of North Carolina. Its greatest diameters are 35X26 mm. In discussing the Eocene series, Emmons (1858:102-107; 1860:211) gave special attention to its occurrence in the western part of Craven i J _ l _1 CM Figure 3.—Distal portion of Eocene sirenian rib from Cra¬ ven County, North Carolina (USNM 329063), sternal end up. (Illustration reproduced from Emmons, 1858:212, fig. 34.) County, on the south side of the Neuse River some 32.2 km (20 miles) above New Bern, in the vicinity of Biddle Landing, Ft. Barnwell, and Core Creek (now generally called Cove Creek), on the Ayden 15 minute quadrangle, USGS, where plantation owners (the Biddle brothers, William Wadsworth, and others) were beginning to use the Eocene marl extensively for agricultural purposes. They were at the same time in com¬ munication with Emmons, which suggests that the sirenian specimen(s) probably came from this area. Wilson (in Kier, 1980:12) felt that all ref¬ erences to the Eocene of Craven County by Em¬ mons pertain to the Wadsworth plantation, but it would seem difficult to exclude the nearby Biddle lands. Occurrences of Uncertain Age.— John T. Thurmond, then of Birmingham-Southern Col¬ lege, has sent four fragments of sirenian ribs collected by Ronald Rhoads from the lower part of the Citronelle Formation in sand and gravel pits west of Jackson, Clarke County, Alabama, along the Tombigbee River, in Sec. 8, T. 6 N, R. 2 E. These specimens are thought by Dr. Thur¬ mond to be reworked from Eocene or Oligocene deposits. A fragmentary sirenian rib was collected by Gerard Case in 1965 at “Marianna, Florida” and said to be of Eocene age, from the “Ocala lime¬ stone.” It is slender, oval in cross-section, and 32X20 mm in diameter. The specimen was do¬ nated to the Smithsonian Institution, where it is catalogued under USNM 244028. If the specimen is indeed of Eocene age, it undoubtedly came from the Crystal River Formation (Jacksonian), but in view of the widespread development of the immediately overlying, and lithologically similar, Marianna Formation (Oligocene, Vicksburgian), as well as still younger limestones, the record cannot be regarded as unequivocally Eocene in age without more precise field data. Reves (1961) provides a useful introduction to the distribution and exposures of these beds, which are potentially productive sources for fossil sirenians. 18 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Old World Records The extensive literature of Eocene sirenians in the Old World spans well over a hundred years, and it is not our purpose here to review it or to discuss Old World materials in any detail, except insofar as it is useful and feasible to compare distribution records and materials with those from North America. The earliest records of the Sirenia reported to date consist of fragments of ribs and vertebrae from the early Eocene (Ypresian) of Hungary (Kretzoi, 1953) and tooth and postcranial frag¬ ments from the early Eocene of India (Sahni and Kumar, 1980; Sahni et al., 1980). As regards the Indian records, however, it should be noted that the mammal-bearing part of the Subathu For¬ mation is more generally regarded as late middle Eocene (Gingerich et al., 1979:113; West, 1980:509), and that the fossils in question may not be sirenian (Savage, pers. comm., 1981; Gingerich and Russell, 1981:237). Savage (1977:344-346) and Savage and Tewari (1977) provide convenient summaries of middle and late Eocene distribution, to which may be added a record of Protosiren fraasi from the middle Eocene of India (Sahni and Mishra, 1975) and new rec¬ ords from Hungary (Kordos, 1977, 1978, 1979, 1980). We have mapped world distribution (Fig¬ ure 2) in order to show visually the striking Tethyan distribution pattern of the Sirenia dur¬ ing Eocene time. The documentation for mapped occurrences in the Old World seems adequately and most succinctly presented in tabular form, which we have done in Table 1, where the num¬ bered entries correspond to the numbered locali¬ ties on the maps (Figures 1 and 2). Florida Eocene Sirenians Description The Eocene sirenian fossils from the Wacca- sassa River fauna are represented solely by iso¬ lated elements and were mixed on the river bot¬ tom with Mio-Pliocene and late Pleistocene ver¬ tebrate remains, the latter including the West Indian manatee Trichechus manatus. Among the 100 or so identifiable sirenian bones in the Wac- casassa River collections, most are totally unlike corresponding elements in T. manatus , and no evidence at all has been found of Metaxythenum or other late Tertiary forms. The majority of the Waccasassa River specimens, together with those from two other Florida Eocene localities (listed below), seem to represent a single taxon of prim¬ itive sirenian. Locality 10 (Waccasassa River, Levy County, Florida) Parietal-supraoccipital skullcaps: UF 14222-14226, 16439, 16441, 16443, 16598, 16607, 21610, 21613; TRO 510,511,513-515,520-524, 529-531, 535-537,539, 540 Partial left maxillae: TRO 516, 538 Sphenopterygoid complexes: UF 16441, 21612 Upper molars: TRO 518 (left M 3 ); UF 24806 (left), 25708 (right) Partial mandibular symphyses: UF 21611; TRO 509, 529 Vertebrae: UF 25706, 25707, TRO 526, 527 Rib fragments: UF 16442, 18440, other uncatalogued fragments; TRO 541 Locality 12 (Dunnellon Phosphate Company pit no. 5, Citrus County, Florida) Parietal-supraoccipital skullcaps: UF/FGS V2282, V2288 Locality 13 (Withlacoochee 4A Site, Marion County, Flor¬ ida) Parietal-supraoccipital skullcap: UF 24786 Locality unknown Parietal-supraoccipital skullcap: UF 3965 Parietal-Supraoccipital Skullcaps (Figures 4-10).—The sirenian sample from the Avon Park Limestone and Inglis Formation consists primar¬ ily of skullcaps, 34 in all. These vary in state of preservation, completeness, and to some extent size, although size is most likely age related. Disregarding broken, waterworn, and immature specimens, the remaining skullcaps are relatively uniform in morphology. As in other sirenians, they consist of the ankylosed parietals and su- praoccipital. These were not firmly co-ossified with the frontals and exoccipitals, as most speci¬ mens lack the frontals and all lack the exoccipi¬ tals. This may indicate that most of these individ¬ uals were subadult. Two specimens (TRO 510, 521) retain the posterior portions of the frontals, including most of the frontoparietal suture. NUMBER 52 19 Table 2.—Measurements (mm) of Florida skullcaps (x = mean, OR = observed range, SD — standard deviation, CV = coefficient of variation) Variate N X OR SD CV Maximum width of lambdoidal crest 9 76.1 73.9-80.7 4.5 5.90 Maximum width of parietals an¬ terior to lambdoidal crest 10 63.0 59.1-67.4 2.7 4.31 Minimum width of parietals (just posterior to frontoparietal su¬ ture) 10 41.4 35.7-46.4 3.1 7.42 Minimum width at parietal-squa¬ mosal suture 13 46.6 42.6-48.9 2.3 4.90 Maximum thickness of supraoc¬ cipital 13 20.0 16.5-24.1 2.4 12.13 Maximum thickness of parietals at internal occipital protuber¬ ance 13 38.9 35.1-46.6 3.0 7.70 Maximum thickness of parietals at anterior margin 13 27.0 25.0-29.3 1.4 5.05 Angle at which supraoccipital meets parietals, rounded to nearest 5° 23 125° 120°-135° 5.0° 3.98 Parietals.— Rather long and narrow in gen- renians, but slight contact may have existed in eral outline, widest at lambdoidal crest (mean=76 several specimens. Pronounced indentation of pa- mm; Table 2) and narrowest just posterior to rietals present anterior to lambdoidal crest for frontoparietal suture (mean=41 mm). Most skull- reception of dorsal margin of squamosal, which caps measure approximately 100± 10 mm in total seems to have reached level of skull roof. Roof of length; noticeably more in several specimens, in cranium slightly convex transversely in most spec- particular those retaining fragile anterolateral imens but nearly flat in some. Parietals have processes of parietals. In most nearly complete narrow depression along dorsal midline just pos- specimen (TRO 510), parietals (= total length of terior to frontoparietal suture, in a few instances skullcap) 127 mm in length. Posterior margin of reaching nearly to nuchal line (TRO 531). Dorsal parietals marked by prominent lambdoidal crest, surfaces of parietals, roughly midway between diverging posterolaterally from midline, and frontals and supraoccipital, in some specimens strongly concave posteriorly. Lambdoidal crest pierced by 1 or 2 prominent emissary foramina seems formed primarily from thickened posterior on 1 or both sides, well away from midline (UF border of parietals, although supraoccipital may 21610: TRO 522, 536, 539, 540). Temporal crests form portion of ventral and lateral borders of poorly developed or absent: where weakly pres- crest. Parietals and supraoccipital so firmly co- ent, they appear to arise from anterior edge of ossified that sutural contact between them is at parietosquamosal suture and may have had their best doubtfully discernible. Location of this su- origin on dorsal processes of squamosals, none of ture below lambdoidal crest was reported by Abel which are preserved in the Florida specimens. (1912) in Eotheroides aegyptiacum. Posterolateral Temporal crests follow dorsolateral borders of prolongation of parietals as lambdoidal crest ap- skullcaps, becoming weaker on anterolateral pears to have precluded contact between squa- processes of parietals. Two specimens which re- mosals and supraoccipital observed in most si- tain small portions of frontals indicate that these NUMBER 52 21 Figure 5.—Parietal-supraoccipital skullcaps (including portions of frontals) of Protosiren species from Florida in dorsal aspect: a, TRO 521; b, TRO 510. crests were continued anteriorly by overhanging dorsolateral margins of frontals. Lateral walls of braincase meet skull roof at nearly right angles. Internal structure of parietals varies to greater extent than do external features. Braincase bounded anteriorly by ventral process of frontal, posteriorly by supraoccipital and tentorium os- seum. Triangular internal occipital protuberance slightly pointed, either broadly confluent with supraoccipital or separated from it by transverse ' '••tfc *2»- • - y-,< 24 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY A 0 3 I-!-1 ) CM Figure 8.— Parietal-supraoccipital skullcaps of Protosiren species from Florida in ventral (internal) aspect: a, TRO 531; b, UF 3965; c, UF 14222; d, UF 16607. NUMBER 52 25 Figure 9.— Parietal-supraoccipital skullcaps (including portions of frontals) of Protosiren species from Florida in ventral (internal) aspect: a, TRO 521; b, TRO 510. SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Figure 10.—Parietal-supraoccipital skullcaps of Prolosiren species from Florida in posterior aspect: a, TRO 531; b, UF 3965; c, UF 14222; d, UF 16607; e, TRO 521; f, TRO 510. NUMBER 52 27 sulcus. Cerebral hemispheres separated along midline by prominent bony falx cerebri which extends forward from internal occipital protuber¬ ance to frontals. Bony falx cerebri sharp, narrow ridge in most specimens but broader and less convex in others; lowest and broadest just abaft frontoparietal suture, where it typically shows median groove. Prominent ridge of bone (tento¬ rium osseum) located perpendicular to bony falx along posterior margin of parietals. Deep trans¬ verse sulcus lies just posterior to tentorium osseum and parallel to it, corresponding to suture be¬ tween parietals and supraoccipital. Transverse sulcus formed in most species by 2 deep lateral concavities and shallow median saddle (internal occipital crest) separating internal occipital pro¬ tuberance from supraoccipital. Transverse sulcus shallow laterally in several specimens and almost nonexistent along midline, in which case protu¬ berance is broadly confluent with supraoccipital. Bony falx, tentorium osseum, and internal occip¬ ital crest typically form strikingly cruciform pat¬ tern (cruciate eminence). Portions of parietals and supraoccipital which were in contact with brain smooth, but pierced by numerous tiny nu¬ trient foramina. Dorsal aspect of frontoparietal suture V-shaped, extending anterolaterally at ap¬ proximately 45° angles to midline. On lateral wall of braincase, frontoparietal suture slopes pos- teroventrally at about 45° angle. On internal surface of braincase, frontoparietal suture located 20-30 mm posterior to its position on dorsal surface and aligned perpendicular to midline. Parietals further distinguished by extreme thick¬ ness of bone (maximum thickness 47 mm along posterior border of parietals, 30 mm farther for¬ ward near contact with frontals; N=14). Supraoccipital.— Supraoccipital firmly fused with parietals, as in other sirenians. Together with parietals, supraoccipital forms prominent lambdoidal crest. Parietosupraoccipital suture obliterated in most Waccasassa River skullcaps, but supraoccipital appears to form lateralmost portion of crest in some specimens and ventral portion of crest in all. In dorsal view, only parie¬ tals seem to be present on cranial roof. Deep median notch in lambdoidal crest visible in dorsal view; notch variably developed and not equally pronounced in all specimens, but in no skullcap does supraoccipital project onto cranial roof. Su¬ praoccipital as broad as or slightly broader than parietals, depending on which bone is interpreted to form lateralmost portion of lambdoidal crest. Overall shape of supraoccipital extremely vari¬ able in the Florida sample, although its external surface is broadly concave in all cases. Supraoc¬ cipital roughly hexagonal in most specimens; in others it may appear pentagonal, trapezoidal, or almost lozenge-shaped. External occipital protu¬ berance usually weakly developed. In several specimens, weak median ridge extends short dis¬ tance ventrally from protuberance. As mentioned above, supraoccipital may have contacted squa- mosals in some specimens. (The large sample of skullcaps from the Florida sirenian, however, in¬ dicates that contact between supraoccipital and squamosals is probably not an important mor¬ phological feature, as it is dependent on several variable factors, including size of the dorsal process of the squamosal and degree of develop¬ ment and configuration of the lambdoidal crest. The large Florida sample also suggests that these features vary with the size and the age of the animal.) Lateral portions of lambdoidal crest bear prominent rugosities for insertion of semispinalis capitis muscles; these enhance posterad concavity of crest as a whole. Although not so thick as parietals, supraoccipital reaches maximum thick¬ ness of 24 mm; on its continuous sutural surface with exoccipitals it is thickest in midline and thins laterad, indicating that supraoccipital did not reach foramen magnum. Frontals (Figures 5, 7, 9, 11). — Posteromost portion of frontals preserved in 2 specimens (TRO 510, 521). Lateral edges of frontals begin to diverge laterad immediately anterior to their dorsal contact with parietals. This is true in most sirenians; however, in this form widening of skull in this region is particularly noticeable due to extreme narrowness of parietals just posterior to frontals. Posterior and median part of frontal skull roof markedly concave; concavity confluent 28 0 4 1 I _i_i_1 CM Figure 11.—Skullcap of Protosiren species from Florida in anterior aspect (TRO 521), showing broken cross-section through frontals. with narrow groove in parietal midline, which extends posteriorly from frontoparietal suture. Farther anteriorly, dorsal surface of frontals be¬ comes flatter, and in 1 specimen in which this area is preserved, there is a slight frontal boss. Interfrontal suture extends 30-35 mm on cranial surface. Frontal projects ventrally as triangular process, on whose anterior sloping surface is lo¬ cated a deep, transversely oriented elliptical pit, probably for reception of spina mesethmoidalis. This pit appears to approach no closer than 2.5 cm to dorsal surface of frontals. Dorsolateral and anterior to this pit, each frontal bears large con¬ cavity which was separated from temporal fossa by thin bony wall; this formed a posterolateral recess of nasal cavity which may have accom¬ modated parts of the ethmoturbinalia (Figure 8). In summary, the most striking features of the Florida skullcaps are: elongate general outline; deep constriction directly posterior to the fronto¬ parietal suture; deeply incised suture for the dor¬ sal process of the squamosal; prominent, rugose, posterolaterally extended and posteriorly concave lambdoidal crest; slightly convex cranial roof; weak to absent temporal crests; prominent bony falx cerebri, tentorium osseum, and internal oc- SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY cipital protuberance forming a cruciform pattern; transverse sulcus with deep lateral pits; supraoc- cipital meeting cranial roof at angle of about 125°; and exceptionally thick bone of the parie- tals and supraoccipital. The important features of the preserved portions of the frontals include a sharply V-shaped frontoparietal suture diverg¬ ing from the midline at a 45° angle, a relatively long interfrontal suture on the cranial surface, and prominent posterolateral recesses of the nasal cavity. Maxilla (Figure 12).—One of the 2 available edentulous maxillae (TRO 516) is slightly more nearly complete, including alveoli for ?P 4 -M“ and the entire zygomatic-orbital process. The less nearly complete maxilla (TRO 538) retains al¬ veoli for ?P 4 -M 3 and the posterior half of the zygomatic-orbital process, although the alveoli for ?P 4 and M 3 are only partially preserved. The only observable differences between the two spec¬ imens are in thickness of posterior border of zy¬ gomatic-orbital process and shape of M 2 alveolus. In TRO 516, posterior border of process 28 mm thick and essentially flat; in TRO 538 it is only 21 mm thick and more rounded. M 2 alveolus in TRO 516 differs from that in TRO 538 in lacking septa between roots of tooth. These slight differ¬ ences are easily attributable to individual varia¬ tion. Zygomatic-orbital process arises directly lat¬ eral to point between M 2 and M' 3 alveoli and extends anteriorly 60 mm (in TRO 516) to point slightly beyond anteromost alveolus in preserved toothrow. Process thickened along posterior bor¬ der (28 mm maximum), thins to 8 mm at anterior border. Ventral side of process lies at level of alveolar rim posteriorly and rises anterodorsad at angle of about 20° to palatal plane. Anterior border of process forms floor of infraorbital fora¬ men, which is relatively small (15 mm in medio- lateral diameter) in the 1 specimen (TRO 516) in which it is partially preserved. Located approxi¬ mately 10 mm medial to infraorbital foramen on dorsal side is prominent palatine foramen, which enters rostrum passing anteroventrad. Lateral to toothrow, flattened zygomatic-orbital process slopes gently dorsad to maxillary-jugal suture, mmm mm. Figure 538); b, 516). Ai 30 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY which is more or less parallel to midline and 65 mm in length. Sutural surface very broad poste¬ riorly, narrow more anteriorly, and slightly broadened again at anteromost end; irregularly concave throughout length, with a particularly deep pit in center of broadest (posterior) part, and horizontal interdigitations along posterior edge. Dorsal surface of zygomatic-orbital process forms floor of orbit and is gently concave both anteroposteriorly and mediolaterally. Just medial to maxillary-jugal suture, on dorsal side, rather prominent ridge of bone extends parallel to suture for most of its length. Toothrow rather short; ?P 4 - M 2 length = 52 mm in TRO 516, 44 mm in TRO 538. Anteromost premolar and M 3 both relatively near (M 3 perhaps slightly closer to) midline, toothrow in between being bowed outward in gentle arc. Five alveoli present in each maxillary fragment, although TRO 516 retains only ante¬ rior wall of M 3 alveolus. Both specimens have 2 small anterior alveoli, which presumably con¬ tained single-rooted premolars. These probably correspond to P 4 and P 5 of a skull referred to Protosiren fraasi by Sickenberg (1934, pi. 3: fig. 6). Behind these alveoli in each specimen is a slight (~ 10 mm) diastema, evidently a space for a DP 5 alveolus now filled with cancellous bone and being encroached upon by the permanent pre¬ molars. A similar condition is observed in speci¬ mens of Eotheroides libycum (holotype, Andrews, 1906, pi. 20: fig. 1a; also YPM 33852) and Hal- itherium schinzi (Krauss, 1862, pi. 6: fig. 2; Lepsius, 1882, pi. 3: fig. 18). TRO 516 retains enough bone in front of P 4 alveolus to demonstrate that there was at least a 20 mm diastema between it and the next anterior tooth. P 4 and P 5 alveoli circular, equal in size, not separated by a dias¬ tema. M 1 and M 2 of TRO 538 appear from the alveoli to have been very similar in shape and size; both were 3-rooted with 2 smaller labial roots slightly flattened anteroposteriorly and a single, larger lingual root circular in cross-section. M 1 alveolus in TRO 516 identical to that in TRO 538, but as described above, M 2 alveolus lacks intra-alveolar septa. M 3 alveolus, though only partially preserved in 1 specimen (TRO 538), appears to have differed from the 2 anterior molar alveoli in having the posterolabial root socket somewhat enlarged, more posteriorly and lin- gually placed, and curved forward apically. Sphenopterygoid Fragments (Figure 13).— Two examples of the fused basisphenoid-ali- sphenoid-pterygoid complex are known. In the larger and better preserved (UF 21612), sphen- occipital suture about 34 mm wide; a single ridge descending after side of each pterygoid process begins in contact with edge of suture, unlike Trichechus or Metaxythenum. No pterygoid fossa present on after side of process. Pterygoid pro¬ cesses converge slightly posteriorly and are at least 40 mm long anteroposteriorly. Alisphenoid canal absent. Optic foramen lies at level of top edge of sphenorbital foramen, which is confluent with large foramen rotundum. Anterodorsal to these on cranial surface, and close to midline at lower edge of cribriform plate, a smaller foramen passes anteroventrad, converging on optic canal. A possible trace of a tiny sinus canal may be discerned dorsolateral to foramen rotundum. Me¬ dial sides of foramina rotunda separated by about 23 mm. Figure 13. —Partial sphenopterygoid complex of Protosiren species from Florida in posterior aspect (UF 21612). NUMBER 52 31 Table 3.— Dental measurements (mm) of Florida Eocene sirenians (L = crown length, AW = anterior width, PW = posterior width) Tooth L AW PW Right upper M (UF 25708) 11.1 13.2 _ Left upper M (UF 24806) 15.4 - 12.4 Left M 3 (TRO 518) 17.5 14.8 10.6 Left M 3 (TRO 518, Figure 14, Table 3).— Tooth very slightly worn. Anterior interdental wear facet present midway between lingual and labial margins. Precingulum consists of promi¬ nent noncuspidate transverse crest arising from anterolabial base of paracone and extending lin- gually slightly more than half breadth of tooth to the point where deep cleft separating protoconule and protocone descends anterior face of proto- loph. Precingulum highest at this point, from which it descends at sharp angle toward lingual margin. Lingually, precingulum represented by 2 tiny cuspules at anterolingual base of protocone. Paracone highest cusp on tooth, only very lightly worn but already nearly confluent with protocon¬ ule, which is only slightly lower. Paracone and protoconule both compressed anteroposteriorly. Although equidistant from paracone and proto¬ cone, protoconule is separated from protocone by prominent cleft which extends down anterior and posterior faces of protoloph to bottom of trans¬ verse valley. Protocone largest cusp on tooth, blunt, circular in outline, and slightly lower than paracone, protoconule, hypocone, and metacone. Deep transverse valley open labially but lingually closed by cingulum which extends from postero- lingual base of protocone to anterolingual base of hypocone. Lingual cingulum a low ridge consist¬ ing of 5 tiny cuspules anteriorly and 1 relatively large cuspule posteriorly, which is separated from hypocone by a cleft. Metacone and metaconule anteroposteriorly compressed and connected to form broad V-shaped ridge with its apex anterior. Metaconule much closer to hypocone than to metacone but separated from former by shallow cleft. Hypocone smallest of the major cusps, blunt, circular in outline, and located near lingual border. From posterolingual base of hypocone a sharp ridge descends posteriorly. This lingual por¬ tion of postcingulum is separated from a promi¬ nent, sharp cusp located at posteromost point of tooth by a deep cleft extending down lingual margin almost to base of enamel crown. Postcin¬ gulum extends anteriorly from large posterior cusp to posterolabial corner of metacone. This labial portion of postcingulum is a low ridge composed of 2 tiny cuspules. Postcingulum and metaloph surround deep circular basin. Tooth has 2 small roots anterolingually and anterolabi- ally and a larger root on posterior margin midway between lingual and labial borders. Small anter¬ olabial root markedly compressed anteroposte¬ riorly; anterolingual root triangular in outline; large posterior root triangular in outline with very deep groove down posterolingual side. Left Upper Molar (UF 24806, Figure 15, Table 3).—This tooth is similar in size and pres¬ ervation to TRO 518. Labial side of protoloph missing; tips of all major cusps pitted by moderate wear; roots broken. Precingulum confined to la¬ bial part of anterior side of crown. Protoloph straight. Slight cingulum present at each end of transverse valley. Metaloph consists of 3 sube¬ qual, evenly spaced cusps and is slightly convex forward. Rounded spur of metaconule extends anterolabiad into transverse valley to touch base of protoconule, while another, sharper and bicus- pulate spur of metaconule extends posterolin- guad, hardly diminishing in height, to stop ab¬ ruptly at rear of crown in flat, vertical, posterad- facing surface. (This pair of extensions, which gives a T-shaped aspect to the metaloph, is not exactly paralleled in any other sirenian we have seen.) Low posterolabial cingulum encloses a small basin. Right Upper Molar (UF 25708, Figure 16, Table 3).—Similar in preservation to the above; heavily worn. Precingulum extends across most of anterior side but best developed labially; at¬ tached to protoloph just lingual to midline of tooth. Transverse valley open labially, partly blocked in center by spurs of protoloph and metaloph; lingual end damaged. Metaloph 32 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Figure 14.—Left M J of Prolosiren species from Florida (TRO 518): a, b, occlusal aspect, anterior up (stereophotograph); c, labial aspect; d, anterior aspect; e, lingual aspect; f, posterior aspect. NUMBER 52 33 A B Figure 15. — Left upper molar of Protosiren species from Florida (UF 24806): a, b, occlusal aspect, anterior up (stereophotograph). A B 0 2 i-1_i_i_i CM Figure 16. — Right upper molar of Protosiren species from Florida (UF 25708): a, b, occlusal aspect, anterior to right (stereophotograph). nearly straight. Small postcingulum attached to metaloph lingually, bounds small posterior basin opening labiad. Posterior interdental wear facet present. Large lingual root preserved, labial roots broken. Mandibular Symphyses (Figure 17). —All 3 mandibular symphyseal fragments broken off dorsal to mental foramen and lack anteromost portion of symphysis. Two retain enough of hor¬ izontal ramus so that abrupt anterior downturn¬ ing of ventral ramal border is evident. The 3 specimens differ somewhat in size and robustness, but all appear to represent the same form. In the largest (TRO 525) and smallest (TRO 509), the mandibular rami, although solidly fused, are sep¬ arated ventrally by a cleft. In the third (UF 21611) the rami are fused anteriorly, the cleft persisting only along posteroventral margin of symphysis. Symphyseal regions laterally com¬ pressed anteriorly, widen gradually posteriorly in 2 specimens, the smallest being more noticeably compressed anteriorly and hence widening more abruptly. Posteroventral margin of symphysis as¬ cends vertically at almost 90° angle, then turns backward almost as abruptly to form ventral ramal border. In smallest specimen, cleft between right and left rami persists on this vertical poste¬ rior margin of symphysis, whereas in other 2 this portion of symphysis fused. Large mental fora¬ men located slightly (about 10-20 mm) forward of posterior border of symphysis. Deep, narrow groove extends forward from mental foramen to point about 10 mm from anteromost margin of symphysis, as preserved in most nearly complete specimen (TRO 509). All 3 symphyses have very small accessory mental foramen, which exits from mandibular canal only a few millimeters posterior and ventral to opening of principal mental fora¬ men. Any other accessory mental foramina which may have been present are obliterated by break¬ age. Only smallest specimen retains much of men¬ tal fossa, which is rather shallow and resembles that of North Carolina mandible described below. Thoracic Vertebrae. —Two examples are preserved. UF 25706, much distorted, has broadly heart-shaped centrum 44 mm high anteriorly, about 50 mm wide, 28 mm thick. Transverse process tapers to irregular point without clear facet for tubercular articulation. Broad, deep fis¬ sure occupies ventrolateral surface of transverse process. Neural canal approximately 24 mm wide, 15 mm high posteriorly; apex not slitlike. Antero¬ posterior length of zygapophyses about 50 mm. Postzygapophyses thin, laterally compressed. Centrum of UF 25707 of comparable size and shape; transverse process much more robust, quadratic in dorsal view, with articular surface and without fissure. Neural canal approximately 18 mm wide, 24 mm high; apex not slitlike. Anteroposterior length of zygapophyses greater than 45 mm. Neural spine inclined backward slightly but still lies over centrum; posterior edge Figure 17. — Mandibular symphyses of Protosiren species from Florida in lateral (a, c, e) and ventral (b, d, f) aspects: a, b, TRO 509; c, o, UF 21611; e, f, TRO 525. NUMBER 52 35 only slightly thickened. Both vertebrae composed almost entirely of cancellous bone. Post-thoracic Vertebrae. —In TRO 527, centrum 62 mm wide and 36 mm high anteriorly, indented at top; overall height about 85 mm; neural canal 19 mm wide, 16 mm high. Trans¬ verse processes 36 mm wide, 10 mm thick. Breadth across prezygapophyses 49 mm, antero¬ posterior length of zygapophyses 62 mm. Neural spine thin, inclined backward, its center lying over posterior edge of centrum. TRO 526 similar, less well preserved. Both vertebrae composed al¬ most entirely of cancellous bone. Ribs.— Only fragments preserved; cross-sec¬ tions oval, greater diameter typically 3-5 cm, lesser 2-3 cm. Cancellous bone found only near proximal ends. Distal ends taper abruptly in their last 4-6 cm to truncated subconical tips. Comparisons Although several genera of Eocene sirenians have been described, considerable doubt and con¬ fusion remain concerning their anatomy and proper diagnosis. The three best-known forms—- Protosiren , Eotheroides , and Prototherium —have been studied by able workers, including Andrews (1906), Abel (1912), and Sickenberg (1934), yet confusion persists because of the typically disso¬ ciated condition of the specimens and the conjec¬ tural nature of referrals of isolated elements to a given taxon (Sickenberg, 1934:9, 10). This situa¬ tion is unlikely to improve until relatively com¬ plete, associated skeletons are discovered; cer¬ tainly we cannot attempt the necessary revision in this paper. On the contrary, the fossils reported herein are prime examples of the frustrating in¬ completeness of key specimens that continues to plague Eocene sirenian taxonomy. We only can seek to determine whether the North American forms appear referable to any known taxon. The Florida specimens will be considered here, the North Carolina material in a later section. We regard the diagnostic Florida Eocene spec¬ imens as representing a single taxon, whose more important features include: elongate skull roof; anteriorly constricted, dorsally convex, and ex¬ tremely thick parietals; massive and posteriorly concave lambdoidal crest; cruciform sculpturing of endocranial roof; lack of an alisphenoid canal; lack of a pterygoid fossa; anteriorly compressed mandibular symphyseal region; sharp anterior downturning of ventral ramal border; restriction of well-developed precingulum to labial part of anterior side of upper molar crown; and presence of cuspules or a small cingulum at lingual end of upper molar transverse valleys. We must deter¬ mine whether the combination of these and other characters is to be found in any of the previously described taxa. Prorastomus sirenoides Owen, 1855: The unique type specimen (BM(NH) 44897) of this most primitive known sirenian, from the middle Eocene of Jamaica, reveals many points of con¬ trast with the Florida material. The form of the supraoccipital is quite different (Owen, 1855, pi. 15: fig. 1); an alisphenoid canal is present; and the ventral border of the mandible is nearly straight. (The cranial cavity is unfortunately filled with matrix, concealing details of the brain- case.) Although resembling the Florida specimens in some primitive features such as interorbital constriction and lateral compression of the ante¬ rior mandibular region, Prorastomus is clearly dis¬ tinct from them. Sirenavus hungaricus Kretzoi, 1941: This mid¬ dle Eocene form from Hungary was based on a partial skull and a posterior mandibular frag¬ ment, and was considered confamilial with Pro¬ rastomus by Kretzoi. It resembles the latter and the Florida material in degree of interorbital constriction, and the interorbital region is de¬ scribed as “very massively constructed,” as in the Florida form; however, its possession of a “sagittal crest” and lack of pronounced development of the lambdoidal crest distinguish it clearly from our material. Anisosiren pannonica Kordos, 1979: Another middle Eocene taxon from Hungary, based on a fragmentary maxilla with ?DP 5 -M 3 and an iso¬ lated permanent premolar. The M 3 closely resem¬ bles TRO 518 but is more than a third again 36 larger, measuring 24.0 mm in length and 22.2 mm in breadth. Protosiren fraasi Abel, 1907: This important form from the middle Eocene of the Mokattam Hills, Egypt, is more beset with taxonomic con¬ fusion than any other we have to consider. Abel first published the name in 1904 but without diagnosis, illustration, or designation of a type. In 1906 he designated as type a skull in the Stuttgart collection (apparently no. 10576 = “Stuck V” of Sickenberg, 1934). Also in 1906, Andrews de¬ scribed a skull in the Cairo Geological Museum (CGM C.10171), which Abel then (in 1907) re¬ ferred to P. fraasi and designated as the type in preference to the Stuttgart skull! Abel’s 1906 work contained no description or illustration of the latter, however, and “mention of a . . . specimen in a collection” does not constitute an “indication” in the meaning of the 1964 Inter¬ national Code of Zoological Nomenclature (art. 16(b) (i)). Therefore, we must with Sickenberg (1934:43, 44) accept Abel’s 1907 action as the first valid use of the name and recognize the Cairo skull (CGM C. 10171) as the holotype; how¬ ever, doubtful association of a mandible with this skull (see below) and subsequent referral of other material to this species have clouded the picture further. The holotype (CGM C. 10171), described by Andrews (1906:204-209, fig. 66), closely resem¬ bles the Florida sirenians in morphology of the supraoccipital, parietals, and posterior portion of the frontals. Both have relatively elongate skull¬ caps, massive, strongly concave lambdoidal crests (though these are not rugose in CGM C. 10171), weak external occipital protuberances and tem¬ poral crests, and slightly convex cranial roofs. Several differences are also apparent. In CGM C. 10171, what seems to be a triangular process of the supraoccipital extends onto the skull roof in the midline. This is not matched in any Florida specimen (TRO 531 and UF 16441 come closest) but may be merely an individual variant. The cranial portion of the squamosal fails to extend as high as the level of the skull roof, which it evidently did in the Florida form. The cranial SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY cavity of CGM C. 10171 is unfortunately filled with matrix, leaving in doubt the presence of a bony falx and associated features. An alisphenoid canal, however, is absent from this specimen, as in the Florida sirenians. It further resembles the latter in lack of a pterygoid fossa, in posterior convergence of the pterygoid processes, and in form of the maxilla and toothrow, but differences appear again in regard to M 3 . In CGM C. 10171, a ridge connects the protocone and precingulum; the cuspidate lingual cingulum connecting pro¬ tocone and hypocone, and the deep, enclosed posterior basin are both lacking. The molars from Florida and Egypt are similar in having a large, blunt protocone and smaller paracone and pro- toconule, which are anteroposteriorly compressed and barely separated. Also, both have a blunt hypocone and sharper, more compressed meta¬ cone and metaconule. The mandible thought by Andrews (1906:210- 212, fig. 67) to belong to the same species as this skull, and perhaps to the same individual, and which bears the same number (CGM C. 10171) in the Cairo Geological Museum, more likely rep¬ resents a different form. In its ventral outline it resembles both the Florida mandibles and the Protosiren mandible of Priem (1907) but contrasts with the latter in its broader dorsal symphyseal surface, which more resembles that of Eotheroides aegyptiacum (Abel, 1912, pi. 32: fig. 1). Abel’s original concept of Protosiren , however, appears to have been based on Sickenberg’s “Stuck V” in the Stuttgart collection. Compari¬ son with the latter individual, insofar as its char¬ acters can be retrieved from Sickenberg’s descrip¬ tion of the species as a whole, reveals the follow¬ ing. The degree of development of the lambdoidal crest is not entirely clear from Sickenberg’s illus¬ trations (1934, fig. 7; pi. 1: fig. 2) but may fall within the range of variation of the Florida spec¬ imens. The supraoccipital, however, is wider in proportion to its height than in any Florida spec¬ imen and reaches the foramen magnum, unlike CGM C. 10171 or the Florida form (though Sick¬ enberg considered such differences to be age re¬ lated). The angles formed by the frontoparietal NUMBER 52 37 and the exoccipital sutures, the development of temporal crests, the lack of an external occipital protuberance, and the size of the infraorbital foramen are all points of resemblance. The max¬ illary tooth row (Sickenberg, 1934, pi. 3: fig. 6) is closely comparable to that indicated by the Flor¬ ida maxillary fragments; both tooth rows are slightly curved and have two single-rooted pre¬ molars anteriorly, which are separated from the next anterior tooth by a diastema. The more anterior premolars of “Stuck V” are forward of the level of the infraorbital foramen and would not have been preserved in the Florida material; the most posterior of these, however, may have been absent or more anteriorly located in the Florida form. Sickenberg emphasized the failure of the squamosal to reach up to the temporal crest in Protosiren , but this appears not to have been true in the Florida form. Though there is similarity in the shape of the pterygoid processes and lack of a pterygoid fossa, the presence of a large alisphenoid canal in Sickenberg’s specimens is an important difference—not only from the Florida sirenians but also from the holotype skull (CGM C. 10171), which Sickenberg apparently did not examine. This casts serious doubt on the identity of his referred material. Specimens of Sickenberg’s hypodigm other than “Stuck V” show additional contrasts with the Florida material, chief of which is the absence or weak development of a bony falx cerebri, tentorium osseum, internal occipital protuber¬ ance, and transverse sulcus (Sickenberg, 1934, pi. 1: fig. 4). A similarly smooth-surfaced skullcap, also from the Mokattam area, is in the British Museum (BM(NH) M8154), and another from the Eocene of France was referred by Sickenberg (1934:189, 190) to Protosiren. An endocranial cast (Sickenberg, 1934, fig. 8), however, shows the frontoparietal suture perpendicular to the mid¬ line as in the Florida sirenians. The mandible described by Priem (1907) (Sickenberg’s “Stuck L”) compares quite closely with the Florida man¬ dibles, particularly the smallest one (TRO 509), in all respects. The M 3 of Sickenberg’s “ Protosiren ” (1934:70, 71, fig. 10), however, is quite different from TRO 518; the lingual cingulum is absent and the postcingulum and metaloph are different in form. Its outline is quadrilateral rather than tapered and rounded posteriorly. Sickenberg’s specimen (“Stuck XXXVIII”), however, is de¬ scribed as a “young animal,” and the photograph (1934, pi. 1: fig. 8) suggests that the “M 3 ” is unerupted; hence we may question its identity as a third molar. Though similar in their lack of compact bone, the Florida vertebrae do not other¬ wise closely resemble those attributed to Protosiren by Sickenberg and show smooth ends without obvious reduction of epiphyses such as he de¬ scribes. We conclude that the Florida sirenians most closely resemble the holotype skull (CGM C. 10171) of Protosiren fraasi, described by Andrews (1906), and the mandible described by Priem (1907). These latter may differ specifically, if not generically, from “Stuck V” and other specimens referred to P. fraasi by Sickenberg (1934), which exhibit alisphenoid canals and smooth endocran¬ ial roofs. ?Protosiren minima (Desmarest, 1822) (= ?P. dubia (Cuvier, 1824); see Hooijer, 1952): Based on three isolated teeth from the Eocene of France (Sickenberg, 1934:190, 191, fig. 36), the affinities of this species are indeed dubious. They are some¬ what larger than the Florida teeth, and the uppers lack lingual cingula, though their general shape and pattern are similar. Discovery of good specimens of Eocene sea cows in France was reported by Freudenthal (1969). These were at first tentatively identified as Protothenum or Protosiren; however, examination by Heal (1973) and by one of us (Domning) confirms that they represent not Protosiren but a dugongid, probably Eotheroides. As they have not been formally described and in any case shed no light on the identity of the American specimens, they will not be further discussed here. Eotheroides aegyptiacum (Owen, 1875a): This middle Eocene form from the Mokattam Hills of Egypt was described solely on the basis of a natural endocranial cast; hence all bones and teeth bearing this name are merely referred spec- 38 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY imens (Abel, 1912:308, 309). Abel, however, con¬ sidered the type to be sufficiently diagnostic for “a certain identification” of his series of topotypic specimens, which included additional endocran- ial casts. The endocranial casts illustrated by Owen and by Abel show the frontoparietal sutures diverging anteriorly rather than perpendicular to the mid¬ line as in the Florida sirenians. The parietals and supraoccipital of E. aegyptia¬ cum , though quite variably sculpted, are qualita¬ tively similar to the Florida specimens, but they seem from Abel’s account (1912) to fall short of the latter in degree of development of the massive, concave lambdoidal crest, median obstruction of the transverse sulcus, and (possibly) thickness of the skull roof. In contrast to other sirenians, prob¬ ably (but questionably) including the Florida form, the foramen ovale is completely enclosed by bone. A large pterygoid fossa is present, and an alisphenoid canal is lacking, as in other du- gongids. The ventral outline of the mandibular ramus resembles that seen in the Florida form, but the anterior symphyseal region of the latter (at least in TRO 509) seems too narrow to have supported the relatively broad symphyseal sur¬ face seen in the Egyptian dugongid (Abel, 1912, pi. 32: fig. 1; Andrews, 1906, fig. 67a). The M 3 of E. aegyptiacum shares the triangular shape of TRO 518, but a lingual cingulum and posterior basin are lacking. The vertebrae (Sickenberg, 1934) bear no close resemblance to the Florida speci¬ mens either in shape or in their relatively high proportion of compact bone, and none of the Florida ribs show quadratic cross-sections like those found in Eotheroides. Eotheroides abeli (Sickenberg, 1934): This mid¬ dle Eocene Egyptian species was founded on an isolated M 2 , with a partial skull and other ele¬ ments referred. It differs from E. aegyptiacum in such features as shorter parietals, higher supraoc- cipitals, and larger molars, differences which sep¬ arate it still further from the Florida sirenians. “Eothenum^ majus Zdansky, 1938: This doubt¬ ful species from the Egyptian middle Eocene was based on an isolated upper molar, 22 mm long and 23 mm wide, hence much larger than the Florida specimens. It also lacks a lingual obstruc¬ tion of the transverse valley. Eotheroides libycum (Andrews, 1902): This well- known dugongid from the Egyptian upper Eocene likewise differs clearly from the Florida form. Although the bony falx and tentorium osseum are well developed, the parietal roof is bilaterally convex rather than nearly smooth and lacks posterolateral indentations for the squamo- sals; the lambdoidal crest is not so prominently developed; and the external occipital protuber¬ ance is more prominent. A posteromedian de¬ pression of the frontal roof is lacking. A pterygoid fossa is present, and an alisphenoid canal absent. The mandibular symphyseal region is very broad and robust (Andrews, 1906, pi. 20: fig. 2; also YPM 38213), unlike the Waccasassa River spec¬ imens. The vertebrae show a greater proportion of compact bone, and the neural spines are more massive than in the latter (Sickenberg, 1934). “ Eotherium” stromen Sickenberg, 1934: This late Eocene nominal species from Egypt was orig¬ inally proposed by Abel (1912), who designated a type but provided no diagnosis or illustrations, which were not available until the work of Sick¬ enberg (1934). The latter, therefore, rather than Abel, is the author of this name. The holotype is a skull and incomplete skeleton. The parietal roof is relatively short and wide, but otherwise this form closely resembles the other Egyptian dugon- gids and therein differs from the Florida sirenians. Protothenum veronense (de Zigno, 1875): This dugongid is known from numerous specimens from the upper Eocene of Italy (Sickenberg, 1934) and is closely allied to Eotheroides. Like the latter, it exhibits a bony falx and internal occipital protuberance but lacks the striking development of the lambdoidal crest seen in the Florida skull¬ caps. It resembles these, however, in apparently lacking a pterygoid fossa as well as an alisphenoid canal. The zygomatic-orbital bridge of the max¬ illa lies well above the palatal plane, in contrast to the Waccasassa River specimens. The ventral border of the mandibular ramus does not turn down abruptly at the symphysis, and the sym- NUMBER 52 39 physeal masticating surface is broad rather than anteriorly compressed. M 3 , however, though rel¬ atively large, comes closest to the Florida speci¬ mens in having a precingulum developed only labially; a large, blunt protocone somewhat sep¬ arated from the compressed protoconule and par- acone; closely appressed metaconule and hypo- cone; and (occasionally) a deep posterior basin and lingual cuspules obstructing the transverse valley. Paralitherium tarkanyense Kordos, 1977: This late Eocene form from Hungary, based on a pair of mandibles with vertebrae and ribs, is excluded from close relationship with the Florida sirenians by its larger size, more massive symphyseal re¬ gion, dense vertebrae, and quadratic rib cross- sections. Eocene Sirenians of Transylvania: Fragmen¬ tary sirenian remains have been known for over a century from middle and late Eocene rocks of Transylvania (Siebenbiirgen), Romania, partic¬ ularly from the region of Cluj (Klausenburg, Kolozsvar) (Sickenberg, 1934; Fuchs, 1970, 1973, and references therein cited). They have not been formally named but have been thought most closely comparable to Eotheroides , Prototherium , and Halitherium. The best specimen so far reported, a skull roof (Fuchs, 1970), somewhat resembles the Florida specimens in characters of the supraoccip- ital and the lambdoidal crest but has Halitherium- like temporal crests. The skull roof is only 18 mm thick, roughly half the thickness of the Florida specimens and indeed more comparable to Hal¬ itherium. Post-Eocene Sirenians: Detailed comparisons with Oligocene and later forms are unnecessary, as none bears a close resemblance to the Florida Eocene specimens. Halitherium schinzi from the Oligocene of Europe occasionally shows a similar development of the lambdoidal crest (e.g., MCZ 8829), but thinness of the skull roof, form of the temporal crests, and other features set it well apart from the Florida sirenians. The same is true of Metaxytherium , Trichechus , and all other later forms, with which there is no danger of confusion. We conclude that the sirenians from the Avon Park Limestone and Inglis Formation of Florida most closely resemble, and are probably conge¬ neric with, Protosiren fraasi from the middle Eocene of Egypt. The Florida sirenians are therefore referred to Protosiren species. North Carolina Eocene Sirenians Description The identifiable elements of the North Caro¬ lina sirenians complement those of the Florida sea cows almost to the bone. The only bones common to the two samples are ribs, vertebrae, the symphyseal portion of the mandible, and the skullcap (represented in North Carolina by a single isolated specimen). The North Carolina specimens are as follows: Locality 18 (Martin Marietta quarry, New Hanover County, North Carolina) USNM 307609: parietal-supraoccipital skullcap Locality 21 (Comfort quarry, Jones County, North Carolina) USNM 214596: mandible lacking posterior parts of both rami; right P 3 - 4 , DP 5 , Mi, parts of M 2-3 and left M 1 - 2 ; left M 3 ; left two-thirds of atlas; part of left side of axis; right side of Pfirst thoracic vertebra; 1 complete poste¬ rior thoracic vertebra and fragments of at least 2 more; 5 incomplete post-thoracic vertebrae; 13 rib fragments including parts of at least 3 right and 6 left ribs; fragments of both scapulae; proximal ends of right humerus and ulna; fragment of Pleft innominate USNM 214597: left M 1 or M 2 USNM 244491: distal end of right humerus USNM 244494-5: lateral portions of atlas USNM 244496: partial left periotic USNM 256686: fragment of right squamosal Numerous uncatalogued rib fragments Parietal-Supraoccipital Skullcap (USNM 307609, Figure 18).—Parietals widest (55 mm) at forward edge of parietal-squamosal suture, nar¬ rowest (43 mm) just abaft frontoparietal suture; length in dorsal midline about 66 mm, overall length 81 mm. Lambdoidal crest less prominent than in most of the Florida specimens. Lateral indentations for squamosals deep and conspicu¬ ous; squamosals reached almost to level of skull roof. Skull roof bilaterally convex with median groove extending entire length of parietals. Tern- 40 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Figure 18.—Parietal-supraoccipital skullcap of Protosiren species from North Carolina (USNM 307609): a, right lateral aspect; b, posterior aspect; c, dorsal aspect; d, ventral (internal) aspect. poral crests absent. Lateral walls of braincase meet skull roof at slightly more than right angles, sloping somewhat ventrolaterally. Roof 21 mm thick in midline at frontoparietal suture, 22 mm at parietal-supraoccipital suture. Bony falx cere¬ bri merely a slight median convexity; internal occipital protuberance weak; transverse sulcus and tentorium osseum nearly absent. Internal surface of parietals and supraoccipital pierced by numerous tiny foramina. Supraoccipital 35 to 40 mm high, slightly broader than parietals (esti¬ mated width 58 mm), roughly elliptical in out¬ line. Surface of supraoccipital moderately con¬ cave, with weak external occipital protuberance and median ridge; lateral and ventral borders damaged, not markedly thickened (about 8 mm in ventral midline). Semispinalis muscle scars not distinctly developed. Parietals and supraoccipital form angle of 115°. Squamosal (USNM 256686).—Includes part of root of zygomatic process, area of glenoid facet (which lacks distinct boundaries), and anterior end of socket for periotic. Postglenoid fossa very shallow. Ridge extending anterolaterad from bro¬ ken postglenoid process short and abruptly slop¬ ing. Posterodorsal edge of zygomatic process gently sloping, its posterior end missing. Periotic (USNM 244496, Figure 19).—Ante¬ rior part of pars temporalis (= tegmen tympani) and most of pars petrosa (= pars labyrinthica) missing. Convexity of dorsal surface of pars tem¬ poralis continues posteriorly onto pars mastoidea, ending (as in Dugong ) at a broad, shallow groove passing posterolaterad across dorsal surface of latter. Groove ends in rugose area, which includes a prominent foramen and lies anteromedial to dorsal end of exposed portion of pars mastoidea (processus fonticulus of Sickenberg, 1934:59). NUMBER 52 41 Figure 19.—Partial left periotic of Protosiren species from North Carolina (USNM 244496): a, a', medial aspect; b, b', lateral aspect; c, c', ventral aspect; d, d', posterior aspect. (ar = “arete du rocher,” ef = endolymphatic foramen, fmm = fossa muscularis minor, ft = fovea triangularis, pf = processus fonticulus, pm = pars mastoidea, pp = pars petrosa (= pars labyrinthica), pt = pars temporalis (= tegmen tympani), sf = sulcus facialis.) Latter portion forms low, irregular rugosity, somewhat elliptical in lateral view, its anterior part concave, and its posterior part roughly con¬ vex, and covering posterior half of lateral side of pars mastoidea. Groove on dorsal side of latter paralleled by a distinct ridge (“arete du rocher” of Robineau, 1969, fig. 4a). Posterior side of pars mastoidea irregularly convex laterally; medially a broadly concave triangular area (fovea trian¬ gularis of Abel, 1912, fig. 3) lies beneath over¬ hanging posterior edge of pars petrosa. Ventral end of pars mastoidea curves smoothly inward. Ventrally, pars mastoidea separated from pars temporalis by deep V-shaped cleft, continued onto lateral surface of periotic by a shallow groove. Fossa muscularis minor (for origin of M. stapedius) deep, well marked, somewhat trian¬ gular; a very shallow but distinct sulcus facialis is visible anterior to it, passing anteromediad. Endolymphatic foramen well developed. Pars pe¬ trosa broken away just lateral to promontory. Anterior edge of pars petrosa laterally forms thin horizontal sheet closely applied to dorsal side of pars temporalis. Height of pars mastoidea from “arete du rocher” to ventral end = 26 mm; maximum dorsoventral thickness of pars tempo¬ ralis at broken surface =17 mm. Mandible (USNM 214596, Figures 20-22).— Symphysis solidly fused, but deep cleft persists ventrally. Masticating surface of symphysis very narrow (16 mm at Ii alveolus), rugose, without median ridge, and deflected 35° to 40° from occlusal plane. Anterior half of symphysis strongly compressed laterally; symphysis widens posteriorly but without becoming conspicuously bulbous posteroventrally as in Eotheroides libycum (Andrews, 1906, pi. 20: fig. 2a). Length of sym¬ physis 72 mm, height 55 mm. Ventral outline ascends abruptly at rear of symphysis, initially passing upward perpendicular to direction of ven¬ tral border of horizontal ramus. Large mental foramen (just below canine alveolus) and a single accessory foramen present on each side. At point of divergence of alveolar rows at rear of symphy- seal masticating surface, a median foramen passes anteroventrad along suture. Abaft this, dorsal 42 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Figure 20. — Mandibles of Protosiren in occlusal aspect: a, Protosiren fraasi from Egypt (USNM 244490, cast); b, Protosiren species from North Carolina (USNM 214596). surface of symphysis curves downward to form thick lip overhanging mental fossa. Latter is broad and well developed but only some 4 mm deep, with a gently convex anteroventral floor. Height of horizontal ramus at P 4 alveolus 37 mm. Dentition (USNM 214596, Figure 23, Table 4).—Lower dentition consists of 3.1.5.3. Alveoli of Ii lie 20 mm from tip of mandible and are separated by 5 mm in midline. I2 and I3 alveoli follow after diastemata of about 9 mm each. Canine alveoli closely follow those of I3 and are separated in midline by 16 mm. Alveoli of Pi (or DPi) are likewise close behind those of C but are separated from those of P 2 by 4 mm; subsequent alveoli follow at closer intervals. All alveoli are deep and well developed. I1-P4 are single rooted, DP5-M3 double rooted. M3 is separated by 10 mm from rear of coronoid canal, and dental capsule has atrophied somewhat, indicating (with wear on M3) adulthood; however, X-ray exami¬ nation failed to disclose the presence of an une¬ rupted P 5 on either side. P3: Bears a single high labial cusp, curved inward at tip, and almost unworn. A much smaller anterolingual and a still smaller postero- lingual cusp flank it; between these latter, a crest descends steeply from middle of lingual side of principal cusp and curves past base of posterolin- gual cusp to form low posterior cingulum. Two NUMBER 52 43 Figure 21.—Mandibles of Protosiren in ventral aspect: a, Protosiren fraasi from Egypt (USNM 244490, cast); b, Protosiren species from North Carolina (USNM 214596). tiny cuspules form anterior cingulum. Root sub- circular in cross section. P 4 : Bears a high labial cusp like that of P 3 , enamel slightly breached by wear at tip. A smaller cusp lies directly lingual to it; from tip of latter a ridge descends steeply anterolabiad and then an- terad to join highest point of low anterior cin¬ gulum. Behind these 2 cusps lies a pair of smaller, subequal cusps, the more lingual lying at edge of crown and forming highest point of minutely cuspate posterior cingulum enclosing the more labial cusp. Root oval, elongate anteroposteriorly, in cross-section. DP5: Heavily worn, both lophids and hypo- conulid lophule forming a single confluent lake of dentine. Mi: Likewise heavily worn; broken on both sides. Lingual cusps much higher than labial on all molars, as shown for Protosiren by Sickenberg (1934, fig. lib). M 2 : Broken on both sides. Tips of lingual cusps almost unworn, labial portions of lophids worn. Labiad-descending crest on anterior side of crown (“vorderes Basalband”) resembles that seen in Halitherium chnstoli (Abel, 1904, fig. 20) and Eotheroides abeli (Abel, 1912, pi. 5: fig. 1). 44 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Figure 22. — Mandibles of Protosiren (a, left lateral aspect, reversed; b, right lateral aspect): a, Protosiren fraasi from Egypt (USNM 244490, cast); b, Protosiren species from North Carolina (USNM 214596). Shallow basin present on crest of protolophid between metaconid and protoconid. Transverse valley largely obstructed by indistinctly devel¬ oped crista obliqua. Hypoconulid lophule small, with single cusp at labial end connected to middle of metalophid by short crest. M 3 : Slightly worn. “Vorderes Basalband” and basin on crest of protolophid like those of M 2 . Crista obliqua forms not a distinct crest but a low, bulbous protrusion of metalophid into transverse valley. Hypoconulid lophule promi¬ nent, consisting of a single large pointed cusp with somewhat sharpened lingual edge; con¬ nected to middle of metalophid by low crest. Anterior root anteroposteriorly flattened, like roots of other molariform teeth; 27 mm long. Posterior root much larger, longer fore-and-aft than wide, sigmoidal and slightly tapering api- cally; 24 mm long. Both roots completely closed. M 1 or M 2 (USNM 214597, Figure 24, Table 4): Moderately worn; posterior interdental wear facet present. Anterior cingulum a simple, non- cuspate ridge, highest at center, contacted lin- gually by basal swelling of protocone. Protocon- ule evenly spaced between paracone and larger (but heavily worn) protocone. Transverse valley open. Metaconule appressed to hypocone; from tip of former a ridge (sharpened by wear and forming top edge of anterolabially descending wear facet) descends to intersect anterior side of 45 Figure 23.—Stereophotographs of lower dentition of Protosiren species from North Carolina in occlusal aspect (USNM 214596): a, b, left M 1 -M 3 ; c, d, right P 3 -M 3 . 46 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Figure 24.—Left M 1 or M 2 of Protosiren species from North Carolina (USNM 214597): a, b, occlusal aspect, anterior to left (stereophotograph); c, labial aspect; d, anterior aspect, e, lingual aspect; f, posterior aspect. NUMBER 52 47 Table 4. —Dental measurements (mm) of North Carolina Eocene sirenians (L = crown length, AW = anterior width, PW = posterior width, w = affected by extreme wear, e = estimated, a = alveolar) Tooth L AW PW USNM 214597 Left M 1 or M 2 16.3 16.7 15.1 USNM 214596 Right P 3 9.0 7.4 Right P 4 9.7 8.4 - Right DP 5 12 . lw 8.8 9.2 Left M l 13.4e — 10.0 Left M 2 14.5 11.7 - Right M 2 - - 11.5 Left M 3 18.6 12.2 11.3 Right h.3 34.6a Right C-M 3 96.3a Right Pi-M 3 90.2a Right Mi-M 3 43.5 Left Mi-M 3 45.4 metacone at middle of latter’s height. From tip of hypocone a thick crest descends across rear of crown, enclosing a narrow, shallow posterior ba¬ sin. Two labial roots broken; large lingual root 24 mm long, closed, with prominent groove down lingual side. Atlas. —USNM 214596 (Figure 25a, c): Dorsal arch robust, topped by single rounded protuber¬ ance, without ridges or other sculpturing. Ventral arch thinner, with very weak ventral keel pro¬ longed into irregular posterior process. Dorsal side of arch bears concave depression 21 mm wide for articulation with odontoid process of axis. Ante¬ rior cotyle rather deeply concave, with thick, prominent lateral edges and deeply embayed me¬ dial edge of articular surface; at level of tubercle for transverse ligament, medial edge of surface lies 15 mm anterolateral to tip of tubercle, and surface itself is 16 mm wide at this point. Arterial canal dorsal to anterior cotyle deep but not bridged over. Posterior cotyle flatter, with broad¬ est part ventral to tubercle for transverse liga¬ ment; ventromedial part of articular surface ab¬ ruptly narrows and turns mediad along rear of ventral arch, reaching level of edge of facet for odontoid process. Transverse process robust and knoblike dorsally, compressed in posteroventrad- aligned plane; large vertebrarterial canal passes less steeply posteroventrad. USNM 244494, 244495 (Figure 25b, d): Only lateral portions of atlas preserved; larger than that of USNM 214596. Articular surface of an¬ terior cotyle interrupted by narrow extension of medial embayment, reaching to within 12 mm of lateral edge; total width of cotyle approximately 21 mm at this point (level of tubercle for trans¬ verse ligament). Arterial canal dorsal to anterior cotyle was probably bridged. Posterior cotyle as described above. Transverse process with more rugose edge than in USNM 214596; vertebrarter¬ ial canal absent on both sides. Axis (USNM 214596).—Only left side of cen¬ trum and neural arch preserved; resembles other sirenians in structure. Dorsal and smaller ventral arches of vertebrarterial canal broken at roots; unclear whether canal was bridged. Thoracic Vertebrae (USNM 214596, Figure 26).—An anterior, probably first or second, tho¬ racic lacks left part of centrum and transverse process, and tip of neural spine. Centrum is square in sagittal section, composed of cancellous bone, and bears anterior and posterior demifacets. Neural spine dense, inclined backward slightly, but its bulk lies directly above centrum. Cross- section of neural spine at its base resembles that at tip of T2 of Eotheroides aegyptiacum (Sickenberg, 1934, fig. 3b) but is more elongated, with sharper projections at posterior corners that may answer to the “Lappen- oder Fliigelbildung” described by Sickenberg. Transverse process very massive; greatest transverse breadth of vertebra at level of top of centrum. A complete posterior thoracic (Figure 27) is preserved. Centrum large relative to neural arch, 45 mm thick, with broadly heart-shaped ends (58 mm wide, 44 mm high), wide, flat ventral keel, and concave sides. Epiphyses thin but fully de¬ veloped. Articular facets for capitulum and tu- berculum of rib not quite confluent. Neural canal small (19 mm wide, 17 mm high), apex pointed but not slitlike. Anterior side of neural arch ex¬ hibits broad furrows with thin median crests con- w . K “y H .V $1 K v f / V 1 W Figure 25.—Partial atlases of Protosiren species from North Carolina in anterior (a, c) and posterior (b, d) aspects: a, b, USNM 214596; c, USNM 244495 (left) and USNM 244494 (right); d, USNM 244494 (left) and USNM 244495 (right). verging from either prezygapophysis to join at midline of vertebra (for attachment of ligamen- tum flavum). Neural spine slightly inclined back¬ ward, thicker (14 mm) posteriorly, with sharp anterior and posterior edges, 40 mm long antero- posteriorly. Total height of vertebra 119 mm, transverse breadth 79 mm; anteroposterior length of zygapophyses 61 mm. Isolated neural spines of other vertebrae vari¬ able: one is somewhat swollen posteriorly near dorsal end, with flangelike posterior edges and a low posterior median ridge; another is strikingly swollen near dorsal end, with more prolonged posterior edge on left but no other ridges or flanges on posterior side. Post-thoracic Vertebrae (USNM 214596).— A probable first lumbar vertebra is represented by left side of centrum with incompletely fused transverse process. Prezygapophysis well devel¬ oped; presumed postzygapophysis simple, with¬ out distinct articular surface, and located very low on neural arch (evidently below dorsal side of centrum) as described for Protosiren (Sicken- berg, 1934:85). Transverse process 87 mm long, separated from centrum by broad, deep cleft on posteroventral side; tapers distally and turns sharply posteroventrad at tip. A low ridge, possi- NUMBER 52 49 Figure 26.— Partial anterior thoracic vertebra of Protosiren species from North Carolina in anterior aspect (USNM 214596). bly for tendon attachment, passes posterolaterad across middle of ventral side of process. Other fragments of post-thoracic vertebrae show no dif¬ ferences from typical dugongids; thin epiphyses are present as on thoracic vertebrae, and centra are hexagonal in end view. Ribs (USNM 214596, Figures 28, 29).—Two complete ribs of this specimen are preserved. Neck of left second or third rib inflected posterad; thick dorsoventrally, flattened ventrally, with high semi-circular cross-section. Capitular facets dis¬ tinct but contiguous; anterior facet elliptical, slightly convex; posterior facet more elongated, convex mediolaterally but flatter anteroposte- riorly. Tuberculum elevated, with C-shaped facet opening anteromediad; shallow, irregular liga¬ mentary fossa lies lateral to tuberculum. Shaft Figure 27.—Posterior thoracic vertebra of Protosiren species from North Carolina (USNM 214596): a, anterior aspect; B, posterior aspect; c, right lateral aspect. 1 \ u if J8s | : ,0; Si D 1 &////■ D % V ^ '= SB NUMBER 52 51 somewhat flattened proximally but thick, with anterolateral and posteromedial surfaces broad¬ est. These maintain their widths distally, the latter coming to face more posterad; anterome¬ dial and posterolateral surfaces, however, broaden abruptly beyond middle of shaft to form greatly swollen distal half of rib with quadratic cross-section (greatest diameters 52X49 mm). An¬ teromedial and posteromedial sides flatter than others, hence medial corner of cross-section some¬ what sharper than lateral. Posteromedial side bears 2 prominent foramina proximally and nu¬ merous fine parallel furrows for blood vessels distally. Angle of rib indistinct. Distal end of rib banana-shaped, tapering abruptly to small apical area for attachment of costal cartilage. Neck of right third or fourth rib wider and flatter; capitular facets subequal, roughly semi¬ circular; tuberculum less prominent, with smaller facet; ligamentary fossa deeper. Shaft similar in shape to that described above, but anteromedial side distinctly broadened, even proximal to angle. Greatest diameters 50X44 mm. Posteromedial side bears foramina and furrows as described above. Distal end tapers less abruptly; tip curved backward, bears facet for attachment of cartilage as above. Fragments of proximal ends of remaining ribs all show distinct tubercular facets; ventral sides of necks noticeably flattened; shallow irregular depressions lie lateral to capitular facets on dorsal sides. Distal ends retain swollen, quadratic shape on some ribs abaft those connected to sternum, but more posterior ones have semicircular distal cross-sections with flat medial sides, while pos- teromost ribs become quite slender, flat (27X14 mm), and gently tapering distally (Figure 29). All ribs wholly composed of compact bone. Scapula (USNM 214596). — Glenoid fossa deeply concave fore-and-aft, less so transversely; roughly triangular, apex lateral; 31 mm wide, 38 mm long measured from external sides of its borders. Coracoid process simple, blunt, slightly Figure 29.—Partial posterior rib, lacking head, of Protosiren species from North Carolina, showing cross-sections at posi¬ tions indicated by lines (USNM 214596). 52 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY deflected mediad; distal end of scapula 42 mm long including coracoid process. Spine extends to within 25 mm of glenoid fossa. Supraspinous and infraspinous fossae gently concave; anterior bor¬ der sharp, posterior thicker and blunter. Medial side of scapula flat. Posterior vertebral corner of right scapula shows “teres major protuberance” 4 cm distal to corner and well-developed area for attachment of cartilage along vertebral border. Humerus.— USNM 214596 (Figure 30): Prox¬ imal end of right humerus lacks articular surface of head and summit of greater tubercle; epiphysis of former may have been unfused. Greatest prox¬ imal width 51 mm. Anterolateral surface of greater tubercle smooth, aligned strongly antero- mediad; deltoid crest not pronounced or rugose, very similar to that of juvenile Eotheroides aegyptia- cum illustrated by Sickenberg (1934, fig. 4). Lesser tubercle well developed, irregularly rounded. Bi¬ cipital groove deep, narrow. Angle between cen¬ ter of head and extremities of tubercles at sides of bicipital groove = 35°-40° (cf. Sickenberg, 1934:29, 30). Groove turns posteriorly through sulcus intertubercularis into central cul-de-sac of proximal end, bordered by head and tubercles; distal part of groove and proximal cul-de-sac correspond to “vertical” and “horizontal” parts, respectively, of bicipital groove described by Sick¬ enberg (1934:30). This and the following frag¬ ment both completely of compact bone. USNM 244491 (Figure 31): Distal end of right humerus 64 mm wide, has extremely mas¬ sive entepicondyle reaching as far distad as lateral extremity of trochlea. Trochlea some 41 mm wide anteriorly, inclined at angle of approximately 75° to humeral shaft. Anterolateral surface of trochlea shows severe pitting, perhaps pathological, ex¬ tending into deep coronoid fossa. Olecranon fossa likewise very deep. Tuberosity for insertion of M. pectoralis major, on anterior side proximal to trochlea, extraordinarily large; a smaller protu¬ berance lies proximolateral and posterior to it. Entocondyloid crest broadly rounded, ectocon- dyloid somewhat sharper; sagittal diameter of shaft considerably exceeds transverse diameter. Ulna (USNM 214596, Figure 32).—Proximal end differs strikingly from those of other sirenians; olecranon aligned coaxially with shaft; outline of posterior edge convex at level of semilunar notch; lateral side has broad, deep longitudinal groove from olecranon to middle of shaft. Summit of olecranon expanded, sloped laterad. Medial side faces posteromediad, gently convex on olecranon, gently concave at middle of shaft. Greatest an¬ teroposterior thickness of olecranon (at upper lip Figure 30.— Proximal part of right humerus of Protosiren species from North Carolina (USNM 214596): a, proximal aspect of head, anterior upward: b, anteromedial aspect. NUMBER 52 53 Figure 31.—Distal part of right humerus of Protosiren species from North Carolina (USNM 244491): a, anterior aspect; b, posterior aspect. Figure 32. — Proximal part of right ulna of Protosiren species from North Carolina (USNM 214596): a, anterior aspect; b, lateral aspect; c, posterior aspect. 54 of semilunar notch) 29 mm; thickness at middle of notch 23 mm. Semilunar notch opens anterad rather than anteroproximad; lateral edge very slightly indented. Proximal part of semilunar ar¬ ticular surface evenly convex mediolaterally; cen¬ ter part occupied by slightly raised, mushroom¬ shaped area extending up from between radioul¬ nar articular facets; distal part bears separate, slightly concave medial and lateral facets, medial facet facing anteroproximad, lateral facing more proximad. Radioulnar articular facets likewise dual, contiguous with anterior edges of distal semilunar facets; medial radioulnar facet flat, pitted, faces anterolaterad, sharply raised above anterior surface of ulna; lateral facet low, slightly convex, indistinctly demarcated. Maximum width at semilunar notch 32 mm. Shaft where broken composed of compact bone; cross-section has asymmetrical teardrop shape, convex ante¬ riorly and laterally, more acutely convex poster- olaterally, and gently concave on posteromedial side, with rather sharp anteromedial point. SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY PInnominate (USNM 214596, Figure 33).— What appears to be the middle portion of a left innominate bears no very close resemblance to innominates of other known sirenians; orientation uncertain. Shaft of ilium composed of compact bone, oval in cross-section, slightly compressed mediolaterally (19X18 mm), narrower end of cross-section ventral; expanded somewhat proxi- mally, much more so distally. Height where bro¬ ken proximally = 27 mm. Dorsal third of proxi- mad-facing surface of fragment occupied by distal side of a deep concavity, smooth surfaced except for low median ridge, with rounded dorsal, me¬ dial, and lateral rims; lateral rim more pro¬ nounced, suggesting an anteromediad orientation of concavity when complete. Lateral surface of ilium occupied distally by longitudinal series of low rugosities. No distinct iliopectineal tubercle present. Distally, fragment broadens abruptly in vertical plane to maximum height of 41 mm; cross-section triangular, medial and dorsolateral sides nearly flat, dorsal edge thickened and Figure 33.—Partial Pleft innominate of Protosiren species from North Carolina (USNM 214596): a, lateral aspect; b, medial aspect; c, broken cross-section. NUMBER 52 55 slightly overhanging laterally, ventrolateral side occupied by large rugose cavity, apparently ace¬ tabulum. Proximodorsal rim of latter irregular, very prominent, causing preserved portion of cav¬ ity to face ventrad; cavity separated by approxi¬ mately 1 cm from anteroventral border of innom¬ inate. Comparisons The North Carolina material is here compared with the previously described Eocene sirenians as was done above with the Florida specimens. Ishatherium subathuensis Sahni and Kumar, 1980. This species, said to be from the lower Eocene of India, is based on an isolated M 2 which shows almost no resemblance to that of USNM 214596: a prominent anterior cingulum is pre¬ sent; all four major cusps are conical, separate, and distinct, with almost no tendency toward lophodonty; and the posterior slope of the talonid is even, with no suggestion of a hypoconulid lophule. Both the age and the sirenian identity of the species have been questioned (see “Old World Records”). Prorastomus sirenoides Owen, 1855: Resembles USNM 214596 in the extreme narrowness of its mandibular symphyseal region, but the relatively straight ventral border of its mandibular ramus, its condylarth-like ear region (Savage, 1977), and the form of its atlas rule it out of consideration. Sirenavus hungaricus Kretzoi, 1941: The holo- type has almost no elements in common with the American form, but Kretzoi states that the molars are “not lophodont,” suggesting a different con¬ dition from that observed here. Anisosiren pannonica Kordos, 1979: M 2 of this form closely resembles USNM 214597 but is con¬ siderably larger (21.3X20.8 mm); even its M 1 (17.4X 18.0 mm) is larger than the North Carolina specimen. It is not clear from Kordos’ illustration whether there is a cuspule at the lingual end of the transverse valley of M 2 as there is on M 3 , but at least a cingulum appears to be present, in contrast to USNM 214597. Protosiren fraasi Abel, 1907: The North Caro¬ lina mandible may without hesitation be consid¬ ered congeneric with the mandible referred to this species by Priem (1907:417, 418, pi. 16) and redescribed and re-illustrated by Sickenberg (1934:63, 64, pi. 1: fig. 9; “Stiick L”), a cast of which (USNM 244490) was available to us (Fig¬ ures 20-22). Protosiren alone among known siren¬ ians other than Prorastomus has the anteriorly compressed symphyseal region seen in USNM 214596 and also shares with the latter the same ventral outline of the ramus, same degree of rostral deflection, and essentially the same dental morphology, though it lacks an accessory mental foramen. Although it is unknown whether DP5 was replaced in Priem’s specimen, the absence of an unerupted P5 in USNM 214596 is problemat¬ ical, for the following reason. Another mandible (YPM 24851, Figure 34) from the late Eocene Qasr el-Sagha Formation of the Fayum, Egypt, although matching the above specimens in most features, clearly shows an unerupted P5 on each side. This mandible apparently represents a new species of Protosiren , advanced beyond the middle Eocene species in at least one respect, greater rostral deflection. If it is presumed to be de¬ scended from P. fraasi , P5 can hardly have been lost in the latter. Whether this tooth had yet to develop in USNM 214596, or whether the North American form had eliminated P5 from its den¬ tition, is unclear; however, Rose and Smith (1979) show that the time of appearance of permanent last premolars in at least some Eocene condylarths was sometimes quite late, so it is not unlikely that in the North Carolina sirenian we see merely a similar case of retarded tooth development. An M 2 with mandible fragment from the mid¬ dle Eocene of Hungary was referred to Protosiren cf. fraasi by Kordos (1978). The photographs of the M 2 are difficult to interpret, but there appears to be an indentation anterolabial to the proto- conid which is not well developed in USNM 214596. The fold connecting hypoconid to hypo¬ conulid is present in the American specimen though not in the Egyptian specimen of Priem (1907). The Hungarian tooth is slightly larger than either of the latter. The mandibular cross- section is similar to that of the American speci¬ men; Kordos states that the Egyptian mandible 56 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Figure 34. — Mandible of Protosiren species from the late Eocene of Egypt (YPM 24851): a, occlusal aspect; b, right lateral aspect; c, ventral aspect. is somewhat different, but this could not be con¬ firmed from the cast we examined. The North Carolina periotic (USNM 244496) does not show the hypertrophied processus fonti- culus of Sickenberg’s Protosiren (1934, pi. 1: fig. 6). The upper molar (USNM 214597), however, cor¬ responds well with Sickenberg’s description of those of P. fraasi (1934:69-71). Its dimensions match those of the holotype (Andrews, 1906:209), and it likewise has a ridge connecting the proto¬ cone and precingulum, but the latter extends across the whole front of the tooth, not just its labial part. The atlas of USNM 214596 contrasts with that of one Protosiren (Sickenberg, 1934, fig. 12) in having an unbridged arterial canal above the anterior cotyle, but the isolated atlas (USNM 244494-5) seems to have had a bridged canal, whereas the vertebrarterial canal through the transverse process is entirely absent (a common variation in sirenians). The fragments of thoracic vertebrae do not closely resemble those referred to Protosiren by Sickenberg (1934:79-84) in either overall shape, lack of compact bone, or reduction of epiphyses. A low postzygapophysis on ?L1 seemingly resembles Protosiren , but the presence of caudal epiphyses again conflicts with Sicken¬ berg’s description. Sickenberg had insufficient material to describe the ribs of Protosiren ade- NUMBER 52 57 quately. He did note a peculiarity of this form— absence of tubercular articular facets in the mid¬ dle thoracic region—which is not evident in the present specimens, though the latter are too in¬ complete to allow certainty in this regard. The distal portion of the scapula matches Sick- enberg’s “St. XLVIII,” referred to Protosiren (1934, pi. 3: fig. 2), but USNM 214596 unmistak¬ ably possessed an unossified scapular cartilage and a “teres major protuberance,” contrary to Sickenberg’s account of Protosiren. The “teres ma¬ jor protuberance” is variably developed in siren- ians and may be unreliable here for diagnosis, whereas Sickenberg (1934:92, 93) believed that the scapular cartilage had “already ossified” in St. XLVIII. To our knowledge such ossification has not otherwise been observed in sirenians, but in such primitive forms we should perhaps expect the unexpected. The humerus of the North Car¬ olina form resembles that of Eotheroides more than that of Protosiren , but as Sickenberg’s example of the latter (“St. XLIX”) is immature and his referrals were largely based on the slightly smaller body size of E. aegyptiacum, they seem open to question. In the ulna of USNM 214596 we indeed encounter the unexpected, as it is quite different from those of E. aegyptiacum (“St. XXX”) and later sirenians. Its similarity of preservation to and association with the other elements, however, seems to make its identification as sirenian incon¬ trovertible; besides, no alternative identification is apparent. As no ulna has previously been re¬ ferred to Protosiren , it is permissible to conclude that this genus possessed an ulna fittingly primi¬ tive for its near-basal position in sirenian phylog- eny. The Pinnominate of USNM 214596 differs con¬ siderably from the pelvic bone illustrated by Abel (1904, pi. 7: fig. 1) and referred to Protosiren by Sickenberg (1934:94) on the basis of the mor¬ phology of associated vertebrae. In particular, the acetabulum is well developed and smooth sur¬ faced in the Egyptian form, whereas in the North Carolina specimen it is extremely rough, pitted, and evidently degenerate. In a specimen from India referred to P. fraasi by Sahni and Mishra (1975), the acetabulum also appears to be rela¬ tively smooth. The overall impression given by the North Carolina fragment is that of a more reduced condition than seen in the innominates referred to Protosiren; its outlines are suggestive of more advanced forms such as Eotheroides libycum. The cavity on the dorsal edge of the ilium, how¬ ever, is not seen in other forms and may be a primitive feature; though Abel (1904) describes no such cavity, he shows (pi. 7: fig. 1) a notch of some sort on the dorsal edge of the ilium in about the same place. Hence the fragment from North Carolina can neither be excluded from Protosiren nor referred to it with any confidence. The North Carolina skullcap (USNM 307609), which may or may not represent the same form as the Comfort quarry material, resembles in its smooth interior surface Sickenberg’s “Stuck II” (1934, pi. 1: fig. 4). In other respects it falls within the range of variation of the Florida skullcaps. ?Protosiren minima (Desmarest, 1822) (= ?P. dubia (Cuvier, 1824); see Hooijer, 1952): The ?M 2 (Sickenberg, 1934, fig. 36c) is somewhat larger than that of the North Carolina form but other¬ wise indistinguishable from it, and the same is true of the worn ?M 2 (1934, fig. 36b). Eotheroides aegyptiacum (Owen, 1875a): This form is distinguished from the Comfort sirenian by its broader mandibular symphyseal surface (Andrews, 1906, fig. 67a; Abel, 1912, pi. 32: fig. 1) and the form of its ulna (Sickenberg, 1934, pi. 4: fig. 6); however, it resembles the latter in several respects: quadratic cross-sections of some ribs, relatively slender build of posterior ribs, presence of a “teres major protuberance” and unossified scapular cartilage (Sickenberg, 1934, pi. 3: fig. 8), and form of humerus (1934:29, 30). The skullcap of E. aegyptiacum differs from the North Carolina specimen in having well-devel¬ oped falx cerebri and tentorium osseum, and probably a higher supraoccipital. Eotheroides abeli (Sickenberg, 1934): The hy- podigm of this species includes a skull referred by Abel (1912) to E. aegyptiacum and on which his description of the latter’s otic region was largely based (Sickenberg, 1934:36). The North Carolina 58 periotic (USNM 244496) resembles this specimen in lacking an enlarged processus fonticulus but otherwise is only generally similar to Abel’s illus¬ trations (1912, figs. 3, 4; pi. 30: figs. 3-5). An M 3 (Abel, 1912, pi. 5: fig. 1) referred here by Sick- enberg, though slightly larger, closely resembles that of USNM 214596, and an M 2 (Sickenberg, 1934:38, 39, pi. 5: fig. 3) resembles USNM 214597 but for its smaller size and obstruction of the transverse valley by a lingual cuspule. A thoracic vertebra of USNM 214596 also resembles T17 of E. abeli (Sickenberg, 1934, pi. 4: fig. 3), but on the whole there are too few preserved elements common to the two forms to allow a satisfactory comparison. “ Eotherium” majus Zdansky, 1938: Molar much larger than USNM 214597. Eotheroides libycum (Andrews, 1902): Differs from the Comfort sirenian in its greater size and much broader, more robust, and more deflected mandibular symphyseal region (Andrews, 1906, pi. 20: fig. 2; also YPM 38213). Its M 2 (Sicken¬ berg, 1934:108, pi. 2: fig. 3) resembles USNM 214597 in size and shape, though its M 3 (1934:109, pi. 5: fig. 2) is more complex than the American form. Its atlas (1934, pi. 4: fig. 8 ) resembles that of USNM 214596, but as indicated above, characters of this element are variable and doubtfully diagnostic. The other vertebrae are likewise similar in the two forms, but Sickenberg does not report the occurrence of quadratic rib sections in E. libycum. The expansion of the supra¬ spinous fossa and development of the acromion (1934, fig. 20) are not seen in USNM 214596, though the humeri are similar (1934:127, 128), whereas the ulna of E. libycum is unknown. The ?innominate of USNM 214596 resembles that of E. libycum in its outlines (Andrews, 1906, fig. 68b) but not in the roughness of its acetabular surface (Sickenberg, 1934:129). The skullcap of E. libycum has well-developed falx and tentorium, unlike USNM 307609. “Eotherium” stromen Sickenberg, 1934: Lacks quadratic rib sections but otherwise has too few diagnostic elements in common with the North Carolina material for useful comparison. SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Prototherium veronense (de Zigno, 1875): Apart from greater size, this species differs from USNM 214596 in less abrupt downturning of the ventral border of the mandibular ramus, and the sym¬ physeal tooth rows are said to be widely separated (Sickenberg, 1934:160, 161). The premolars are more complex, as is M 3 , with its more obstructed valley and massive, multicuspate hypoconulid lophule (1934:170, 171, fig. 27). The supraspinous fossa is broader (1934, fig. 30). A bony falx and internal occipital protuberance are present, in contrast to USNM 307609. Paralithenum tarkanyense Kordos, 1977: This form is larger than the North Carolina sirenian, with a more massive and strongly deflected (50°-55°) mandibular symphyseal region, but it does exhibit quadratic rib sections (as do also the rib fragments from Transylvania; Sickenberg, 1934:183, 184). Post-Eocene Sirenians: As with the Florida material, comparison with post-Eocene forms is here unnecessary, as none even remotely resem¬ bles the Comfort sirenian in compression of the mandibular symphyseal region, tooth formula, or form of the scapula and ulna. The Oligocene genus Miosiren lacks a well-developed falx and tentorium (Sickenberg, 1934:298) but otherwise differs in proportions of the skullcap from USNM 307609. We conclude that the North Carolina sirenian most closely resembles specimens referred to the genus Protosiren , and may be referred to Protosiren species. Apparent similarities of many elements to those of Eotheroides may in part reflect the sharing of primitive characters by many Eocene sirenians but to a large extent could be artifacts of erroneous referrals of isolated elements to var¬ ious taxa, including P. fraasi. The conspecificity of the North Carolina and Florida Protosiren , however, seems questionable. The former lacks the small accessory mental fo¬ ramina observed posteroventral to the principal foramen in all three Florida mandibles. The Com¬ fort upper molar lacks a lingual cingulum, and its precingulum extends prominently across the entire front of the tooth. The Florida vertebrae NUMBER 52 59 contain a greater amount of cancellous bone, and quadratic ribs have not been found in the Wac- casassa River sample. The North Carolina skull¬ cap lacks the prominent falx and tentorium of the Florida specimens. Only more nearly com¬ plete and comparable specimens can reveal whether these differences are significant. Sirenian Dental Formulae and the Cladistic Classification of Mammals Although it has been known for more than a century that Eocene sirenians had five instead of the usual eutherian four premolars, this fact has been overlooked by almost all mammalogists. The well-preserved Protosiren mandible from North Carolina reported herein, together with an undescribed specimen from Egypt (YPM 24851), provide additional confirmation of this primitive condition. This evidence has gained new signifi¬ cance from recent interest in revising supraordi- nal mammalian taxonomy. McKenna (1975) out¬ lined a new classification of the higher taxa of mammals based on cladistic analysis. Both pro¬ ponents and critics of this procedure agree that it has the virtue of vulnerability to falsification. McKenna’s cladistic scheme, which did not take into account the presence of five premolars in early sirenians, is in fact partly falsified by these data. McKenna erected a “Magnorder Preptotheria” to include all known eutherians except edentates, macroscelideans, lagomorphs, possibly rodents, and a few extinct forms. The primitive postcanine dental formula from which preptotheres departed was regarded as DPI P2 P3 P4 P5 Ml M2 M3. All preptotheres except Deltatheridia and Pholi- dota were, in turn, placed in a “Superorder To- kotheria,” primitively characterized by the postcanine formula DPI P2 P3 P4 DP5 Ml M2. In other words, McKenna concluded that all tokotheres have lost P5 and M3, and that the teeth traditionally regarded as Ml-3 in most orders of living mammals are really DP5, Ml, and M2. This hypothesis would be falsified by discovery of a tokothere having five premolars together with three molars, as in the pre-prep- tothere condition. In 1875, Sir Richard Owen redescribed Proras- tomus sirenoides , a primitive member of the order Sirenia (hence a tokothere) from the middle Eocene of Jamaica having the dental formula 3.1.5.3. Recent repreparation and restudy of the unique type specimen have confirmed this (Sav¬ age, 1977). In 1934, Sickenberg described speci¬ mens of Protosiren fraasi , a more advanced sirenian from the middle Eocene of Egypt, as having 11-3 Cl PI P2 P3 P4 P5 Ml M2 M3, with a doubtfully present sixth premolar. A mandible of Protosiren species from the late Eocene of Egypt (YPM 24851, Figure 34) shows an identical 3.1.5.3 for¬ mula, including unerupted P 5 S on both sides. Abel (1912) reported Eotheroides aegyptiacum (mid¬ dle Eocene, Egypt) as having PI P2 P3 P4 DP or P5 Ml M2 M3. In the still more advanced forms Eotheroides libycum (late Eocene, Egypt) and Pro- totherium veronense (late Eocene, Italy), Sickenberg (1934) found 11-3 Cl DP or Pl-2 P3 P4 DP5 Ml M2 M3, P5 apparently having been lost. This latter formula is also seen in the mandible from the middle Eocene of North Carolina (USNM 214596) referred here to Protosiren species. Therefore, eight postcanine tooth positions are regularly present in all of the four best-known genera of Eocene sirenians. In at least some of these, both deciduous (molariform) and perma¬ nent (premolariform) teeth are accounted for at the fifth position, ruling out the possibility that the five apparent premolars include both DP4 and P4. Although late retention of a more ante¬ rior DP together with its replacement cannot be absolutely ruled out, the obviously simpler inter¬ pretation is that primitive sirenians merely re¬ tained the ancestral preptothere condition. The likelihood of discovering early eutherians retain¬ ing this condition has, indeed, recently been pointed out by Bown and Kraus (1979:180). Thus, McKenna’s hypothesis of the loss of M3 is falsified for at least the “Grandorder Ungu- lata,” the division of the Tokotheria which in¬ cludes the Sirenia; hence loss of M3 cannot be used as a defining autapomorphy of the Toko- 60 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY theria. The only remaining autapomorphy of¬ fered by McKenna for the Tokotheria is reduction of the incisors to 3/3. This feature is shared by the “Superorder Leptictida,” which includes most members of the “Magnorder Ernotheria,” sister group of the Preptotheria. The Leptictida, how¬ ever, have lost the DP3-P3 locus and show other autapomorphies not seen in tokotheres, so there is no question of uniting the latter with the former. The defining autapomorphy of the Prepto¬ theria, according to McKenna’s figure 3, is reten¬ tion of DP 5/5 in the adult. As shown above, this does not apply to the most primitive sirenians; at best it could be a convergence between delta- theridians and some tokotheres. This leaves the Preptotheria without a defining autapomorphy. Thus, of the proposed preptothere-tokothere autapomorphies—retention of DP 5/5, loss of M 3/3, and reduction of incisors to 3/3—the first two have been falsified, and the third, which defines the tokotheres, is shared (convergently) with leptictidans. Apparently, the Preptotheria are defined only as plesiomorphic to ernotheres and are, therefore, in danger of being paraphy- letic (if not polyphyletic). If “M 1-3” of any tokotheres are really DP5, Ml, and M2, this must have come about separately in one or more clades of tokotheres other than sirenians. We leave it to specialists in other groups to trace the further ramifications of this situation. Suffice it to say that the Preptotheria and To¬ kotheria as characterized by McKenna are not adequately diagnosed. It may be further ob¬ served, however, that we have here a concrete example of the instability which McKenna (1975:22) warned us to expect from the cladistic method. The “phylogenetic repairs,” or at least redefinitions, called for here are at a fairly ancient site, and the classification, if not crumbling, has at least been jarred in some fairly high categories. Whether this is a cause for “rejoicing” will prob¬ ably depend on whether subsequent “iterations toward a stable . . . system” decrease or increase in amplitude. Meanwhile, we see no reason to make formal classification a tool for identifying confusion in our phylogenetic concepts when the same result can be achieved, by cladistic analysis, without putting our common taxonomic lan¬ guage so constantly at risk. History, Biogeography, and Correlation The earliest geological records of sirenians are in early Eocene rocks of Hungary and India, although both age and identity of the latter have been questioned (see “Old World Records”). Middle Eocene records extend from the Carib¬ bean to India, with a late Eocene occurrence in Java (Table 1; Figure 1). Thus, from almost the beginning of their recorded history, the sea cows have had what can justly be called a “pan-Teth- yan” distribution, being found throughout the length of the Tethyan Seaway, which formed the heart of the Paleogene marine tropics. With the exception of Hydrodamalis , they have remained essentially tropical animals to this day. Consideration of the morphology and relation¬ ships of Eocene sirenians allows more detailed conclusions to be drawn about their early patterns of dispersal. As by general consensus the order’s closest affinities are with moeritheres, probosci¬ deans, and other “subungulate” or “tethythere” groups, we may confidently conclude that they arose in the Old World from an unknown con- dylarth stock. The rather advanced grade of bone density seen in some early Eocene remains (Kret- zoi, 1953) suggests that their aquatic adaptation was well under way by that early date, as would be expected from the diversity already evident in the middle Eocene. (Sahni and Kumar (1980), however, stated that early Eocene specimens which they referred to the Sirenia are not com¬ posed of compact bone, possibly a reason to doubt their sirenian identity.) Though the most primi¬ tive, adequately diagnosable sirenian, Prorastomus, does not appear in the record until the middle Eocene, it is obvious that by then prorastomids were already past their prime and well on the way to being superseded by more derived forms. Perhaps they had already been relegated to a relict, marginal distribution, far from the Old NUMBER 52 61 World center of Tethys (unless Sirenavus is in reality a prorastomid). In any case we must place the origin of prorastomid sirenians in pre-Eocene times. It may be doubted whether Prorastomus was even a fully aquatic mammal, rather than am¬ phibious in pygmy hippo- or tapir-like fashion. Sahni and Kumar (1980) stated that limb and girdle elements which they referred to their lower Eocene genus hhatherium (considered by them to be a dugongid yet “ancestral to all later sireni¬ ans”) showed that this form “was equally adapted to the life in coastal, near-shore conditions as well as on land.” They neglected, however, to substan¬ tiate this by describing the postcranial elements. By the protosirenid grade, at least, we are much more certain of having an obligatorily aquatic organism, in which the modern sirenian morpho- type had appeared in final form save, perhaps, for incomplete reduction of the hind limbs. Pro- tosirenids attained a pan-Tethyan distribution and very likely underwent a modest radiation, giving rise to dugongids (already present in the middle Eocene) and probably, through isolation in South America, to trichechids (Domning, 1982) as well as to apparent post-Eocene proto- sirenids such as Miosiren. By the end of the Eocene, however, protosirenids had already yielded to a third sirenian radiation, that of the dugongids, which have dominated the mammalian marine- herbivore niche down to the present. The roles that high vagility, low diversity of potential food plants, and relative lack of effective geographic barriers have played in limiting sir¬ enian diversity have already been noted (Dom¬ ning, 1978:139). Thus the apparent lack of pro¬ vinciality in Eocene sirenian distribution is not surprising; the Paleocene-Eocene continuity of the Caribbean and Old World Tethyan marine tropics (Berggren and Aubert, 1975) would lead us to expect a relatively homogeneous sirenian fauna. Such a situation has obvious biostrati- graphic potential; however, its realization de¬ pends on establishment of a sound taxonomic and phylogenetic foundation and on the discov¬ ery of relatively complete specimens in farflung areas. It is certainly encouraging to find evidence of Protosiren in middle Eocene rocks of Florida, North Carolina, France, Hungary, Egypt, and India, but until there is some basis for reliable species-level identifications, it would be prema¬ ture to announce the discovery of a new tool for precise intercontinental correlation. Paleoecology of Eocene Sirenians and Seagrasses It is well known that modern sirenians in trop¬ ical marine waters eat mainly seagrasses, marine angiosperms of the families Hydrocharitaceae and Potamogetonaceae (Heinsohn et al., 1977; Hartman, 1979; Best, 1981). As nearly all fossil sirenians also appear to have been both tropical and marine, it seems reasonable to conclude that they also depended on seagrasses (Domning, 1977, 1981). What little is known of seagrass evolution does not contradict this; a seagrass flora already existed in the Early Cretaceous, and the distribution of both fossil and Recent seagrasses shows a close association with (and probably origin in) the Tethyan region (den Hartog, 1970; McCoy and Heck, 1976). Cretaceous seagrasses have been reported from Japan, Germany, and the Netherlands (den Hartog, 1970), and there are unpublished records from North America (Ossian, pers. comm., 1981). Thus sirenians most likely fed on seagrasses from the beginning of their entry into the marine environment. Seagrass fossils, however, are unfortunately rather rare, and some authors have sought to identify other, more common organisms that are regularly associated with seagrasses in order to detect the former presence of the latter. For ex¬ ample, Bafuk and Radwanski (1977) and Hoff¬ man (1977) have attempted this with invertebrate faunas, and Brasier (1975) with Foraminifera. Brasier, however, was unaware of the occurrences of seagrasses in Eocene rocks of Florida (see be¬ low), and he concluded that seagrasses (and spe¬ cifically his “Thalassia association”) did not reach Neotropical waters until the Miocene. This would imply that Paleogene sirenians in the Caribbean 62 did not eat marine angiosperms. Brasier’s recon¬ struction was criticized by Eva (1980), who con¬ cluded (again from Foraminifera) that seagrasses did occur in the Eocene Caribbean, though not as far north as Florida. Dixon (1972), Randazzo and Saroop (1976), and Saroop (1977), however, reported abundant remains of seagrasses, referred by them to the living genera Thalassia , Syringodium, Diplanthera (= Halodule), and Ruppia , in the middle Eocene Avon Park Formation of Levy and Citrus coun¬ ties, Florida—the same unit that yielded many of the sirenian fossils reported herein. The seagrass remains include not only leaf-blade impressions but also entire rhizome mats in growth position, and detailed analysis of the associated fauna, flora, and sediments confirms the existence of a well-developed seagrass community closely re¬ sembling the Thalassia-Syringodium communities typical of Florida today. This evidence decisively refutes Brasier’s (1975) conclusion and removes the need to postulate that Caribbean Eocene sirenians ate no seagrasses (as suggested by Savage, 1977). It also refutes Eva’s (1980) supposition that Eocene seagrasses were absent from Florida. Apparently the rela¬ tionship between Brasier’s “ Thalassia association” and certain Foraminifera is less obligatory than the dependence of sea cows on seagrasses, and we propose that the latter relationship is the more reliable indicator of areas where seagrasses, though unrecorded, once existed (see, for exam¬ ple, Domning, 1977, 1978, with regard to Mio¬ cene seagrasses in California). This conclusion is also more consistent with McCoy and Heck’s (1976:207) thesis that sea¬ grasses, mangroves, and hermatypic corals have shared a common (pan-Tethyan) distribution since the Cretaceous. They predicted that “there should be a number of globally-distributed orga¬ nisms associated with the three groups which are likely a remnant of Tethyan distributional pat¬ terns.” Sirenians clearly answer to this descrip¬ tion, in addition to the coral-reef fishes, decapod crustaceans, mollusks, and miscellaneous plants cited by McCoy and Heck. SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY The degree of diversity and pattern of niche partitioning shown by Caribbean Eocene sireni¬ ans remain in doubt, however. Two forms ( Pro- rastomus and Protosiren ) are known to have been simultaneously present. Judging by the rostral deflection model (Domning, 1977), they should have been able to coexist sympatrically; the ros¬ tral deflection (RD) of the New World Protosiren is about 35°-40°, whereas that of Prorastomus is a scant 6°, much less than in any other sirenian. Together with its generally very primitive aspect, this might suggest a rather different, and possibly less than fully aquatic, lifestyle in Prorastomus than in other sirenians. Certainly Protosiren would seem to be the better adapted of the two for feeding on seagrasses. Among modern sea cows, it compares best with Trichechus manatus (mean RD = 38.2°; SD = 4.88, N = 72), which has a mixed diet of freshwater plants and seagrasses (Hartman, 1979). It is not known what floating or emergent freshwater plants were available to Protosiren , but by analogy with T. manatus it seems likely that some were included in its diet. Had it fed exclu¬ sively on bottom plants such as seagrasses, we would expect its rostral deflection to have been greater (even up to 70° as in Dugong ). No Eocene sirenians with Dugong-\ike snout deflections are known, though by the late Eocene, species of Protosiren and Eotheroides with somewhat increased deflections had appeared. As we see from the Avon Park Formation that seagrass beds of modern aspect already existed in the middle Eocene, it is not clear why specialized bottom¬ feeding sirenians with strong snout deflections should not have evolved by that time. It would seem possible for an extreme bottom-feeder to have coexisted with Protosiren and Prorastomus , as Dioplothenum (RD approximately 70°) evidently coexisted with Dusisiren (RD approximately 41°) in the Miocene of California (Domning, 1978). Such a specialized middle Eocene form may yet be discovered. Conclusions 1. Worldwide distribution of Eocene sirenian fossils is limited to the area of the former Tethyan seaway. NUMBER 52 63 2. Identifiable specimens of Eocene sirenians known from the New World represent Prorastomus sirenoides from Jamaica and Protosiren species from Florida and North Carolina. Protosiren is also re¬ ported from France, Hungary, Egypt, and India. All the above records are middle Eocene, though Protosiren also evidently occurs in the upper Eocene of Egypt. Clarification of the species-level taxonomy of these and other Eocene sirenians may permit their use in intercontinental bio- stratigraphic correlation. 3. All adequately known Eocene sirenians pos¬ sessed a dentition including five premolar posi¬ tions, evidently the primitive eutherian number. They are the latest-surviving eutherians known to have retained this formula. This necessitates revision of some current ideas on the cladistic relationships of the mammalian orders. 4. Distribution of fossil sirenians is evidently a more reliable guide to the past presence of sea- grasses than are the distributions of Foraminifera or other organisms whose relationship with sea- grasses is less obligatory. Literature Cited Abel, O. 1904. Die Sirenen der mediterranen Tertiarbildungen Osterreichs. Abhandlungen der Kaiserlich-Koniglichen Geologischen Reichsanstalt (Wien), 19(2): 223 pages, 26 figures, 7 plates. 1906. Die Milchmolaren der Sirenen. Neues Jahrbuch fur Mineralogie, Geologie und Palaontologie , 2:50-60, 1 figure. 1907. Die Morphologie der Huftbeinrudimente der Ce- taceen. Denkschriften der Mathematisch-Naturwissen- schaftlichen Klasse der Kaiserlichen Akademie der Wis- senschaflen (Wien), 81:139-195, 56 figures. 1912. Die eocanen Sirenen der Mittelmeerregion, Erster Teil: Der Schadel von Eotherium aegyptiacum. Palaeontographica, 59:289-360, 5 figures, plates 30- 34. Andrews, C.W. 1902. Preliminary Note on Some Recently Discovered Extinct Vertebrates from Egypt (Part III). Geolog¬ ical Magazine, new series (Decade IV), 9:291-295, 3 figures. 1906. A Descriptive Catalogue of the Tertiary Verlebrala of the Fayum, Egypt, Based on the Collection of the Egyptian Government in the Geological Museum, Cairo, and on the Collection in the British Museum (Natural History), London, xxxvii + 324 pages, 101 figures, 26 plates, frontispiece. London: British Museum (Natural History). Applin, E.R, and L. Jordan 1945. Diagnostic Foraminifera from Subsurface Forma¬ tions in Florida. Journal of Paleontology, 19(2): 129- 148, 2 figures, plates 18-21. Applin, P.L., and E.R. Applin 1944. Regional Subsurface Stratigraphy and Structure of Florida and Southern Georgia. Bulletin of the American Association of Petroleum Geologists, 28(12): 1673-1753, 38 figures, 5 plates. Arata, A.A., and C.G. Jackson, Jr. 1965. Cenozoic Vertebrates from the Gulf Coastal Plain-I. Tulane Studies in Geology, 3(3): 175-177, 1 plate. Auffenberg, W. 1974. Checklist of Fossil Land Tortoises (Testudinidae). Bulletin of the Florida State Museum, Biological Sciences, 18(3): 121-251, 15 figures. Ball, O.M. 1931. A Contribution to the Paleobotany of the Eocene of Texas. Bulletin of the Agricultural and Mechanical College of Texas, Professional Paper, fourth series, 2(5): 173 pages, 8 figures, 48 plates. Bafuk, W., and A. Radwanski 1977. Organic Communities and Facies Development of the Korytnica Basin (Middle Miocene; Holy Cross Mountains, Central Poland). Acta Geologica Polon- ica, 27(2) :85—123, 6 figures, 12 plates. Bartolomei, G. 1969. Rinvenimento di un Sirenio nei Colli Berici (Vi¬ cenza) . Atti della Accademia Nazionale dei Lincei, series 8 (Rendiconti, Classe di Scienze Fisiche, Mate- matiche e Naturali), 47(2):39, 40, 2 plates. Bataller, J.R. 1956. Contribucion al conocimiento de los Vertebrados terciarios de Espana. Cursillos y Conferencias del In¬ stitute “Lucas Malladaf 3:11-28, 1 figure, 8 plates. Baum, G.R., J.S. Collins, R.M. Jones, B.A. Madlinger, and R.J. Powell 1980. Correlation of the Eocene Strata of the Carolinas. South Carolina Geology , 24(1): 19-27, 3 figures. Baum, G.R., W.B. Harris, and V.A. Zullo 1978. Stratigraphic Revision of the Exposed Middle Eocene to Lower Miocene Formations of North Carolina. Southeastern Geology , 20(1): 1-19, 5 figures. Berggren, W.A., and J. Aubert 1975. Paleocene Benthonic Foraminiferal Biostratigra¬ phy, Paleobiogeography and Paleoecology of At- lantic-Tethyan Regions: Midway-Type Fauna. Palaeogeography, Palaeoclimatology, Palaeoecology, 18(2):73— 192, 20 figures, 19 plates. Best, R.C. 1981. Foods and Feeding Habits of Wild and Captive Sirenia. Mammal Review, 11(1):3—29, 2 figures. Bizzarini, F., B. Bizzotto, and G. Braga 1977. Resti di sirenio ( Protolhenum ) nella marna di Pos- sagno (Eocene superiore) Trevigiano Occidentale. Memorie degli Istituti di Geologia e Mineralogia delPUniversita di Padova, 30:1-15, 5 figures, 2 plates. Bown, T.M., and M.J. Kraus 1979. Origin of the Tribosphenic Molar and Metather- ian and Eutherian Dental Formulae. In J.A. Lil- legraven, Z. Kielan-Jaworowska, and W.A. Cle¬ mens, editors, Mesozoic Mammals: The First Two- Thirds of Mammalian History, pages 172-181, 5 fig¬ ures. Berkeley: University of California Press. 64 NUMBER 52 65 Brasier, M.D. 1975. An Outline History of Seagrass Communities. Pa¬ laeontology, 18(4):681-702, 10 figures. Cole, W.S., and E.R. Applin 1964. Problems of the Geographic and Stratigraphic Distribution of American Middle Eocene Larger Foraminifera. Bulletins of American Paleontology, 47 (212): 1-48, 11 plates. Cook, T.D., and A.W. Bally, editors 1975. Stratigraphic Atlas of North and Central America, [vi] + 272 pages, many maps and stratigraphic sec¬ tions. Princeton: Princeton University Press. Cooke, C.W. 1943. Geology of the Coastal Plain of Georgia. United States Geological Survey Bulletin, 941:vi + 121 pages, 1 figure, 1 plate. Cope, E.D. 1869. Synopsis of the Extinct Mammalia of the Cave Formations in the United States, with Observa¬ tions on Some Myriapoda Found in and near the Same, and on Some Extinct Mammals of the Caves of Anguilla, W.I., and of Other Localities. Proceedings of the American Philosophical Society, 11(82): 171-192, 5 plates. Crusafont-Pairo, M. 1973. Mammalia Tertiaria Hispaniae. In Fossilium Ca¬ talogue, P. Ammalia, 121: iii 4- 198 pages. Cuvier, G. 1824. Les Ossemens de Reptiles et le resume general. In Recherches sur les Ossemens fossiles, ou I’on retablit les caracteres de plusieurs animaux dont les revolutions du globe ont detruit les especes, second edition, 5(2): 547 pages, 33 plates. Paris: G. Dufour et E. D’Ocagne. [Not seen; citation based on third edition, 1825, a re-issue of second edition.] den Hartog. See Hartog, C. den Desmarest, A.G. 1822. Mammalogie ou description des especes de Mammiferes. Part two, pages i-viii, 277-556. Paris: Mme. Veuve Agasse. de Zigno. See Zigno, A. de Dixon, F.S. 1972. Paleoecology of an Eocene Mud-flat Deposit (Avon Park Formation, Claibornian) in Florida, v + 44 pages, 7 figures. Master’s thesis, University of Florida, Gainesville. Domning, D.P. 1974. Fossil Seacows of the Southeast. Rocky Echoes, The Official Bulletin of the Mississippi Gem and Mineral Society, 14(7): 7-9. 1977. An Ecological Model for Late Tertiary Sirenian Evolution in the North Pacific Ocean. Systematic Zoology, 25(4):352-362, 5 figures. [December 1976 issue, mailed 8 February 1977.] 1978. Sirenian Evolution in the North Pacific Ocean. University of California Publications in Geological Sci¬ ences, 118: xi -I- 176 pages, 37 figures, 18 plates. 1981. Sea Cows and Sea Grasses. Paleobiology, 7(4):417— 420. 1982. Evolution of Manatees: A Speculative History. Journal of Paleontology, 56(3):599-619, 9 figures. Emmons, E. 1858. Agriculture of the Eastern Counties; together with Descriptions of the Fossils of the Marl Beds. In Report of the Norlh-Carolina Geological Survey, xvi + 314 pages, more than 256 figures. Raleigh: H. D. Turner. [Reprinted in part, 1969, Bulletins of Amer¬ ican Paleontology, 56(249):57-230, with new index.] 1860. Manual of Geology , Designed for the Use of Colleges and Academies, i-xii + pages 13-290, 218 figures. Phil¬ adelphia: Sower, Barnes and Company. Ericson, D.B. 1945. The Gulf Hammock Formation in Florida. Science, 102(2644) :234. Eva, A.N. 1980. Pre-Miocene Seagrass Communities in the Carib¬ bean. Palaeontology, 23 (1): 231—236, 1 Figure. Fischer, A.G. 1951. The Echinoid Fauna of the Inglis Member, Moodys Branch Formation. Florida Geological Sur¬ vey, Geological Bulletin, 34(11) :45—101, 18 Figures, 7 plates. Flower, W.H., and J.G. Garson 1884. Class Mammalia, Other Than Man. In Catalogue of the Specimens Illustrating the Osteology and Dentition of Vertebrated Animals, Recent and Extinct, Contained in the Museum of the Royal College of Surgeons of En¬ gland, part 2, xliii + 779 pages. London: J. & A. Churchill. Freudenthal, M. 1969 [1970]. Fossiele zeekoeien in het Eoceen van Tau- lanne. Experimenteel Geologisch Onderwijs, 1969/ 70:64, 65. Fuchs, H. 1970. Schadelfragment einer Sirene aus dem Eozan von Cluj, SR Rumanien. Geologie (Berlin), 19(10): 1185-1191, 3 Figures, 1 plate. 1973. Contributiuni la Cunoasterea Sirenidelor Fosile din Bazinul Transilvaniei (IV): Asupra unui frag¬ ment de humerus din Cheia Baciului (Cluj). Studia Universitatis Babe$-Bolyai, Series Geologia-Mineralogia, 18(2): 71-77, 1 Figure. Gibbes, R.W. 1845. Description of the Teeth of a New Fossil Animal Found in the Green Sand of South Carolina. Proceedings of the Academy of Natural Sciences of Phil¬ adelphia, 2 (9): 254-256, 1 plate. 66 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Gingerich, P.D., and D.E. Russell 1981. Pakicetus inachus, a New Archaeocete (Mammalia, Cetacea) from the Early-Middle Eocene Kuldana Formation of Kohat (Pakistan). Contributions from the Museum of Paleontology, University of Michigan, 25(11): 235-246, 5 figures. Gingerich, P.D., D.E. Russell, D. Sigogneau-Russell, J.-L. Hartenberger, S.M. Ibrahim Shah, M. Hassan, K.D. Rose, and R.H. Ardrey 1979. Reconnaissance Survey and Vertebrate Paleontol¬ ogy of Some Paleocene and Eocene Formations in Pakistan. Contributions from the Museum of Paleontol¬ ogy, University of Michigan, 25(5): 105-116, 3 figures, 3 tables. Grigorescu, D. 1967. Asupra prezen(ei unor fragmente scheletice de Sirenide din Paleogenul de la Albe§ti-Muscel. An- alele Universitatii Bucuresti Sena Stiintele Naturii Geo- logie-Geografie, 16( 1):73—78, 2 plates. Hartman, D.S. 1979. Ecology and Behavior of the Manatee ( Tnchechus manatus) in Florida. The American Society of Mam- malogists, Special Publication , 5: viii -I- 153 pages, 40 figures. Hartog, C. den 1970. The Sea-Grasses of the World. Koninkhjke Neder- landse Akademie van Wetenschappen Verhandelingen, Afdeling Natuurkunde, series 2, 59(1): 1-275, 31 plates. Hay, O P. 1923. The Pleistocene of North America and Its Verte- brated Animals from the States East of the Mis¬ sissippi River and from the Canadian Provinces East of Longitude 95° Carnegie Institution of Wash¬ ington Publication, 322: viii + 499 pages, 25 figures, 41 maps. Heal, G.J. 1973. Contributions to the Study of Sirenian Evolution. Part 1, ix 4- 245 pages, 41 figures, 30 plates; part 2, 40 pages, 5 figures, 2 plates. Doctoral disserta¬ tion, University of Bristol, England. Heinsohn, G.E., J. Wake, H. Marsh, and A.V. Spain 1977. The Dugong ( Dugong dugon (Miiller)) in the Sea- grass System. Aquaculture , 12:235-248, 4 figures. Hoffman, A. 1977. Synecology of Macrobenthic Assemblages of the Korytnica Clays (Middle Miocene; Holy Cross Mountains, Poland). Acta Geologica Polonica, 27(2):227-280, 42 figures. Hooijer, D.A. 1952. Fact and Fiction in Hippopotamology (Sampling the History of Scientific Error). Osiris, 10:109-116. Huddlestun, P.F., and J.H. Hetrick 1979. The Stratigraphy of the Barnwell Group of Geor¬ gia. Georgia Geologic Survey Open File Report, 80-1: iv + 89 pages, 15 figures. Huddlestun, P.F., W.E.. Marsalis, and S.M. Pickering, Jr. 1974. Tertiary Stratigraphy of the Central Georgia Coastal Plain. In Geological Society of America, South¬ eastern Section, Twenty Third Annual Meeting, Guide¬ book, 12(Field Trip 2): iv 4- pages 2-1 to 2-35, 16 figures. Atlanta: Georgia Geological Survey. Hunter, M. 1976. Mid-Tertiary Carbonates, Citrus, Levy, and Mar¬ ion Counties, West-Central Florida. Southeastern Geological Society (Tallahassee, Florida), Field Trip Guide Book, 18:2-14, 3 figures. Kellogg, R. 1925. A New Fossil Sirenian from Santa Barbara County, California. Contributions to Palaeontology from the Carnegie Institution of Washington, Publication, 348(111) :57—70, plate 9: figure 3, plates 10 and 11. 1936. A Review of the Archaeoceti. Carnegie Institution of Washington Publication, 482: xv -I- 366 pages, 88 figures, 37 plates. 1966. Fossil Marine Mammals from the Miocene Calvert Formation of Maryland and Virginia, 3: New Species of Extinct Miocene Sirenia. United States National Museum Bulletin, 247:65-98, figures 32-38, plates 33-43. Kellum, L.B. 1926. Paleontology and Stratigraphy of the Castle Hayne and Trent Marls in North Carolina. United States Geological Survey Professional Paper, 143: iv + 56 pages, 1 figure, 11 plates. Kier, P.M. 1980. The Echinoids of the Middle Eocene Warley Hill Formation, Santee Limestone, and Castle Hayne Limestone of North and South Carolina. Smithson¬ ian Contributions to Paleobiology, 39: iv H- 102 pages, 26 figures, 22 plates. Koenigswald, G.H.R. von 1952. Fossil Sirenians from Java. Koninkhjke Nederlandse Akademie van Wetenschappen Proceedings, series B, 55(5):610—612, 1 figure. Kordos, L. 1977. A New Upper Eocene Sirenian ( Paralithenum tar- kanyense n.g.n.sp.) from Felsotarkany, NE Hun¬ gary. Magyar Allami Foldtani Intezet Evi Jelentese az 1975 Evrol, 1977:349-367, plates 1-5. [Pages 349, 350 in Hungarian.] 1978. Major Finds of Scattered Fossils in the Palaeo- vertebrate Collection of the Hungarian Geological Institute (Communication No 3). Magyar Allami Foldtani Intezet Evi Jelentese az 1976 Evrol, 1978:281- NUMBER 52 67 290, 1 plate. [Pages 281-284 in Hungarian.] 1979. Major Finds of Scattered Fossils in the Palaeo- vertebrate Collection of the Hungarian Geological Institute (Communication No 4). Magyar Allami Foldtani Intezet Evijelentese az 1977 Evrol, 1979:313- 326, 1 figure, 2 plates. [Pages 313-316 in Hungar¬ ian.] 1980. Contribution to the Knowledge of Sirenians from the Hungarian Eocene. Magyar Allami Foldtani In¬ tezet Evi Jelentese az 1978 Evrol , 1980:385-397, 1 figure, 2 plates. [Pages 385-389 in Hungarian.] Krauss, C.F.F. 1862. Der Schadel des Halitherium Schinzi Kaup. Neues lahrbuch fiir Mineralogie, Geologic, und Palaeontologie, 1862:385-415, plates 6, 7. Kretzoi, M. 1941. Sirenavus hungaricus n.g. n. sp., ein neuer Prorasto- mide aus dem Mitteleozan (Lutetium) von Fel- sogalla in Ungarn. Annales Musei Nationalis Hungar- ici, Pars Mineralogica, Geologica et Palaeontologica, 34:146-156, 1 figure, plate 6. 1953. A legidosebb magyar osemlos-lelet. Foldtani Koz- lony, 83(7-9):273-277. [In Hungarian; Russian and French abstracts.] Lepsius, G.R. 1882. Halitherium Schinzi , die fossile Sirene des Mainzer Beckens. Abhandlungen des Mittelrheinischen Geolo- gischen Vereins, 1: vi 4- 200 + viii pages, 10 plates. Lydekker, R. 1887. Catalogue of the Fossil Mammalia in the British Museum, {Natural History) Cromwell Road, S. W. Part V, xxxvi + 345 pages, 55 figures. London: British Museum (Natural History). Lyell, C. 1850. A Second Visit to the United States of North America, Volume II. Second edition, xii 4- 385 pages, figures 7-14. London: John Murray. Macfadyen, W.A. 1952. Note on the Geology of the Daban Area and the Localities of the Described Nautiloids. In O. Haas and A.K. Miller, Eocene Nautiloids of British Somaliland. Bulletin of the American Museum of Nat¬ ural History, 99(5):347-349. Maldonado-Koerdell, M. 1953. Segundo hallazgo de sirenidos fosiles en Mexico. Ciencia, 13(7-8): 146-148, 1 figure. McCoy, E.D., and K.L. Heck, Jr. 1976. Biogeography of Corals, Seagrasses, and Man¬ groves: An Alternative to the Center of Origin Concept. Systematic Zoology, 25:201-210, 2 figures. McKenna, M.C. 1975. Toward a Phylogenetic Classification of the Mam¬ malia. In W.P. Luckett and F.S. Szalay, editors, Phytogeny of the Primates, pages 21-46, 3 figures. New York: Plenum Publishing Corporation. Meeder, J.F. 1976. The Mollusca from the Inglis Formation (Upper Eocene, Florida) and Their Zoogeographic Impli¬ cations. vii 4- 104 pages, 4 figures. Master’s thesis, University of Florida, Gainesville. Miillerried, F.K.G. 1932. Primer hallazgo de un sirenido fosil en la Repub- lica Mexicana. Anales del Instituto de Biologia Umv- ersidad Nacional de Mexico, 3(1):71-73, 2 figures. Owen, R. 1855. On the Fossil Skull of a Mammal {Prorastomus si- renoides, Owen), from the Island of Jamaica. Quar¬ terly Journal of the Geological Society of London , 11:541 — 543, plate 15. 1875a. On Fossil Evidences of a Sirenian Mammal {Eo- thrium aegyptiacum, Owen) from the Nummulitic Eocene of the Mokattam Cliffs, near Cairo. Quar¬ terly Journal of the Geological Society of London, 31:100- 105, plate 3. 1875b. On Prorastomus sirenoides (Ow.), Part II. Quarterly Journal of the Geological Society of London, 31:559-567, plates 28, 29. Palmer, W. 1917. The Fossil Seacow of Maryland. Journal of the Washington Academy of Sciences, 7(4): 120. [Reprinted in Science, new series, 45(1162):344.] Piccoli, G. 1966. Segnalazione di un frammento di sirenio {Proto- theriurri) nello Stratotipo del Priaboniano. Bollettino della Societa Geologica Italiana, 85 (2): 349-353, 1 fig¬ ure. Priem, F. 1907. Sur des Vertebres de l’Eocene d’Egypte et de Tu- nisie. Bulletin de la Societe Geologique de France, series 4, 7:412-419, 2 figures, plates 15 and 16. Puri, H.S., and R.O. Vernon 1959. Summary of the Geology of Florida and a Guide¬ book to the Classic Exposures. Florida Geological Survey Special Publication, 5: viii + 255 pages, 11 figures, 11 plates. Rainwater, E.H. 1955. Type Localities Field Trip: Tertiary Type Local¬ ities (Vicksburg, Oligocene; Jackson, Upper Eocene; Claiborne, Middle Eocene; Wilcox, Lower Eocene; and Midway, Paleocene) of Mis¬ sissippi and Alabama. In R. J. Russell, editor. Guides to Southeastern Geology, pages 428-459. New York: Geological Society of America. Randazzo, A.F., and H.C. Saroop 1976. Sedimentology and Paleoecology of Middle and Upper Eocene Carbonate Shoreline Sequences, 68 SMITHSONIAN CONTRIBUTIONS TO PALEOBIOLOGY Crystal River, Florida, U.S.A. Sedimentary Geology, 15:259-291, 12 figures. Ray, C.E. 1975. The Relationships of Hemicaulodon effodiens Cope 1869 (Mammalia: Odobenidae). Proceedings of the Biological Society of Washington, 88(26): 281-304, 6 plates. Reinhart, R.H. 1959. A Review of the Sirenia and Desmostylia. Univer¬ sity of California Publications in Geological Sciences, 36(1): 146 pages, 19 figures, 14 plates. 1971. Fossil Sirenia of Florida. The Plaster Jacket, 15: 10 pages, 5 figures. Gainesville: Florida State Mu¬ seum. 1976. Fossil Sirenians and Desmostylids from Florida and Elsewhere. Bulletin of the Florida State Museum, Biological Sciences, 20(4): 187-300, 39 figures. Renick, B.C., and H.B. Stenzel 1931. The Lower Claiborne on the Brazos River, Texas. The University of Texas Bulletin, 3101, Contributions to Geology, 1931:73-108, figures 9-11. Reves, W.D. 1961. The Limestone Resources of Washington, Holmes and Jackson Counties, Florida. Florida Geological Survey Geological Bulletin, 42: x + 121 pages, 27 figures. Richard, M. 1946. Les Gisements de Mammiferes tertiaires: Contri¬ bution a l’etude du bassin d’Aquitaine. Me'moires de la Sociele Geologique de France , new series, 24(l)(memoire 52): 380 pages, 52 figures. Richards, H.G., and K.V.W. Palmer 1953. Eocene Mollusks from Citrus and Levy Counties, Florida. Florida Geological Survey Geological Bulletin, 35: vi + 95 pages, 13 plates. Robineau, D. 1969. Morphologie externe du complexe osseux tem¬ poral chez les sireniens. Me'moires du Museum Na¬ tional d’Histoire Nature lie. Sene A, Zoologie, new series, 60(1): 32 pages, 17 figures. Rose, K.D., and B. H. Smith 1979. Dental Anomaly in the Early Eocene Condylarth Ectocion. Journal of Paleontology, 53(3):756-760, 2 figures. Sahni, A., and K. Kumar 1980. Lower Eocene Sirenia, Ishalhenum subathuensis, gen. et sp. nov. from the Type Area, Subathu Forma¬ tion, Subathu, Simla Flimalayas, H.P. Journal of the Palaeontological Society of India, 23 & 24(for 1978- 1979): 132-135, 3 figures. Sahni, A., K. Kumar, and B.N. Tiwari 1980. Lower Eocene Marine Mammal (Sirenia) from Dharampur, Simla Himalayas, H.P. Current Sci¬ ence, 49(7):270, 271, 1 figure. Sahni, A., and V.P. Mishra 1975. Lower Tertiary Vertebrates from Western India. Palaeontological Society of India Monograph, 3: 48 pages, 6 figures, 6 plates. Said, R. 1962. The Geology of Egypt, xvm -1- 377 pages, 71 figures, 10 plates. Amsterdam, New York: Elsevier Pub¬ lishing Company. 1963. Note on the Biostratigraphy of the Middle and Upper Eocene Sections in Egypt. Revue de Tlnslitul Franqais du Pe'trole et Annales des Combustibles Liquides, 18(11): 1500-1503. 1965. Egitto (Republica Araba Unita). In Enciclopedia del Petrolio e del Gas Nalurale, 4:8-76. Rome: Editore Carlo Colombo. Sanders, A.E. 1974. A Paleontological Survey of the Cooper Marl and Santee Limestone near Harleyville, South Caro¬ lina (Preliminary Report). South Carolina State De¬ velopment Board, Division of Geology , Geologic Notes, 18(1):4— 12, 4 figures. Saroop, H. 1977. Callianassid Burrows in the Avon Park Formation of Florida. Florida Scientist, 40(2): 160-166, 5 fig¬ ures. Savage, R.J.G. 1969. Early Tertiary Mammal Locality in Southern Libya. Proceedings of the Geological Society of London, 1657:167-171. 1971. Review of the Fossil Mammals of Libya. In C. Gray, editor, Symposium on the Geology of Libya, pages 215-225, 1 figure. Tripoli: University of Libya. 1977. Review of Early Sirenia. Systematic Zoology>, 25(4):344-351, 2 figures. [December 1976 issue, mailed 8 February 1977.] Savage, R.J.G., and B.S. Tewari 1977. A New Sirenian from Kutch, India. Journal of the Palaeontological Society of India, 20(for 1975):216— 218, 3 figures. Savage, R.J.G., and M.E. White 1965. Exhibit: Two Mammal Faunas from the Early Tertiary of Central Libya. Proceedings of the Geolog¬ ical Society of London, 1964-5(1623) :89-91. Scott, J.C. 1966. Geology of Geneva County, Alabama. On Geologic Map of Geneva County, Alabama. In Geological Survey of Alabama, map 54 [ 1 sheet, with text], Sickenberg, O. 1934. Beitrage zur Kenntnis tertiarer Sirenen. Me'moires du Musee Royal d'Histoire Nature lie de Belgique, 63: 352 pages, 52 figures, 11 plates. Siler, W.L. 1964. A Middle Eocene Sirenian in Alabama. Journal of Paleontology , 38(6): 1 108, 1 109. NUMBER 52 69 Simons, E.L. 1968. Early Cenozoic Mammalian Faunas, Fayum Prov¬ ince, Egypt, Part I, African Oligocene Mammals: Introduction, History of Study, and Faunal Succession. Bulletin of the Peabody Museum of Natural History, Yale University , 28:1-21, 103-105, 1 figure. Stenzel, H.B. 1938. The Geology of Leon County, Texas. University of Texas Publication , 3818: 295 pages, 61 figures, 1 plate. Vernon, R.O. 1951. Geology of Citrus and Levy Counties, Florida. Florida Geological Survey Geological Bulletin, 33: xii + 256 pages, 40 figures, 2 plates, von Koenigswald. See Koenigswald, G.H.R. von Voorhies, M.R. 1969. An Eocene Sea Cow Tooth from Twiggs County, Georgia. Bulletin of the Georgia Academy of Science, 27(2):93, 94. Ward, L.W., B.W. Blackwelder, G.S. Gohn, and R.Z. Poore 1979. Stratigraphic Revision of Eocene, Oligocene, and Lower Miocene Formations of South Carolina. Geologic Notes, 23(1): 2-32, 10 figures. [Published by South Carolina Geological Survey.] Ward, L.W., D.R. Lawrence, and B.W. Blackwelder 1978. Stratigraphic Revision of the Middle Eocene, Oli¬ gocene, and Lower Miocene-Atlantic Coastal Plain of North Carolina. Contributions to Stratigra¬ phy, United States Geological Survey Bulletin, 1457- F:F1-F23, 3 figures. West, R.M. 1980. Middle Eocene Large Mammal Assemblage with Tethyan Affinities, Ganda Kas Region, Pakistan. Journal of Paleontology, 54(3):508-533, 1 figure, 5 plates. Weyl, R. 1973. Die palaogeographische Entwicklung Mittelamer- ikas. Zentralblatt fiir Geologie und Palaontologie, 1973 (part 1) (5-6):432-466, 11 figures. Woodring, W.P. 1966. The Panama Land Bridge As a Sea Barrier. Pro¬ ceedings of the American Philosophical Society, 110(6):425-433, 3 figures. Zdansky, O. 1938. Eothenum majus sp. n., eine neue Sirene aus dem Mitteleozan von Agypten. Palaeobiologica, 6(2): 429-434, 1 figure. Zigno, A. de 1875. Sirenii fossili trovati nel Veneto. Memone del Reale Istituto Veneto di Scienze, Leltere ed Arti, 18:1-30, plates 1-5. 1880. Nuove osservationi suWHalilherium veronense Z. Memone del Reale Istituto Veneto di Scienze, Lettere ed Arti, 21:291-296, plate 4. 1881. Nuove aggiunte alia fauna eocena del Veneto. Memorie del Reale Istituto Veneto di Scienze, Lettere ed Arti, 21:775-789, plate 15. REQUIREMENTS FOR SMITHSONIAN SERIES PUBLICATION Manuscripts intended for series publication receive substantive review within their originating Smithsonian museums or offices and are submitted to the Smithsonian Institution Press with approval of the appropriate museum authority on Form SI-36. Requests for special treatment—use of color, foldouts, casebound covers, etc.—require, on the same form, the added approval of designated committees or museum directors. Review of manuscripts and art by the Press for requirements of series format and style, completeness and clarity of copy, and arrangement of all material, as outlined below, will govern, within the judgment of the Press, acceptance or rejection of the manuscripts and art. Copy must be typewritten, double-spaced, on one side of standard white bond paper, with 1 %" margins, submitted as ribbon copy (not carbon or xerox), in loose sheets (not stapled or bound), and accompanied by original art. Minimum acceptable length is 30 pages. Front matter (preceding the text) should include: title page with only title and author and no other information, abstract page with author/title/series/etc., following the establish¬ ed format, table of contents with indents reflecting the heads and structure of the paper. First page of text should carry the title and author at the top of the page and an unnum¬ bered footnote at the bottom consisting of author’s name and professional mailing address. Center heads of whatever level should be typed with initial caps of major words, with extra space above and below the head, but with no other preparation (such as all caps or underline). Run-in paragraph heads should use period/dashes or colons as necessary. Tabulations within text (lists of data, often in parallel columns) can be typed on the text page where they occur, but they should not contain rules or formal, numbered table heads. Formal tables (numbered, with table heads, boxheads, stubs, rules) should be sub¬ mitted as camera copy, but the author must contact the series section of the Press for edito¬ rial attention and preparation assistance before final typing of this matter. Taxonomic keys in natural history papers should use the alined-couplet form in the zoology and paleobiology series and the multi-level indent form in the botany series. If cross-referencing is required between key and text, do not include page references within the key, but number the keyed-out taxa with their corresponding heads in the text. Synonymy in the zoology and paleobiology series must use the short form (taxon, author, yearpage), with a full reference at the end of the paper under “Literature Cited.” For the botany series, the long form (taxon, author, abbreviated journal or book title, volume, page, year, with no reference in the "Literature Cited”) is optional. Footnotes, when few in number, whether annotative or bibliographic, should be typed at the bottom of the text page on which the reference occurs. Extensive notes must appear at the end of the text in a notes section. If bibliographic footnotes are required, use the short form (author/brief title/page) with the full reference in the bibliography. Text-reference system (author/year/page within the text, with the full reference in a "Literature Cited” at the end of the text) must be used in place of bibliographic footnotes in all scientific series and is strongly recommended in the history and technology series: “(Jones, 1910:122)” or . . Jones (1910:122).” Bibliography, depending upon use, is termed "References,” “Selected References,” or "Literature Cited.” Spell out book, journal, and article titles, using initial caps in all major words. For capitalization of titles in foreign languages, follow the national practice of each language. Underline (for italics) book and journal titles. Use the colon-parentheses system for volume/number/page citations: "10(2):5-9.” For alinement and arrangement of elements, follow the format of the series for which the manuscript is intended. Legends for illustrations must not be attached to the art nor included within the text but must be submitted at the end of the manuscript—with as many legends typed, double¬ spaced, to a page as convenient. Illustrations must not be included within the manuscript but must be submitted sepa¬ rately as original art (not copies). All illustrations (photographs, line drawings, maps, etc.) can be intermixed throughout the printed text. They should be termed Figures and should be numbered consecutively. If several "figures” are treated as components of a single larger figure, they should be designated by lowercase italic letters (underlined in copy) on the illus¬ tration, in the legend, and in text references: "Figure 9b,” If illustrations are intended to be printed separately on coated stock following the text, they should be termed Plates and any components should be lettered as in figures: "Plate 9b," Keys to any symbols within an illustration should appear on the art and not in the legend. A few points of style: (1) Do not use periods after such abbreviations as "mm, ft, yds, USNM, NNE, AM, BC.” (2) Use hyphens in spelled-out fractions: "two-thirds." (3) Spell out numbers "one” through "nine” in expository text, but use numerals in all other cases if possible. (4) Use the metric system of measurement, where possible, instead of the English system. (5) Use the decimal system, where possible, in place of fractions. (6) Use day/month/year sequence for dates: “9 April 1976.” (7) For months in tabular list¬ ings or data sections, use three-letter abbreviations with no periods: "Jan, Mar, Jun,” etc. Arrange and paginate sequentially EVERY sheet of manuscript —including ALL front matter and ALL legends, etc., at the back of the text—in the following order: (1) title page, (2) abstract, (3) table of contents, (4) foreword and/or preface, (5) text, (6) appendixes, (7) notes, (8) glossary, (9) bibliography, (10) index, (11) legends.