GEOLOGY The person charging this material is re- sponsible for its return to the library from which it was withdrawn on or bef6re the Latest Date stamped below. fTorefd,„7r,0,10n' ?" UndeH,ni"« »' "••k* - .on, Z £££? oc"on ond "■* — • - — — *- To renew coll Telephone Center, 333-8400 UNIVERSITY OF ,..,nq,s 1IHAB AT MMNA-CHAMPAI6N FEf 2 8 1988 t J^ O-1096 THE VERTEBRATE FAUNA OF THE SELMA FORMATION OF ALABAMA PART I. INTRODUCTION PART II. THE PLEURODIRAN TURTLES RAINER ZANGERL FIELDIANA: GEOLOGY MEMOIRS VOLUME 3, NUMBERS 1 AND 2 Published by CHICAGO NATURAL HISTORY MUSEUM APRIL 30, 1948 GtOLOtf UfcRNf THE VERTEBRATE FAUNA OF THE SELMA FORMATION OF ALABAMA PART II. THE PLEURODIRAN TURTLES THE VERTEBRATE FAUNA OF THE SELMA FORMATION OF ALABAMA PART II. THE PLEURODIRAN TURTLES RAINER ZANGERL CURATOR OF FOSSIL REPTILES FIELD IANA: GEOLOGY MEMOIRS VOLUME 3, NUMBER 2 Published by CHICAGO NATURAL HISTORY MUSEUM APRIL 30, 1948 PRINTED IN THE UNITED STATES OF AMERICA BY CHICAGO NATURAL HISTORY MUSEUM PRESS CONTENTS PAGE List of Illustrations 21 Introduction 23 Description of Podocnemis alabamae, sp. nov 25 Description of Materials 26 Plastron 33 Shoulder Girdle and Humerus 34 Pelvis 34 Measurements 35 Systematic Relationships 39 Summary 52 References 55 19 LIST OF ILLUSTRATIONS PLATES 4. Carapace, plastron, and sacral part of shell of Podocnemis alabamae. TEXT FIGURES PAGE 3. Carapace and plastron of Podocnemis alabamae 27 4. Peripheral plates of Podocnemis alabamae 29 5. Nuchal and pygal plates, with adjoining peripherals, of Podocnemis ala- bamae 32 6. Ventral aspect of middle portion of carapace of Podocnemis alabamae . 35 7. Ventral aspect of anterior portion of carapace of Podocnemis alabamae . 43 8. Longitudinal and cross sections through first neural bone of Podocnemis alabamae 44 9. Ventral view of posterior portion of carapace of Podocnemis alabamae . 45 10. Outer view of left hyo- and mesoplastron of Podocnemis alabamae . . 46 11. Visceral view of xiphiplastral end of plastron of Podocnemis alabamae . 47 12. Lateral view of right ischium and pubis of Podocnemis alabamae ... 49 13. Right humerus of Podocnemis alabamae 49 14. Morphological type patterns of pelomedusid, pelusiine, and pelomedu- sine turtle shells ( 50 15. Morphological affinities among pelomedusine turtles 51 16. Morphological affinities among pelomedusid turtles 53 21 The Pleurodiran Turtles INTRODUCTION Among the turtle remains collected in the lower Selma marls there are a few specimens readily distinguishable from all the rest by the thickness of the plates. All but one of these specimens were found exposed on the surface, broken into many more or less bleached fragments (pi. 4, fig. 1). It was recognized in the field that these bones belonged to a pleurodire turtle, with the xiphiplastra showing pronounced ischio-pubic scars and the hyo- and hypoplastra separated in the region of the bridge by mesoplastral plates. The fragments of three of the specimens proved well enough preserved to permit the reassembling of fairly complete shells. The other individuals are either badly weathered or consist of only small portions of the skeletons. Isolated bones of the shell of this species are rare in comparison with the occurrence of fragmentary remains of other forms. Associated with some of the specimens were parts of the shoulder girdle, the humerus, the pelvis, and several unidenti- fiable pieces of limb bones. Unfortunately, not a single skull was found with any of these remains and it can be stated with little probability of error that none of the isolated finds of turtle skulls, lower jaws, humeri, and scapulae in the collection belong to this species. Altogether, parts of seven specimens were collected. All of them were found in the region of Harrell Station, but this is probably due to the limited extent of the outcrops elsewhere rather than to accidental preservation of an aggrega- tion, or to restriction of the species to a particular stratigraphic horizon within the lower Selma beds, although these possibilities cannot be excluded. As soon as the preparation of the better specimens approached completion it became evident that these forms are very similar to the materials collected by Mr. Barber in the Brownstown marl of Arkansas 'and subsequently de- scribed by Schmidt (1940) as Podocnemis barberi. In spite of the great degree of similarity, there are a number of constant differences by which the Alabama form can readily be distinguished from P. barberi, thus requiring the erection of a new species. Certain differences between the Alabama specimens, although rather considerable, can be explained as due to individual variation and age differences. 23 Order TESTUDINATA Suborder PLEURODIRA Family PELOMEDUSIDAE Podocnemis alabamae sp. nov. Type.— C.N.H.M. No. P27370, a nearly complete shell with pelvis. Col- lected by C. M. Barber, May, 1946. Referred specimens. — C.N.H.M. No. P27369, a nearly complete carapace, most of the plastron and other skeletal fragments. Specimen is somewhat larger than type. Collected by C. M. Barber, May, 1946. C.N.H.M. No. P27372, a smaller specimen consisting of most of the cara- pace and part of the plastron. Collected by R. Zangerl and W. D. Turnbull, May, 1946. C.N.H.M. No. P27331, posterior portion of the carapace of a very small individual. Collected by Mrs. A. Zangerl, November, 1945. C.N.H.M. No. P27406, largest, but very fragmentary specimen, badly weathered. Collected by C. M. Barber, May, 1946. C.N.H.M. No. P27405, fragmentary, badly weathered specimen, about the size of P27369. Includes a humerus and parts of both carapace and plastron. Collected by Mrs. A. Zangerl, November, 1945. C.N.H.M. No. P27419, two peripherals of a very small individual, about the size of P27331. Collected by C. M. Barber, October, 1946. Horizon and Locality. — Lower, marly member of the Selma formation, late Cretaceous. All specimens were collected in exposures of the Harrell Station area (see pi. 3), southeast of Marion Junction, Dallas County, Alabama. Diagnosis. — Proportions of shell as a whole and of constituent parts closely resembling Podocnemis barberi; number of neurals varying from six to seven. Scars of spinous processes of ninth and tenth vertebrae 'separated by shallow, saddle-shaped ridge, not united as in P. barberi; low, rounded, transverse ridges between free proximal ends and buttress scars of first and second costal ribs more pronounced than in P. barberi; anterior edges of epiplastra blunt, sometimes with broadly rounded swelling on either side of the sagittal suture; scales probably wider in comparison to length than in P. barberi. 25 26 FIELDIANA: GEOLOGY MEMOIRS, VOLUME 3 Description of Materials In publishing literature dealing with fossil turtles many authors neglect to furnish a clear description of the extent of the preserved bones. Such informa- tion is highly desirable, since it is often impossible to tell from the illustrations which parts of the skeleton are preserved in situ and which have been recon- structed. The following account gives an adequate description of the preserved parts of all individuals of the series. The type specimen is a fairly large shell, on the whole well preserved, but some of the plates have been crushed, others distorted. During preparation no attempt was made to compensate for obvious or assumed distortions; thus, the outline of the shell (pi. 4, fig. 1) is probably different from what it was in life. The carapace, well arched both transversally and longitudinally, is distinctly higher than in the type specimen of P. barberi and all other individuals of the Alabama species. As a rule, turtle shells tend to flatten out during fossili- zation, rather than to become more highly arched. Thus it appears likely that the curvature of the carapace of the type specimen resembles most closely the original state. The shell lacks right peripherals 3, 4, 10, and most of 11. On the left side the eighth marginal is missing and some of the remaining peripherals are crushed or damaged. The sixth neural is absent. The three posterior, left costal plates are more or less incomplete. The posterior margin of the suprapygal is broken off, so that the over-all length of the shell cannot be determined with certainty. The space between the broken posterior end of the suprapygal and the anterior margin of the pygal (now filled with plaster) is probably some- what too wide. It was determined by the sutural contacts of left peripherals 9-11 and the pygal. Peripheral 9 made good sutural contact with the distal ends of the sixth and seventh costal plates. Some amount of distortion probably accounts for the fact that the pygal now lies in a slightly asymmetrical position. The plastron is essentially complete except for the posterior tips of the xiphiplastra. The constituent bones evidently became disarticulated before burial. The sutures are completely filled with matrix that could not be entirely removed and consequently had to be left widely open when the plastron was assembled. Besides the shell, the type specimen furnishes the dorsal portion of the right ilium, the right ischium, and both pubes. The anterior free processes of the pubes and the medial process of the ischium that meets its mate in the midline are broken off (fig. 12). The acetabular surfaces have suffered so badly that they can scarcely be recognized. The carapace of P27369 is fairly complete; peripherals 5 and 8 as well as minor parts of 1, 3, 4, 6, 7 and 10 of the right side are missing. On the left side the fourth and parts of the third, fifth, sixth, seventh, eighth, and tenth marginal plates are lacking. Furthermore, the pygal is not quite complete; the fifth and seventh neurals, and parts of the right third, fourth, fifth, and sixth and the left second, third, and sixth costal plates were not recovered. The peripherals of the bridge region are crushed dorso-ventrally. In the plastron the bones remained intact during burial. A large portion of it is lacking, but fortunately none of the more important regions. Associated with this specimen are parts of both scapulae. The left one is the more complete. It lacks parts of both the dorsal and medial processes and the glenoidal surface is weathered. The best-preserved bones belong to the carapace of P27372. It lacks the right costals 3, 5, and 6, all but the first and fourth neurals, the nuchal and pygal plates and peripherals 1 to 3 and 10 and 11 on the right side and 1 and 2 on the left. The peripherals do not show any visible signs of distortion. The plastron consists of both hyoplastra with the left meso- plastron attached, and the entoplastron. P27406, the largest individual of the series, consists of fragments of all regions of the shell, but most pieces are weathered and of little value. Among the fragments are pieces of the shoulder girdle and of limb bones, including a left scapula with both processes broken 27 28 FIELDIANA: GEOLOGY MEMOIRS, VOLUME 3 away and the proximal end of the right humerus. About half of the caput humeri is preserved, but the radial and ulnar processes are broken off. Another piece was tentatively identified as the proximal end of the right tibia. P27405, although very fragmentary, is of interest because the seventh neural and parts of the sixth and adjoining costals are preserved, besides sizable portions of the plastron. Associated with this specimen is an exceptionally good humerus. P27331 consists of costal plates 5 to 8 on both sides, and P27419 of the left eighth and ninth peripherals and some unidentifiable fragments. The anterior margin of the carapace is notched, as in P. barberi. In the type specimen the notch is almost as distinctly expressed as in that species, but in P27369 it is very shallow. The nuchal plate of P27369 is hexagonal and its proportions are nearly identical with those of P. barberi. In the type specimen, however, the nuchal is relatively much wider and shorter (fig. 3; pi. 4, fig. 1). The medial aspect of the nuchal of P27370 along with a sagittal section are shown in figure 5. The number of neural plates can definitely be determined as varying between six and seven. P27369 and P27405 have seven neurals; the type, P27372, and P27331 have definitely six only. The first neural is four-sided, with the sides rounded. It is slightly longer than wide. The remaining neurals are hexagonal except the last, which is pentagonal. In P27369 the seventh neural is missing, but all the adjoining bones were recovered, so that its former presence is certain. In P27405 the seventh neural is preserved, together with portions of the adjoining bones. Its shape is exactly as indicated by the surrounding plates in P27369. The visceral surfaces of the neural bones are very well preserved in the type specimen and in P27372. Slightly distorted neurapophyses (belonging to vertebrae 13-15) remain attached to the third, fourth, and fifth neurals in the type and to the first and fourth neurals of P27372. The first neural gave attachment to two vertebrae, the ninth and tenth, as, for example, in the living P. expansa. The upper ends of the spinous processes of these vertebrae are preserved and appear firmly fused to the neural plate, permitting a few details to be given. The spinous process of the ninth vertebra is small and oval in cross section. Its place of fusion with the neural is separated from that of the tenth vertebra by a saddle-shaped, sagittal elevation on the ventral surface of the plate (figs. 6, 7, 8). The spinous process of the tenth vertebra is long and narrow in cross section; its area of fusion with the neural amounts to about 77 per cent of the length of this bone and extends backward to the hind margin of the latter. In P. expansa the antero-posterior extent of the attachment of the equivalent spinous process does not reach the posterior margin of the neural and amounts to only about 32 per cent of its length (figs. 6, 7, 8). In P. barberi the areas of fusion of the ninth and tenth spinous processes are so close together as to form one attachment scar. It cannot be made out with certainty whether the tenth spinous process in this species reached the posterior end of the neural plate or was short as in P. expansa, although the present state of preservation favors the latter interpretation. The length of the spinous processes of the other vertebrae preserved cor- responds both in P. barberi and P. alabamae to the length of their respective Fig. 4. Third left and fourth to ninth right peripheral plates of Podocnemis alabamae, C.N.H.M. P27372. Bones seen in ventral aspect; cross sections indicated by arrows. 29 30 FIELDIANA: GEOLOGY MEMOIRS, VOLUME 3 neural bones, whereas, in P. expansa, the spinous process of the eleventh vertebra is only about half as long as the second neural, that of the twelfth vertebra a little more than half as long as the third neural, and the following two about three-fourths of the length of the fourth and fifth neurals, respectively. The spinous processes are thin fore and aft and very low, as in all observed members of the genus. Just back of mid-length the neurapophyses form strong diapophy- seal processes, which, together with the parapophyseal processes of the centra, articulate with the free rib-heads. The dorsal portion of the neural canal is well preserved in the neurapophyses belonging to vertebrae 13 and 14. It is only 4 mm. wide and a delicate crest runs lengthwise in the mid-dorsal line of the canal. The suprapygal plate is essentially complete in P27369. It is triangular as in P. barberi, but is relatively smaller and narrower than in that species. The costal plates compare very closely with those of P. barberi. The but- tresses are located below the first and fifth costals, and the three plates between them are narrow for their length. Costal 3 is fairly straight in the plane of the carapace, costal 2 is concave anteriorly and costals 4 to 6 are convex anteriorly (fig. 3; pi. 4, fig. 1). The structures on the visceral surfaces of these plates are well preserved in the type specimen and in P27372 and call for description, particularly since the corresponding surfaces in the type of P. barberi are con- siderably weathered. The buttresses are unequal in size, the axillary one being the stronger. The latter is attached to the entire lateral half of the first costal plate and distinctly lies in the path of the second carapacial rib. The inguinal (hypoplastral) buttress occupies only the outer third of the fifth costal. As is usual in the genus Podocnemis (as far as could be observed), the first and second carapacial ribs are thoroughly fused with the first costal plate. In P. alabamae the first and second ribs are, at least near their proximal ends, of equal size. In P. barberi the first rib slightly exceeds the second in strength, while in P. expansa the first rib is distinctly smaller than the second. The sacral region (figs. 6, 9) is particularly interesting. The state of pres- ervation in the type specimen of P. barberi leaves much to be desired, but P27370, P27372, and P27331 from Alabama permit an accurate interpretation of the preserved elements of this relatively complicated area if they are compared to the equivalent structures of a modern form. In a medium-sized specimen of the recent P. expansa (pi. 4, fig. 3) the morphology of the sacral region of the shell is as follows: Four vertebrae, numbers 16 to 19, should be considered as sacral vertebrae as their ribs take part in the sacral attachment of the ilium. The ribs of the sixteenth vertebra underlie the seventh costal plate, those of vertebrae 17, 18, and 19 underlie the eighth costal. The ribs of vertebra 17 are thoroughly fused with the eighth costal plates, and the distal ends of the small eighteenth and nineteenth pairs of ribs are fused together, but not with the costal plates. The upper ends of the ilia are joined by suture mainly to the eighth costal pair, and, dorso-medially, to the four pairs of ribs just described. If the ilia become detached by maceration, the areas of their sutural connection with the carapace remain well marked by rough scars (pi. 4, fig. 3). The neurapophy- ZANGERL: VERTEBRATE FAUNA OF THE SELMA FORMATION 31 ses of the sacral vertebrae, lying behind the last neural plate, are attached by coarse sutures to the medial margins of the sixth and seventh costal bones. In those specimens that have a seventh neural plate, the neurapophysis of the first sacral vertebra (the sixteenth in the series) is attached to it. This is shown by P27405, in which the scar extends the full length of the neural and is 5 mm. wide at mid-length. The last sacral vertebra is free from the carapace. The above description applies essentially to the condition in P. alabamae (fig. 6) where, in the type specimen, the distal ends of the four pairs of sacral ribs and the dorsal end of the right ilium are exceptionally well preserved. The cara- pacial areas covered by the iliac scars differ greatly in P. barberi and P. alabamae on the one hand and the available specimen of P. expansa on the other. This difference is due to the difference in shape and extent of the eighth pair of costals and the suprapygal. The narrow suprapygal and the great lateral extent of the eighth costal plates in this specimen of P. expansa (pi. 4, fig. 3) seem to be an individual variation that does not reflect the basic podocnemid pattern. An X-ray photograph of a juvenile P. unifilis reveals the shapes of the bones in question to be very nearly as in the fossil form here described. In P. alabamae the iliac scars extend laterally onto the seventh costal plates and postero- medially onto the suprapygal for a short distance. The peripherals are ade- quately illustrated in figures 4 and 5. In the type specimen the bridge peripherals form a sharp lateral edge, but this is probably due to intense dorso-ventral crushing. Shield furrows are visible on the carapace bones of specimens P27370, P27372, and P27331, but not even an indication of them can be found on the carapace of P27369. Numerically, the scale pattern is identical with that of P. barberi. Comparison of the shape of the various scales in these two species is complicated by the well-known fact that the shape of the shields changes with age. Thus, in order to arrive at a conclusion as to whether the shapes of the scutes are the same or not, individuals of approximately the same age or size should be available. This is not the case. P27370 is noticeably smaller than the type of P. barberi, and P27369, which is but slightly smaller, lacks any shield impressions. In P27370, as would be expected, the vertebral shields are rela- tively wider and shorter than in P. barberi. In the two smaller specimens from Alabama these scales, in agreement with the rule, are exceedingly short and broad. This is particularly noticeable in P27331, where1 the fourth vertebral shield is relatively wider than in P27370 (see measurements, p. 38). If the rate of change in the vertebral scale proportions during growth from relatively wide and short to relatively long and narrow remains the same in these forms, then there would be no difference in this respect between P. alabamae and P. barberi, and the apparent difference in shape would merely reflect a difference in age. But it appears unlikely that the scale shape changes as much in large individuals as it does in juvenile ones. Until the contrary has been proved, it is better to consider the scale proportions of the two forms as different. The question cannot be definitely answered, however, on the basis of the materials now avail- Fig. 5. The nuchal plate with adjoining first and second peripheral bones and the pygal with adjoining eleventh and tenth peripherals of Podocnemis alabamae, C.N.H.M. P27370 (holotype). All bones shown in ventral aspect; cross sections indicated by arrows. 32 ZANGERL: VERTEBRATE FAUNA OF THE SELMA FORMATION 33 able for study. A slight difference in the position of some of the scale furrows exists between P27370 and P27372. In the latter specimen the posterior margin of the fourth vertebral shield and the anterior third of the sulcus between the fourth costal scute and the fifth vertebral lie on the seventh pair of costal plates, as in P. barberi; in the type specimen these furrows lie on the eighth pair of costal bones. In P. alabamae the lateral edges of the first vertebral scale are almost straight (pi. 4, fig. 1), whereas in P. barberi they are strongly indented. Irregularities in the scale pattern appear to be fairly common in the Alabama species. In the type specimen there is such an irregularity in the region of the right bridge. The eighth peripheral bone shows on its posterior half the shield furrow in its proper position, but in the anterior portion of the bone it divides to outline a scale that is not normally present in that position. No imprints are visible on the preceding peripheral bone, so that the full extent of the irregularity cannot be made out. In P27372 a similar accessory scute lay between the peripheral and costal shields on the left side of the shell on peripheral 6 and pre- sumably on peripheral 7, which is lacking. There is little evidence of pathological deformations, except in the carapace of P27372. The anterior point of the right first costal plate is greatly deformed, probably as a result of a parasitic infestation (fig. 7). The adjoining first and second peripheral bones are missing, so that the full extent of the injury cannot be determined. There are three deep, oblong depressions cut into the dorsal surface of the costal; anteriorly they merge to form what appears to be a large hole piercing the plate completely. As the erosion grew deeper, bone was deposited on the visceral side of the plate, thus forming a large, somewhat irregular nodule. Similar phenomena can sometimes be observed on shells of modern turtles, for example, in a specimen of Pelusios sinuatus, C.N.H.M. 12699. The formation of bone on the inside of plates that are beset with parasitic erosions on the outer surface, is not a very common occurrence. Normally the shell simply becomes pierced. Two such holes penetrate the shell of P27372, one between the left tenth and eleventh peripherals and the other between the fourth and fifth costals. Plastron The general shape of the plastron is very similar to that of P. barberi. The anterior lobe is short and broadly rounded, and the posterior lobe is about as wide as long, with the xiphiplastral end deeply notched (fig. 11; pi. 4, fig. 2). The plastron is thickest near the midline at the level of the inguinal buttresses. The anterior edge of the epiplastron is blunt, in contrast to P. barberi, where it is acute. In P27369, where no sign of crushing is visible, the epiplastra form swollen lips on either side of the median line and the edge between these lips is straight (pi. 4, fig. 2). These details are not so well shown in the type specimen because the epiplastra have been compressed dorso-ventrally. The remaining plastral elements are much as in P. barberi, but the mesoplastron is slightly wider transversely in the Alabama form. As in P. barberi, both the ischial and 34 FIELDIANA: GEOLOGY MEMOIRS, VOLUME 3 the pubic scars on the xiphiplastra are elevated and coarse. The ischial scars do not extend quite to the midline (fig. 11). The plastral shield sulci are present neither in the type of P. barberi nor in that of P. alabamae, but in P27369 they are clearly discernible. As in the modern South American species, for example, P. expansa, there are the usual six pairs of plastral scutes plus an unpaired inter- gular shield. The general arrangement of the pattern is very similar to that of P. expansa; the intergular and gular shields are relatively shorter in P. alabamae and the pectoro-abdominal furrows cross onto the mesoplastra in the latter, but not in the former species. In P. expansa the hypo-xiphiplastral sutures lie, laterally, about halfway between the abdomino-femoral and the femoro-anal sulci; in P. alabamae they are distinctly closer to the femoro-anal furrows (fig. 3). The inner surface of the preserved portion of the entoplastron of P27370, instead of being flat as in other specimens, seems to have formed a strong boss or a sagittal crest that has become intensified by crushing. The entoplastron of P27369 is longer than wide as seen from the visceral surface, wider than long on the outer surface (fig. 3; pi. 4, fig. 2), a difference due to growth. The suture between the entoplastron and the hyoplastra is nearly vertical in young individ- uals, but, as the shell becomes larger, the entoplastron grows in length faster on the inside than on the outside. Consequently, the angle between the suture and the plane of the plastron becomes smaller. Shoulder Girdle and Humerus Of the shoulder girdle only parts of the scapulae are known; these are similar to equivalent parts in P. expansa, but a detailed comparison cannot be made. A right humerus found with P27405 is of particular interest. The over-all length of the bone is 97.5 mm. The shaft is curved, as in P. expansa, but instead of an entepicondylar foramen there is a shallow, distinct groove (fig. 13). The relative sizes of the caput humeri and the radial and the ulnar processes are similar to those of P. expansa. In the latter there is, on the ventral side, a broadly V-shaped ridge connecting the radial and ulnar processes. This ridge is not developed in P. alabamae. On the ventral face of the shaft, a moderately acute ridge, originating at the base of the ulnar process, runs toward the distal end of the humerus. In P. expansa it runs approximately parallel with the shaft; in P. alabamae it crosses over to the radial side and ends below the entepicondylar groove (fig. 13). Probably because of the differences in the region of the ent- epicondylar canal, the humerus is relatively narrower at its distal end than in P. expansa. Pelvis The dorsal part of the ilium is shown in figure 6. It corresponds fairly well to the corresponding part in P. expansa, but the depression on the antero- medial face of the bone is much deeper in the Alabama form. The right ischium and pubis are illustrated in figure 12. Their sutural ventral ends fit well into the corresponding scars on the xiphiplastron. The height of these bones is about ZANGERL: VERTEBRATE FAUNA OF THE SELMA FORMATION 35 Fig. 6. Ventral aspect of middle part of carapace of Podocnemis alabamae, C.N.H.M. P27370 (holotype). Dorsal end of ilium and several neurapophyses pre- served in situ (cf. pi. 4, fig. 1). 50 mm. They resemble closely those of P. expansa, but their state of preserva- tion does not justify close comparison. Measurements In the table below, the actual meas- urements of the type and referred speci- mens of P. alabamae and of the type of P. barberi are listed, together with their relative values calculated according to the following formula: Measurement of character xlOO Maximum width of nuchal plate1 The figures indicate rather clearly that the over-all geometric patterns of the Alabama specimen P27369 and the type of P. barberi (the two individuals are most nearly of the same size) are very similar. The indices worked out for the type of P. alabamae (P27370) and the much smaller individual (P27372) do not fit as well into the pattern of P27369 or into that of P. barberi. Three factors, at least, are responsible for the irregular index pattern of the more complete specimens of P. alabamae: individual variation, differences in age of the members of the sample, and an un- determinable amount of distortion that not only affects each specimen in dif- ferent degrees but also the various parts of each specimen. If, the amount of variation in a uniform sample of a living species of Podocnemis were known, to- gether with the proportional changes that take place during growth, it would be possible, by analogy, to properly evaluate the present material and to determine approximately the degree of 1 This measurement was used because it could be obtained accurately in the better- preserved specimens. in 10 o ■8 S H N lO fl> H (O T* 1ft r-t eo oo oo oo eo ■ oo ■* a 00 00 CM CO CD OS Tt oo-* eoco 6 •* 00 i-H CO CMCOOSCMOCO • Tf CO CO i-1 00 CM OS tj< co ^ eo co cm CD OS 00 WoOXOOiH o in o oo co oo os t> t> ■<* O lO TC CO CO CM CM CM ■** in o CO in CO CM iH lO in c^Tjicot^-^iooq os in cm coa>o o co co m o os o -j co in oq co t^ t-; o i-HCDinocMcoco © os° ■>* o CO CM CM CM i-h in CO CO CO CM coco 00 00 CO "* •>* CD t-; C- CM O ■* OS 00 t> CO -* -^ CO CO CO CM lO lO o U5 (M Oh m oo oo CD CO ■^ i— I t— OS OS OS co m m ■<* t* co C- i-H Tjl t- os in in co o^f co mm ■* CO CM CM CM CM OS CO CM CO ■* O os eo Ont-NWN m m Tt -^ •>* t* fa fa CO < fa eo . • • oo eo • • • CO p : : : ::::::: -h Cm -H o OS CO o t- o CM OS CO m m m t- ^ >-i co t)< l os COCMCOCOCMCM . ■** -H-H -H-H • t-oscomcocM . cm i— i in eo ■*# co co co .coco ICh CM t- oo o m co ■H ■^t-ooosi-i-^co co oo m coTt-^co-^cocM co t- m cmooosio^ CM i-H i-H H t-I e— e— i— I CO OS OS i-H CM CO CM r- 1 r-l CM CM CD 00 ' Tf CM 00 00 ' -^ t- CD ,-H : -H -H^ OS OO C- OS o CM CO CO CM CM CO ■H-H -ti-H OOOCO CO t- CM -* co eo co co -^ -H -H oOHO^t-ooo oo eo gs os CO CO CO CM * ■<* tj< eo -H-H -H a> 0. it Si o C3 — a ■~ a o J5 os ft"S "2 "os 3 • a> • a --a 3 J5 ,3^ S olo e.5 ^5 5 co a "a "3 Ho >. p, ad C CO O O — -S S "3 a 3 C C "— — o o -4-> -W -o -a be bo a c cu cu fc t/2 H fc fe Cfl t/2 J3 x .2 .2 - !- 2 2 C 'C 2 8 c c . .J2 •-O X . -u S2 8-= S45-S > .3 cu 'c or3.2 cp fe CC E-1 fe &h cQ C/J S >. A d E — So o -J XI — JS ftp, 2 "3 2 S 3 cu c c o o J= .3 -u -t-> -O -3 O .3 .3 3 < B 3 I c3 a! fc CO H fa * co co T. 36 fe US • .o © in *- co CO 9 Ph 00 i-H 00 CO m CO Oi 05 CO CO CO 00 CO CO CO CO >-H CO CO CO CO I-H .-H 00 CO ^H CJ * CO 05 T* •* CO od oi oi io oo cd oo r-5 1> co oo oo od oo co o oj o oo t> ci in" ■* ^tj< in co t> c- oo co io in Hccw^'i|'^^|in(Oin'^ ■^>t-oot>t>co^H(Nooinoo iniointninincoococoin a E S z o CO • (MiMoooooim • in .-h ■* •* ,-h ■* in oo . coco -^ Tj< • ■* as cq as '. CO t> ■* t> . o in in in C- t-CMCOC~00 c- (N-* mm ' OSCO OSCN1 ,10 10'*'* t>ooiT-H-*coooiOiinoo co'oma>t>comcdcoojo mmm*Tj co co c— m ■* m •^ to 00 N pO - •e i s CO H s- t- o 1 CO E- z Ph s •M £ CO < a eo W S a •M Ph P e 99 X in 3 Ph H s CD fc- co CO locet-^TjiNiooiOHoo »ioioi>ooooofflt-t- mrHoooo'*co'*ooooo NiOt-lOlOlOiOt-ISt-O CMt>00COt>mtMtr-r-it~00 t-t>t>t~t>[~ooasasoot> ascoaicoaicMasoO'*'* .cMcoco-*'*m'*cococo co •*■«* t-oSPw •*-<*■* co gj-t- . m> i-h co co .[-t>lOiO C— SS-CN- C- OHHIOHIOHHIO .mcocococo'*m'*co m cmcocmo^co :ot>ooo mio®*^-* .ocomm I>IOOS ! i— i as in as ©asooas : o ■ :Ercoo as in cm coco CO . . . .OSCO COCO co-*coco . . . .^~m m iot-oo Iph •CO !aiooHooioio*H .•***-*Tjascooococoinasast-OOt-iOCC -H -C A r- C 37 o co PL. Tf»NNoqcc»i>cooq»NOitDo; COCM CM rH rH rH CM^CTJ05COC^a5rHOUSTtf.rt CM • -100501 ■ Iocmcm ■ • -us • CO ' t-mio t> rH oqoo -cNjcOrHoao ■ 0 ooo OOrHCl-tfCOCOCOrHrH rH rH rH rH • ocot>co •** 10 1- t> t- 1~ t>oo a S a - cm co ■<* t-; us as t(< o co ■* us co US CM 00 00 CO rH NNH rH ■-rJH C0O5 ; c-^ 0606 . a>c-oo • rH Ol •<# CM t> ! CO US r-i CO •<* . CO OCO -rj< CO rH oi eg 05 oa 00 CO COCO CD O CO l>CO C~ O rH TtNcooOTjjoocooqcMcooiCvjascqt-; odiOrHO-^rHoo'ooaicocooiooaio rHcooocooicoooocoooocoioas CONNHH rH rH rH rH CM * * # COOl-tf GOOVOOOrHO^^liH'CfO^rHCOt- Ci CM 00 - CM t- us O 1— I CN. C-. rH US OJOOOOUSrHOOOOCOOCMrHOO C-00-«t(MC0CM00CMCMc0'H> ph a mco .JSco'cn CO >> !>, >>P3P3 J=ftft &6 S . d >> >> co >, CO !« bO _ O u o O 43 o rt H h TzMcocotchSS+j °! -c8cJo"Sft» a£ O-P 3 » C! «S 0"S »i2 B.O'C -8> s.S"Js c P-,-5 a o S C3 'E o>p3p3 x &g-C g ss II co a o CO CO O O CO -eg §| C CO S5.a . us CO l>Xl< OOrH 'CO t^oo .COCMl> US ' C~ 00 CO us 00 . CO US US CO rH ■M-ti .CO ' OH COrH CO coco t~ OlC-Ot^COUSOi^OcMOlOOiOOl rHCMCOrHUSt-t>-^^t CO T3 > O (^ CO JS CO +J JSJ3 ES I § H x c3 c3 •§ ws -C—J co co ro i S a3 £? -O tn CO co 53 -^ 1-. ►* co •pOl »0'3 !T3 r/l j_i rCC CO 33 .c- CO 0) J3 J=PH +-) -u +j teS'E'S'S t* ~ :- ~ .2.2.2 'E 'E'E co co co d q e u CO L co^, +3 PC *^ c C r* 55 r» fe CO ,_, L> _ CO p O l hi! co g.H 3S CR ffl+JSHtC J3 » 'E 2.o-^ E" cu ^000000' r J3 pC -C pC J5 j3 ► g +j +j -ip> +p -u +j . 43 be m bo co be 00 co 5 c S a d c O CO CO CO CO CO CO Oh hJ PH J J 1-5 J 8 38 ZANGERL: VERTEBRATE FAUNA OF THE SELMA FORMATION 39 distortion suffered by the fossils. To my knowledge, however, such information is not available at present. In order to make statistical treatment possible at a later date, the complete set of measurements is herewith published. Systematic Relationships Fossil turtles definitely referable to the Pelomedusidae were first found in the London Clay and described by Owen and Bell (1849), but it is doubtful that these forms belong to any of the living genera. In the early part of this century, however, a number of fossil representatives of all three living genera of the family were discovered in Egypt and described by Andrews (1900, 1901, 1903, and 1906) and von Reinach (1903). Dacque (1912) published an excellent review of the fossil turtles of the Tertiary of Egypt and attempted an outline of the phylogenetic relationships that appeared to exist between the fossil Egyptian pleurodiran turtles and the living forms. Since much of the fossil turtle material from Egypt belongs to the existing Podocnemis and to the related, extinct Stereogenys, a discussion of the probable relationships of these fossils to the Recent species of Podocnemis was of necessity included as of particular interest in view of the striking dis- continuous distribution of the latter in northern South America and in Mada- gascar. At that time no fossil representatives of the genus Podocnemis were known from the New World. Dacque, obviously with reservations in his mind (he points out the fact that no fossil American species had yet been described), assumed that the genus had originated in the Old World, particularly in the region of Egypt. In late Eocene beds of this area there were thus known to occur species definitely belonging to Podocnemis (P. stromeri von Reinach, P. antiqua Andrews) as well as two species belonging to the closely related Stereogenys (S. cromeri Andrews and S. podocnemoides von Reinach) . Both genera were also RECENT madagascariensis American types PLIOCENE MIOCENE aegyptiaca OLIGOCENE fajumensis •/■ S. libyca EOCENE antiqua stromeri''- podocnemoides- -Stereogenys s.s. ? Podocnemis s.s. known from Oligocene deposits (P. fajumensis Andrews and S. libyca Andrews). P. aegyptiaca Andrews had been recorded from beds of Miocene age. Dacque clearly recognized that all of these species of Podocnemis were closer morphologically to the living P. madagascariensis than to the New World forms. This interpretation was based chiefly on the scale pattern of the anterior plastral 40 FIELDIANA: GEOLOGY MEMOIRS, VOLUME 3 lobe. In P. Madagascar iensis, the intergular shield is small and triangular; both the gular and humeral scales meet at the midline. In the New World species the intergular is noticeably larger and the gular shields do not meet in the sagittal plane, but are more or less small, triangular scales on either side of the intergular. The diagram (p. 39) illustrates the morphological affinities of the extinct Egyp- tian and the living species according to Dacque's interpretation (1912, p. 13). He derived the various forms from a morphological type (called by him Podo- cnemis s.s.) in which the scale pattern of the anterior plastral lobe was considered to be essentially that of the living Madagascar species. From this morphological type, Podocnemis stromeri was derived by a displacement of the gulo-humeral sulci forward to the posterior tip of the intergular. The fajumensis-aegyptiaca- madagascariensis type results from the stromeri-pattem by diminution of the intergular shield. In Stereogenys podocnemoides, the intergular shield is relatively large, the gulars are relatively small, triangular scales, and the posterior tips of the humeral scutes meet at the hind end of the intergular (fig. 15, B). By decrease in length of the intergular, the pattern of the living South American species of Podocnemis could be arrived at (fig. 15, L). An additional forward displacement of the humero-pectoral sulcus would lead to the Stereogenys s.s. pattern (fig. 15, C). Dacque regarded Pelusios as a morphological derivative from the line leading from Stereogenys podocnemoides to the living South American Podocnemis spp.1 The discovery of representatives of Podocnemis in the late Cretaceous of both North and South America,2 reopens the question of relationships among the various pelomedusid pleurodires. It is now obviously no longer possible to derive the recent South American species from an Old World form such as Stereogenys podocnemoides; a restudy of the morphological affinities of the mem- bers of the genus Podocnemis and of the Pelomedusidae as well is necessary. The modern pleurodiran turtles fall readily into two groups, the Pelomedu- sidae and the Chelyidae. Of these families the former includes the more gener- alized living pleurodires, Pelusios Wagler, Pelomedusa Wagler, and Podocnemis Wagler. The plastron includes mesoplastral elements; these either meet at the midline and separate the hyoplastral and hypoplastral plates entirely (Pelusios), or are small, more or less circular plates, laterally wedged between the hyo- and hypoplastra (Pelomedusa and Podocnemis; fig. 14). In Pelusios the anterior lobe of the plastron is movable, a hinge being present between the hyo- and mesoplastra (fig. 14, B). In the carapace there is always a series of neural plates varying from eight to six in number. In the Chelyidae mesoplastral bones are absent and there is a distinct tendency toward reduction of the neuralia. In 1 This view is, in my opinion, quite untenable, since Pelusios has large mesoplastral plates (fig. 14, B) that meet in the midline, much as in the Pleurosternidae, for example, Glyptops, a condition here considered as primitive in view of its general occurrence among the phylogenetically older Amphichelydia. 2 Further preparation of the type and only known specimen of Podocnemis olssoni Schmidt from the Eocene of Peru (Schmidt, 1931) revealed that this species does not belong to Podocnemis, but to the genus Taphrosphys (fig. 15, F), so far known mainly from the Green Sand deposits of New Jersey. A redescription of this specimen was recently published (Zangerl, 1947). ZANGERL: VERTEBRATE FAUNA OF THE SELMA FORMATION 41 Rhinemys there are but four neurals; in Chelodina oblonga, Platemys platycephala, and Emydura macquariae neural plates are totally lacking. Among the fossil turtles there are a number of forms that can definitely be assigned to the family Pelomedusidae. All three of the modern genera are represented in the fossil record, and, in addition, the genera Stereogenys Andrews from the Tertiary of Egypt and Taphrosphys Cope from New Jersey and Peru (see above) must on the basis of present knowledge be included in the family. Amblypeza Hay is poorly known. It is close to Taphrosphys, from which it may not be generically distinct. Naiadochelys Hay is based on a small xiphiplastral fragment, and the Patagonian materials referred to this genus by Staesche (1929) on what appear to me to be very uncertain grounds scarcely add to the knowledge of this form. The meager fragments, however, indicate a turtle close to Podocnemis. The recognized genera and species of the Pelomedusidae stand as follows: Family PELOMEDUSIDAE Genus Pelusios Wagler P. niger Dumeril and Bibron West Africa P. sinuatus Smith Africa P. nigricans Donndorff Madagascar P. derbianus Gray Africa P. adansoni Schweigger West Africa P. gambonensis A. Dumeril Angola P. dewitzianus von Reinach Egypt P. blanckenhorni Dacque" Egypt Genus Pelomedusa Wagler P. galeata Schoepff Africa, Madagascar P. progaleata von Reinach Egypt Genus Platyckeloides Haughton P. nyasae Haughton Nyasaland Genus Podocnemis Wagler P. cayennensis Schweigger Northern South America P. dumeriliana Schweigger Northern South America P. expansa Schweigger Northern South America P. lewyana Dumeril Northern South America P. sextuberculata Cornalia Northern South America P. unifilis Troschel Northern South America P. vogli Muller Northern South America P. madagascariensis Grandidier Madagascar P. aegyptiaca Andrews Egypt P. bramlyi Fourtau Egypt *P. lata Ristori Malta P. fajumensis Andrews Egypt *P. dehmi Bergounioux Germany P. antiqua Andrews Egypt P. stromeri von Reinach Egypt P. indica Lydekker India P. bowerbanki Owen and Bell England *P. congolensis Dollo Congo *P. harrisi Pacheco Sao Paulo P. brasiliensis Staesche Sao Paulo P. barberi Schmidt Arkansas P. alabamae sp. nov Alabama Recent Recent Recent Recent Recent Recent Pliocene Miocene Recent Oligocene Age uncertain ?Cretaceous Recent Recent Recent Recent Recent Recent Recent Recent Miocene Miocene Miocene Oligocene Oligocene Eocene Eocene Eocene Eocene Paleocene Late Cretaceous Late Cretaceous Late Cretaceous Late Cretaceous 42 FIELDIANA: GEOLOGY MEMOIRS, VOLUME 3 Genus Stereogenys Andrews S. libyca Andrews Egypt Oligocene S. cromeri Andrews Egypt Eocene S. podocnemoides v. Reinach Egypt Eocene Genus Taphrosphys Cope *T. sukatus Leidy New Jersey ?Eocene T. longinuchus Cope New Jersey ?Eocene T. leslianus Cope New Jersey ?Eocene *T. strenuus Cope New Jersey ?Eocene T. molops Cope New Jersey ?Eocene *T. dares Hay Not known Not known *T. nodosus Cope New Jersey ?Eocene T. olssoni Schmidt Peru Eocene Genus Amblypeza Hay* *A. entellus Hay New Jersey ?Eocene Genus Dacochelys Lydekker* D. delabechei Bell England Eocene Genus Naiadochelys Hay* *N. ingravata Hay ?New Mexico ? *N . patagonica Staesche Patagonia Late Cretaceous *N. major Staesche Patagonia Late Cretaceous Genus Shweboemys Swinton* |S. pilgrimi Swinton Upper Burma Pliocene * Poorly known or doubtful genera and species, t Skull only. If the pelomedusid plastron is compared to that of the Pleurosternidae, where one pair of mesoplastra, meeting at the midline, is suturally joined to the hyo- plastra, and to that of the Baenidae, where the mesoplastra are either completely developed as in the Pleurosternidae, or medially reduced in varying degrees, it becomes clear that the morphological type from which the pelomedusid condition can be derived, is, in principle, that of the pleurosternid plastron. The scale pattern of the pelomedusid plastron presents a much less clear-cut picture, but a few basic features are apparent. The gular shields are almost always separated anteriorly by a relatively large intergular scale. Only in Taphrosphys molops Cope does the intergular appear to be displaced caudad so that the gular scutes meet in front of it (essentially as in Chelodina oblonga Gray). In Podocnemis fajumensis, P. aegyptiaca, and P. madagascariensis the intergular is small and very narrow (fig. 15, H, I, K). In most Tertiary and Recent species of Podoc- nemis the humero-pectoral sulci meet in the anterior part of the entoplastron ; in the Cretaceous species in the posterior half of the entoplastron; in Pelusios along the posterior margin of the entoplastron; and in Pelomedusa, the Pleuro- sternidae, and Baenidae far behind the entoplastron. The pectoro-abdominal shield furrows follow the hyo-mesoplastral hinge in Pelusios. In Pelomedusa they lie close to the hyo-hypoplastral sutures and extend onto the lateral meso- plastra. The same condition prevails in Podocnemis alabamae. In the remaining forms the sulci lie farther forward, bending backward laterally, but do not extend onto the mesoplastra. In the Pleurosternidae and Baenidae they extend ZANGERL: VERTEBRATE FAUNA OF THE SELMA FORMATION 43 across the mesoplastra. The abdomino-femoral and femoro-anal shield furrows do not exhibit any significant positional relationships in the forms here considered. These comparisons permit the deduction that the basic plan of organization of the pelomedusid plastron includes a pair of medially joining mesoplastral bones. The intergular shield is relatively large, separating the gular scales. The humero-pectoral sulci lie behind the entoplastron, and the pectoro-abdominal furrows lie on the mesoplastra (fig. 14, A). The pelomedusid carapace presents a fairly generalized picture. On the whole, it compares favorably with that of the Pleurosternidae and Baenidae. FlG. 7. Ventral aspect of anterior portion of carapace of Podocnemis alabamae, C.N.H.M. P27372. The rugosity at the anterior edge of the right first costal is a parasitic deformation. Nu, nuchal plate (missing); lc, first costal plate; 2Ne, second neural plate (missing). The major differences lie in the tendency toward reduction of the neural plates and in the absence of a nuchal scale (except in Amblypeza) in the Pelomedusidae. Pelusios derbianus has eight neurals, P. sinuatus (C.N.H.M. 12699) has six, Pelomedusa seven, and Podocnemis six or seven. The various Egyptian species of Podocnemis as well as the Recent P. madagascariensis have six, the living South American species have seven (occasionally there is a specimen with only six, for example, a shell of P. expansa, C.N.H.M. 16075), the late Cretaceous P. barberi has six, P. brasiliensis seven, and P. alabamae six or seven. Stereogenys and Taphrosphys have seven neurals. In most cases the first neural plate adjoins the nuchal bone, but in Stereogenys it appears to be displaced caudad, permitting the first pair of costal plates to join sagittally. This character was observed in several individuals and may be characteristic of the genus.1 In Pelomedusa 1 The carapace referred to S. podocnemoides by von Reinach may not belong to this species. It can scarcely be distinguished from Podocnemis stromeri (Dacque\ 1912, p. 24). 44 FIELDIANA: GEOLOGY MEMOIRS, VOLUME 3 galeata the first neural is either a long, slender plate adjoining the nuchal or a small, round plate widely separated from it (C.N.H.M. No. 17160). A condition similar to that of Stereogenys thus occurs in this species as an individual variation, but here it is clearly a reduction of the anterior part of the first neural, whereas in Stereogenys there seems to be a caudad segmental displacement of neural plates 1 to 5, whereby Nos. 6 and 7 come to occupy the space held by the seventh neural alone in Podocnemis. Within the family Pelomedusidae there is a dif- ference in relationship of the first vertebral scute to the first and second marginal Fig. 8. Longitudinal and cross sections through first neural bone of Podocnemis alabamae, with parts of neurapophyses of vertebrae Nos. 9 and 10 attached. C.N.H.M. P27372. scales. In the living species of Podocnemis and in Stereogenys the lateral shield furrow of the first vertebral ends at the first marginal scale; in all other pelo- medusid turtles it ends at the second marginal. In the Pleurosternidae with a wide nuchal scale, for example, Glyptops, the furrow in question ends at the first marginal scute. In the Baenidae this region is exceedingly variable. The morphological type of the pelomedusid carapace can be characterized as follows: It greatly resembles that of the Pleurosternidae, with eight neural plates, the first adjoining the nuchal bone. There is no nuchal shield and the lateral shield furrows of the first vertebral end at the second marginal scales (fig. 14, A). The attempt may now be made to establish the morphological affinities between the various pelomedusid turtles with regard to the mutual structural plan of the shell, and based on those features that have been discussed. If the basic type pattern of the pelomedusid shell (fig. 14, A), arrived at above, is compared with the actual representatives of the family it becomes clear that the latter fall into two major groups, one represented by Pelusios, the other ZANGERL: VERTEBRATE FAUNA OF THE SELMA FORMATION 45 by Podocnemis, Pelomedusa, Stereogenys, and Taphrosphys. The two groups are considered as subfamilies, Pelusiinae and Pelomedusinae, diagnosed on page 48. In the Pelusiinae (fig. 14, B) the sutural union of the hyoplastron with the mesoplastron is replaced by a hinge, the pectoro-abdominal sulci run next to the hinge, the humero-pectoral furrows are farther anterior, being situated at the hind margin of the entoplastron, and the intergular and gular shields are smaller than in the morphological type. In the Pelomedusinae (fig. 14, C) the medial portions of the mesoplastra are reduced; the pectoro-abdominal scale lines lie Fig. 9. Ventral view of posterior portion of carapace of Podocnemis alabamae, C.N.H.M. P27372, showing extent of ilial attachment scar, sacral rib stubs on costal plates 8 and neurapophyseal rugosity along medial edges of costal plates 7 (cf. pi. 4, fig. 3). farther forward, laterally transgressing onto the mesoplastra; the intergular is large, more or less triangular; and the gular shields are relatively smaller than in the family type pattern. In the carapace the neural series is reduced to seven plates. Podocnemis, Pelomedusa, Stereogenys, and Taphrosphys can be morphologi- cally derived from the pelomedusine type pattern (figs. 14, C, and 15, A). Stereogenys can be derived from the pelomedusine pattern by diminution of the humeral scutes so that they become totally separated by the intergular, as in S. podocnemoides. A further diminution of the humeral shields leads to the condition in Stereogenys cromeri Andrews where the posterior point of the inter- gular partially separates the pectoral scutes. A decrease in size of all the anterior plastral shields leads in turn to S. libyca Andrews. In the carapace, the caudad displacement of neural plates 1 to 5 is combined with a very narrow first vertebral scale. Taphrosphys is a parallel form to Stereogenys. It can be thought of as derived from the pelomedusine pattern by general broadening of the shell, enlargement of the intergular and diminution of the gular shields. The humeral scales are separated by the intergular (fig. 15, F). The abdominal scute is narrow medially, but greatly expanded at the side, almost reaching from the axillary to the inguinal buttresses. In the carapace, the neural series is made up of seven plates, the first being suturally connected with the nuchal. The first vertebral 46 FIELDIANA: GEOLOGY MEMOIRS, VOLUME 3 shield is broad, as in the pelomedusine type (fig. 14, C). Within the genus Taphrosphys there appears to be a line represented by T. molops, in which the gular shields are enlarged to meet at the midline in front of the intergular. The scale pattern of the carapace in Amblypeza, an extremely broad form, apparently close to Taphrosphys, curiously enough includes a nuchal shield, a character not observed in any other pelomedusid. Two lines within the genus Podocnemis can be derived from the pelomedusine pattern (fig. 15, G-K and L). In one the intergular and the gular scales are considerably smaller and the former particu- FlG. 10. Outer view of left hyo- and mesoplastron of Podocnemis jilabamae, C.N.H.M. P27372. The pectoro-abdominal shield furrow is plainly visible. larly narrow (P. alabamae and P. brasiliensis) . The humero-pectoral sulci lie far forward about at mid-length of the entoplastron. The pectoro-abdominal furrows are either slightly advanced craniad and transgress onto the mesoplastra laterally (P. alabamae) or are considerably farther craniad on the hyoplastral plates (P. brasiliensis). The carapace includes seven neural plates in P. brasi- liensis, six in P. barberi, and six or seven in P. alabamae. In all three of these species the lateral furrows of the first vertebral scute extend to the second marginal scales. This line, in which Tertiary intermediates are unknown, seems to lead to the living South American species. In these forms the normal number of neural plates is seven (rarely six), and the first vertebral shield is relatively narrow, extending laterally only to the first marginal scutes. If belief in a pro- gressive reduction of the neurals in Pleurodira is correct, it appears that the Cretaceous North American forms are early side branches of the main line that have undergone reduction of the seventh neural plate. ZANGERL: VERTEBRATE FAUNA OF THE SELMA FORMATION 47 In the second group comprising the Old World forms, the intergular becomes shorter, no longer than the gular shields, the humeral scales make no contact with the intergular, and the pectoro-abdominal furrows lie anterior to the meso- plastral plates. The neural series is reduced to six (P. stromeri). The conditions observed in P. fajumensis, P. aegyptiaca, and P. madagascariensis result from further diminution of the intergular so that the gular shields meet in the midline. P. bramlyi Fourtau from the Miocene of Egypt is peculiar in that the scale Fig. 11. Visceral view of xiphiplastral end of plastron of Podocnemis alabamae, C.N.H.M. P27370, showing shape and placement of ischial and pubic scars. pattern on the anterior lobe of the plastron is essentially that of the New World representatives of the genus. Its geographic position and age, however, suggest (more complete knowledge of its organization is wanting) that it is derived from the P. stromeri pattern that paralleled the New World forms. It seems clear that the Old and New World lines of Podocnemis have been separate at least as far back as the late Cretaceous. The living genus Pelomedusa, represented in the Oligocene of Egypt by P. progaleata von Reinach, exhibits only slight modifications of the pelomedusine type. One of the outstanding features of this form is the thinness of the plastron, 48 FIELDIANA: GEOLOGY MEMOIRS, VOLUME 3 which led, as far back as the Oligocene, to the formation of a central fontanelle between hyo- and hypoplastra. The shield pattern of the plastron is essentially that of Podocnemis stromeri, except that the humero-pectoral furrows lie, towards their medial ends, far behind the entoplastral bone. The pectoro-abdominal sulci are exactly as in the pelomedusine pattern. Save for proportional dif- ferences, the carapace compares very well with that of Podocnemis stromeri. In the Recent species Pelomedusa galeata a tendency toward reduction of the first neural plate is apparent. Thus, morphologically, Pelomedusa stands closer to the stem-line of Podocnemis than do Stereogenys and Taphrosphys. Haughton (1928) has described a turtle of uncertain age from Nyasaland (see also Stromer, 1934) as Platycheloides nyasae (fig. 15, D). This form shows a great deal of similarity to Pelomedusa, and would undoubtedly have been referred to this genus were it not for the lack of a central fontanelle in the plastron. Such a vacuity, however, must be regarded as a secondary development; apparently this form represents the most primitive member of the Pelomedusa line known to date.1 In conclusion, it may be stated that, on the basis of evidence now at hand, the members of the family Pelomedusidae can be divided into two groups repre- sented by the Pelusiinae and the Pelomedusinae. The latter can be subdivided into four major branches: Stereogenys, Taphrosphys, Podocnemis, and Pelomedusa. Family PELOMEDUSIDAE Pleurodiran turtles with eleven plastral bones, mesoplastra being present. Neural plates always present, varying from six to eight. Ilium in intimate contact with carapace; ischia and pubes firmly attached to plastron. A bony temporal arch; no parieto-squamosal arch; palatine bones in contact; no nasals; pre- frontals in contact. Subfamily Pelusiinae Mesoplastra large, in contact at midline; anterior lobe of plastron movable, a hinge being present between hyo- and mesoplastra. Africa; Miocene to Recent. Genus Pelusios. Subfamily PELOMEDUSINAE Mesoplastra small, laterally located; anterior lobe of plastron not movable. Africa; Cretaceous (?) North and South America; Late Cretaceous. Congo; Paleocene. Egypt, England, India, New Jersey, Peru; Eocene. Egypt and Germany; Oligocene. Egypt, Malta; Miocene. Burma; Pliocene. Africa, Madagascar, South America; Recent. Genera: Podocnemis, Pelomedusa, Platycheloides, Stereogenys, Taphrosphys. Tentatively referred genera: Amblypeza, Dacochelys, Naiadochelys, Shwe- boemys. 1 The time of the segregation of the genera Pelomedusa and Podocnemis is not known but Podocnemis s.s. was already quite widely distributed in the late Cretaceous. The con- siderable degree of specialization of the genera Stereogenys and Taphrosphys in the Eocene suggests that the separation of all of these forms from the pelomedusine stem-line appears to have taken place before the end of the Cretaceous. Fig. 12. Lateral view of right ischium and pubis of Podocnemis alabamae, C.N.H.M. P27370. The acetabular cavity is badly crushed. FIG. 13. Right humerus of Podocnemis alabamae, C.N.H.M. P27405. A, posterior view; B, dorsal view; C, ventral view. 49 Fig. 14. A, morphological type pattern of pelomedusid shell; B, pelusiine type pattern; C, pelomedusine type pattern. The conditions illustrated in B and C can be morphologically derived from A. e= epiplastron ent= entoplastron hyo= hyoplastron m= mesoplastron hypo= hypoplastron x= xiphiplastron Sp#g=suprapygal 1, 2, etc.=neurals i = intergular shield G=gular shield H= humeral shield P= pectoral shield A= abdominal shield F= femoral shield An= anal shield h= hinge between mesoplastron and hyo- plastron 50 FlG. 15. Illustration of morphological affinities among pelomedusine turtles. A. Pelomedusine type pattern. B. Stereogenys podocnemoides, and C. Stereogenys cromeri, representing the Ste- reogenys branch. D. Platycheloides nyasae, interpreted as a primitive member of the Pelomedusa branch. E. Podocnemis alabamae, illustrating the Podocnemis branch. F. Taphrosphys olssoni, representing the Taphrosphys branch. G. Podocnemis stromeri. H. Podocnemis fajumensis. I. Podocnemis aegyptiaca. K. Podocnemis madagascariensis. L. Pattern of living South American species of Podocnemis. 51 52 FIELDIANA: GEOLOGY MEMOIRS, VOLUME 3 The morphological affinities (fig. 16) among the pelomedusid turtles, as out- lined above, are worked out on a sufficiently general basis so that major changes are not likely to be necessary in the future. More detailed relationships within the Podocnemis line cannot yet be accurately analyzed, because many forms are insufficiently known. This is true of the osteology of almost all Recent species, of Podocnemis lata, of P. dehmi, of P. indica, of Bantuchelys congolensis ( = Podoc- nemis? congolensis Dollo), and of Podocnemis harrisi. Podocnemis bowerbanki and Dacochelys delabechei are relatively well-preserved specimens, but some of the features considered in the discussion above are unknown in both forms. It may be pointed out that the interpretation of the mesoplastra as primitive elements of the turtle shell is at variance with Hay's (1908) idea that they, together with the peripheral plates, had been secondarily acquired by the Amphichelydia, Cryptodira, and Pleurodira. The early Mesozoic turtles have two pairs of mesoplastra, one of which disappears in the later Mesozoic forms. Among the latter, the mesoplastral elements may be retained or they may become partially or completely reduced, and without a doubt this process took place in different groups independently, for example, in the Cryptodira, in the Chelyidae, and in the Baenidae. Other cases of parallel modifications in the structural organization of the pelomedusids can readily be observed. Taphrosphys and Stereogenys resemble each other in the scale arrangement on the anterior plastral lobe. In an occasional specimen of the South American Podocnemis dumeriliana Schweigger the shield arrangement on the anterior lobe of the plastron duplicates that of P. madagascariensis (Siebenrock, 1902 and 1903), indicating a trend in the former species that, if permanently established, would constitute a parallel to the Old World line. The implied assumption of a closer genetic relationship of P. dumeriliana and the Madagascan species (Siebenrock, 1902; Miiller, 1935) on the basis of this character and of some minor (and variable) features of the head, can, in view of the fossil record, scarcely be upheld. Yet another case of obvious parallelism can be observed in the numerical reduction of the neural series (see discussion above). In the light of these observations it becomes increasingly necessary, particu- larly for the student of turtle morphology, to base his conclusions concerning structural affinities on as many anatomical features as are available for com- parison, a challenge in line with a demand recently made by several morphologists who see the organism in its entirety — i.e. structure, physiology and behavior — instead of regarding it as a mosaic of more or less independent "characters." SUMMARY 1. A new species of Podocnemis (P. alabamae) from the late Cretaceous of Alabama is described. It is closely related to Podocnemis barberi from deposits of approximately the same age in Arkansas. 2. The basic pattern of the pelomedusid shell can be derived from the morphological type of the Pleurosternidae. Mesoplastra are regarded as primi- tive elements of the turtle shell. >/> .o <5 « M 1- z 1 1 i l UJ 1 6 *» Q o UJ oe ! 1 «. o,- -^ "^ UJ 1 1 1 V) z 1 UJ .ft o c o ^ _1 ft 0. 1 i UJ z J 1 " UJ £ ■» 5 o 1 1 # X <=> 1 ^ S -1 Q 1 o _l 2 ■o 5 g QC ■5 o O ae o I ft; 1 ft; CV. 5 1 , 1 5 Ilia $ i UJ i 1 i -i Ul 1 z 5 o, O z Ul o o P * \ s o _l o \ ' ! \ i 1 \ i % UJ z ' 1 V Ul > <5 *» r. •C^ o o Ul 5 5 1 1 ? 1 1 I ! ■ ' 1 '. / i PALEOCENE * ' 1 \ 1 1 ' , V / i 1 \ \ % \ / Pbrosiliensis P. barber/ / / i \ \ o \ P. a/a bam ae i \ \ % \ i ID \ \ \ x \ v \ •<, Platycheloides g Jf1 O UJ o fS UJ \ »v J? '* IE o \ \ \ ,' X V v / Pelusune pattern Pelomedusine pattern Pelomedusid type pattern 3 a] a 0) O. a •s o a Xi be C O s a m a> o '5b _o o J3 a bo c o JS £ d fa 53 54 FIELDIANA: GEOLOGY MEMOIRS, VOLUME 3 3. An analysis of the morphology of the shell in the pelomedusid turtles results in the recognition of two principal groups within the family Pelomedusi- dae: the Pelusiinae and the Pelomedusinae. Stereogenys, Pelomedusa, Podoc- nemis, and Taphrosphys are included in the pelomedusine group. There is a closer morphological affinity between Pelomedusa and Podocnemis than between either of these forms and Stereogenys or Taphrosphys. 4. Two lines can be recognized within the genus Podocnemis. One, well documented in the Tertiary of Egypt, leads to the living Madagascan species; the other, known from the American Cretaceous but lacking Tertiary intermedi- ates to date, leads to the Recent South American species. 5. Platycheloides nyasae Haughton of uncertain age, from Nyasaland, is interpreted as a primitive relative of Pelomedusa. 6. Podocnemis olssoni from Peru belongs to the genus Taphrosphys, hitherto known only from the Green Sand beds of New Jersey. REFERENCES Andrews, C. W. 1900. On a new species of chelonian (Podocnemis aegyptiaca) from the lower Miocene of Egypt. Geol. Mag., Dec. 4, 7, pp. 1-2, pi. 1. 1901. Preliminary note on some recently discovered extinct vertebrates from Egypt. Part II. Geol. Mag., Dec. 4, 8, pp. 440-444, 4 figs. 1903. On some pleurodiran chelonians from the Eocene of the Fayum, Egypt. Ann. Mag. Nat. Hist., (7), 11, pp. 115-122, pis. 7, 8. 1906. A descriptive catalog of the Tertiary Vertebrata of the Fayum, Egypt. British Museum (Natural History), London, xxxvii+319 pp., 98 figs., 26 pis. Bergounioux, F. M. 1932. Cheloniens fossiles conserves au Museum d'Histoire Naturelle de Munich. Bull. Soc. Hist. Nat. Toulouse, 64, pp. 523-544, figs. 1-5, pis. 65-66. BOULENGER, G. A. 1889. Catalog of the chelonians, rhynchocephalians and crocodiles in the British Museum (Natural History), London, x+311 pp., 73 figs., 6 pis. Dacque, Edgar 1912. Die fossilen Schildkroten Aegyptens. Geol. Palaeont. Abhandl., (n.s.), 10, No. 4, pp. 275-337, figs. 1-23, pis. 1-2. Dollo, Louis 1913. Podocnemis congolensis tortue fluviatile nouvelle du Montien (Pal^ocene inf&- rieure) du Congo et 1'eVolution des cheloniens fluviatiles. Ann. Mus. Congo Beige, Bruxelles, (3), 1, pp. 47-65, pi. 7. 1924. Bantuchelys, genre nouveau de tortues, decouvert dans le Pal^ocene du Congo. Bull. Acad. Roy. Belgique, (5), 10, pp. 613-626, pi. 1. Fourtau, R. 1920. Contribution a l'&ude des vertebras miocenes de l'Egypt. Cairo (Government Press), pp. 29-34, figs. 21-23. Haughton, S. H. 1928. On some reptilian remains from the dinosaur beds of Nyasaland. Trans. Roy. Soc. South Africa, 16, pp. 67-75, figs. 1-2, pis. 2-5. Hay, O. P. 1908. The fossil turtles of North America. Pub. Carnegie Inst. Wash., 75, iv+568 pp., 704 figs., 113 pis. Lydekker, R. A. 1887. Eocene chelonians from the Salt-Range. Palaeont. Ind., Calcutta, (10), 4, pp. 59-65, pi. 2. 1889. Catalog of the fossil Reptilia and Amphibia in the British Museum (Natural History). Part III: Chelonia. London, 3, xviii+235 pp., 53 figs. Lydekker, R. A., and Boulenger, G. A. 1887. Notes on Chelonia from the Purbeck, Wealden and London-Clay. Geol. Mag., (3), 4, pp. 270-275. 55 56 FIELDIANA: GEOLOGY MEMOIRS, VOLUME 3 MUller, Lorenz 1935. Ueber eine neue Podocnemis-Art (P. vogli) aus Venezuela nebst erganzenden Bemerkungen iiber die systematischen Merkmale der ihr nachstverwandten Arten. Zool. Anz., 110, pp. 97-109, figs. 1-3. Owen, Richard 1850. Monograph on the fossil Reptilia of the London Clay. Part I: Chelonia, Supplement No. 1. Palaeontogr. Soc, 6, pp. 1-4, pi. 29. Owen, Richard and Bell 1849. Monograph on the fossil Reptilia of the London Clay. Part I: Chelonia. Palaeontogr. Soc, 6, pp. 1-76, figs. 1-4, pis. 1-28. Pacheco Joviano, A. d'Amaral 1913. Notas sobre a geologia do valle do rio Grande a partir da foz do rio Parto at£ a sua confluencia com o rio Parahyba. In JOSE DOS DOURADOS, Exploracao do rio Grande e seus affluentes. Sao Paulo, Comm. Geogr. Geol. Sao Paulo, pp. 33-38, pis. 1-4. von Reinach, Albert 1903. Schildkrotenreste aus dem agyptischen Tertiar. Abh. Senckenberg Naturf. Ges., 29, pp. 1-64, pis. 1-17. Ristori, G. 1894. Di un nuovo chelonio fossile nel Miocene dell'isola di Malta. Att. della Soc. Toscana Sci. Nat., Mem., 14, pp. 1-17, pi. 1. Schmidt, K. P. 1931. A fossil turtle from Peru. Field Mus. Nat. Hist., Geol. Ser., 4, No. 8, pp. 251-254, pis. 46-47. 1940. A new turtle of the genus Podocnemis from the Cretaceous of Arkansas. Field Mus. Nat. Hist., Geol. Ser., 8, No. 1, pp. 1-12, figs. 1-5. Siebenrock, Friedrich 1902. Zur Systematik der Schildkroten-Gattung Podocnemis Wagl. Sitz.-ber. K. Akad. Wiss. Wien, Math.-Nat., Ill, pp. 1-14, pi. 1. 1903. Schildkroten von Madagaskar und Aldabra. Abh. Senckenberg Naturf. Ges., 27, pp. 241-259, pis. 33-35. 1909. Synopsis der recenten Schildkroten, mit Beriicksichtigung der in historischer Zeit ausgestorbenen Arten. Zool. Jahrb. Suppl., 10, pp. 427-618. Staesche, K. 1929. Schildkrotenreste aus der oberen Kreide Patagoniens. Palaeontographica, 72, pp. 103-123, figs. 1-17, pis. 15-18. 1937. Podocnemis brasiliensis n. sp. aus der oberen Kreide Brasiliens. Neues Jahrb. Min. Geol., Palaeont, (B), Beilage, 77, pp. 291-309, figs. 1-2, pis. 16-20. Stromer, Ernst 1934. Ergebnisse der Forschungsreisen Prof. E. Stromer's in den Wiisten Aegyptens. II. Wirbeltiere der Baharije-Stufe (unterstes Cenoman) No. 14. Testudinata. Abh. Bayer. Akad. Wiss. Math.-Nat., 25, pp. 1-26, pi. 1. Swinton, W. E. 1939. A new fresh-water tortoise from Burma. Records Geol. Surv. India, 74, pp. 548-551, figs. 1-2. Zangerl, Rainer 1947. Redescription of Taphrosphys olssoni, a fossil turtle from Peru. Fieldiana, Geol., 10, pp. 29-40, figs. 10-13. EXPLANATION OF PLATE 4 Fig. 1. — Carapace of Podocnemis alabamae, C.N.H.M. P27370 (holotype). FlG. 2. — Visceral view of plastron of Podocnemis alabamae, C.N.H.M. P27369; m= mesoplastron. Fig. 3. — X-ray photograph of posterior (sacral) part of shell of Podocnemis expansa, C.N.H.M. 16075, for comparison with text figures 6 and 9. The similarity in the arrange- ment of the sacral ribs between this species and P. alabamae is evident. The last sacral vertebra and its ribs are broken off. On the left side of the plate the sacral ribs are outlined in ink, the last one reconstructed. The distal end of the fourth sacral rib is visible on the right side of the plate. Fieldiana: Geology Memoirs, Volume 3 Plate 4 ^oUN°«> M