LIBRARY OF THE UNIVERSITY OF ILLINOIS AT URBANA-CHA^1PAIGN GEOLOGY UNfVERSITYOF 'LLINOfS LIBRARY AT URBANA-CHAMPA/GN GEOLOGY - FIELDIANA • GEOLOGY Published by CHICAGO NATURAL HISTORY MUSEUM Volume 14 September 26, 1960 No. 2 A New Specimen of Desmatochelys lowi Williston A Primitive Cheloniid Sea Turtle From the Cretaceous of South Dakota Rainer Zangerl Curator of Fossil Reptiles AND Robert E. Sloan Dbpartubnt of Geology and Mineralogy, University op Minnesota In 1953, in the course of stripping overburden from the Dakota Rose Granite quarry in the Milbank-Ortonville granite district in eastern South Dakota and western Minnesota, a Cretaceous marine vertebrate fauna was found. The quarry is located in Grant County, South Dakota (SW. H, Sec. 18, T. 120 N., R. 47 W.), and is operated by the Dakota Granite Company, Milbank, South Dakota. The fossils were found in an arkosic chalky marl that was depos- ited on top of the early Precambrian Ortonville granite described in detail by E. H. Lund (1956). The depositional environment of the Cretaceous sediment was a marine boulder beach on one of a series of islands in the Milbank-Ortonville area on the west flank of the Sioux Arch. The Cretaceous sediment is thin, not over five feet in thickness, and varies greatly from place to place. The granite sur- face was deeply weathered prior to the transgression of the Creta- ceous sea from the west. Joint cracks were enlarged and widened by weathering to notches 3-4 feet wide and as deep, and were filled with the arkosic sediments as the seas transgressed over the islands, eventually submerging them. Resting directly on the granite is a j basal, marine beach boulder conglomerate with the maximum diam- eter of the boulders about eight feet (average diameter about two feet). The boulders are all well rounded and are composed of the underlying granite. Finer sediments from coarse sands to marls are found between and above the boulders. Large angular fragments of orthoclase feldspar and quartz derived from the underlying granite are common in these sediments, in places amounting to 50 per cent Library of Congress Catalog Card Number: 60-1 USUt No. 897 7 ilJE LIBRARY OF THE OCT 19 1960 GEOLOGY LIBRARY 8 FIELDIANA: GEOLOGY, VOLUME 14 of the sediment, particularly in the widened joint cracks. Sharks' teeth and small teleost vertebrae are sufficiently abundant in some places to make up a significant proportion of the rock, up to about 5 per cent. The clay component of the marl seems to be a kaolinitic weathering residuum of the granite similar to that described by Goldich (1938) from the vicinity of Redwood Falls, Minnesota. Thiel (1944) reports similar material to be common, as the basal Cretaceous sediment, where Cretaceous rocks rest on granite in southwestern Minnesota. The surfaces of the exposures have been glacially scoured and the finer-grained Cretaceous sediments are to be found only on the lee side of the beach boulders, in the widened joint cracks, or in other protected areas. Similar Cretaceous sediments, most exposures of which contained sharks' teeth and small teleost vertebrae, were found at the Cold Spring Granite Company quarry, one-fourth of a mile west of the Dakota Rose quarry (SE. M, Sec. 13, T. 120 N., R. 48 W.), in Grant County; the Melrose Granite Company quarry and the D. G. Kad- dez and Company quarry (NE. %, Sec. 17, T. 120 N., R. 47 W.), in Grant County, South Dakota; and in the series of quarries and out- crops 3 miles north of the town of Bellingham in Lac Qui Parle County, Minnesota (fig. 2). Any Cretaceous sediments in the other active quarries in the area or at other outcrops were removed at some time prior to the deposition of the glacial drift. Five miles west of the granite outcrops, in Milbank, a well re- ported by the South Dakota Geological Survey (Petsch, 1948) pen- etrated 41 feet of glacial drift, 188 feet of Carlile shale, 43 feet of Greenhorn limestone and at least 43 feet of Graneros shale. The Milbank-Ortonville granite area was a topographic high during early Cretaceous time, as shown by geophysical studies of the surface of the granite by the South Dakota Geological Survey. In Cretaceous sediments, very slight dips prevail throughout the area of outcrop in southwestern Minnesota, and the only form of deformation re- corded is that due to compaction. These facts and the elevations of the outcrops in the Milbank-Ortonville granite district suggest a correlation of the conglomerate and arkosic marl of the latter area with the Carlile shale, the upper formation of the Benton Group. FAUNA Teeth of a small species of the shark genus Isurus are the most common fossils; next in abundance are the teeth of another shark. s Se^O i O 3/ ZANGERL AND SLOAN: CHELONIID SEA TURTLE R48W R47W R46W 3 4 5 6 SCALE OF MILES R45W .I.UVK WALL OF MINNESOTA RIVER VALLEY * GRANITE OUARRV 'I...- AREA OF CRETACEOUS EXPOSURES Fig. 2. Map of Milbank-Ortonville granite district in eastern South Dakota and western Minnesota. Squalicorax. Those of the ptychodont shark Ptychodus janewayii are rare. Isolated vertebrae of small clupeoid fishes are very numerous and a species of Ichthyodectes is present. Teleost fish scales up to an inch in diameter are fairly common. A few amphiplatyan vertebral centra (possibly plesiosaurian) and isolated reptile teeth have been collected. The most important reptile remains so far discovered is a large portion of the skeleton of a marine turtle, Desmatochelys lowi Williston, which will be described below. Pyritized microfossils and carbonized scraps of gymnosperm wood are relatively abundant. CIRCUMSTANCES OF COLLECTION OF THE TURTLE SKELETON i In the process of stripping the overburden off the top of the gran- ite quarry the turtle skeleton was exposed among rubble loosened by 10 FIELDIANA: GEOLOGY, VOLUME 14 explosives. Several pieces of sediment containing bones were col- lected by personnel of the Museum of the South Dakota School of Mines, Rapid City. Another piece found its way into a bar in Orton- ville, Minnesota, where it was handed to Dr. Lloyd A. Wilford of the Department of Anthropology, University of Minnesota, who in turn brought it to the attention of the junior author. The late Dr. James D. Bump kindly gave us the pieces in the collection of the South Dakota School of Mines and permitted deposition of the specimen at Chicago Natural History Museum. The pieces thus reunited fit together perfectly, but part of the skeleton was evidently lost in the blast and some damage has been inflicted on the parts that have been recovered. As preparation of the skeleton neared completion, it became possible to identify the turtle as Desmatochelys lowi Williston, originally described from the Benton Group near Fairbury, Nebraska. The bones of the type ma- terial of this species are crushed to some extent, whereas the new skeleton is not. For this reason numerous minor differences may be noted between the two specimens, but these reflect, we feel confident, mostly differences in the preservation of the bones. The identifica- tion is further strengthened by the similar stratigraphic position of the finds and the fact that the new material cannot be placed in any other group of Cretaceous sea turtles, such as the Protostegidae, the Toxochelyidae or the other members of the Cheloniidae. Since Desmatochelys lowi is inadequately known, the description of the new find substantially increases our knowledge of this form and sheds new light on the question of origin of the sea turtles dur- ing Cretaceous time. Order Chelonia Suborder Cryptodira Family Cheloniidae Desmatochelys lowi Williston Referred material. — CNHM-PR385, a partial skeleton, lacking most of the skull and the posterior part of the shell. Locality and horizon.— Cretaceous overburden on top of Dakota Rose Granite quarry (SW. H, Sec. 18, T. 120 N., R. 47 W.), Grant County, South Dakota. Near top of Benton Group, probably Car- lile shale equivalent. Description. — 1. The skull: Only a single skull element, the right postorbital bone, was recovered. It is essentially complete except for ZANGERL AND SLOAN: CHELONIID SEA TURTLE 11 most of the posterior rim (fig. 3), of which only a very small section next to the suture with the parietal was preserved. This, however, shows the position of the posterior edge of the skull roof, very deeply excavated at this point in Desmatochelys (Williston, 1894, 1898). As nearly as can be determined from Williston's illustration, the postorbital bone of the present specimen is virtually identical with Fig. 3. Desmatochelys lorn (CNHM-PR385), postorbital bone in dorsal (left) and ventral (right) views. that of the type specimen, even including the slight curve in the course of the parieto-postorbital suture. Williston illustrated both the dorsal and palatal views of the skull; these are correct except for a few details.^ The lateral aspect of the skull of the type specimen (fig. 4) shows an unusual development of the cheek bones with a very prominent quadratojugal and a jugal restricted in its dorsal extent to the lower margin of the orbit. 1 Dr. Theodore H. Eaton, Jr., has kindly examined the skull of the type speci- men of Desmatochelys lowi and has rendered drawings (fig. 4). We wish to express our sincere gratitude to Dr. Eaton for his meticulous re-examination; with his kind permission we quote from his account: "The dorsal view is traced over Wil- liston's (1894) pi. II, which is mainly correct except for minor differences in the sutures (esp. prefrontal) and the length of the occipital spine. His palatal view (pi. Ill) I have not redrawn; it is correct as far as it goes, but there is in addition a median suture separating the pterygoids as well as the palatines, and the vomer comes back to a point almost half the distance between the choanae and the trans- verse suture. With more preparation the posterior part of the cranium would give many details of value, although a hole }/2 inch in diameter has been drilled verti- cally up into the basicranial region. The actual length of the skull to the point where the occipital spine is broken off is 215 mm., not 205, and its probable length when complete would be about 240. The scale of figs. II and III is approximately ^ rather than %. "In lateral view the squamosal, quadratojugal and jugal are plainly visible. The quadratojugal reaches the margin of the orbit on the right side but does not quite do so on the left. In the articular region on both sides there is some doubt, because the quadrate seems to have been pushed up and what was probably an oval tympanic cavity is a narrow slit, as shown. I think the sutures are correct, as given, but possibly one or two of them are cracks." 12 FIELDIANA: GEOLOGY, VOLUME 14 Pmx Fig. 4. Sketch of skull of Desmatochelys lorn, type specimen, in dorsal and side views (sketch kindly provided by Dr. Theodore H. Eaton, Jr.; see also com- ments in footnote, p. 11). 2. Vertebral column: Of the neck vertebrae the arches of the atlas and the major portions of the fourth and sixth through the eighth are preserved (pis. 1 and 2). The most striking feature of these cervical vertebrae is their shortness compared to those of all other known sea turtles, with the possible exception of the proto- stegids (fig. 5). Aside from the fact that vertebrae 6 to 8 were artic- ulated with the shell vertebrae in the matrix prior to preparation, the overall morphology of these elements is sufficiently similar to that of the modern sea turtles to permit the determination of the numerical position within the neck region. In almost every detail, however, there are obvious differences. The articulation pattern of the vertebral centra agrees with the common condition in the Denna- w T T si a o (3 B o a O U5 13 Oo '^•<- II »^^ C! > eo C _ Si o W W o J-! -2 cs ^^ S ? -^ '^^^ JS W . W (M S • '-' -B in ec to Z Ql 13 14 FIELDIANA: GEOLOGY, VOLUME 14 temydidae and the Chelydridae (Williams, 1950) and the normal condition in the Toxochelyidae (Zangerl, 1953) as follows: 3 (,(4), 5 ,)6),)7),)8) This condition occurs relatively rarely in modem cheloniine sea turtles. The atlas arches compare quite well with those of Caretta except that the postero-lateral processes are sharply set off from those por- tions of the arches that are attached to the atlas centrum ; in Caretta the two areas are connected with a sheet of bone. Furthermore, the antero-dorsal processes are relatively higher than in Caretta and the facets of the postzygapophyses are round instead of oval as in the compared genus. The latter difference probably reflects the relative shortness of these elements. The fourth vertebra is very short and possesses a weak neura- pophysis whose suture with the centrum lies above the transverse processes. A pair of low ridges not present in Caretta connects the lateral edges of the prezygapophyses with the base of the postzyga- pophyses. Two shallow pits are located medial to the mentioned ridges and these appear to have served as areas of attachment for muscles or ligaments (pi. 1, fig. d) ; comparable pits are only vaguely indicated in Caretta. Vertebrae 6, 7, and 8 differ notably from those in Caretta, but this is largely due, we think, to the fact that the artic- ulation of the centra is of the normal and very probably primitive ball-and-socket type rather than the transversally expanded, flat- jointed (between centra 6 and 7), or double-jointed (between 7 and 8) condition that constitutes the norm in modern sea turtles. The greater range of movability that exists between these vertebrae in Desmatochelys (and the shortness of the elements; see above) may be reflected in the details of differentiation of the neural arches as com- pared with the condition in modern sea turtles. The postzygapophy- ses of vertebra 6 are separated by a deep, acute, V-shaped notch with near- vertical walls descending to the neural canal. In Caretta (see pi. 31, B Zangerl, 1960) this area is flat and the notch is shallow. The postzygapophyseal joint surfaces of vertebra 8 face strongly postero-dorsad and laterad (pi. 1, fig. g). In Caretta the transverse processes of vertebrae 6, 7 and 8 tend to divide into dual processes, the upper, small one being located at the base of the prezygapophyses. These smaller, dorsal processes are absent in Desmatochelys. The major proportions of the neck vertebrae as graphically pre- sented in figure 5 show clearly primitive relations in vertebra 8, ZANGERL AND SLOAN: CHELONIID SEA TURTLE 15 where the centrum is not significantly shortened, as in modem sea turtles, and where the neurapophysis is shorter than the centrum, as in Chelydra; in Recent sea turtles and in toxochelyids (Zangerl, 1953) it is longer. There are other peculiarities in this element. The hypapophysis is not a single ventral sheet of bone as it is in the anterior neck verte- brae and in all the cervicals (save the atlas centrum) of Caretta; in- stead there are two sharp, longitudinal ridges, diverging slightly in posterior direct on, along the mid-ventral face of the centrum (pi. 2, fig. g) . A similar condition of the hypapophysis of the eighth cervical vertebra is present in Dermockelys (Volker, 1913, pi. 31). The post- zygapophyses, with their characteristic dorsal projection that articu- lates with a knob on the under side of the nuchal plate in modem and some fossil sea turtles, are less expanded but higher than in Caretta. This seems to be correlated with the lack of a nuchal knob in Desma- tochelys, and with the fact that in this form the eighth vertebra con- tinues the gentle curve formed by the shell vertebrae from the apex of the carapace forward, instead of forming a pronounced angle, in side view, with the first shell vertebra. These features indicate that the neck of Desmatochelys was even less retractable than that of modem sea turtles. The anterior seven shell vertebrae are preserved in articulation with the carapace (fig. 6). The centra are of the normal cheloniine construction, ventrally rounded rather than angular and provided (near mid-length and close to the dorsal border) with small foramina for the exits of the spinal nerves, as in modem cheloniines. The third centrum is the longest. The centrum of the first vertebra re- sembles, particularly in ventral aspect, the succeeding centra; the anterior, procoelous end faces straight forward, rather than forward and downward as in modem cheloniines (see also discussion of eighth cervical vertebra). The reduced ribs (first pair of shell ribs) are rela- tively longer than in cheloniines and are only distally attached to the antero-ventral face of the first pair of costal plates; apparently there was no proximal connection with the free ends of the second shell ribs. The neurapophyses form a continuous sagittal ridge along the ventral faces of the neural plates from the second one backward; the neura- pophysis of the first shell vertebra is modified and has a long, rela- tively thick dorsal spine, connected to the anterior end of the first neural plate by connective tissue only. Posteriorly it is suturally connected with the second neurapophysis. The prezygapophyses are strongly developed. The first neural plate is suturally connected to the neurapophysis of the second shell vertebra (figs. 6 and 8). All Fig. 6. Desmatochelys lowi (CNHM-PR385), last cervical vertebra and under side of articulated part of carapace. 16 ZANGERL AND SLOAN: CHELONIID SEA TURTLE 17 this differs from the situation in modem cheloniid turtles in but minor details. 3. The carapace: The carapace is elongated, but only vaguely cordiform. The antero-lateral margins above the forelimbs are some- FiG. 7. Desmatochelys lorn (CNHM-PR385), nuchal plate and first periph- erals. A, dorsal, and B, ventral views. what excavated in typical sea turtle fashion. The nuchal margin is not incised. There are large lateral fontanelles extending from the peripherals about halfway to the midline and from the nuchal plate to the pygal. Postnuchal fontanelles are present also (fig. 13). The nuchal plate (fig. 7) forms an obtuse-angled arch from side to side and lacks a mid-ventral boss. Near the midline it is quite thick (13.5 mm.) and gradually becomes thin toward the posterior margin, except for the usual thick ridges that run in the shape of a V back to the suture with the first pair of costal plates. The sutural connection with the first neural is very weak. 18 FIELDIANA: GEOLOGY, VOLUME 14 The anterior six neural plates (fig. 8) are long, unkeeled and fairly- flat in front, and gently arched from side to side farther back. The first and second neurals are narrower than the following four, as sometimes in Chelonia. The first neural is broken, and part of its posterior half is crushed down upon the neurapophysis. Its antero- lateral edges are extremely thin, which indicates the presence of postnuchal fontanelles. Posteriorly the first neural is about 5 mm. thick. Five pairs of costal plates are retained in articulation (figs. 6 and 8); in addition there are fragments of the sixth and seventh and nearly the whole eighth (fig. 9) of the left side. The costal plates are flattened and slightly warped so that they could not be mounted in perfect sutural union. The ribs beneath the costal plates are broad and flat and there is no sharp distal delimitation between costal plates and ribs much as in Chelosphargis among the protostegids (Zangerl, 1953) and a num- ber of cheloniids. At the distal ends the ribs are subcircular in cross section. The only peripherals (figs. 7 and 10) that are definitely determin- able as to position in the carapace are the first pair, part of the right second in sutural connection with the third, and the posterior half of the left third. There remain six whole or partial elements whose positions are somewhat open to question. We have interpreted them as the eighth of the right side and 9, 10, and 11 of the left side (be- cause of color differences in the matrix on the right and left sides of the skeleton) . Williston (1894, 1898) figured the pygal and the right eleventh peripheral. His description of the latter corresponds gen- erally with our eleventh peripheral except for a notable difference in width; this is almost certainly due to crushing in the type specimen. The cross sections of the peripherals interpreted as posterior ele- ments have the shape of airfoils, with the thicker edges facing the carapace disc and the blade-like sides forming the peripheral rim of the shell. Two broken elements are certainly bridge peripherals. They are triangular in cross section; the widest faces bear the rib pits. The pattern of the epidermal shield cover is usually imprinted upon the bones of the shell by sulci formed at the junction of adja- cent shields. In the specimen here described there are no sulci; in- stead, the shield boundaries are indicated by ridges (figs. 7 and 8) and are not equally visible in all parts of the carapace. The signifi- cance of this is not presently understood. The cervical shield is large as in modem cheloniines and the pattern as a whole (fig. 13), in so far Fig. 8. Desmatochelys lovri (CNHM-PR385), last cervical vertebra and dorsal side of articulated part of carapace. 19 s^ Fig. 9. Desmatochelys loivi (CNHM-PR385), eighth left costal plate in dorsal and ventral views. bridge ptripheral Fig. 10. Desmatochelys lovxi (CNHM-PR385), isolated peripheral plates in dorsal and ventral views. 20 ZANGERL AND SLOAN: CHELONIID SEA TURTLE 21 as can be determined, shows no unusual features; it conforms to that of the modern Chelonini (Zangerl, 1958) and all of the fossil chelo- niids, toxochelyids and protostegids (Zangerl, 1953). 4. The plastron : The bones of the plastron are more fragmentary than those of the carapace. Most of the left epiplastron and part of Fig. 11. Desmatochelys lorn (CNHM-PR385), epiplastra and entoplastron in dorsal view. the right one, the entoplastron, and portions of the left hyo- and hypoplastra are preserved. All parts were pressed in more or less natural position onto the ventral side of the carapace, except the lateral fragments of the hyo- and hypoplastra, which were loose and had no matrix contact with the rest of the specimen. The xiphi- plastra are missing. Epi- and entoplastra present no problems (fig. 11). Hyo- and hypoplastra, on the other hand, are quite incompletely preserved, and the reconstruction (fig. 12), while it very probably represents the correct relationships, remains open to some doubt as far as the exact shapes of the plates are concerned. The loose fragment, read- ily identifiable as the central portion of the hypoplastron, unques- tionably belongs to the medial fragment that was left in place. Its dorsal and ventral surfaces are clearly distinguishable and its curva- ture (ventrally convex from side to side) leaves no doubt that it be- longs to the left side; its exact position with regard to the medial fragment, however, cannot actually be determined; figure 12 shows what we believe to be the most likely position. 22 FIELDIANA: GEOLOGY, VOLUME 14 The fragment here identified as part of the lateral wing of the left hyoplastron obviously presents greater difficulties. The position in- dicated in figure 12 conforms to the surface characteristics of the fragment, its thickness and its natural edges; apparently it cannot be placed anywhere else, but its exact position in the area indicated remains, of course, doubtful. The plastron as presently interpreted (fig. 13) is clearly of chelo- niid character with a plastral index (Zangerl, 1953 and 1958) esti- mated at about 105. The anterior plastral lobe is very short and blunt and there are large medial and lateral fontanelles. The epiplastra are flat, blade-like bones anteriorly expanded and were suturally attached to the antero-lateral edges of the entoplas- tron. Prominent ridges follow the mentioned sutures along the dor- sal faces of the bones (fig. 11). The entoplastron is approximately T-shaped; the lateral processes face outward and backward (fig. 11). The hyoplastron has a broad antero-medial and three clearly sep- arated medial prongs; the preserved fragment is thin except near the center of the plate where it is 9 mm. thick. The lateral fragment measures 13 mm. in thickness along the break facing the center of the plate; its anterior margin is thick and crested; posteriorly its thickness is 7 mm. and the edge is blunt. The hypoplastron has three strong medial and an undeterminable number of smaller pos- tero-medial prongs. The sutural notch for the union with the xiphi- plastron is well developed (fig. 12). It reaches its greatest thickness, 13 mm., near the center of ossification. 5. Shoulder girdle : The left half of the shoulder girdle is perfectly preserved in uncrushed condition; on the right side parts of the bones are missing. The scapula (fig. 14) is remarkable in that its dorsal and ventral processes are of nearly equal length as in Corsochelys (Zangerl, 1960) and in Dermochelys coriacea. The two processes stand at an angle of about 103° to each other, a much wider angle than in the type specimen (88°) (see Williston, 1894, 1898). From Willis- ton's figure it is evident that the type scapula suffered crushing — a process that might well have reduced the value of the angle. The scapula is a relatively heavy bone, not nearly as much flattened in antero-posterior direction as in modern cheloniines. The coracoid (fig. 15) is long, slender and only moderately expanded at its poste- rior end. There is a very close resemblance between this bone and its homologue in Eretmochelys. The bone labeled coracoid by Wil- liston seems to be a badly crushed proximal fragment. 6. Humerus and flipper: The proximal half of the left humerus was articulated with the scapula; the distal end was broken off in Fig. 12. Desmatochelys lorn (CNHM-PR385), parts of hyo- and hypoplastron of left side in ventral view. 23 Oh I o d •a c3 24 ventral limb Fig. 14. Desmatochelys lowi (CNHM-PR385), left scapula. Fig. 15. Desmatochelys Ixmi (CNHM-PR385), left coracoid. 25 26 FIELDIANA: GEOLOGY, VOLUME 14 the blasting operation and a small portion of the shaft was lost. In the right humerus a larger portion of the shaft was lost. In the left humerus the two portions were united according to the dimensions of the break surfaces and the surface configuration and we feel con- fident that the length of the bone and the position of the distal versus the proximal ends are very nearly correct (fig. 16). Comparison of the present material with the type specimen (Williston, 1898, pi. 75) shows very notable differences, which are, however, not necessarily of a systematic nature. The type humerus is obviously flattened by severe crushing, the ends more so than the middle of the shaft, which consists of denser bone. Accordingly, the ends, especially the prox- imal half of the bone, are spread in the plane in which the bone was preserved and thus give the impression of far greater massiveness than was true in life. Besides differences clearly due to preservation there are others that indicate a difference in the individual ages of the two specimens; these are the same differences that can be ob- served, for example, in the humeri of mature and very old individuals of modern cheloniines : in the latter we observe full osteogenetic dif- ferentiation with very pronounced, rough muscle attachment scars and ridges, a long ulnar process lacking an area occupied by a carti- lage cap, an ectepicondylar canal entirely enclosed by bone, and the distal end surface devoid of a cartilage cap except for a thin layer covering the actual joint surface. Accordingly, the new specimen must have been somewhat younger than the type specimen. Taking these factors into account, we may characterize the hu- merus as follows: It is massive and stout, provided with a very pro- nounced ulnar and an unusually big radial tuberosity. This large proximal region is followed by a short, slender shaft and a normally expanded, but unusually thick (32 mm. in the present specimen) dis- tal end (fig. 16; Williston, 1898, pi. 75). The degree of marine speciali- zation (angles a and /3, Zangerl, 1953, p. 165) of the humerus falls in the range of modern cheloniine sea turtles. Since the shaft is very short, the radial tuberosity lies close to mid-length between caput humeri and distal joint surface; in its position it is thus intermediate between the condition in modern cheloniines and that in the proto- stegids. The attachment surface of the radial tuberosity is an equi- lateral, triangular area (fig. 16); it contains a pit of "finished" bone surface that is located near the dorsal side of the triangle. In chelo- niine turtles the radial tuberosity forms a V-shaped attachment ridge with the apex of the V pointing toward the caput. The pit within the triangular area in Desmatochelys thus corresponds to the depres- sion between the limbs of the V-shaped ridge in cheloniines. The 27 28 FIELDIANA: GEOLOGY, VOLUME 14 joint surface of the caput humeri is nearly hemispherical, more exten- sive than in cheloniines, and this indicates, in conjunction with the more pronounced radial tuberosity, that the flipper was capable of greater axial rotation in this form than in modem cheloniines. A steeper forward incline of the flipper surface during the downstroke would result in quicker acceleration. A fair portion of the right flipper was preserved very nearly in proper articulation (fig. 17) . The carpal elements were not entirely freed from matrix because of negative surfaces in the matrix that indicate the position of elements that have been lost, and because it seemed advisable to retain the relative position of the carpal ele- ments as preserved. Part of the shaft of the radius and the distal ends of metacarpals IV and V are missing, but the missing portions were present as negatives in the matrix and could thus be accurately cast in plaster. Of the right ulna only the proximal end is available; the other end is indicated by a matrix imprint on the proximal side of the intermedium. Only the distal portion of the ulnare was pre- served. The radius is relatively long (figs. 17 and 22), notably curved, distally expanded and less angular in cross section than is the case in modern cheloniines. There is, furthermore, no distal contact rugosity by means of which the radius was attached to the ulna in the charac- teristic fashion of modem cheloniines. The shape of the ulna can only be described in general terms based mostly on indirect evidence. Combination of the right proximal fragment with the left shaft frag- ment and the matrix imprint of the distal end suggests that the ulna was notably curved and somewhat flattened in dorso-ventral direc- tion. Its distal end was reconstructed (fig. 18) in an unexpected fashion because of the nature of the carpus, which seems to suggest the presence of such a peculiarity. In the carpus, as preserved (fig. 17), there are two larger proximal elements as in all sea turtles, the intermedium and the ulnare. The intermedium is entirely preserved; the ulnare lacks the proximal end. In modern cheloniines both intermedium and ulnare are notably elon- gated bones (see Zangerl, 1958, fig. 13). In Desmatochelys the inter- medium is short, about as long as wide; in shape it agrees fairly closely with the distal half of its homologue in Recent cheloniines. An intermedium of virtually identical shape is present in Corsochelys (Zangerl, 1960), and with it a complete, though somewhat fiattened ulnare. This element agrees, in so far as comparison is possible, with the ulnare fragment in Desmatochelys. For this reason it seems reasonable to assume that intermedium and ulnare had the same ZANGERL AND SLOAN: CHELONIID SEA TURTLE 29 shapes and relationships in the two forms. Since the ulnare in Corso- chelys is elongated to much the same degree as it is in modern cheloniines we must assume that it extended farther proximad than the intermedium and that the ulna attached itself to both elements Fig. 17. Part of right flipper of Desmatochelys lorn (CNHM-PR385) as pre- served; dorsal view and views of proximal ends of radius and ulna. in the manner suggested in figure 18. Phylogenetically speaking, this condition is thus intermediate between a primitive situation in which intermedium and ulnare are not enlarged and the specialized condition in modern cheloniines where both elements are notably elongated at the expense of the ulna. Distal to the intermedium in the present specimen there is an element, standing on end, which probably belonged next to the intermedium and which has the shape of the centrale in modem cheloniines. Distal to the ulnare there are a small and a fairly large Fig. 18. Outline drawing of flipper of Desmatochelys lorn (interpreted on the basis of CNHM-PR385 and another Cretaceous cheloniid; see text). 80 ZANGERL AND SLOAN: CHELONIID SEA TURTLE 31 carpal bone and these can be identified readily as the third and fourth distal carpals. The remaining carpal elements and the pisiforme are missing. Metacarpals II to V are elongated as in modern cheloniids; the fifth one, however, is relatively longer in Desmatochelys than in the modem forms. The basal phalanx of the thumb and the claw phalanx differ from those of modern cheloniine sea turtles only in very minor details. 7. Pelvis: Noteworthy portions of the pelvis are known both in the type and in the present specimen (fig. 19). Such areas as are available in both show, we believe, differences in preservation only. The preserved portion of the ilium is more slender and more angular than in cheloniines; in these respects it is intermediate between the toxochelyid condition and that of the cheloniines. The relatively large ischium, well preserved in both known individuals, is provided with a big, posterior spur as in the toxochelyids, chelydrids (and many other fresh-water turtles) and in some protostegids. In the cheloniids, in advanced protostegids, and in dermochelyids this spur is either very much reduced (as in modem cheloniines) or absent. This is clearly a primitive feature probably related to the type of locomotion in fresh-water turtles, and it is therefore to be expected among the more generalized members of all the sea turtle families, as well as among those forms in which the hind limbs are not pri- marily steering devices. The pubis, although incomplete, shows the essential character of this bone. There was an antero-lateral process of about the same shape and size as in toxochelyids (Zangerl, 1953, fig. 65) ; the base of the antero-medial process likewise indicates toxo- chelyid condition. 8. Hind limb: Of the hind limb only the left femur and tibia and a tarsal element are preserved. The femur (fig. 20) was broken and a small section of the shaft was missing; its length as restored is very probably correct; it could not have been shorter than restored be- cause the central parts of the break surfaces are virtually in contact with each other, and it probably was not longer because of the surface configuration of the adjacent break edges. The femur is relatively long compared to the humerus (fig. 22) ; it is longer than in any of the modern cheloniines, more slender, distally less expanded and the head is more nearly spherical. The trochanters are individual processes as in fresh-water turtles, but they are connected basally by a well- developed ridge, a condition almost perfectly intermediate between the chelydrid turtles and the modem cheloniines. On the distal end of the femur the tibial side of the joint surface is more bulbous than the fibular side, another feature in which Desmatochelys resembles the 1 Fig. 19. Desmatochelys loivi (CNHM-PR385), pelvic girdle in dorsal and medial views. Fig. 20. Desmatochelys lorn (CNHM-PR385), left femur. 82 Fig. 21. Desmatockelys lorn (CNHM-PR385), left tibia. Wmmmm!^^^ ^B^^^^^EE-^ WMmmmmM- 1 '//////////mj/m////////////m Dermochelys coriacea after Wieland 1906 Protostega gigas after Wieland 1906 Chelonia mydas 22066 Chelonia mydas 56 Ct)elonia mydas 42217 Erelmoctielys imbricata 31009 Eretmochelys imbricata Z 72 Eretmochelys imbricata 63263 Caretta caretta 31023 Caretta caretta 51676 Lepidochelys olivacea 31334 Lepidochelys olivacea UNO 6187 Lepidoctieiys olivacea Z 71 Desmatochelys lowi t Macrochelys femminckii O B Q O O O Q Ci O Fig. 22, Bar diagram to show relative sizes of stylopodial and zeugopodial bones in a number of sea turtles and a chelydrid. H (=100), humerus; R, radius; F, femur; T, tibia. The shaded columns within the humerus bars indicate in per- centages the absolute size range of the humeri in relation to that of Chelonia mydas (CNHM 22066;= 100). 33 34 FIELDIANA: GEOLOGY, VOLUME 14 fresh-water turtles rather than the cheloniines. The tibia (fig. 21) is relatively short (fig. 22) compared to that of the cheloniines and very much shorter than in the chelydrids. Its proximal end is very bul- bous but the distal end compares rather closely with that of the cheloniines.